text
stringlengths 0
1.19M
|
|---|
= 0.24515 while the one given by the exact PLS criteria is mpl, |
n, |
mz, |
y |
= 0.04726, which means that the value obtained with the EC3 criterion is 5.2 times larger than the exact solution. The difference between these two results for mpl, |
n, |
mz, |
y corresponds to 17% of the reduced plastic moment mpl, |
y show that the results given by the EC3 criterion for biaxial bending with axial loading may present considerable differences from the exact solutions obtained by the semi-analytical method described in Nevertheless, the main cause for these differences, between the exact solutions and the EC3 plastic criterion, is not entirely in Eqs. themselves, but on the evaluation, by means of Eqs. , of the resisting plastic moments Mpl, |
n, |
y and Mpl, |
n, |
z considered in Eqs. , are based on some approximations proposed in Ref. , may lead to significant differences between their solutions and the corresponding exact values of Mpl, |
n, |
y and Mpl, |
n, |
z which, being at both the denominators of Eqs. , affect the results given by these equations.A possible solution to overcome this issue consists on replacing the use of Eqs. in the evaluation of the reduced values of Mpl, |
n, |
y and Mpl, |
n, |
z. The results given by the modified criterion, based on Eqs. , proposed in this work, are compared in with the exact solutions obtained by the semi-analytical method described in , for the HD 400 × 1086 and HE 1000 × 584 sections, respectively.These figures show a much better agreement between the results obtained by these two methods; those obtained with the modified EC3 criterion are usually slightly conservative, although the differences from its results to the exact solutions may increase when the value of n is smaller than αb/kA.On the other hand, in few cases, the modified EC3 criterion may lead to a slight overestimation of the cross-section resisting internal forces, as shown in . Yet, the differences are, in this case, much smaller than those shown on and, as far as the studies carried out for other different sections have shown, this occurrence will be relatively rare.In general, the results given by this modified EC3 criterion, based on the use of Eqs. , seem to be in better agreement with the exact solutions than those obtained with Eqs. ; therefore, this modified version might be considered as a possible alternative for the plastic interaction criterion adopted on the current version of the Eurocode 3 This paper presents a set of non-dimensional criteria for the analysis of steel I-sections subjected to axial forces and biaxial bending moments, at the elastic or the plastic limit states (as long as buckling phenomena are not involved). Both the elastic and the plastic criteria are expressed in a set of reduced (normalised) variables, which are independent from the steel yield strength and from the cross-section dimensions, and, therefore, from the system of units in which they are expressed.In a first step, the basic principles of the analytical model, followed on the formulation of these interaction criteria, are described, and the cross-section main geometrical and mechanical characteristics are identified. Then, the reduced variables used in the non dimensional form of the interaction criteria are defined.Afterwards, the analytical expressions for the elastic limit state and the plastic limit state criteria are presented. In order to simplify, in practice, the evaluation of the resisting cross-section internal forces at its plastic limit state, the plastic interaction criteria are presented, in a first step, for some particular combinations of those internal forces, such as axial loading with bending about a main axis, and biaxial bending without axial loading.In these cases, the plastic criteria are given by exact equations (within the frame of the hypotheses adopted in this study). They are all compared with the corresponding plastic criteria adopted in the Eurocode 3 (EC3).This comparison shows that the EC3 results are usually conservative in the case of I-sections subjected to biaxial bending without axial loading. The results given by the EC3 in the case of axial loading with bending about the I-section strong axis are usually conservative, except for small values of the axial force where the EC3 overestimates the I-section plastic resistance, due to the linearization of the simplified criterion adopted for this loading scenario. In the case of axial loading with bending about the weak axis, the EC3 overestimates the I-section plastic resistance.Finally this paper deals with the general case of biaxial bending with axial loading.First, a semi-analytical method, based on a mixed analytical and numerical integration procedure is presented; this method is applied to obtain exact solutions for the problem, which are used as reference values for the appraisal of the results obtained by other means. Then, a simplified global criterion is proposed for the simultaneous combination of an axial force and bending moments about both the main axes of inertia.This simplified criterion and the corresponding plastic criterion adopted in the EC 3 are compared with the exact solutions; this comparison shows that this new simplified criterion usually leads to results closer to the exact solutions than the EC3 criterion.Finally, some suggestions are presented to improve the results given by the EC3. A modified EC 3 criterion is proposed, resulting from the use of the equations proposed by the author for the cases of axial loading with bending about one of the I-section main axes, as an alternative to the simplified equations adopted by the EC3 for this purpose.The comparison of the results given by this modified EC 3 criterion with the exact solutions show a better agreement than the one found between the actual EC 3 criterion and the same exact solutions; therefore, this modified version might be considered, in a future revision of the Eurocode 3, as a possible alternative for the plastic interaction criterion adopted in its current version Effect of sp2/sp3 bonding ratio and nitrogen content on friction properties of hydrogen-free DLC coatingsHydrogen-free diamond-like carbon (DLC) coatings display markedly lower friction coefficients under a lubricated condition owing to the adsorption of oiliness agents and their decomposed components. This study examined the effect on friction properties of the sp2/sp3 bonding ratio of DLC coatings and the addition of nitrogen for improving surface wettability. Evaluations of friction properties under a lubricated condition showed that DLC coatings with a lower ratio of sp3 bonds and nitrogenated DLC coatings displayed higher friction coefficients. Results obtained by electron spin resonance revealed that adding nitrogen decreased the number of dangling bonds of DLC coatings in proportion to the sp3 bond ratio. Assuming that such dangling bonds are oiliness agent adsorption sites, it is concluded that adding nitrogen is not conducive to reducing friction coefficients.► DLC coatings with different sp3 ratio and different nitrogen content were prepared. ► Their friction properties were evaluated under a lubricated condition. ► Low sp3 ratio or high nitrogen content of DLC causes high friction coefficients. ► The addition of nitrogen on DLC reduces sp3 ratio and number of dangling bonds. ► Higher friction arises by less number of dangling bonds as adsorption sites on DLC.Diamond-like carbon (DLC) coatings display a low friction coefficient under non-lubricated conditions, similar to molybdenum disulfide (MoS2) and graphite, and also have outstanding wear resistance, which is an issue with solid lubricants. Additionally, they also exhibit excellent anti-seizure properties owing to their surface hardness, which greatly surpasses that obtained in the heat treatment and plating treatment processes traditionally applied to steel. Research activities have been proceeding to develop DLC coatings for protective use in many different fields. On the other hand, one drawback of DLC coatings has been their low level of adhesion strength to the substrate due to high internal stress, which is an issue specific to hard thin films. As one way of addressing this issue, metal elements have been added cyclically in an effort to form laminated bonds for the purpose of relieving internal stress All types of DLC coatings display remarkably low friction coefficients equal to those of solid lubricants under non-lubricated conditions, but their friction properties in the presence of engine oil or other types of lubricating oil are not necessarily as admirable. Research results have shown that it is important to select DLC coating specifications that suitably match the additive package of the lubricating oil Based on this mechanism, it should be possible to reduce friction coefficients even further by changing the coating specifications to enhance adsorption of an oiliness agent. One factor that can be cited for changing the coating specifications is the bonding state of the carbon atoms composing the coating; for instance, the ratio between the sp3 bonds that bind diamond crystals and the sp2 bonds that bind graphite crystals. By varying the coating deposition conditions, the surface hardness of DLC coatings can be controlled over a relatively wide range, which is attributable to changes in the sp2/sp3 carbon bonding ratio. However, there are no examples of studies in the literature that have specifically investigated the influence of the carbon bonding ratio on the friction coefficient under lubricated conditions, especially in the presence of an oiliness agent. Another factor that can be cited is the addition of various elements to DLC coatings, which can be done via different methods. Metal elements can be mixed into coatings by sputtering them in parallel during the deposition process or by introducing them as a gas into the ambient atmosphere during coating deposition. With the CVD process, the elements desired to be added can be included in the gases that are used.Among the reports concerning the addition of elements besides carbon, there is a study by Grischke et al. Therefore, this study examined two factors that might possibly reduce the friction coefficient further, using hydrogen-free DLC coatings that show markedly low friction coefficients under lubricated conditions, including the presence of an oiliness agent. One factor investigated was the sp2/sp3 bonding ratio indicating the bonding state of carbon atoms on the DLC coating surface. The other factor was the addition of nitrogen, which has been reported to be effective in improving surface wettability presents the specifications of the DLC coatings used in the friction tests. Specifically, it shows the coating deposition conditions, including the nitrogen partial pressure and substrate temperature, as well as the substrate materials and coating properties such as the nitrogen content, density, sp3 bonding ratio and surface hardness as found by a nanoindentation technique. DLC coatings in which the sp3 bonding ratio was varied by changing the nitrogen content and substrate temperature were deposited by laser-induced pulsed arc deposition using graphite as the vaporization source material The nitrogen content of the DLC coatings prepared by laser-induced pulsed arc deposition was controlled by varying the nitrogen gas pressure in the furnace from 0.001 Pa to 0.3 Pa. The nitrogen content of the DLC coatings was calculated using the relationship with the nitrogen partial pressure proposed by Wienss et al. Arc deposition processes generally produce DLC coatings with a high sp3 bonding ratio and coatings deposited with such methods are classified as ta-C that two of the DLC coatings evaluated in this study had sp3 bonding ratios of 20% or lower. They were both classified as ta-C:N on the basis of the deposition method used. As indicated in the table, nitrogenated coatings are designated as ta-C:N and non-nitrogenated ones as ta-C. The coating deposition method is indicated after the coating type as I for laser-induced pulsed arc deposition, as II for DC arc deposition, and as III for HIPIMS. These designations are used in the following discussion to indicate the different types of DLC coatings examined.The substrates of the ta-C:N I, ta-C I and ta-C II coatings for use in the friction tests were made of SCM420 steel (equivalent to BS 708M20) and were disk-shaped, measuring 31 mm in diameter by 2 mm in thickness. The steel was first carburized to a surface hardness of 60 HRC and then lapped to a surface roughness of Ra 0.01 μm to prepare the base surface for coating deposition. One of the DLC coatings prepared by DC arc deposition and those prepared by sputtering were deposited on Si wafers to facilitate measurement of the number of dangling bonds (unpaired electrons) in the coatings by electron spin resonance (ESR).The mating pins used in the friction tests were made of bearing steel (JIS SUJ2, equivalent to AISI 52100 or DIN 100Cr6) with a surface roughness of Ra 0.05 μm and measured 5 mm in diameter and 5 mm in length.The lubricating oil used in the friction tests was a PAO-based prototype oil with dynamic viscosity of 18.0 mm2/s at 40 °C and 4.0 mm2/s at 100 °C. It contained 1 wt% of the GMO additive that displays markedly low friction properties in combination with hydrogen-free DLC coatings. The reason for selecting this oil is explained here.The purpose of this study was to research and develop an ultralow friction coating for application to automotive parts. Therefore, it was desirable to use a lubricating oil formulation close to actual engine oil in the coating evaluations. PAO is a single-component synthetic oil that is used as the base oil of engine oils with outstanding viscous properties. Because its viscosity undergoes little change relative to temperature and it possesses thermal oxidative stability, PAO is suitable as a base oil for use in friction assessments. As the additive, GMO was selected which is used as a friction modifier in engine oil. Numerous analyses conducted to date The friction coefficients of the DLC coatings were evaluated in pin-on-disk friction tests conducted with the test specimens and test apparatus shown in . Three uncoated steel pins were in sliding line contact with a disk specimen coated with a DLC coating. The pins were secured with jigs to create a condition of complete sliding contact with the disk. The pins and disk were submerged in an oil bath and the disk was rotated while an equal normal load was applied to the pins. The reaction force generated in the jigs holding the pins was measured during the sliding contact, and the friction coefficient was calculated from the measured results.The temperature of the lubricating oil was kept constant at 80 °C, and a combined load of 50 kgf was applied to the three pins, which was equivalent to Herztian contact pressure of approximately 0.7 GPa. The sliding speed was set at 0.03 m/s and a sliding friction test was conducted for 1 h. A comparison was made of the mean friction coefficient during the last 5 min of the test period when the friction coefficient was approximately in a steady-state condition.A laser-based surface acoustic wave (SAW) method was used to measure Young's modulus and the atomic density of the DLC coatings deposited by laser-induced pulsed arc deposition In order to estimate the surface free energy of the coatings, the contact angle was measured in dripping tests using two types of liquids, namely, purified water and methylene iodide (CH2I2). After first washing the DLC coating surface in methanol, approximately 2 μL of water or 1 μL of methylene iodide was dripped on the coating surface and the contact angle was measured after 1 min of dripping. Based on the contact angles measured for both liquids, the values of the polar and dispersive components of the surface free energy were calculated for each coating. The contact angle was measured with an FTA188 contact angle and surface tension analyzer (First Ten Angstroms).The sp3 ratios of ta-C I # 5, 6, 7 and 8 were estimated from data in the references, and those of the other coatings were estimated from the results of an EELS analysis. Cross sections of the DLC coatings were extracted by a microsampling technique using a focused ion beam (FIB) system, and a bonding analysis was performed with a scanning transmission electron microscope (STEM). The device used was a field emission TEM (JEM2100F, JEOL, Ltd.) fitted with a Gatan GIF Tridiem energy filter having energy resolution of approximately 1 eV full width at half maximum (FWHM). The measurement conditions were an accelerating voltage of 200 kV, specimen absorption current of 10−9 |
A, and a beam spot diameter of 1.0 nm. EELS spectra were acquired by STEM spectrum imaging The sp3 ratio was calculated based on its relationship with the plasmon peak as proposed by Camps et al. The DLC coating hardness were calculated from measurements obtained with a nano-indentation hardness tester (ENT-1100a nano-indenter, Elionix Inc.) that was driven into the specimen at a load of 2.94×10−3 |
N.The surface properties of ta-C:N I #4 prepared by laser induced pulsed arc deposition and ta-C:N III #12 deposited by sputtering were investigated by using X-ray photoelectron spectroscopy (XPS) to analyze the surface layer elements and their bonding state. The XPS analysis was performed with a PHI Quantum 2000 Scanning XPS Microprobe, and the measurement conditions were the use of a monochromated Al Kα source, 1486.6 eV, 40 W and a photoelectron take-off angle of 200 μm. As a pre-test treatment condition, the top layer was removed to a depth of several nm by argon-ion sputtering.Additionally, for ta-C:N I #4, the distribution of elements containing hydrogen was analyzed in the depth direction by Rutherford backscattering spectrometry (RBS) and hydrogen forward scattering spectrometry (HFS). A Model 3SDH Pelletron tandem accelerator (National Electrostatics Corporation) was used in the analysis. The analysis conditions were injection of a 4He++ ion beam, injection energy of 2.3 MeV, beam diameter of 2 mm, RBS normal detection angle of 160° and HFS grazing angle of 30°.An ESR analysis was performed to measure the number of dangling bonds (unpaired electrons) in the DLC coatings deposited on Si wafers, inasmuch as such measurements cannot be made for coatings on steel substrates. Measurements were made for ta-C II #10 deposited by DC arc deposition and ta-C III #11 and ta-C:N III #12, both of which were deposited by sputtering. A Bruker ESP350E spectrometer was used in the analysis. The number of dangling bonds was measured under conditions of a measurement temperature of 10 K, central magnetic field of 3377 G, magnetic field sweep span of 200 G, modulation of 100 kHz, 3 G, microwave wavelength of 9.46 GHz, 0.25 mW, sweep time of 83.89 s×4 times, and time constant of 163.84 ms. The number of dangling bonds per unit volume was found by first calculating the volume of a DLC coating from its surface area and depth. The number of dangling bonds per unit mass was also calculated using the atomic density of each coating. The atomic densities of the DLC coatings in the ESR analysis were estimated from the results of an X-ray reflectometry (XRR) analysis conducted with a PANalytical X’Pert PRO MRD system. The measurement conditions were voltage of 45 V, current of 40 mA, and X-ray wavelength of CuKα1, and Ge (220) dispersive crystals were used in the spectrometer. Monochromatic X-rays were obtained and the divergence angle was suppressed by using double-diffracted Ge (220) profiles.The influence of the different coating deposition conditions on the sp3 bonding ratio was investigated using the ta-C:N I coatings having different nitrogen partial pressures and the ta-C I coatings the substrate temperature of which was varied during the deposition process. The influence on the coating density measured by the SAW method , the sp3 bonding ratio of the coatings increased with decreasing deposition temperature. In contrast, the sp3 bonding ratio markedly declined when the nitrogen partial pressure in the furnace was increased while keeping the deposition temperature constant. shows the atomic density of the ta-C:N I and ta-C I coatings along the vertical axis in relation to the sp3 bonding ratio along the horizontal axis. The results indicate that the coating atomic density tended to increase in proportion to the sp3 bonding ratio, which decreased as the nitrogen partial pressure was increased. The fact that the results are plotted along nearly the same line indicates that the coating atomic density tended to be strongly dependent on the sp3 bonding ratio regardless of the nitrogen content. shows Young's modulus of the ta-C:N I and ta-C I coatings along the vertical axis in relation to the sp3 bonding ratio along the horizontal axis. Like the atomic density, Young's modulus of all the coatings showed a strong correlation with the sp3 bonding ratio. The tendency for Young's modulus to decline with an increasing nitrogen content of DLC coatings is consistent with the findings reported by Wines et al. shows the friction test results for ta-C:N I (denoted by ), for ta-C I (denoted by □), and for ta-C II (denoted by ○). The friction coefficients are plotted as a function of the sp3 bonding ratio along the horizontal axis. Two tests (n=2) were conducted on ta-C:N I and ta-C I and one test (n=1) on ta-C II. The friction coefficient of ta-C:N I tended to increase with decreasing sp3 bonding ratio (with increasing nitrogen content), contrary to initial expectations, though the values were widely scattered. The friction coefficients of ta-C:N I were clearly higher than the value of the nitrogen-free ta-C II.In case of ta-C I, the friction coefficients tended to increase with a lower sp3 bonding ratio, similar to the results of ta-C:N I, though again the values were widely scattered. Even when ta-C II #9 (○) with an sp3 bonding ratio of approximately 80% is included in the comparison, the download tendency seen for the friction coefficient relative to the sp3 bonding ratio does not change.While the friction coefficients of the DLC coatings deposited with these deposition methods showed the overall tendencies described above, an examination of the individual friction coefficients revealed the presence of several problematic points. For example, the friction coefficients of coating samples in the same group differed considerably, and ta-C I #6 displayed a friction coefficient much lower than the values of coatings having roughly similar sp3 bonding ratios. The surface roughness of the pins and the DLC coating before and after the friction test did not show any definite correlation with the friction coefficient. As a result of visually examining the sliding marks on the DLC coatings after the friction test, signs of rather strong contact were found on the inner diameter side of the pins at both locations tested, especially for ta-C I #6. In cases where a central load is applied on the inner diameter side, the friction coefficient has a low apparent value because friction force is measured from the resultant moment. Related factors here include warping due to internal stress following coating deposition and the stiffness or misalignment of the jigs supporting the pins and imparting force to them uniformly. Unfortunately, the influence of these factors could not be isolated, identified and verified quantitatively in this study.With the friction test conditions used in this study, the total test time and the interval in which the friction coefficient was in a steady-state condition (i.e., the last five minutes of the 60-min test) were too short as wear conditions. Consequently, none of the coating samples after the test showed a significant bathtub wear curve, which precluded a comparison of the anti-wear performance of the coatings.The surface free energy of the DLC coatings was then investigated in order to examine the friction coefficient results in more detail. shows the water contact angles measured for all the coatings: ta-C:N I, ta-C I, ta-C II, ta-C III and ta-C:N III as well as the surface free energy calculated from the contact angles found for distilled water and methylene iodide. Surface free energy is shown as the polar component and dispersive component. The test results obtained for the ta-C:N I and ta-C:N III coatings differed substantially from the results reported by Grischke et al. that the water contact angle of the ta-C:N I coatings did not decrease with a decreasing sp3 bonding ratio (i.e. increasing nitrogen content) but rather increased between 55 to 52% sp3 (i.e. 0.005–0.01 Pa) and then leveled off after that. Additionally, the surface free energy did not show any distinct change as the nitrogen content was increased. The surface free energy of ta-C I #6 was approximately 45 mn/m at the same deposition temperature of 117 °C. Since the values in are at approximately the same level, it is inferred that the addition of nitrogen had virtually no effect on surface free energy. The ta-C II coatings showed a water contact angle of 70° The biggest difference in comparison with the results reported by Grischke et al. One example of a report concerning hydrogen-free nitrogenated DLC coatings is the work done by Tetard et al. also shows the measured water contact angle and calculated surface free energy for the ta-C I coatings and whose sp3 bonding ratios differed as a result of varying the substrate temperature during coating deposition. The water contact angle and surface free energy did not show any distinct tendency to change relative to the increase in the sp3 bonding ratio (i.e. decrease in the substrate temperature during deposition). It is observed that the dispersive component of the surface free energy tended to decrease slightly. A comparison with the results for the nitrogenated DLC coatings indicates that the polar component's share of the total energy increased slightly, but no clear relationship was seen with the deposition temperature. Moreover, compared with the results reported by Grischke et al. The surface of ta-C:N I #4 with nitrogen doping at a nitrogen partial pressure of 0.3 Pa was subjected to an XPS analysis to analyze the elemental composition and the bonding states of carbon and nitrogen. This analysis was performed in order to examine in more detail the surface free energy results obtained in the dripping tests. Since the results obtained by XPS analysis might be affected by surface contamination, the same analysis was performed on ta-C:N III #12 and the results for the two coatings were compared to ensure the reliability of the analysis. shows the percentages of carbon, nitrogen and oxygen found in the coating surface composition. ta-C:N I #4 showed a nitrogen content of 12.5 at%, which was close to the value of 14 at% that was estimated from the literature show the results of a C1s analysis and the absorption peak separation results for ta-C:N I #4 and ta-C:N III #12, respectively, and show the corresponding results of an N1s analysis and peak separation results.In the C1s spectra, a search was made for similar absorption peaks in reference to the peak values of the bonds reported by Shi , respectively, and the values are given as a footnote in the figures. The mole ratios (mol%) were calculated as the product of the peak area ratios and the element percentages shown for the types of carbon bondings in the tables in , respectively, and the values are given as a footnote in the figures. The mole ratios indicated that 51% were C1: C–C (sp2) bonds and 21% were C2: C–C (sp3) bonds, resulting in an sp3 bonding ratio, i.e., sp3/(sp2+sp3), of 0.29. The same calculation performed for C and N bonds yielded an sp3 bonding ratio of 0.231, indicating a tendency for a slight increase in sp2 bonds in the case of C and N bonds. The same tendency was observed for ta-C:N III #12 deposited by sputtering. The sp3 bonding ratio for C atoms was 0.313, whereas that for the bonding of C and N atoms was 0.273, indicating a tendency for sp2 bonds to increase.The same operation was then performed in the N1s spectra. A search was made for similar absorption peaks in reference to the sp2 bonds and sp3 bonds of N and C reported by Shi The –C–NH2 bonds pointed out by Tabbal et al. The –C–NH2 bonds discussed by Tabbal et al. Hydrogen was not used in the coating deposition process in this study, but hydrogen might have entered the coatings due to diffusion from the ambient air following deposition. Therefore, HFS/RBS analyses were conducted on ta-C:N I #4 to investigate the distribution of elements, including hydrogen, in the depth direction of the coating. presents the results of the component analysis in the coating depth direction. As a general rule for the coating depth (horizontal axis), 1015 |
atoms/cm2 corresponds to approximately 0.1 nm. Hydrogen content of 0.6 at% was observed in the very uppermost layer, but it decreased rapidly in the depth direction to a value considerably below 0.05 at% and subsequently become unmeasurable. As was expected, hydrogen was present in the uppermost layer as an adsorbed layer of contaminants and tended to decline steadily from the surface to a depth of 1000×1015 |
atoms/cm2. This implies that hydrogen was not introduced during coating deposition, but rather diffused from the surface afterward. The foregoing results suggest that the source of hydrogen forming amino groups in the case of DLC coatings deposited from graphite is limited to diffusion from the surface. Accordingly, it is inferred that the addition of nitrogen to the hydrogen-free DLC coating does not have the effect of increasing the polarity of the surface or improving wettability at least in the case of nitrogen content up to 14%.As a factor influencing the surface free energy of magnetron-sputtered DLC coatings containing nitrogen, Shi The sp3 bonding ratio of ta-C II #10 was estimated from the plasmon peak in the EELS analysis to be 81%. Although somewhat higher, this value is close to the figure of 71% reported by Barros et al. The number of dangling bonds in ta-C II #10, ta-C III #11 and ta-C:N III #12 was then investigated by ESR analysis, which could be conducted on these coatings because they were deposited on Si substrates. The ESR analysis results are presented in . In addition to calculating the number of dangling bonds per volume, the number per unit mass was also calculated from the coating density measured by XRR analysis, and the results are shown in the table as well. The DLC coating density was found to be 3.11 g/cm3 for ta-C II #10 that had an sp3 bonding ratio of 81%, 2.51 g/cm3 for ta-C III #11 with an sp3 bonding ratio of 28%, and 2.25 g/cm3 for ta-C:N III #12 with an sp3 bonding ratio of 20%. These results indicate that there is nearly a proportional relationship between the sp3 bonding ratio and atomic density, which is consistent with the findings reported by Libashi et al. Similar to the results seen for the sp3 bonding ratio, the number of dangling bonds found for ta-C:N III #12 was noticeably reduced by the addition of nitrogen. shows the number of dangling bonds as a function of the sp3 bonding ratio. The results indicate that the number of dangling bonds increased with an increasing sp3 bonding ratio. However, the number of dangling bonds found from an ESR analysis of DLC coatings deposited by different processes cannot be interpreted unequivocally on the basis of the sp3 bonding ratio alone because it is also affected by differences in the deposition conditions, in addition to the sp3 bonding ratio. Nonetheless, it is noted that Lee et al. , it is estimated that the number tends to increase with an increasing sp3 bonding ratio. The plots show a good correlation between the number of dangling bonds and the sp3 bonding ratio. The addition of nitrogen to DLC coatings reduces the sp3 bonding ratio and leads to a reduction in the number of dangling bonds. Therefore, when considering the effect of the number of dangling bonds on surface free energy, the foregoing results suggest that the addition of nitrogen to DLC coatings is more of a concern with respect to the effect on reducing surface free energy.The results of our previous analyses of the friction reduction effect of hydrogen-free DLC coatings under a lubricated condition have shown that markedly lower friction coefficients can be obtained together with the use of the GMO oiliness agent terminated by –OH polar groups There are several reports concerning the reactions of other elements with DLC coatings, including the findings of research on protective coatings for hard disk drives (HDDs). Yanagisawa Most of the dangling bonds detected in the ESR analysis were presumably in a steady-state condition inside the coating and not on the surface. Though chemically active dangling bonds on the surface become contaminated and disappear during exposure to the air following coating deposition, we assume that GMO, –H and –OH were able to react with dangling bonds in the friction tests because the surface layer of the coating was removed by friction and rubbing. The dangling bonds in the coating were thus exposed at the newly created surface, enabling GMO and its decomposed products to react with them. GMO is adsorbed on the surface because it is terminated by –OH radicals and is chemically active. Moreover, the H molecules at the ends diffuse into the coating and the O and subsequent molecules are able to form covalent bonds with the dangling bonds It is our view that dangling bonds in DLC coatings are a crucial and indispensable element of the friction reduction mechanism. In this regard, Kasai et al. From a macroscopic point of view, avoiding direct contact between two mating surfaces reduces friction under boundary lubrication conditions, while from a microscopic point of view, reducing friction at the contact points results in a lower overall level of friction. With regard to the former approach, Fan et al. The results of the friction tests and analyses conducted in the present study are applied here to the proposed friction reduction mechanism. The tendency seen for the friction coefficient to increase in the case of DLC coatings having a smaller sp3 bonding ratio can be explained in terms of the decrease in the number of dangling bonds, which is consistent with the proposed mechanism. Additionally, the results for the sputtered DLC coatings showed that the addition of nitrogen reduced the number of dangling bonds, indicating a correlation between the sp3 bonding ratio and the number of dangling bonds. Therefore, the slight worsening of the friction coefficient seen for the DLC coatings deposited by laser-induced pulsed arc deposition due to the addition of nitrogen can also be explained in terms of a decline in the number of dangling bonds accompanying a marked decrease in the sp3 bonding ratio of the coatings.The friction coefficients of the DLC coatings evaluated in this study and the numbers of dangling bonds found by ESR analysis are plotted in a pseudo ternary map in in relation to the sp2 bonding ratio, sp3 bonding ratio and nitrogen content. The sp3 bonding ratios used in this map are the values given in . The values estimated from the plasmon peak were used in the calculation of the number of dangling bonds based on the EELS analysis. The values used for the friction coefficient are the averages of the friction coefficients plotted for each coating in The ta-C:N I coatings displayed a distribution that shifted toward the nitrogen side of the map as the sp3 bonding ratio decreased markedly. These coatings had the same substrate temperature as ta-C I #6, but they were not distributed toward the apex of the triangle, representing a nitrogen content of 100%, while retaining a constant sp3 bonding ratio. The ta-C:N III coating also displayed this same tendency. The friction coefficient was the lowest and the number of dangling bonds was the largest for ta-C II #9, which was prepared by DC arc deposition and had the highest sp3 bonding ratio. In contrast, on the line for zero nitrogen content, the friction coefficient tended to increase as the sp3 bonding ratio decreased. This tendency was the same for the nitrogenated DLC coatings; it is seen that their friction coefficient also tended to increase with a decreasing sp3 bonding ratio. The two sputtered DLC coatings (ta-C III #11 and ta-C:N III #12) that had the smallest sp3 bonding ratio were positioned closer to the sp2 bonding content of 100% in this data distribution. The numbers of their dangling bonds were markedly lower than that of ta-C II #9, and notably the numbers were less than one-half the values.There are still several unexplained issues remaining in the proposed friction reduction mechanism, but the significance of the map in is that it suggests that a hydrogen-free DLC coating with a higher sp3 bonding ratio is effective in reducing the friction coefficient under a lubricated condition that includes an oiliness agent. The reason is that, assuming the dangling bonds of the DLC coating are the adsorption sites of the oiliness agent, a DLC coating with a higher sp3 bonding ratio is desirable from the standpoint of the number of dangling bonds. With respect to the wettability of the coating surface, the possibility of reducing friction by adding nitrogen was also considered, but that effect was not observed in terms of the surface free energy, at least not for hydrogen-free, nitrogenated DLC coatings with a nitrogen content of less than 14%. Moreover, the addition of nitrogen causes a pronounced decline in the sp3 bonding ratio, resulting in a substantial decrease in dangling bonds. Accordingly, adding nitrogen is presumably not conducive to the adsorption of an oiliness agent.The ta-C coatings prepared by laser-induced pulsed arc deposition with an sp3 bonding ratio in a low range of 42–69% and the ta-C:N coatings with a nitrogen content of approximately 14 at% both showed higher friction coefficients under a lubricated condition than the ta-C coatings prepared by DC arc deposition with an sp3 bonding ratio of approximately 81%. The ta-C:N coatings in particular were expected to exhibit improved wettability, but neither the water contact angle nor the surface free energy showed any tendency to improve as a result of adding nitrogen. Moreover, compared with the ta-C coatings deposited at the same substrate temperature, a marked decline was observed in the sp3 bonding ratio. The results of EELS and ESR analyses performed on a ta-C coating deposited on a silicon substrate by DC arc deposition and a ta-C coating and a ta-C:N coating containing approximately 11 at% nitrogen, both of which were deposited on Si substrates by high-power impulse magnetron sputtering, showed the existence of a correlation between the sp3 bonding ratio and the number of dangling bonds. These results and a comparison of the friction coefficients obtained under a lubricated condition suggested that a nitrogen-free DLC coating with a large number of dangling bonds is effective in reducing the friction coefficient.Local and distortional buckling behaviour of back-to-back built-up aluminium alloy channel section columnsThe use of aluminium alloy channel sections is becoming popular in lightweight structures, especially for pillars of curtain wall systems and brace and chord members in roof trusses. In cases where increased axial strength is required, it is popular to use built-up sections, instead of single channel sections. Back-to-back built-up aluminium alloy channel sections as compression members can be used to achieve higher strengths and longer spanning capabilities. In such an arrangement, intermediate fasteners at discrete points along the length can prevent the individual channel sections from buckling independently. However, no investigative research is available on the axial strength of back-to-back built-up aluminium alloy channel sections. This paper presents both experimental and numerical investigations on the behaviour of screw fastened back-to-back built-up aluminium alloy stub columns under compression. In total, the results from 15 axial compression tests are reported. For all test specimens, initial imperfections were measured using a laser scanner. A nonlinear elasto-plastic finite element (FE) model was then developed and validated against the test results. Thereafter, a comprehensive parametric study was conducted using the validated FE model to investigate the effects of modified slenderness, screw number and section thickness on axial strength of back-to-back built-up aluminium alloy channel sections. In total, 232 FE models were analysed. Axial strengths obtained from the tests and FEA were used to assess the performance of current design guidelines offered by the Aluminium Design Manual (ADM), Australian/New Zealand Standards (AS/NZS), Eurocode 9 (EC9), Eurocode3 (EC3) and the American Iron and Steel Institute (AISI) standards. It is shown that the design, in accordance with these standards, is conservative within 20%, with the exception of the allowable stress design method mentioned in AS/NZS (1664.2:1997). The ADM, and AISI & AS/NZS (4600:2018) provide highly accurate predictions; being conservative to the experimental results by 6% and 10% on average, respectively.Screw spacing from the AISI & AS/NZS (4600:2018);Area of the edge stiffener from the EC3;Effective width of the plate from the EC3;Buckling constant from the ADM, AS/NZS (1664:1997);Buckling constant from the ADM, AS/NZS (1664:1997);Buckling constant from the ADM, AS/NZS (1664:1997);Buckling constant from the ADM, AS/NZS (1664:1997);Buckling constant from the ADM, AS/NZS (1664:1997);Buckling constant from the ADM, AS/NZS (1664:1997);Least of the elastic flexural, torsional, and flexural–torsional buckling stress;Compressive yield strength from the ADM;Least of the elastic flexural, torsional, and flexural–torsional buckling stress;Characteristic value of 0.2% tensile proof stress from EC9;Limit state stress from the AS/NZS (1664.1:1997);Nominal buckling stress as per the AISI & AS/NZS (4600:2018);Effective second moment of area of the edge stiffener from the EC3;Effective length/buckling factor from the AS/NZS (1664.2:1997), EC3;Coefficient for compression members from the AS/NZS (1664:1997);Spring stiffness per unit length from the EC3;Overall Slenderness from the AISI & AS/NZS (4600:2018);Total length of the back-to-back built-up aluminium alloy channel sections;Limit state design from the AS/NZS (1664.1:1997);Allowable stress design from the AS/NZS (1664.2:1997);Linear variable displacement transducers;Factor of safety from the AS/NZS (1664.2:1997);Factor of safety from the AS/NZS (1664.2:1997);Axial strength from the AS/NZS (1664.2:1997);Axial strength from the AISI & AS/NZS (4600:2018);Axial strength from the finite element analysis;Axial strength from the AS/NZS (1664.1:1997);Elastic critical force for global buckling mode from the EC3;Minimum radius of gyration from the AISI & AS/NZS (4600:2018);Nominal thickness of the channel section;Factor for the allowed weakening effects from the EC9;Non-dimensional slenderness ratio as per the AISI & AS/NZS (4600:2018);Equivalent slenderness ratio from the ADM;Stress ratio of an individual plate from the EC3;Strength reduction factor from the AS/NZS (1664.1:1997);Capacity factor from the AS/NZS (1664.1:1997);Elastic local buckling stress of the plate;Reduction factor on the plate width from the EC3;Reduction factor for the distortional buckling mode from the EC3;Aluminium alloy members are a popular choice for structural engineering applications shows the general arrangement and cross-sectional details of the built-up columns investigated in this study.Current design guidelines in accordance with the Aluminium Design Manual (ADM) In terms of the cold-formed carbon steel built-up section columns, significant research is available in the literature. Ting et al. The use of stainless steel built-up columns has become increasingly common. This is mainly due to their aesthetic appearance, high corrosion resistance, ease of maintenance and convenience of assembly and construction In terms of aluminium alloy single channel section columns, limited research is available in the literature. Feng et al. As mentioned previously, this paper presents the results of 15 new experimental tests on the axial strength of back-to-back built-up aluminium alloy channel sections. Prior to compression tests, initial geometric imperfections were measured using a laser scanner. Tensile coupon tests were also conducted to determine the material properties of the aluminium alloy channel sections. A nonlinear elasto-plastic FE model was then developed and validated against the experimental results, both in terms of failure loads and deflected shapes. Using the validated FE model, a parametric study involving 216 models was conducted to investigate the effects of modified slenderness, screw number and section thickness on axial strength of such columns. The axial strengths obtained from the tests and finite element analysis (FEA) were used to assess the performance of current design standards, including the ADM In this study, a total of 15 back-to-back aluminium alloy built-up channel sections were tested to failure under axial compression. Nominal cross-sections of test specimens considered in this paper are shown in , which shows that two different cross sections were used in the experiments: BU150×65×25 and BU240×45×20. The nominal thickness of section BU150 and BU240 were 1.56 mm and 1.98 mm, respectively. summarizes the dimensions of test specimens. Test specimens were subdivided into two different column lengths: stub (300 mm) and short (500 mm).In the experimental test programme, the following longitudinal spacings for screw spacings (S) were considered:Column length of 300 mm; screw spacing of 50 mm, 100 mm and 200 mmColumn length of 500 mm; screw spacing of 100 mm, 200 mm and 400 mmThe spacings of fasteners were designed to cover the spacing requirement of AISI The test specimens were labelled such that the depth of web, longitudinal spacing between fasteners, and nominal length of section and specimen number were indicated by the label (). For example, the label “BU150-S50-L300-1” could be interpreted as follows:The number 150 refers to the nominal depth of web in millimetres i.e. d=150 mm.“L300” is the nominal length of the specimen in millimetres i.e. L=300 mm.“S50” represents the screw spacing in millimetres i.e. S=50 mm.The last number “1” indicates the specimen number for a repeated group.Tensile coupon tests were conducted to determine the material properties of the specimens. The coupons were obtained from the centre of the web plate in the longitudinal directions of the untested specimens ((a)), in accordance with the British Standard for Testing and Materials (b). The tensile coupons were tested in a 100 kN Instro machine, while applying a displacement rate of 2 mm/min ((c)). The stress–strain curves for BU150×65×25 and BU240×45×20 are shown in . The corresponding material properties obtained from the tensile coupon tests are listed in , the average values of yield strengths for BU150×65×25 and BU240×45×20 are 108.40 MPa and 150.50 MPa, respectively.A universal testing machine of 500 kN capacity was used to apply the axial load to the aluminium alloy built-up columns. The load was applied through the centre of gravity (CG) of the specimens under pin-ended boundary conditions. In order to ensure there was no gap between the two pin-ends and end plates of the specimen, all columns were loaded initially up to 25% of their expected failure load and then released. A photograph of the test setup is shown in shows the photograph of the pin support used in the test setup. Displacement control method was used to apply the axial loads on the columns. The load was applied at a constant loading rate of 0.1 mm/min.An external load cell was placed at the top of the built-up columns. A total of three linear variable differential transformers (LVDTs) were used to record the displacements. The axial shortening of the specimens was recorded from the readings of LVDT-1 and the lateral displacements were recorded from the readings of LVDT-2 and LVDT-3 at mid-height of the back-to-back built-up aluminium alloy channel sections. Four different strain gauges (SG1, SG2, SG3 and SG4) were used to measure the strain values at mid-height of the aluminium alloy built-up channel sections. shows the locations of the strain gauges. The axial load and the readings of the transducers were recorded by a data acquisition system at regular intervals during the tests.Imperfections in aluminium alloy channel sections can occur as a result of transportation and fabrication processes. Initial geometric imperfections significantly affect the stability of aluminium alloy members under compression. Therefore, the magnitude and shape of imperfections for each test specimen were recorded before undertaking the compression tests., a laser scanner was used to measure the initial imperfections of all test specimens. The platform was designed to have a precision shaft in the transverse (2500 mm) direction which guided a moveable laser scanner. The laser scanner was used to measure imperfections along six longitudinal lines on aluminium alloy sections, as shown in . The laser scanner recorded readings at every 0.1 mm.The local imperfections were calculated by subtracting the average reading along lines W-1 and W-3 from the readings taken along the lines W-2 ((a)). The overall imperfections were calculated as the average value of the readings recorded along the lines W-2 at mid-height of the columns ((b)). Distortional imperfections were calculated as the maximum reading along the lines F-1 and F-2 (, a typical imperfection profile is plotted against the length of BU240-S200-L500. lists the maximum local, distortional and global imperfections of all test specimens. The values for initial geometric imperfections were used in the FE modelling described in Section of the current paper. A similar procedure was also used by Chen et al. summarizes the column dimensions and experimental failure loads (PEXP) for all 15 test specimens. As can be seen from , BU150-S50-L300 and BU240-S50-L300 were both tested with three repeats. The corresponding coefficient of variations (COVs) of axial strength for these two sections in tests were 0.03 and 0.00, respectively. (a) also showed that all stub and short columns failed through local and distortional buckling for BU150. plots the axial load versus lateral displacement graphs at mid-height of the sections BU240-S50-L300-3 and BU240-S100-L300. shows the deformed shapes of the 300- and 500 mm-long BU150 and BU240.load–axial shortening behaviour for stub and short columns is plotted in . It was observed how the load–axial shortening behaviour was linear up to a load of 64.79 kN. This was approximately 73.43% of the ultimate failure load for BU150-S100-L300. After that, nonlinear behaviour was noticed until the failure load was reached, which occurred at 88.23 kN. Similar observations were made for other screw spacings of back-to-back built-up aluminium alloy channel columns.Four strain gauges, two each on tension and compression sides, were used to determine the axial–strain at mid-length from both ends of the back-to-back built-up aluminium alloy channel sections. The load–axial strain relationship for the BU240-S200-L500 section is plotted in The effect of the screw spacing on axial strength was investigated and is shown in both , when the spacing was increased from 50 mm to 100 mm, axial strength was reduced by 2.44% on average for stub columns. Upon doubling the screw spacing from 100 to 200 mm, a reduction of 2.95% in axial strength was observed for stub columns.The specimen BU240-S50-L300 with five fasteners spaced at 50 mm failed through local and distortional buckling. Back-to-back built-up aluminium alloy channel sections remained integral at failure, showing some plastic deformation near the bottom or top end of the stub columns, as shown in . For most of the stub columns, local and distortional buckling was observed. When the ultimate load was reached, localized deformation was noticeable near the compression side of the columns. The deformed shapes for stub and short columns of back-to-back built-up aluminium alloy channel sections are shown in An elastic–plastic model was used for modelling the overall geometry of the built-up columns. In order to define the isotropic yielding and plastic hardening of the steel, the von Mises yield surface was used in the classical metal plasticity model. The material properties were taken from the results of tensile coupon tests and included in the FE models. As per the ABAQUS manual Where, E is the Young’s modulus, and σtrue and εtrue(pl) are the true stress and strain, respectively. σ and ε are the engineering stress and strain, respectively in ABAQUS The back-to-back built-up aluminium alloy channel sections were modelled using the S4R shell elements available in ABAQUS The back-to-back built-up aluminium alloy channel sections considered in this study were pin-ended columns. In order to simulate the upper and lower pin-end supports, the displacements and rotations (boundary conditions) were assigned to the upper and lower end plates through reference points. The applied boundary conditions in the FE model are shown in (a) for BU150-450-L300. To simulate the experimental boundary conditions, the translation in the x and y were restrained, while the vertical translation in the z direction was not restrained at the top reference point (loading point). For the bottom reference point (reaction point), the translation in the x, y and z were restrained. It should be noted that two ends were free to rotate along the minor axis. A multi-point constraint (MPC) beam connector element available in the ABAQUS “Surface to surface” contact was used for modelling the interaction between the webs of two channel sections. Surface-to-surface interaction was used to model the intersection of the edges of channel sections and the top surfaces of the base plates. It should be noted that the web surface of one channel was considered as a master surface. The other was considered as a slave surface. The edges of the channel section were modelled as the slave surface, while the top surfaces of the end plates were considered as the master surface. The normal behaviour of the surface was defined as “hard”, indicating that no penetration of the surfaces into each other was allowed.). The contours of local and distortional buckling modes are shown in The comparison of axial strengths from the tests and FEA are shown in shows how the FEA strengths are close to experimental strengths. shows the deflected shapes of stub and short columns obtained from the FE analysis, which presents good agreement with the experimental failure modes. load–axial shortening behaviour from both the FEA and experiments is plotted in , for stub and short columns. Comparison results showed that the differences between the FE model prediction and the test results are very small. The mean value of the ratio between the experimental and numerical results (PEXP∕PFEA) is 1.06, with a COV of 0.08. The load versus lateral displacement curves are plotted in for BU240-S50-L300, BU240-S100-L300 and BU240-S350-L1500 sections and demonstrate good comparisons between the FEA and test results. The axial load versus axial–strain relationship from both the FEA and experiments are shown in It should be noted that most of the load–displacement curves obtained from the tests and FEA, were in good agreement, and only some of the curves did not show a good match. This is due to the localized slip which was observed between the loading supports and test specimens in some experiments. Also, this might be a consequence of the friction between two webs of back-to-back aluminium alloy channel sections, which was simplified in the FE models. This explains such differences between the numerical and experimental results for some curves, particularly at the initial parts (elastic) of the load–displacement curves. However, most of the results obtained from the numerical analysis were in good agreements with the experimental results, both in terms of axial strengths and deformed shapes.A parametric study comprising 232 FE models was conducted in order to investigate the effect of screw spacing on the axial strength of back-to-back built-up aluminium alloy channel sections. A wide range of slenderness, covering stub and short columns was considered. Column lengths from 300 to 800 mm were considered, as shown in . As can be seen, three different screw number were considered in the parametric study: 3, 5 and 9. FE results from the parametric study are presented in , showing the variation of axial strength against the screw number and section thickness for BU150 and BU240. shows the axial strength of back-to-back built-up aluminium alloy channel sections with varying modified slenderness (KL/r)m. The results showed in demonstrate how the axial strengths were reduced by 5.17% and 10.80% on average, when the average modified slenderness ((KL/r)m) was increased from 3.61 to 11.21 and 4.24 to 14.22 for BU150 and BU240, respectively.The effect of number (n) of screws on axial strength was also investigated. From (a), it can be seen that with the increase screw number, the axial strength for BU150 and BU240 increased slightly by 0.56% and 0.87%, respectively. This indicates that the effect of screw number on the axial strength of the columns was limited.The effect of channel thickness on axial strength was also investigated. From (b), it can be seen that the channel thickness plays a significant role in determining the axial strength of back-to-back built-up aluminium alloy channel sections. For BU150, the axial strength increased significantly by 109.84% on average when the section thickness increased from 1.6 to 2.6 mm. On the other hand, for BU240, axial strength increased by 67.72% on average when the section thickness changed from 1.9 to 2.9 mm.The calculation of axial strengths for aluminium alloy single channel section columns is available in the current design standards, including ADM The design strengths of aluminium alloy single channel section columns can be predicted using design guidance from Aluminium Design Manual (ADM) The design compressive strength (ϕcPn1) and the allowable compressive strength (Pn1∕Ωc) in the Aluminium Design Manual (PADM−1) uses the lowest of the available strengths for the limit states of member buckling (Pn1), local buckling (Pn2), and the interaction between the member buckling and local buckling (Pn3).The member buckling strength (Pn1) is available in chapter E.2 of ADM) for the inelastic buckling condition (λ1≺λ≤λ2): for the elastic buckling condition (λ≥λ2): The weighted average local buckling strength (Pn2) was calculated in accordance with Chapter E.4.1 of ADM The strength of interaction between the member buckling and local buckling (Pn3) was determined as per the design rules given in chapter E.5 of ADM The direct strength method (DSM) for aluminium alloy single channel section columns is available in chapters B.5.4.6 and E.3.2 of ADM For the inelastic buckling condition (λ3≺λeq≺λ4): For the elastic buckling condition (λeq≥λ4): Where, λ3=(Bp−fc)∕Dp, which is the slenderness for the intersection of yielding and inelastic buckling; λ4=0.35Bp∕Dp, which is the slenderness for the intersection of inelastic buckling and elastic buckling; λeq is the equivalent slenderness ratio for flexural or axial compression; fe is the elastic local buckling stress of the cross section; and k2=2.04, is the post buckling constant.The design strengths (PLSD) of aluminium alloy single channel section columns can be determined from the design equations given in the AS/NZS Where, λ=kLr(lπ)Fcy∕E; Dc∗=πDcE∕Fcy; S1∗=Bc−Fcy∕kcDc∗; S2∗=CcπFcy∕E. Where, FL is the limit state stress, ϕ is the strength reduction factor, ϕFL is the limit state stress, ϕcc is the capacity factor; A is the gross cross-section area, E is the elastic modulus, Fcy is the compressive yield stress (0.2% tensile proof stress), kc=1.12 is the coefficient for compression members, r is the radius of gyration, and Bc, Cc and Dc are the buckling constants.The calculation of design strengths (PASD) for aluminium alloy single channel section columns in AS/NZS Where, S1=Bc−nuFcykcnyDc; S2=Cc; k is the effective length factor by rational analysis. k should be taken as greater than 1.65 or equal to unity unless rational analysis justifies a smaller value; L is the unsupported length; r is the radius of gyration of the column about the axis of buckling; Bc, Cc and Dc are the buckling constants; nu, ny=1.95, and 1.65, respectively.Where, Aeff is the effective section area based on the reduced thickness allowing for local buckling, and for this paper, it is equal to the gross section area of the column; κ is the factor for the allowed weakening effects, in this paper κ is taken as 1; χ is the reduction factor for the relevant buckling mode; f0 is the characteristic value of 0.2% tensile proof stress (σ0.2).Where, Aeff is the effective section area; fy is the yield stress.A brief discussion is offered below to explain the EC3 (a). This effective area should be assumed to resist the full bending action applied to the section. The effective widths of compressive elements are given as: For internal compression elements,ρ=beb=1λp1−0.055(3+ψ)λpFor external compression elements,ρ=beb=1λp1−0.188λpWhere, ρ is the reduction factor on the plate width, while b and be are the total width and effective width of the plate, respectively. The slenderness ratio λp relates the material yield stress fy to the elastic local buckling stress of the plate σcr and ψ is the stress ratio of an individual plate, for axial load calculations, ψ=1. k is the buckling factor.Distortional buckling of a single channel section is linked to any buckling mode causing a distortion of the shape of the cross-section, but excludes those deformations related to local buckling ((b)). As a result, distortional buckling is always associated with the displacement of one or more of the fold-lines of the section being out of original position. The distortional buckling behaviour can be studied by considering an equivalent strut on an elastic foundation, as shown in (c). The distortional buckling strength of a single channel section can be calculated from Eq. Where, K is the spring stiffness per unit length, and As and Is are the area and effective second moment of area of the edge stiffener. The reduction factor χd, for distortional buckling resistance should be obtained from the relative slenderness λd¯=fyb∕σcr,sFor0.65≺fyb∕σcr,s≤1.38,χd=1.47−0.723fyb∕σcr,sFor members in compression, the global buckling resistance can be calculated based on a non-dimensional slenderness ratio, as given in Eq. Where, Pcre is the elastic critical force for global buckling mode based on the gross cross-sectional properties and Aeff is the effective area of the cross-section.Axial strengths determined from the tests and FE analyses were also compared against the design strengths calculated in accordance with the design rules of carbon steel built-up columns as per the AISI The critical elastic buckling stress (Fn) was determined using Eqs. Where non-dimensional critical slenderness (λc) was determined using Eq. Where, fy is yield stress and foc is the least of elastic flexural, torsional and flexural–torsional buckling stress calculated in accordance with sections C3.1.1 of AS/NZ was used in all calculations to determine the axial strength of back-to-back built-up columns. Where, (KL/r)o is the overall slenderness ratio; a is the intermediate fastener or spot weld spacing; and ri is the minimum radius of gyration of full unreduced cross-sectional area of an individual shape in a built-up member. Eq. Axial strengths obtained from the experiments and FEA were compared against the design strengths calculated in accordance with the design standards of ADM (a) for BU150 and BU240, respectively. As can be seen, the design strengths calculated from the ADM , the axial strength of built-up columns determined from the FEA was compared against the design strengths calculated in accordance with the current ADM (b) compared the design strengths with the FEA results of the parametric study, and it was found that the design strengths were conservative and within 20%. The design rules in AISI (d)), being only 5% conservative for most of the columns.(c) plotted the tests, FE and design strengths of the BU150 and BU240 against the section thickness. Axial strength of the columns with 3, 5 and 9 screw number are shown in (c) for BU150 and BU240, respectively. As can be seen, for all three screw configurations (3, 5 and 9 screw number), the FEA strengths for BU150 demonstrated good agreement with the design strengths predicted in accordance with the AISI The mean values for PEXP∕PADM−1 for BU150 with the lengths of 300 and 500 mm were 1.03 and 1.11, respectively with the COVs of 0.03 and 0.02. For BU240, the values were 0.99 and 0.96, respectively with the same COV of 0.02. The average values of PEXP∕PAISI&AS∕NZS for the BU150 stub and short columns were 1.12 and 1.25, respectively, with the COVs at 0.04 and 0.03 ((a)). The corresponding mean values of PEXP∕PAISI&AS∕NZS for BU240 were 0.96 and 0.90, respectively with the same COV of 0.02 ((b)). Considering the parametric study results, the AISI (d), the effect of screw number is plotted against the axial strengths of built-up columns. For stub and short columns, increasing the number of fasteners from 3 to 9, had little effect (less than 5%) on the axial strength of back-to-back built-up aluminium alloy channel sections.This paper has presented a detailed experimental study on the axial strength of back-to-back built-up aluminium alloy channel sections under axial compression. The results of 15 new tests are reported. Prior to compression tests, the material properties of aluminium alloy channel sections were determined from the tensile coupon tests and the initial imperfections were measured using a laser scanner. The failure modes, load–axial shortening, load–lateral displacement and load–axial–strain relationships were discussed. The effects of modified slenderness, screw number and section thickness were investigated in the experimental study.A non-linear FE model was then developed which included the modelling of screws, material non-linearity, and initial geometric imperfections. The results from the FE model were compared against the experimental results which showed a good match, both in terms of axial strength and failure modes. The validated FE model was further used to conduct a parametric study comprising 216 models to investigate the effects of modified slenderness, screw number, and section thickness on the axial strength of back-to-back built-up aluminium alloy channel sections. From the results of the parametric study, it was found that the section thickness could significantly affect the axial strength of such columns. For BU150, the axial strength increased by 109.84% on average when the section thickness increased from 1.6 to 2.6 mm. On the other hand, for BU240, the axial strength increased by 67.72% on average when the section thickness increased from 1.9 to 2.9 mm. However, the effect of modified slenderness for stub and short columns was not noticeable. It was shown that axial strengths of BU150 and BU240 were reduced by 5.17% and 10.80% with the increase of modified slenderness from 3.61 to 11.218 and 4.24 to 14.22, respectively. The effect of screw number had a limited impact on the axial strength of back-to-back built-up aluminium alloy channel sections. The axial strength of columns increased by 0.56% and 0.87% on average for BU150 and BU240, respectively, when the screw number changed from 3 to 9.Axial strengths obtained from the experiments and FEA were compared against the design strengths calculated in accordance with the Aluminium Design Manual (ADM), Australian/New Zealand Standards (AS/NZS), Eurocode 9 (EC9), Eurocode 3 (EC3) and American Iron and Steel Institute (AISI) standards. The ADM, AS/NZS (1664.1:1997 and 1664.2:1997), EC9 and EC3 only cover the design of an aluminium alloy single channel section column. Therefore, the design calculations assumed the strength of built-up aluminium alloy channel sections to be twice the strength of a single channel section. Furthermore, based on the design rules of carbon steel built-up sections, the axial strengths of aluminium alloy built-up sections were also calculated from the AISI and AS/NZS (4600:2018). From the comparison of design strengths against the test and FEA strengths, it was found that the AISI and AS/NZS (4600:2018) were conservative by 10% and 3% on average for columns failed mainly as a consequence of local and distortional buckling. The results also showed that the ADM, AISI & AS/NZS (4600:2018) standards were conservative for calculating the experimental axial strengths of back-to-back built-up aluminium alloy channel sections by around 6% and 10% on average, respectively.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Special Issue Article: The First International Symposium on Mine Safety Science and Engineering► Transmission laws in pipeline gas explosion were studied. ► The maximum pressure value lowers firstly near the explosion point. ► Then it rises to a peak, and finally drops gradually. ► Two waves divide the pipeline space into three sections. ► The simulation result is basically consistent with the experiment result.The frequently occurring gas accidents, especially extraordinarily serious gas explosion accidents in coal mines caused great threat to the safety of state property and people, and have badly impacted the healthy development of coal industry. According to statistics, among the serious accidents occurred in coal mines of our country, over 70% are gas explosion accidents (Due to this, the author carried out numerical simulation on gas explosion process with FLUENT, and provided information hardly observed in experiment. The numerical simulation is based on the theoretical model of gas explosion transmission, uses continuous phase method to simulate the transmission of gas explosion in the pipeline. Given that the model is too long and the computation is too large, numerical simulation research was just carried out with gas optimal explosion concentration 9.5%.The major constituent of mine gas is methane, a kind of combustible gas. The essence of mine gas explosion is that explosive gas mixture consisted of methane and air forms drastic chemical reaction under fire source inducement, accompanied by large amount of heat (The chemical reaction equation is as follows:FLUENT is now a prevailing commercial CFD software package in the international market, through which, engineering problems, as long as concerning fluids, heat transfer, and chemical reactions, all could be explained. Therefore, this paper chose it to carry out simulation of pipeline gas explosion (The simulation calculation of the gas explosion process has adopted flow field method, so geometric modeling and meshing calculation is indispensable pretreatment work which was done based on Gambit.Geometric modeling of pipeline gas explosionSo given the complexity of places where gas explosions happen, this simulation used mixed grid of both structured and unstructured grid. In actual calculations, the ignition area was handled with irregular quadrilateral grid and the transmission area with rectangular one. Structural grid could provide better structural boundary compatibility, so that the boundary area of the roadway adopted structural meshing to the greatest extent. Simulation initial conditions and boundary conditions settingInitial pressure condition: excessive pressure of ignition area is set to be P |
= 1000 Pa; Other areas are P0 = 0 Pa; Initial temperature condition: Ignition area: T0 = 1600 K; Other areas: T0 = 300 K; Initial speed condition: the initial speed of the whole area is 0, i.e. V0 = 0 m/s; Initial components condition: to simplify the problem, air components are 22% of oxygen and nitrogen in 78% by volume. The experimental pipe wall was designed as typical non-slip, non-invasive boundary with a closed left end and the right end as explosive vent. presents the whole process going from the pressure wave begins to spread as spherical wave since the ignition of gas, to gradually becomes plane wave after multiple overlaps. that the pressure wave spreads around in spherical shape from the ignition point, after reflected by the pipe wall, the reflection waves will overlap with each other, and lead to the significant pressure increase, as the shown red area ⑤ and ⑥. The continuously overlapped spherical waves will finally become plane waves and will spread to both ends of the pipe. The pressure wave on the left side, reflected by the end of the pipe, would produce a reflection wave which will promote the rightward pressure wave, lead finally to a gradual increase of the right pressure wave.Compare the simulation result with the experimental result at the same concentration of different measured points, shown as that there is difference between simulation result and experimental result. At the range of 10–50 m and 55–65 m, the simulation result is significantly higher, maybe because the wall conditions set for simulation is less tough than real pipe wall, due to which the energy lost during gas explosion transmission has lowered the pressure value. While the simulation value lowered gradually near the exit, exactly opposite to the experimental value. This is because in the experiment, the exit was sealed which could cause influence to pressure nearby and the increase of pressure value, but the exit in simulation has not been processed, so the pressure value lowered after decompression. From the variation trend of the curve, it can be seen that the maximum pressure value of gas explosion pressure wave in the pipeline always decreased near the start site of the explosion, then rose to a peak, after that, it will decrease again. is the variation trends comparison of maximum pressure presentation time on different measured points between simulation and experiment. We can see that the simulated variation trend of maximum pressure value is basically identical with the experimental result, i.e., the presentation time of maximum explosion pressure at measured points steadily increase along with its distance to the explosion source. It is easy to see that the maximum explosion pressure of measured points near the explosion source have shorter intervals between presentation time, while the farther the distance with the explosion source is, the longer the interval is. This is basically the same with the experimental result, but the presentation time of measured points is shorter than that of the experiment which demonstrates that simulation pressure wave transmits faster than that of experiment, due to the smooth wall and non-sealing exit in simulation. that the average pressure amplitude speed in simulation is faster than that in experiment. When the gas’s concentration was 9.5%, the experimental maximum explosive pressure was 0.970 MPa, the initial pressure was 0.492 MPa, and the average pressure amplitude speed was 7.71 MPa/s; while in simulation with the same concentration, the corresponding values were 1.056 MPa, 0.524 MPa and 9.33 MPa/s. The simulation results apparently were larger than experimental ones. The reasons may be the pipe wall in simulation is smooth and the exit is open without sealing, while in real experiment, the pipe wall is tough and the exit is sealed with plastic film. The film not only seals the exit but also vents the explosion. Aging and corrosion of the pipe wall could also influence the experiment result.As the gas concentration is increasing, the average pressure amplitude speed in explosion decreases on the contrary; the average pressure amplitude speed of gas explosion in simulation is faster than that in the experiment. When the gas concentration was 8.8%, the average pressure amplitude speed was 10.95 MPa/s, while when the concentration was 9.5%, the speed was 7.71 MPa/s. And when the speed was 9.33 MPa/s, the simulation concentration reached 9.5%.There is difference between simulation values and experimental values. Mostly, the simulation values are larger than experimental values. This may relate to the different setting of conditions of the pipe wall. In numerical simulation, the pipe wall was set to be smooth while the toughness of the real wall may cause some loss of energy, making the pressure value lower. However, the simulation values decreased near the exit, contrary to the experimental trend, and this is because in experiment, the exit was sealed with film so the pressure wave was influenced and the pressure value increased thereof. Instead, in simulation there was no such handling, so the pressure value decreased after decompression.Effect of food/oral-simulating liquids on dynamic mechanical thermal properties of dental nanohybrid light-cured resin compositesThe purpose of this work was the study of the effect of food/oral simulating liquids on the dynamic mechanical thermal properties (viscoelastic properties) of current commercial dental light-cured resin composites characterized as nanohybrids. These nanohybrids were Grandio, Protofill-nano and Tetric EvoCeram.The properties were determined under dry conditions (1 h at 37 °C after light-curing) and also after storage in dry air, distilled water, artificial saliva SAGF® or ethanol/water solution (75 vol%) at 37 °C for up 1, 7, 30 or 90 days. Dynamic mechanical thermal analysis tests were performed on a Diamond Dynamic Mechanical Analyzer in bending mode. A frequency of 1 Hz and a temperature range of 25–185 °C were applied, while the heating rate of 2 °C/min was selected to cover mouth temperature and the materials’ likely Tg.Storage modulus, loss modulus and tangent delta were plotted against temperature over this period. The Tg of composites was obtained as the temperature indicated by tanδ peak. Moreover, the maximum height of tanδ peak, the width at the half of tanδ maximum and a parameter known as “ζ” parameter were determined. All composites analyzed 1 h after light-curing and 1 day in air or in food/oral simulating liquids showed two Tg. All composites stored for 7, 30 or 90 days in any medium showed unique Tg value. Also among the various properties studied the most sensible in the structural changes of composites seems to be the Tg.Storage of composites in dry air at 37 °C which is very close to their Tg (40 °C) for 1 or 7 days caused post curing reactions, while storage for 30 or 90 days has no further effect on composites. Storage in water or artificial saliva 37 °C for 1 or 7 days caused post curing reactions, while storage for 30 or 90 days seems to cause plasticization effect affecting some parameters analogously. Storage in ethanol/water solution (75 vol%) 37 °C for 1 or 7 days caused also post curing reactions, while storage for 30 or 90 days caused plasticization and/or probable oxidation/hydrolysis of polymeric network.Hybrids composites in general were marketed in early 1980s and the ongoing effort for smaller particles with greater resistance resulted in nanohybrids in 2000s. Nanohybrid composites were developed from microhybrid composites by the use of nanofillers, a step which led to a significant increase in the filler content of the materials and a considerable improvement in their physical properties The food/oral simulating liquids have been recommended by the Bureau of Food, FDA. Some of them do exist in the mouth, others can simulate its conditions (ingredients, pH, viscosity). Their use enables us to see a fast wear of the materials in short time and furthermore to take in consideration processes like chemical affinity, elution or bonding. It goes without saying that physical/mechanical properties are influenced profoundly by the presence of the FSLs Recently we studied the dynamic mechanical thermal properties of two nanofilled resin composites after curing and after immersion in some food/oral simulating liquids for a certain time This work aimed at the study of dynamic mechanical thermal properties of three current commercial dental light-cured resin nanohybrids. The nanohybrids studied are Grandio (GR), Protofill-nano (PR) and Tetric EvoCeram (TEC). They have different structural characteristics (content and structure of organic matrix and filler) and it was found interesting to study the effect of these characteristics on dynamic mechanical properties after photo-polymerization and after storage in some food/oral simulating liquids.Three commercially available light-cured nanohybrid composites were studied; TEC (Ivoclar-Vivadent, Schaan, Liechtenstein), GR (VOCO, Cuxhaven, Germany) and PR (Germany). Their specifications are listed in The SAGF® medium was chosen as artificial saliva in this work and its composition is given in For DMA tests, bar specimens of rectangular geometry were prepared by filling a Teflon mold (2 mm × 2 mm × 40 mm, as recommended by DMA manufacturer's instructions) with unpolymerized material, taking care to minimize entrapped air. The upper and lower surface of the mold was overlaid with glass slides covered with a mylar sheet (thickness 0.05 mm) to avoid adhesion with the uncured material. The completed assembly was held together with spring clips and irradiated by overlapping, using a XL 3000 dental photocuring unit (3 M Company, St Paul, USA). This source consisted of a 75 W tungsten halogen lamp, which emits radiation between 420 and 500 nm and lighting power at 200 mW/cm2 measured by Hilux curing light meter (Benlioglu Dental INC, Serial No. 9080935). This unit was used without the light guide in direct contact with the glass slides. The samples were irradiated for 60 s on each side applying overlaps. The assembly was dismantled, the composite was carefully removed by flexing the mold and the specimens were then slightly scratched to remove the material excess.DMA is a powerful technique where a sinusoidal stress is applied and the resultant strain is measured. The properties measured under this oscillating loading are storage or Young modulus (E′), loss or viscous modulus (E″) and tanδ (ratio of loss to storage modulus). The storage modulus (the in-phase component of the modulus) represents the stiffness of a viscoelastic material and is proportional to the energy stored during a loading cycle. The loss modulus (the out of phase component) is related to the amount of energy lost due to viscous flow. The tanδ is also the tangent of the phase angle and is called damping. Damping is a dimensionless property and also a measure of how well the material can disperse (absorb or emit) energy throughout its mass. A high tanδ indicates high molecular mobility, while a low tanδ indicates less mobility into the material. As temperature increases and the material approaches the rubbery state, the tanδ value and the molecular mobility increases. The tanδ value goes up to a maximum as the polymer undergoes the transition from the glassy to the rubbery state. The glass transition temperature (Tg) can be determined as the position of the maximum on the tanδ vs. temperature plot.The bar-shaped specimens used for DMA measurements divided into groups of three samples each. The first group consisted of dry samples measured 1 h after curing. During this first hour they were placed in a desiccator at 37 ± 1 °C. The other groups consisted of samples which had been stored in dry air, distilled water, ethanol/water solution (75 vol%), or artificial saliva SAGF at 37 ± 1 °C for 1, 7, 30 or 90 days-time.DMA tests were performed on a Diamond DMA dynamic mechanical analyzer (Perkin-Elmer) in bending mode. A frequency of 1 Hz was applied (approximately average chewing rate), bending force of 4000 mN and amplitude of 10 μm. A temperature range of 25–185 °C and a heating rate of 2 °C/min were selected to cover mouth temperature and the materials’ likely glass-transition temperature (Tg). Storage modulus (E′), loss modulus (E″) and tangent delta (tanδ) were plotted against temperature over this period. After the DMA run was completed, the sample was allowed to cool naturally at room temperature and the values of E′, E″ at 37 °C and the Tg were noted. Then the process was repeated with the same specimen and the re-run values for E′, E″ and Tg were determined as well. This method was used for each of the samples and the mean values were calculated.The values reported in all following Tables represent mean values ± standard deviation of three replicates. One-way analysis of variance (ANOVA) test, followed by a Tukey's test, for multiple comparisons between means to determinate significant differences was used at a significance level set at p |
≤ 0.05, for analysis of the results (software package OriginPro 8, by OriginLab Corporation run on Microsoft Windows). the storage modulus, loss modulus and tanδ, versus temperature curves are demonstrated correspondingly for TEC material, as representative. The sample has been kept for 1 h at 37 °C before the record. After the first scan the specimen was cooled down to 25 °C and a second scan was recorded from 25 to 185 °C. The curves obtained from the second run are shown also in (a–c). The double decrease of elastic modulus observed during the 1st run became a single decrease and the values of elastic modulus recorded during the 2nd run were higher than the corresponding ones recorded during the first run. The values of viscous modulus in the second run remained almost the same. The tanδ peak at low temperature was disappeared, while that at higher temperature remained also the same. These differences in the storage modulus and tanδ curves during the 1st and 2nd run confirmed the fact that additional curing reaction took place during the first run. When light-curing at room temperature is over, unreacted vinyl bonds are left behind. The lifetime of residual free radical at room temperature could reach several days or months, depending on crosslinking density and storage conditions Each of the two maxima in the tanδ versus temperature curves corresponds to one of the two drops in the storage modulus curves and represents the two Tg values for the dry samples. The first Tg is the apparent one of the light-cured specimen, which still contains unreacted methacrylate CC bonds within the vitrified network, whereas the second Tg is due to the specimen cured during the DMA test, as said before. Two Tg values were also observed in the tanδ curve of homopolymer resins and in commercial light-cured composites. In all these cases the two Tg values were attributed to the additional thermal curing occurring.It is reported that the increase of filler particles in the composite's composition introduces a broadening of the tanδ curve The structural heterogeneity of a polymer network formed by photopolymerization of dimethacrylates can be expressed qualitatively by the so called ‘ζ’ parameter. The values of storage modulus of the material drop uniformly with the temperature increase, until they reach a plateau (rubbery region) (). The inverse ratio of the modulus in the rubbery region (rubbery modulus) to temperature at which the modulus was measured was called ‘ζ’ parameter and was generally considered to be inversely correlated to the crosslink density of the polymer network. Higher ‘ζ’ values correspond to lower crosslink density summarizes the initial behavior of the composites, shortly after curing. The effect of composites storage in dry air, distilled water, artificial saliva SAGF or ethanol/water solution (75 vol%), at 37 ± 1 °C for up 1, 7, 30 or 90 days on their viscoelastic properties is shown correspondingly in The studied nanohybrid composites show differences in the composition and the content of organic matrix and the inorganic filler (). Currently, there is no consensus on the positive effect of a certain particle size and shape (spherical or irregular) of filler on the mechanical properties. According to Rueggeberg et al. All composites showed two maxima in the tanδ versus temperature curves which represent two Tg values. They showed statistically the same value for the first Tg at about 40 °C, which is attributed to the light-cured specimens containing unreacted methacrylate CC bonds within the vitrified network. The second Tg was statistically the same for GR and PR (137 and 135 °C correspondingly) and higher than that of TEC (114 °C) and is due to the specimen cured during the DMTA test. Two Tg values were also observed in two commercial light-cured resin composites studied previously This environment is chemically inert and dry while the constant temperature of 37 °C, which is about 15 °C above the environmental at which curing occurred and only about 3 °C below the Tg, may cause post-curing reactions that improve the mechanical properties of the composites.It is interesting to note that Truffier-Boutry et al. C bonds was observed. The Tg of their samples after photopolymerization (t |
= 0) was 55 °C and the samples remained for 24 h at much lower temperature (25 °C) in which polymer chain mobility is restricted. So, it was proposed by the above authors that during the photopolymerization, which is a very fast reaction, a large excess of free volume is trapped in non-equilibrated samples. As they have no time to return to an equilibrium state free volume should decrease below Tg and samples do physically shrink during the storage for 24 h. As a consequence free radicals can also come in contact and undergo termination reactions (post-curing).In our work we have another situation since we studied composite materials which showed a Tg about 40 °C. Also the composites remained in air 37 °C, a temperature very close to their Tg where polymer chains have sufficient mobility. Storage for 24 h caused a significant decrease of tanδ and increase of Tg due to post curing reactions between the double bonds CC of methacrylate groups and/or between free radicals.All composites showed also a second Tg at higher temperatures indicating that additional curing reactions occurred during the DMTA test too. Storage of composites for 7 days caused a further increase of Tg and ΔT revealing that additional post-curing reactions occurred. These samples showed only one Tg which indicates that no more free radicals are present in the system and so there is no further curing reactions as temperature increases Polymer networks are considered to be heavily insoluble structures with relatively high chemical and thermal stability. However, these networks may absorb water and chemicals from the environment. In turn, the network may release components to its surroundings. The phenomena of sorption and solubility may exist as precursors to a variety of chemical and physical processes that create biological concerns as well as producing deleterious effects on the structure and function of the polymeric material. These effects may include volumetric changes such as swelling, physical changes such as plasticization and softening, and chemical changes such as oxidation and hydrolysis Polymer networks based on Bis-GMA is highly susceptible to chemical softening There is concern that the effects of solvent uptake and hydrolytic degradation may lead to a shortened service life of dental restorations. it is concluded that storage of composites in water 37 °C for 24 h caused post curing reactions, which is mostly relevant from a significant increase of Tg. The samples showed also a second Tg at higher temperatures due to additional curing reactions during the DMTA test.Storage for 7 days caused further increase of Tg and the samples did not show a second Tg. Storage of GR and PR for 30 days caused a significant decrease of Tg due to the plasticization effect of water, while storage for 90 days has not further effect. Storage of TEC for 30 or 90 days has no any significant effect on Tg.Storage of composites in artificial saliva SAGF 37°C (The effect of saliva on composites was analogous to that observed for water. Storage for 1 and 7 days seems to cause post-curing reactions. Storage of GR for 30 or 90 days seems to cause plasticization of polymer network, while storage of PR and TEC for 30 or 90 days has no further effect.Storage of composites in ethanol/water solution (75vol%) 37°C ( show that storage of composites in ethanol/water solution (75 vol%) 37 °C for 1 or 7 days caused post curing reactions, while storage for 30 or 90 days caused plasticization of organic matrix.It is interesting to note that a significant increase of “ζ” parameter observed after storage of TEC for 7 days (7.78 × 10−7 |
K/Pa) and mainly after 30 days (18.91 × 10−7 |
K/Pa) may be due to hydrolysis/oxidation of polymer network.All composites showed two Tg values. They showed statistically close first Tg values around 40 °C which is attributed to the light-cured specimens containing unreacted methacrylate CC bonds within the vitrified network. The second Tg is due to the specimen thermally cured during the DMTA test. A general conclusion observed is that all composites stored in any medium for one day showed in DMTA two Tg, while those stored for 7, 30 or 90 days one Tg. Also among the various properties studied the most sensible in the structural changes of composites seems to be the Tg. GR with the highest filler content showed the lowest value of tanδ at 37 °C. The lower the tanδ the quicker the material will respond to load returning faster to its original shape. GR showed also the lowest value for the maximum of the tanδ curve (hmax) and lowest “ζ” parameter followed by PR and then by TEC with the lowest filler content.Storage of composites in dry air at 37 °C which is very close to their Tg (40 °C) for 1 or 7 days caused post curing reactions, while storage for 30 or 90 days has no further effect on composites. Storage in water or artificial saliva 37 °C for 1 or 7 days caused post curing reactions, while storage for 30 or 90 days seems to cause plasticization effect affecting some parameters analogously. Storage in ethanol/water solution (75 vol%) 37 °C for 1 or 7 days caused also post curing reactions, while storage for 30 or 90 days caused plasticization and/or probable oxidation/hydrolysis of polymeric network. Among the composites studied GR seems to have the best behavior.Rheology of graphene oxide suspended in yield stress fluidThis work investigates the rheology of graphene oxide (GO) suspension on a yield stress model fluid. The effects caused by the variation of the GO concentration in the suspension and by the amount of oxygenated groups pendant present on the nanosheets surface, on the rheology of the suspensions were evaluated. The dispersant used was an aqueous solution of Carbopol® Ultrez 10. This is a transparent, nontoxic elastic yield stress fluid. GO nanosheets were produced from synthesis of graphite oxide (GrO) by modified Hummers method, and the nanosheets were characterized by XRD, Raman, FTIR, TEM and AFM techniques. The rheological behavior of these suspensions was characterized by oscillatory and steady-state flow experiments. The GO structure characterization shows that oxygenated functional groups were incorporated in its graphite surface in different levels after using two distinct times of oxidation. GO oxidized for 96 h (GO 96 h) showed greater interplanar distance and also presented few layers when compared with GO oxidized for 2 h (GO 2 h). The GO exfoliation process directly into the aqueous dispersion of Carbopol® showed to be an effective method. As for the pure Carbopol solutions, the GO suspensions were well modeled by the Hershel-Bulkley equation. The increase of GO concentration in the suspensions impairs the level of fluid structure and leads to a decrease in viscosity, yield stress, and elasticity. When compared, the GO 96 h promoted a lower decrease in viscosity, yield stress and elasticity than the GO 2 h suspension. For the suspension with a higher concentration of GO 96 h, it was observed the appearance of hysteresis at low shear rates. These results show that small changes in the GO nanosheets surface can significantly influence the rheological responses of a non-Newtonian fluid.Since the isolation of graphene in 2004 by Geim and Novoselov ]. A disadvantage of the use of graphene is the difficulty to form nanosheets solutions in water, organic solvents and polymer systems, due to the hydrophobic nature of graphene.As alternative, graphene oxide (GO) is being widely investigated, since it provides a more stable dispersion in several solvents with relatively simpler, cheaper, and more productive synthesis, maintaining the desired characteristics for many applications Due to its structure, GO is a 2D amphiphile with a hydrophilic perimeter and largely hydrophobic center The rheological behavior of aqueous and other Newtonian fluids GO suspensions has been widely analyzed in the literature These GO unique properties make it a promising material to be used to improve fluid properties for several potential applications such as thermoeletric conductive paints A comparison of aqueous GO dispersions with melted GO/Poly(methyl methacrylate) (PMMA) composites was performed by Vallés et al. Another mechanism of interaction is analyzed by Zhong et al. As it was demonstrated by the stated studies, the mechanisms of interaction between GO and the non-Newtonian matrices are strongly related to the rheological behavior of the GO dispersions in complex structures. In this context, the objective of this work is to contribute to the analysis of the effects of GO in the rheology of yield stress fluids. In this sense, the main goal of the present work is to investigate the effects of GO nanosheets dispersed in a model yield stress fluid. More specifically, we focus our attention on the effect of the GO oxidation degree and its concentration on the suspension rheology.To this end, we investigated the rheological behavior of GO suspensions in aqueous dispersion of Carbopol®, trade name of a polyacrylic acid. This dispersion produces a yield stress fluid, which is widely used in several industries, and as a model fluid among rheologists The graphene oxide (GO) was obtained from the exfoliation of graphite oxide (GrO). The synthesis of graphite oxide was performed using the modified Hummers method The polyacrylic acid used in the composition of the base fluid was Carbopol®, grade Ultrez 10 (Lubrizol Corporation). The base fluid preparation followed the procedure given by the supplier To prepare the base fluid, Carbopol® (pKa 4.0 ± 0.5) was hydrated in deionized water for 10 min, and then the mixture was dispersed by stirring in a mixer with an anchor-type shovel at 300 rpm for 1 hour. In the next step, the Carbopol dispersion was neutralized by equimolar amount of sodium hydroxide (NaOH) addition. Finally, the suspension remained under mechanical agitation with an anchor-type shovel at 300 rpm for three days to achieve homogeneity.The Carbopol supplier indicates that a highest viscosity plateau and good stability behavior for an aqueous dispersion of Carbopol Ultrez 10 are found with pH between 6 and 9 Carbopol suspensions the pH was kept inside this range. The pH values were measured using the pHmeter Quimis (precision of ±0.02), and are given in . Despite the GO nanosheets produced have many different chemical groups linked in its structure, such as carboxylic acid (pKa 4.3 – 6.6), phenol (pKa 10), among others Three preparation methods were evaluated to prepare the suspension of GO in Carbopol. In preparation method #1, the graphite oxide (GrO) was directly exfoliated into the base fluid (pH 7.0) by 2 h of ultrasonic bath at a maximum temperature of 40 °C. This direct liquid exfoliation promoted the separation of GrO layers, resulting in graphene oxide nanosheets (GO).In method #2, a slightly amount of NaOH was added into the base fluid to promote a slight increase of pH, until pH 7.5. Then, the GrO was added, and the same procedure of method #1 was done.In preparation method #3, the GrO was first exfoliated in deionized water to obtain the GO aqueous suspension. Then Carbopol powder was added, and the preparation procedure for the base fluid preparation was followed.The X-ray diffraction technique (XRD) was used to evaluate the influence of oxygenated groups on the interplanar distance of the GO layers. A Rigaku diffractometer, model Miniflex II, was used, equipped with copper monochromator (radiation λCuKα = 1,54 Å, 5° ≤ 2θ ≥ 60° e ∆2θ = 0,02), operating with a voltage of 30 kV and a current of 15 mA.The thermogravimetric analysis quantitatively indicates the oxidation level of GOs oxidized for 2 and 96 h. The equipment used for the analysis was the SDT-Q600 from TA Instrument. The temperature range analyzed was from 25 °C to 1000 °C, with a variation rate of 10 °C/min and in an inert atmosphere of nitrogen gas (N2).For structural characterization of the GOs oxidized for 2 and 96 h, firstly the respective graphite oxides were exfoliated in deionized water (concentration of 1 mg/mL) in an ultrasonic bath for two hours. After drying the resulting material, the following techniques were applied:From the Raman spectroscopy it was possible to analyze the quality of the GO particles and obtain information on their oxidized groups levels, analyzing the intensity of the D and / or G bands. A Witec confocal microscope, Alpha 300R, 50X objective lens and the 532 nm laser were used. The samples were prepared by diluting 1 mg of GO in 1 mL of deionized H2O, dripping a drop of the solution onto a silicon oxide substrate.Fourier transform infrared spectroscopy (FTIR) reveals the chemical bonds present in the materials. In addition to the GOs oxidized for 2 and 96 h, the spectrum of pure Carbopol® was analyzed. The equipment used to obtain the results was the Bruker spectrometer, model Velex 70, -Latgs detector, MIR source, 10 kHz. The absorption spectra were obtained from suspensions of graphene oxides prepared with a concentration of 1 mg/mL in deionized H2O and pH> 5. The suspension was dripped onto a silicon oxide wafer (SiO2) and the samples were analyzed after dried. Each sample was measured in triplicate.Transmission Electron Microscopy (TEM) was used to analyze the nanosheets morphological characteristics. TEM micrographs were obtained in a Quanta 200 equipment, model FEG-FEI 2006. The measurement occurred under vacuum with the 200 kV electron beam (tungsten filament), with acceleration voltage between 5 and 30 kV. The GO suspensions in water were dropped onto a grid of carbon and copper (Holey Carbon Copper Grids) for TEM studies. Prior to the measurements, the oxidized samples were exfoliated in water using an ultrasonic bath with low power and frequency of 37 Hz, for 2 h at a maximum temperature of 40 °C, later they were dripped in copper grid and evaporated in air at room temperature for 24 h.Atomic Force Microscopy (AFM) was performed to determine the number of GO's layers. The image obtained at AFM relates the thickness of the sample to a color scale. The nanosheets present in the sample are identified and associated with their respective thickness, and from there a thickness distribution profile is established. It was used a Bruker microscope, model Icon Dimension, with a resolution of 512 lines with 512 points in the areas of each image, captured with Scan Asyst mode. The samples were made from the dilution of 1 mg of GO in 1 mL of deionized H2O, and from this suspension, a further dilution of 1:200 with a total volume of 1 mL was prepared. The analyses were performed after complete drying of the samples.Rheological tests were performed on an AR-G2 (TA Instruments) tension controlled rotational rheometer. The geometry used was a cross-hatched parallel plates with diameter of 60 mm and a gap of 1 mm, to avoid wall slip. All tests were conducted at room temperature (25 °C) and pressure. To ensure homogeneity, the suspensions were placed in an ultrasonic bath for 16 min prior to each test. All tests were repeated in triplicate, and a good repeatability was obtained. The suspensions rheology was analyzed for steady-state and oscillatory flow tests. Flow curves were obtained with shear rates ranging from 1000 to 0.01 s−1. Before each test, the sample was kept at rest for 10 min so that thermal equilibrium is obtained, and the fluid microstructure is allowed to rebuild after loading. Moreover, each point was recorded to guarantee that steady state was obtained, considering that the time to reach steady state is of order of the inverse of the applied shear rate The chemical and physical characteristics of the nanosheets is extremely important to understand the rheological responses of the suspensions. Thus, X-ray diffraction (XRD) was used to characterize the GOs oxidized for 2 and 96 h (GO 2 h and GO 96 h, respectively). (a) shows the resultant crystalline patterns of these nanostructures.The XRD graphite pattern has characteristic peaks in 2θ = 26,5° and 2θ = 54,5° to the planes (002) and (004), respectively (a) it is possible to observe that the peaks related to the plane (002) of the GOs shifted to smaller angles, 2θ = 10.4° for GO 2 h and 2θ = 9.7° for GO 96 h. Applying Bragg's Law to the plane (002) the -spacing obtained for graphite is 0.336 nm -spacing for GO 96 h is higher is an indication that this material has a greater amount of oxygenated functional groups introduced between its plans.The characteristic Raman spectrum of graphite presents a high intensity band (G band) at 1575 cm−1, a low intensity band (D band) around 1355 cm−1, besides a 2D band at 2700 cm−1(b) shows the Raman spectra of GO oxidized for 2 and 96 h (GO 2 h and GO 96 h, respectively). The G bands present in the spectra are intense and broad and, unlike graphite [], D bands are also intense and broad, indicating that graphene oxides have a greater number of defects, resulting from the introduction of oxygenated groups in surfaces and edges of its sheets. When calculating the intensity ratio of the ID/IG peaks, it is possible to characterize the defects level of the GOs. GO 2 h presented ID/IG ratio = 0.96, while GO 96 h presented ID/IG = 1.05, which means that GO 96 h was more affected by structural defects.Fourier Transform infrared spectroscopy allowed to analyze which functional groups and bonds integrate the GO samples. (a)-(b) shows the FTIR spectra for GO 2 h and GO 96 h, respectively.Both spectra show vibrations associated with oxygenated functional groups, further evidence that the synthesis of GO was successful. The small band that appears for the frequency of 588 cm−1 in the GO 2 h spectrum indicates the presence of C = O out-of-plane deformation vibrations. The 967 cm−1 peak in the GO 96 h spectrum is associated with the presence of out-of-plane deformation of the carboxylic acid group O stretching, which can be related to alcohols. The peaks at 1218 and 1282 cm−1 also determine the presence of stretching CO bonds, and can be related to carboxylic acids, esters and ethers (epoxy) H deformation vibrations, while the peaks for the frequencies of 1619 and 1615 cm−1 are associated with the C = C stretch, and at 1578 and 1591 cm−1 they are related to the fact that the C = C bonds are arranged in aromatic rings. The bands at 1724 and 1720 cm−1 in (a) and 2(b), respectively, determine the stretching of C = O bonds, which may be related to the presence of carboxylic acids and ketones. The broad bands between the frequencies of 3600 and 2500 cm−1 in both spectra are associated with OH in the samples. The peaks at 3354 cm−1 for the GO 2 h spectrum and 3190 and 3050 cm−1 for the GO 96 h arise due to the -OH stretch from carboxylic acids and hydrogen bonds derived from moisture in the samples. However, the fine peaks that appear at the frequencies of 2917 and 2849 cm−1 in the GO 96 h spectrum indicate symmetric and asymmetric CH stretching that can be related to the appearance of local defects in the GO nanosheets H agrees with the results of Raman spectroscopy, which presented a higher ID/IG ratio for GO 96 h.The evaluation of the oxidation degree of the GO samples was carried out through thermogravimetric analysis. The TGA and DTG graphs for GO 2 h and GO 96 h are shown in In the thermogram it is possible to identify three stages in which loss of mass occurs in the samples. The first stage is correlated to the moisture evaporation that occur in the GOs. For GO 2 h, the highest evaporation rate occurred at 59 °C and the total lost mass was equal to 19%, while for GO 96 h the maximum evaporation rate occurred at 60 °C and the lost mass was equal to 15%. The second stage of mass loss concerns of the degradation of the oxygenated groups introduced in the graphite structure, that is, it indicates the degree of oxidation of the GO. The temperatures where the maximum rates of degradation occurred were close, equal to 206 °C for GO 2 h and 200 °C for GO 96 h. Regarding the mass of degraded functional groups, GO 2 h showed a loss of approximately 26.6% and GO 96 h lost around 35% of its total mass. In other words, the GO oxidized for 96 h has 8.4% more functional groups than the GO 2 h, showing that the exposure time of graphite exerts influence to the oxidation process. Finally, the third stage of mass loss comes from the degradation of the GOs graphitic structure.The morphology and structure of the nanosheets were analyzed through images obtained by TEM and AFM. shows the TEM images obtained for the GOs, where it is possible to see wavy, crumpled and translucent nanosheets with lateral size up to 10 nm, which translate as sheets with few layers of GO for both samples. The images suggest that the exfoliation process promoted the breaking of van der Waals interactions present between the graphical nanosheets The different interplanar distances, number of defects and number of functional groups present between the GOs sheets can influence the exfoliation level of the particles. AFM images allow to estimate the distribution of the nanosheets thickness.a shows the AFM image of GO 2 h, where it is possible to estimate the distribution of the nanosheets thickness. The GO 2 h nanosheets thickness presented an average value of ~17 nm while the GO 96 h showed an average thickness of ~ 6 nm. Thus, it is observed that the exfoliation process was more efficient for this GO 96 h sample, as confirmed by the diffractogram of (a). The small red portion of the topography shown in the AFM image may be related to the presence of artifacts from salt residues from the GrO synthesis process, or by the presence of carbonic debris contained in the sample Since the interplanar distance of the GOs is known from XRD results, it is possible to estimate the number of sheets of the GO samples from the thickness distribution. This result can be confirmed by the thickness distribution profiles for GO 2 h and GO 96 h, respectively, shown in (b) shows that about 71% of the GO 2 h nanosheets produced are up to 20 nm thick. Consequently, the largest percentage of nanosheets has up to 19 layers, and this can be considered a multilayer graphene oxide (mGO). Regarding the GO oxidized for 96 h, (c) exposes that approximately 52% of the nanosheets are up to 6 nm thick and 83% of the sample is up to 10 nm. Thus, slightly more than half of the nanosheets have up to 4 layers (few-layer graphene oxide - FLGO), and more than 80% of the sample is composed of mGO of up to 9 layers. Therefore, the exfoliation of GrO 96 h produces nanosheets with a lower number of layers compared to the GrO 2 h, due to the greater number of functional groups introduced in the sheet surface. It is worth noting that, for technical reasons, the thickness distribution determined for the GO 96 h was processed using an image with a 20 × 20 µm scale, while the GO 2 h results were obtained using an image of 5 × 5 µm. In other words, even on a larger scale the number of layers present in the GO 96 h nanosheets is lower. Having analyzed the chemical and structural characteristics of the nanosheets, the rheological effect of their suspensions on the aqueous dispersion of Carbopol is discussed below.The flow curves for the suspensions of 1 mg/ml of GO 2 h in 0.15 wt% of Carbopol aqueous dispersion obtained for the three preparation methods described in . The experimental data in all the plots show the mean value of the three data measured for each sample.The pure Carbopol®'s aqueous dispersion is a model yield stress fluid that is usually modeled by the Herschel-Bulkley equation []. Below the yield-stress, the Herschel-Bulkley model predicts that the shear-rate is identically zero. Above the yield stress the shear stress is given by:Three inherent parameters of the fluid govern the Herschel-Bulkley model, namely, the consistency index (k), the flow index (n) and the yield stress (τy). This behavior is commonly observed in dispersions above a certain polymer concentration. These fluids maintain an irregular internal order forming three-dimensional structures with sufficient rigidity to withstand an external stress lower than τy, offering flow resistance although allowing elastic deformation shows that the three GO suspensions, prepared with different methods, present similar behavior, and can be modeled by the Herschel-Bulkley equation, similar to the pure Carbopol®'s aqueous dispersion.The rheological parameters that describe the behavior of the three suspensions as well as their respective pH values are presented in The results of stress sweep tests of the same suspensions are shown in (b). It can be noted that the results are similar for the three cases. Three regions can be identified, which are also characteristic of pure Carbopol dispersions: the linear viscoelastic region (LVR) in the range of low stress, where the storage and loss moduli, G’ and G” respectively, are independent of stress []. The stresses applied in this region are not sufficient to break the bonds that make up the microstructure of the fluid The results show that the suspension prepared by method #3 has a slightly lower G’. Since in method #3 the GrO goes through the exfoliation process first, it is possible that the polymer chains did not open in the same way as for the other suspensions leading to weaker elasticity The effect of GO concentration in the suspension rheology is presented in (a) and (b). The suspensions were prepared with concentrations of 0.1, 1 and 5 mg/mL of GO oxidized for 96 h in a 0.30% wt Carbopol® aqueous dispersion. The flow curves obtained experimentally, with their respective Herschel-Bulkley curve fittings, are shown in (a). The base fluid presents a slightly higher viscosity than the 0.1 and 1 mg/mL suspensions of GO. As the concentration increases to 5 mg/mL of GO, a more pronounced decrease of the suspension viscosity is noted, while the Herschel-Bulkley equation continues to model the fluid behavior. Therefore, the increased concentration of GO promotes the decrease of the suspension viscosity, as it will be discussed in the follow.The rheological characteristics of the aqueous dispersion of Carbopol come from polymer chains ionization. After dispersion neutralization, the polymer backbone is negatively charged, and, consequently the chain uncoils and absorbs water, developing the swelling and forming a microgel unit. It turns out that for dispersions with Carbopol concentrations above a certain critical value, these units are so close to each other that repulsive interactions restrict the motion among them, and a jammed structure (reversible network) is formed [], being necessary to apply a stress above the critical (yield) value to break the structure and allowing the hydrogel to flow. Since graphene oxide has carboxylic acids, as it was demonstrated on the chemical characterization, it becomes negatively charged in water, and it will interact with Carbopol. However, these repulsive electrostatic interactions between GO and the hydrogel are weaker than the polymer-polymer interactions, which will form weaker bonds within the material ], and their interactions will dominate within the system, causing a perturbation and developing at the end a weaker microgel network ] developing a significant viscosity decrease, as it is observed for the higher concentration in Stress sweep results for the same suspensions are shown in (b). It can be observed that all suspensions have G’ higher than G” in LVR, which indicates the behavior of a gel The effect of GO oxidation time on fluid rheology is presented in (a) shows the flow curves of the suspensions with 1 and 5 mg/mL of GO oxidized for 2 and 96 h. shows the values of the rheological parameters of the suspensions prepared for the two types of GO, as well as their pH values, which are within the pH range for the high constant viscosity plateau []. The results show that the suspensions prepared with GO oxidized for 2 h show higher drop in rheological behavior than the suspensions prepared with GO oxidized for 96 h. This behavior is addressed in the following discussion.Although both GOs contribute to the drop of the rheology of the suspensions, less oxidized GO leads to higher decrease of rheological parameters. This behavior is another indication that the relatively smaller number of ions present in the GO interface is an important factor for the weakness of the microstructure of the Carbopol dispersion. TGA results showed that the GO oxidized for 96 h has higher degree of oxidation (8.4% more functional groups) in relation to the GO oxidized for 2 h, as well as the XRD have shown higher -spacing for GO 96 h, suggesting the increase of the effect of oxidation time. Such increase in oxidation degree probably is sufficient to balance the repulsive electrostatic interactions between GO 96 h and Carbopol, and developed a tougher hydrogel when compared to the GO 2 h. Thus, these results lead the suggestion that there is a competition between interactions that affects the rheological behavior of the Carbopol with different oxidation degree of GO. If from one side, the increase in negative charges from the carboxylic acids of the GO turns the hydrogel weaker, from other side the increase in oxygenated groups could enhance the interactions by hydrogen bonding The responses of the storage and loss moduli as a function of the stress amplitude are correspondent to a gel, with G’ higher than G” in the LVR, as shown in (b). Stress sweep tests also showed significant differences between GO 2 and 96 h. The curves for the suspension made with GO oxidized for 2 h show lower level of elasticity, with G’ smaller than GO oxidized for 96 h. These results suggest that the competition between interactions discussed for the steady shear behavior, also influence the viscoelastic behavior of the Carbopol. The suspension with GO 96 h, which has more oxygenated groups, behaves as a tougher gel than the GO 2 h, which the crossover point occurs at lower stress amplitudes. This is another indication, that GO oxidation degree influences the rheological behavior of the Carbopol, due to different interactions between GO and the hydrogel.Another interesting result concerns the hysteresis of suspensions made with particles of different levels of oxidation. shows the results of hysteresis tests performed to evaluate time effects, for the base dispersion of 0.3% wt of Carbopol® with the suspension of 5 mg/mL of GO oxidized for 2 h, and with 5 mg/mL of GO oxidized for 96 h.Thixotropic behavior can be related to weak attraction forces, which allows the fluid microstructure to change reversibly when submitted to a moderate stress The rheological behavior of graphene oxide suspensions in the aqueous dispersion of Carbopol® was investigated. The modified Hummers method adopted culminated in the effective oxidation of the graphite particles. The use of a longer oxidation process time resulted in a GrO with greater interplanar distance and higher number of oxidized groups. These defects incorporated on the sheet surface, which is a characteristic that facilitated the exfoliation, and promotes the achievement of a GO with fewer layers.As for the pure Carbopol® solutions, the suspensions of GO were also modeled by Herschel-Bulkley equation, meaning that the material continues to behave as a yield stress fluid. However, the addition of GO nanosheets in the Carbopol®, have led to the development of a weaker gel, with the decrease in shear viscosity and yield stress. This behavior was suggested as a result from the weaker repulsive electrostatic interactions between GO and Carbopol, which have caused the perturbation of the hydrogel network. Moreover, the increase in concentration could lead to agglomerates and result in micro-phase separation, turning in a weaker gel. The viscoelastic properties were also investigated, and it was observed that elasticity also decreases as an influence from GO incorporation, with the crossover of G’ and G” occurring at smaller stress amplitudes.The influence of GO oxidation degree in Carbopol suspensions was also studied, by analyzing two different oxidations times, 2 h and 96 h. The lower oxidation time leads to higher decrease in shear viscosity, yield stress and elasticity, which suggests a competition between interactions that affects the rheological behavior of the Carbopol with different oxidation degree of GO. The increase in oxygenated groups seems to lead the hydrogel with GO 96 h to be a tougher gel, when compared to the suspension with the GO oxidized for 2 h.Finally, hysteresis curves were performed and a time-dependent response was observed for the suspension with 5 mg/mL of GO 96 h, mostly due to the presence of more oxygenated groups and higher interactions between the hydrogel and GO. The increase in oxidation time have led to increase the interplanar distance between the GO sheets, and the obtaining of less stacked layers, increasing the dispersion. Thus, a better interaction with the hydrogel, would have required a different time to the reorganization of the developed microstructure during the flow.The results obtained have shown the influence of the oxidation degree on the rheological behavior of the suspensions of GO in Carbopol. However, more formulations and more studies should be performed with the aim to understand the fundamental chemical and physical interactions between the GO and the fluid. With that aim, a study is on progress to investigate the effect of charges by using a functionalized GO nanosheet and to visualize the fluid microstructure during the flow.Behaviour of beam–column joints made of recycled-aggregate concrete under cyclic loading► Focus on mechanical behaviour of recycled-aggregate concrete. ► Use of recycled concrete aggregate. ► Study of cyclic loading behaviour of beam–column joints.In this work, a recycled-aggregate concrete (RAC) was prepared by replacing 30% virgin with recycled concrete aggregate coming from an industrial crushing plant in which concrete from building demolition is suitably treated. A reference concrete was also prepared by using 100% virgin aggregate with the same workability and strength class. Concrete specimens were manufactured for evaluating compressive, tensile, flexural and bond strengths with reinforcing steel bars, as well as static elastic modulus to assess RAC suitability for structural use. The behaviour of beam–column joints (scale of 2–3) made of reinforced concrete and subjected to cyclic loading was also studied. The results obtained gave experimental evidence of the suitability of RAC for structural use.Concerning the concept of sustainable development, 21st century technologies should take into account their social and environmental costs Recycled-aggregate concrete (RAC) was prepared and tested in which coarse virgin aggregate, extracted from quarries, was replaced with 30% recycled aggregate coming from treatment (crushing, cleaning, sieving) of concrete from building demolition. This operation enables reduction of the consumption of non-renewable resources and also the disposal of waste from demolition in landfills Results developed by several researchers In this work, RAC was characterized from a mechanical point of view, and RAC structural behaviour was tested by means of beam–column joints (scale of 2–3) under cyclic loading to test their toughness and ability to dissipate energy. Control specimens were also made of ordinary virgin-aggregate concrete, to compare the behaviour of RAC.Previous results, concerning beam–column joints, obtained by the authors on concretes prepared with 100% recycled aggregates replacing virgin aggregate A commercial portland-limestone blended cement, Type CEM II/A-L 42.5R according to the European Standards EN-197/1 Virgin aggregate, gravel (22 mm maximum size), fine gravel (12 mm maximum size), sand (5 mm maximum size) and fine sand (4 mm maximum size) were used.A recycled aggregate fraction (12 mm maximum size) was also used as a partial replacement for fine gravel. This fraction came from an industrial crushing plant in Villa Musone (Italy), in which rubble from building demolition is obtained for recycling in new concrete mixtures. Such rubble is checked for proper quality in order to avoid hazardous materials such as asbestos or chalk, then ground, cleaned, sieved, and graded. Its average composition is 84% concrete, 13% masonry, and 3% miscellaneous (mainly bitumen, but also glass and wood) by volume.Bulk specific gravities (in saturated surface dry condition) and water absorption values of the aggregate fractions evaluated according to UNI EN 1097-6 , and their grain size distribution curves, evaluated according to UNI EN 933-1 A water-reducing admixture was added to mixtures in all cases. It was a carboxylic acrylic ester polymer superplasticizer in the form of 30% aqueous solution.Concrete mixture proportions are reported in . Both control as well as RAC concretes were prepared with the same water to cement ratio of 0.53 and the same fresh concrete workability (slump value of 180 mm), evaluated according to EN 12350-2 In order to optimize the grain size distribution of the solid particles in the concrete, the aggregate fractions were suitably combined according to the Bolomey particle size distribution curve The dosages of cement (350 kg/m3), water, superplasticizing admixture (1.0% by mass of cement), fine sand (20% of the total aggregate volume), sand (20% of the total aggregate volume), gravel (30% of the total aggregate volume) were kept constant. On the other hand, the fine gravel fraction was dosed at 30% of the total aggregate volume in the control mix, while it was completely substituted by the recycled aggregate fraction (characterized by a similar grain size distribution, see ) in the recycled-aggregate mixture. The different dosages (in kg/m3) between fine virgin gravel and recycled aggregate is due to their different specific gravity values (see ). Moreover, the recycled-aggregate fractions were added to the mixture after water-soaking, in SSD condition. In fact, it is believed that pre-soaked aggregates can be more effective in order to create an internal water supply that is able to reduce drying-shrinkage, as well as to avoid mixing water absorption in aggregates during mixing and, consequently, to maintain concrete workability 100 mm cubes, were manufactured for compression and splitting tension tests, according to EN 12390-1 Moreover, three specimens were made for each mix type for pull-out testing. Ribbed steel bars were placed in concrete cubes (150 mm in size) with a constant embedment length of five times the bar diameter (16 mm) according to CEB Recommendations RC 6 The beam–column joints (scale of 2–3) were made for each concrete mix, according to “Recommendations for Design of Beam–Column Joints in Monolithic Reinforced Concrete Structures” reported by ACI-ASCE Committee 352 shows one of the test specimens. The beam–column joint specimens were cured in air at room temperature of about 20 °C for 28 days to simulate the real conditions at the building site.The reinforcement arrangement was designed in order to concentrate damage in the beam portion close to the joint by using ribbed bars made of FeB38k steel according to Italian Ministerial Decree D.M. 9 1996.The control joint was built with 220 × 220 mm column (2050 mm high) reinforced by 4φ16, stirrups φ6 every 50 mm, and 180 × 280 beam (1650 mm long) reinforced by 4 + 4φ12, stirrups φ6 every 140 mm.On the other hand, the RAC joint was built with 250 × 280 mm column (1950 mm high) reinforced by 8φ12, stirrups φ6 every 50 mm in the most critical zone (and φ6 every 100 mm for the remaining portion), and 200 × 250 beam (1650 mm long) reinforced by 3 + 3φ12, stirrups φ6 every 50 mm in the most critical zone (and φ6 every 100 mm for the remaining portion). The RAC beam–column joint design was made in order to take into account the suggestions of Eurocode 8 Compressive strength was evaluated according to EN 12390-3 Due to a higher degree of compaction (i.e. ratio between densities of actual specimen and fully compacted specimen, 0.98 against 0.97), the RAC mix had slightly higher early age strength as compared to the control mix.Tensile strength was evaluated according to EN 12390-6 Tests were carried out after 3, 7 and 28 days of curing and results obtained are reported in . Results confirm that for similar compressive strength, recycled-aggregate concrete (RAC) is approximately 10% weaker than virgin-aggregate concrete (control mix), as already stated in the literature Flexural strength was evaluated according to EN 12390-5 The flexural behaviour of the specimens was evaluated after 3, 7 and 28 days of curing. The results obtained are reported in . Results confirm data of splitting tension testing: the MOR of RAC is around 0.83 of that of concrete made with virgin aggregate, except for early age RAC (in this case MOR is about 0.70 of that of concrete made with virgin aggregate).The stress–strain curve and modulus of elasticity were determined by testing a 100 mm diameter cylinder 300 mm long in compression according to Italian Standards UNI 6556 Results obtained after 28 days of curing are reported in . The data confirms the results obtained by Xiao et al. The pull-out bond strengths for monotonic loading of the two concretes were evaluated after 28 days of curing. Pull-out tests were carried out according to the CEB Recommendations RC 6 Results obtained after 28 days of curing are reported in . Specimens made from recycled-aggregate concrete had similar bond strength as compared to those made with control concrete. In fact, bond strength seems to be influenced by the quality of the cement paste more than by the kind of aggregate used for preparing concrete. Indeed, data reported in the literature show that, in case of equal compressive strength, bond strength between RAC and ribbed steel rebars is equal, if not higher, than for normal concrete The bond behaviour of these concretes was also studied in order to compare them by means of low-cycle loading. Such loading condition can correspond to loading of real structures subjected to earthquake or high wind and is characterized by a load history containing few cycles (generally less than 100) but having a large range of bond reversal stresses Two pins at the upper and lower ends of the column restrained the beam–column joint and the column was vertically preloaded with 200 kN. In this way, the joint bore roughly the same load that it would have supported if included in a typical reinforced concrete framework.In addition to the static loading, the presence of dynamic loads, such as earthquakes or high winds, was simulated by applying a displacement (±v) at the free end of the beam (see Fifteen cycles were applied: three cycles with an amplitude of ±25 mm, three further cycles with an amplitude of ±50 mm, then with increasing amplitude of ±75 mm, ±100 mm and finally ±125 mm. The corresponding drift ratios of such amplitudes were 0.015, 0.030, 0.045, 0.061, 0.076 and 0.091 respectively for increasing amplitudes. In the case of the joint prepared with RAC a further cycle with an amplitude of ±150 mm was carried out.A dynamometer was attached at the upper end of the column in order to monitor the axial load on the column itself (Q). Eight hydraulic jacks were placed under a basement carrying the beam–column joint in order to jack it up.The free end of the beam was alternatively displaced by means of two hydraulic jacks connected in series, to secure a maximum displacement up to 200 mm. One of these jacks was equipped with an inductive half bridge in order to measure the displacement (v, mm) and with a dynamometer in order to monitor the load necessary to impress it (Q, kN). Eight inductive half bridges were placed on the joint, four for each side (see Both joints showed a ductile way of rupture, typical of ‘strong column – weak beam’ design criterion, with cracking appearance on the beam portion close to the joint due to flexural stresses, followed by concrete cover spalling (see ). Columns never showed cracking during the test.Results obtained by cyclic loading are shown in , in which the applied load (Q) is reported vs. the vertical displacement (v) at the end of the beam. Hysteretic loops were quite wide in both cases, showing high energy dissipation capacity, coherently with the ductile mechanism of failure visually observed (see ). In particular, the beam–column joint prepared with RAC concrete showed a very stable behaviour up to 125 mm cycles and the loss in strength was very little cycle by cycle. the values of supplied and dissipated energies as well as their ratios are reported, concerning cycles with amplitude of 75 mm and 125 mm, respectively.In particular, the RAC joint subjected to 75 mm cycles showed higher energy dissipation with respect to REF joint (see ). The reason probably lies in the lower loss in stiffness detected for the RAC joint (characterized by an initial lower elastic modulus of concrete, see ). Only during the last cycle before the collapse (i.e. the third with an amplitude of 125 mm), for the RAC joint a certain reduction of both supplied and dissipated energies was detected, due to the failure of one of the steel reinforcements of the beam.However, REF and RAC joints showed very similar values of the supplied to dissipated energy ratios during the last three cycles before the collapse (see the envelop curve of the loading cycles for the RAC joint is shown. It was obtained following the indications reported by Kim and LaFave ), displaying the most distinct stiffness changes. The first change in stiffness (A key point) is due to the first concrete cracking, the second (B key point) is related to the initial yielding of longitudinal beam reinforcements, the third (C key point) corresponds to the maximum shear/stress response and, finally, the fourth (D key point) corresponds to the initiation of joint failure. the values of load and displacement (positive and negative) corresponding to the four key points of the envelop curves are reported for both REF and RAC joints. On the basis of the displacements values after initial steel yielding (labelled Δy) and just before shear failure of joint (labelled Δfin), an estimation of the joint ductility can be carried out by calculating the ratio Δfin/Δy. The values of Δfin/Δy obtained for both positive and negative displacements are reported in as well. RAC joint generally showed slightly higher values of Δfin/Δy with respect to REF joint confirming a good ductile behaviour. These values are confirmed by Xiao et al. the values of the maximum momentum experimentally obtained (Mmax) and those of the design momentum (Mdes) are reported for both REF and RAC joints, by keeping into account their different geometry. On the basis of the ratios Mmax/Mdes, clearly higher than 1.0 for both joints, it can be observed that the value of the design momentum was widely overcome during experimental texts in every case, particularly for the REF joint.In conclusion, by comparing the results obtained for the two concrete mixture, it can be noted that, when recycled aggregate instead of fine gravel is used (at 30% by volume), almost the same mechanical performances can be achieved. In particular, at the same level of compressive strength, lower tensile strength, flexural strength and static elastic modulus (generally −10%) are found for the concrete with RAC. On the other hand, about the same bond strength between concrete and ribbed steel reinforcing bar is obtained independently of the kind of aggregate used.In addition, on the basis of the results obtained through cyclic loading tests of beam–column joints made of either REF or RAC concrete, evaluated by means of parameters such as cracking patterns, supplied and dissipated energy, ductility and design values, the joint made of RAC showed adequate structural behaviour. It should be kept in mind that, however, RAC joint must be properly designed in order to achieve safe structural performance, by suitably considering the actual RAC shear strength and stiffness.Synthesis and characterization of boron carbon nitride films by radio frequency magnetron sputteringBoron carbon nitride (BCN) films were deposited on silicon substrates by radio frequency (r.f.) (13.56 MHz) magnetron sputtering from hexagonal boron nitride (h-BN) and graphite targets in an Ar–N2 gas mixture of a constant pressure of 1.0 Pa. During deposition, the substrates were maintained at a temperature of 400°C and negatively biased using a pulsed voltage with a frequency of 330 kHz. Different analysis techniques such as X-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES), Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy, X-ray diffraction (XRD) and scanning Auger electron microscopy (SAM) were used for characterization. In addition, the mechanical and tribological properties of the films were investigated by nano-indentation and micro-scratching. The carbon concentration in the films could be adjusted by the coverage area of a graphite sheet on the h-BN target, and decreased with increasing bias voltage. It was found that the ternary compound films within the B–C–N composition triangle possessed a less ordered structure. BN, BC and CN chemical bonds were established in the films, and no phase separation of graphite and h-BN occurred. At zero bias voltage, amorphous BC2N films with atomically smooth surface could be obtained, and the microfriction coefficient was 0.11 under a normal load of 1000 μN. Hardness as determined by nano-indentation was usually in the range of 10–30 GPa, whereas the Young’s modulus was within 100–200 GPa.Boron nitride (BN) films have attracted much attention due to their highly desirable mechanical, thermal, electrical and optical properties Up to now most of the BCN compounds were prepared by chemical vapor deposition (CVD) methods from different organic precursors In the present experiment, BCN films were deposited by radio frequency (r.f.) magnetron sputtering of a complex target composed of h-BN and graphite, and the structure as well as the mechanical and tribological properties of the films were studied in more detail.Boron carbon nitride films were deposited by r.f. magnetron sputtering at a frequency of 13.56 MHz from graphite and hexagonal boron nitride targets. The experimental setup has been described elsewhere In order to reduce the contamination by oxygen and moisture, the vacuum chamber was evacuated to a background pressure lower than 2.0×10−7 Pa using a Cryo-Torr 8 cryogenic pump (CTI-Cryogenics) after a 150°C bake for 36 h. The residual gases in the chamber could be monitored in situ by a quadrupole mass spectrometer (RGA 200, Standard Research Systems). The substrates were single crystal Si(100) wafers with a thickness of 500 μm. They were cleaned ultrasonically in acetone, methanol and deionized water in sequence, and then dipped in a 5% HF aqueous solution, and finally dried by nitrogen blowing. Prior to deposition, the substrates had been baked at ∼600°C in a vacuum for outgassing (∼1 h), and then sputter-cleaned in an Ar discharge at a bias of −300 V for 30 min. The magnetron target was presputtered with the shutter closed at 80 W for 30 min. All the depositions were carried out in a gas mixture of Ar/N2=4/1 at a total pressure of 1.0 Pa adjusted through a MKS multigas controller.The BCN films were characterized for composition, chemical bonding, crystalline and phase structure, as well as mechanical and tribological properties, etc. The chemical bonding states and compositions of the films were investigated by X-ray photoelectron spectroscopy (XPS) (Quantum 2000, Physical Electronics) with a source of monochromatized AlKα. Before analyzing, the films had been sputtered by Ar ions at 2 kV for 3 min to remove the impurities absorbed on the film surface. Auger electron spectroscopy (AES) (PHI 670, Physical Electronics) was used to detect the depth profiles of the constitutional elements by Ar+ sputtering (2 kV, 1 μA). The high-resolution images of the film surface could be obtained by scanning Auger electron microscopy (SAM) when the AES system was equipped with a scanning device. The X-ray diffraction (XRD) spectra of the films with CuKα radiation were recorded by a Philips X’Pert diffractometer using a grazing incidence mode (incidence angle 2°, 40 kV, 50 mA). Fourier transform infrared spectroscopy (FTIR) (FT-IR 1600, Perkin-Elmer) was used to characterize the phase structure in both transmission and reflection modes in the 600–2000 cm−1 wave number range. The background spectra of an uncoated Si substrate had been subtracted. In addition, Raman spectra were recorded at room temperature in backscattering geometry over a range of 1000–2000 cm−1, using a Renishaw 2000 spectrometer with an Ar-ion laser operating at a power of 20 mW. The surface imaging was performed by atomic force microscopy (AFM) (Autoprobe CP, Park Scientific Instruments), from which the root mean square (rms) roughness could be determined. Nano-indentation was carried out with the Triboscope nanomechnical test system (Hysitron Inc.) equipped with AFM under ambient conditions (20–24°C, 45–55% RH). The hardness and elastic modulus were evaluated from the loading–unloading curves a–c shows the typical XPS spectra of B1s, C1s and N1s core level electrons for the films deposited at a temperature of 400°C with the substrate grounded. As determined by the compositional analysis, the atomic ratio of boron to nitrogen (B/N) was approximately 0.93, and the carbon concentration in the films was 49.2 at.%, whereas the oxygen content was below 2 at.%. This analysis result indicates that the composition of the films is quite close to that of BC2N compound, a material which has attracted much attention due to its prospective applications a, the position of B1s peak at 190.2 eV is considered to be attributed to BN bonding in h-BN a. Therefore, the oxidation of the films can be negligible, although the reactivity between B and O is high.The spectrum for C1s electrons shown in b has a peak energy of 284.4 eV, which can be assigned to CC bonding. The broadening of the spectrum (with the FWHM of 2.7 eV) indicates also the existence of other chemical bonds. The binding energies of C1s electrons have been reported to have peak values at 283.0 and 284.3 eV for B4C and BC3.4, respectively The XPS spectrum for N1s electrons is shown in c. A peak energy of 398.5 eV is in accordance with BN bonding. The FWHM was approximately 2.3 eV. The shoulder at higher energy side implies the formation of CN bonds, since the binding energy of N1s electrons in CNx films has been reported to be approximately 399 eV AES sputter depth profiles revealed that the composition of the films was uniform throughout the thickness. The oxygen content in the films was lower than 2 at.%. shows the carbon concentration incorporated into the films and the B/N atomic ratio as a function of negative bias voltage (Vb). It can be seen that the carbon concentration decreases abruptly at bias voltages higher than 100 V, and the atomic ratio of B/N decreases with increasing magnitude of the voltage, from 0.93 at 0 V to 0.65 at 300 V. If the bias voltage is larger than 300 V, the resputtering effect becomes much pronounced, giving a very low deposition rate of the films.The grazing incidence XRD (incidence angle 2°) spectra were recorded in a wide scanning range (2θ=20–100°). However, no diffraction peak could be detected for all the films prepared in this experiment, suggesting that the films possess a less ordered structure. Previous experiments have shown that BCN compounds have an amorphous structure in most cases, and their crystallization temperature can be up to 1470 K High resolution SAM revealed that the films are smooth and have a featureless morphology, similar to the amorphous diamond-like carbon (DLC). Other experiments with the aid of SEM and TEM also confirmed the amorphous structure in BCN films shows FTIR spectra of BCN films deposited at different negative bias voltages. At a low bias voltage (<100 V), two distinct absorption bands at approximately 780 and 1380 cm−1 are displayed, which can be interpreted as the out-of-plane B–N–B bending vibration and the in-plane B–N stretching vibration, respectively ). The non-stoichiometric films cannot satisfy the prerequisite for the growth of the cubic phase. shows the Raman spectra of BCN films deposited at different negative bias voltages. When the voltage is lower than 100 V, the spectra are similar to those of amorphous carbon, with two broad bands centered at approximately 1380 and 1560 cm−1, corresponding to the D and G bands of disordered sp2 carbon, respectively . In addition, the absence of Raman peaks, corresponding to the TO (1057 cm−1) and LO (1306 cm−1) modes of c-BN as well as the E2g mode of h-BN (1367 cm−1), suggests the poor crystallinity of the films The nano-indentation was performed under ambient conditions. Each experimental value was calculated by averaging the data of at least 10 different indentations. shows the nano-hardness and Young’s modulus of the BCN films as a function of negative bias voltage. At a lower bias voltage (≤100 V), the hardness is in the range of 11–14 GPa, which is comparable to that of h-BN films (∼12 GPa) as well as BCN films by CVD methods ). In addition, the formation of BC and CN bonds may make a contribution to the higher hardness.For the determination of a friction coefficient (μ), the microscratching was carried out in this experiment. A conical diamond tip scratched the film surface over a small length (4 μm) under a constant normal force within 30 s. shows the microfriction coefficient as a function of normal load for the amorphous BC2N films deposited at zero bias voltage. The microfriction is in the range of 0.08–0.19 at different normal loads ranging between 300 and 2500 μN, and increases almost linearly with increasing normal load. The depth of scratch at a typical load of 1000 μN is approximately 4 nm. Compared with the macrofriction coefficient of BCN films by dual cathode magnetron sputtering (μ=0.42–0.63) shows the root mean square (rms) roughness and the microfriction coefficient of the BCN films as a function of negative bias voltage. At zero bias voltage, the film surface is atomically smooth as the rms roughness is only 0.2 nm, which will be attractive for solid lubricant application. When the bias voltage is higher than 100 V, the rms roughness becomes much larger (normally in the range of 2–3 nm). Correspondingly, the microfriction coefficient under a normal load of 1000 μN also has an abrupt increase at bias voltages above 100 V. Therefore, the friction coefficient should be associated with the surface roughness, especially in the case that the ploughing mechanism is the major contributor BCN films have been deposited on Si(100) substrates by r.f. magnetron sputtering from graphite and h-BN targets in an Ar–N2 gas mixture. The films were characterized with different techniques in more detail. The experimental results indicate that BN, BC and CN chemical bonds are established in the synthesized BCN compound films located in the B–C–N composition triangle. The films possess an amorphous structure with B–C–N hybridization, rather than a simple mixture of graphite and h-BN. The bias voltage can tailor the film composition as well as the structure and properties. However, no evidence for the formation of sp3-bonded phase can be found from both FTIR and Raman spectra, even if the films were deposited at a high bias voltage up to 300 V, suggesting that the BCN films prepared in the present experiment have a highly disordered structure. At zero bias voltage, amorphous BC2N films with an atomically smooth surface can be obtained, and the microfriction coefficient is low (μ=0.08–0.19) when the normal load is within 300–2500 μN. The microfriction is also related to the roughness of the film surface. As determined by nano-indentation, the hardness is usually in the range of 10–30 GPa, whereas the elastic modulus is within 100–200 GPa.Experimental study and numerical analysis of a composite truss jointThis paper presents a model test and numerical finite element analysis (FEA) on the mechanical behavior of a composite joint in a truss cable-stayed bridge. The model test with the scale of 1:2.5 for the truss joint was conduct to fully understand the safety and serviceability. In the experiment, stress distribution, crack resistance ability and shear resistance of headed studs were carefully measured to investigate the mechanical performance, force transmission of the joint part. The maximum strain of the steel plate and concrete chord remained in the linear elastic region until 1.7 times the design load, which means there is a significant safety margin for such composites. On the basis of the experimental results of composite truss joints, three-dimensional finite element models are established. The results of the finite element analysis are in good agreement with those of the tests in terms of strength and stiffness. It is also expected that the results presented in this paper would be useful as references for the further research and the design of composite truss bridges and composite joints.► Experimental study on mechanical performance of composite truss joint. ► Finite element analysis on composite truss joint. ► Mechanical behavior of headed studs in composite structure.) is the first composite truss cable-stayed bridge in China. The total length is 1212 m, including a main span of 708 m and side spans of 63 m. It is also a double-deck highway bridge, with eight lanes on the upper and six lanes on the lower deck (). It has been fully operational since 2010.The cross section of the main span is composed of orthotropic steel deck and steel truss. While the cross section of the side spans consist of steel reinforced concrete (SRC) chord, steel truss web members, steel inclined brace and prestressed concrete deck The composite joint is one of the most important parts for such a bridge structure, and the mechanical behavior of the composite joint is very complex and difficult to analyze in practical design. As shown in , the composite truss joint includes the steel vertical web member, the steel diagonal web member, the steel inclined brace, the shaped steel chord and the concrete chord. The steel vertical web member connects the steel diagonal web member with gusset plates conventionally. Headed studs are used to resist shear forces in the interface between concrete chord and other steel members. This innovative composite truss structure was initially used in a large cable-stayed bridge in China, that needs experimental and numerical analysis to fully understand the safety and serviceability. Constructability and durability also should be considered in the design of the composite joint.With the development of composite truss bridges, some studies on various issues concerning the design and construction of similar structures have been reported. Pison Udomworarat et al. Only a little research has been carried out on the behavior of the composite joint. Also there is very little literature on the design and construction of the joint. Therefore, a full understanding of the mechanical behavior of such a joint is necessary for its application in composite truss bridges.One of the typical composite joints, E5, as shown in , is selected as an objective to be tested. A finite element analysis is performed to facilitate the interpretation of the test results.A test specimen with the scale of 1:2.5 is designed for the typical composite joint E5, as shown in . All the steel members in the specimen were made of Q345qD. The material properties of the steel members have been determined by tensile coupon tests as prescribed by the relevant standards. The results of steel coupon tests with different thickness are summarized in . The concrete filling the chord of specimen is C50 concrete. The material properties including compressive strength, tensile strength and Young’s modulus of the cube at 28 days after the casting of the concrete are given in . The specification and properties of the headed studs connected the concrete and steel members in the specimen are shown in According to the similarity principle, the ratios between the test model and the prototype are listed in . As the similarity ratio of stress between the prototype and the model is 1:1, the stresses that result from the model test will be the stresses for the prototype.The test setup should have enough strength and stiffness to subject loads of opposite directions. The specially designed test setup is as shown in . The main components of the test setup were the reaction girders. Reaction girders 1 and 2 were connected with finely-rolled threaded bars to form a self-equilibrium structure. The reaction girder 3 was connected with the diagonal web member through the hinge. The loading was applied through hydraulic jacks. When the jacks pushed, the vertical web member was in compression and the diagonal web member reacted axially in tension. Reaction girder 2 was supported to the reaction wall to transfer horizontal compression forces. The forces in the diagonal web member were transferred to the reaction wall through the reaction girder 3. In this paper, positive signs are taken for tensile forces while negative signs are taken for compression forces.After the drying shrinkage of the specimen had stabilized, the preloading should be taken into account to check the good contact between the support and loading equipments, the reliability of all the test equipments and the workability of all the measurement instruments.In the test, a stepped loading method was used. At each loading step, 10% of the design load was applied and the maximum load applied was 170% of the design load. There were seventeen loading steps. Ten minutes were allowed for the specimen to stabilize at the maximum loading step and five minutes at the other loading steps. Values of the applied loads were determined from the analysis of bridge structure using an FEA model and the worst-case load. The maximum loads applied at each member are shown in One of the main tasks of the model test was to measure strains in the joint model under the worst-case load. Strain gauges were installed on the surfaces of the model.A total number of twenty 3-dimensional strain rosettes were mounted on the gusset plate to measure the strain distributions at these hot spots. To determine the axial forces on the chord and gusset plate and verify the loading condition of the tests, additional strain gauges were mounted on the chord, gusset plate and reinforcement steel bar. shows the location of strain gauges for the composite joint. The strain gauges installed on the surface of steel, concrete and reinforcement steel bar are show in (a)–(c), respectively. All the information obtained from the transducers, gauges, and load cells were automatically recorded by a data acquisition system at regular intervals during the tests.To extend the interpretation of the results and observations obtained in the test and gain a better understanding of the behavior of composite joint, a numerical study on joint specimen was carried out using the finite element analysis program.The three-dimensional FEA model is shown in . The steel plate was represented by SOLID45 elements, the concrete of chord was modeled with SOLID45 elements, and spring elements COMBIN14 were chosen to simulate the headed studs. In this study, the corresponding nodes of the concrete and steel at the same position were connected by spring elements in the tangential directions and coupled in the axial direction. The stiffness of the spring element was derived from load-slip curves obtained by headed studs’ push-out tests. The change of the headed studs’ stiffness during the design life was not taken into account. The finite element model did not include the influence of cracks in concrete. At the interface of steel and concrete, the concrete surfaces were represented by CONTA173 elements and steel surfaces were represented by TARGE170 elements. Target and contact elements that make up a contact pair were associated with each other via a shared real constant set. The contact elements can transfer the pressure but not the tension. To simulate the experiments, the same loading conditions and constraints as the experiments were used in the finite element analysis.The normal stresses of steel gusset plate and concrete chord by finite element analysis are shown in . From the stress contours, maximum stresses on the gusset plate are located at the intersection between the vertical web member and the diagonal web member.Under 170% design load, the maximum compressive and tensile stresses on the gusset plate are 264.8 and 210.9 MPa and the maximum compressive stress in concrete is 22 MPa. The maximum compressive stress in steel is less than the design strength of steel (310 MPa) and far less than the yield strength of steel (345.0 MPa). The maximum compressive stress of concrete is less than the design compressive value of concrete (50 MPa). The stress results indicate that there is a significant safety margin for such composite joint under 1.7 times of the design load.The relationship between load and strain for each parts of the composite joint are shown in . From these figures, it can be found that the measured strains increased linearly with the applied loads, indicating that the material remains in a linear elastic state and the linear elasticity theory can be applied to analyse the joint under 1.7 times design load.Some stress results for 170% of the design load from both the model test and FEA are presented in . A good agreement between the measured stresses and FEA results is achieved. As the stress measured at the center axis of the steel members or the concrete chord, the value represents the average one of the section. The details of the stress distribution need to refer to verified finite element analysis results to improve the performance of such joint. Therefore, the finite element models can be used to provide some guidance in the design of the composite joints.The position of headed studs is shown in and the shear forces of headed studs under 1.7 times design load are presented in . The maximum shear force in the headed studs is about 10.8 kN in the diagonal web member. According to the headed studs’ characteristic, the headed studs remain in the purely elastic range at all load steps. The shear forces in the first few rows of headed studs are greater than others. The shear forces in the rear headed studs decrease gradually. The average shear forces in the second row are 1.49 kN and 0.92 kN less than that in the first row for the diagonal and vertical web member direction, respectively. The trend of reduction slows down in the rear rows of the headed studs., are selected to study the load-carrying ratios by steel and concrete in the chord. The axial load-carrying ratios under 170% design load are listed in . The load-carrying ratios by steel are 10.8% and 10.7% in the sections A–A and B–B, respectively. According to the test and the FEA, the load-carrying ratio is constant along the entire length of the steel chord. The results show that the load is mainly borne by the concrete chord.A comprehensive study on the composite truss joint in Minpu Bridge has been conducted and the following conclusions can be made based on the model test and finite element analysis. The results show that under 170% design load, the maximum compressive stress in steel (264.8 MPa) was less than the design strength of steel (310 MPa) and far less than the yield strength of steel (345.0 MPa), and the maximum compressive stress in concrete chord (22 MPa) was less than the design compressive value of concrete (50 MPa). There is a significant safety margin for such composite joint under 1.7 times of the design load.The finite element analysis results showed fairly good agreement with the experimental ones in terms of strength and deformation, which means that the finite element analysis can be used to provide some guidance in the design of the composite joints.The shear forces in the first few rows of headed studs are greater than others on the gusset plate. The shear forces in the rear headed studs decrease gradually. The trend of reduction slows down in the rear rows of headed studs.The load-carrying ratio by the steel chord is about 11% in the steel reinforced concrete chord. The load is mainly borne by the concrete chord.The results from this study would be useful as references for design of composite joints in steel–concrete composite truss bridges.Depth profile analyses and thermal stability of a Cu–Co–Cu three-layer system investigated by polarized neutron reflectometrySpin polarized neutron reflectometry is an appropriate method for the non-destructive determination of concentration profiles in magnetic layer systems. The investigation of the stability of a thin Cu–Co–Cu-layer system after thermal heat treatments up to 400°C was performed. The concentration distribution of the cobalt after production and after heat treatments was measured, with the result, that the concentration profiles already changed at this low temperature. These changes were traced back to diffusion processes using polarization analysis of the reflected neutrons.In multi-layered thin films the interfaces play a dominant role giving rise to physical properties, which are not obtainable in bulk materials. In addition, magnetic thin films play an increasingly important role in advanced technologies.For example, due to surface anisotropies, the magnetic moments in such structures can be oriented perpendicular to the film plane if the thickness of the magnetic layer is sufficiently thin. This moment orientation is required for high density magneto-optic recording applications Another important phenomenon is the large negative magneto-resistance referred to as “giant magneto-resistance (GMR)” in the case of alternated ferromagnetic and non-ferromagnetic layers. This effect is caused by the oscillatory exchange coupling, that is, an oscillatory switching of the relative orientation of the magnetization of magnetic layers between parallel or antiparallel as the thickness of a non-magnetic interlayer is changed. In the antiparallel case a significant enhancement of the electrical resistance due to spin-dependent electron scattering can be induced in such a system. This effect is used for magneto-resistive read-out heads in the hard disk manufacturing industry The physical properties of such layered structures are closely related to their magnetic structure and, thus, the structure of the interfaces. Often this magnetic nanostructure can only be indirectly inferred e.g. from magnetization measurements. With neutrons, however, a direct determination of magnetic structures is possible. In recent years, reflectivity measurements with neutrons have matured and have been established as an important new scattering technique for the study of thin film systems. For instance, in a multi-layer stack consisting of ferromagnetic and non-ferromagnetic layers, any ferromagnetic or antiferromagnetic alignment of the ferromagnetic layers can be uniquely distinguished. Spin polarization of the neutrons and polarization analysis of the reflected neutrons provide additional detailed information on the exact orientation of any magnetic moments in the film plane Another important question is the thermal stability of such thin ferromagnetic layers, especially in the case of hard disks, due to the operating temperature of the hard disk. The theory of decomposition is engaged in supersaturated alloys, which show at certain concentration regions a gap in the range of miscibilities. They build during quenching from single-phased solid solution in the gap precipitation with a second phase. These decomposition processes determine many material parameters like ductility, elasticity, hardness and magnetic properties.Many important questions on the stability and the growth kinetics of small precipitations can be answered only insufficiently with conventional investigations, where in general size, morphology and volume fraction of the precipitations were determined Earlier studies of nucleation in Cu–Co alloys were done A plane Cu–Co–Cu three-layer system can be a model for a single precipitation to determine directly the growth of the precipitation. In this model the thickness of the cobalt layer in the middle corresponds to the diameter of a cobalt precipitation. The two outer layers represent in this model the solid solution phase consisting of a supersaturated copper–cobalt-alloy. Precipitation growth means in this model a growth of the cobalt layer in the middle of the system by diffusion.For these investigations the different layer thicknesses of the system and the distribution of the cobalt concentration have to be measured before and after the isothermal aging processes. The study should be performed in the temperature range up to 490°C, where a diffusion constant of The technique of polarized neutron reflectivity Neutron reflectometry has several advantages compared to the X-ray reflectivity for the investigation of a Cu–Co-layer system. The scattering length density, in contrast to that which occurs for electromagnetic radiation, does not depend on the number of electrons in the atom, and varies randomly between the elements, and even isotopes of the same element. Therefore there is a sufficient contrast between copper and cobalt. By means of the magnetic interaction of the polarized neutrons with ferromagnetic material the cobalt has a spin-dependent refraction index nwhere λ is the wavelength of the incident neutrons, b is the scattering length for coherent scattering and p is the scattering length for magnetic scattering. Small p is also dependent on the magnetizability of the sample. N is the number of nuclei per cm3 and σa the total absorption cross-section at a wavelength λ. σa is very small and, in contrast to X-ray reflectometry, may be neglected for the Cu–Co system considered here.If the spin direction of the neutrons is antiparallel (η− or η↑↓) with respect to the magnetization direction of the sample, the contrast between copper and cobalt is enhanced, whereas if the spin is parallel to the magnetization, the contrast between Cu and Co is close to zero (In the antiparallel case it is possible to study the modification of the cobalt layer after a heat treatment of the system, in the parallel case the total thickness of such a system.For explaining the total reflection and the multiple reflection at internal boundaries, a quantum mechanical treatment of the scattering process is necessary The Schrödinger equation for the wavefunction of a neutron in a solid Ψ(r) can be written in general as can be solved. On grounds of this optical method a characteristic matrix is defined for the jth layer bywhere nj, dj and ϑj are the refractive index, thickness and glancing angle of incidence. The multi-layer is now given by successive multiplication of similar matrices for the previous layers.For the calculation of the reflectivity of a multi-layer the amplitude of the incident neutron wave is normalized to 1, and the reflectivity is given bywhere nU and nS are the refractive indices of the medium of the environment and the substrate. The reflective coefficient r is given by r=|R|2.There is an exact equation for the reflectivity of a multi-layer, depending solely on the number of layers n, the thickness of the layers dn and on the refraction indices nn, the latter containing essentially the average coherent nuclear and magnetic scattering length densities. Since the square of the reflectivity is measured, phase information is lost. Thus by reflectivity measurements alone it is not possible to determine the scattering length density profile giving rise to the measured reflectivity profile.By means of a model it is possible to calculate the corresponding theoretical reflectivity profiles, to compare these calculated profiles with the measured reflectivity profile and then to modify the model until the best fit is obtained. The scattering length density profile and the concentration profile can be calculated from the parameters of the best fit model.Due to interdiffusion of Cu and Co during preparation and heat treatment of the multi-layers at the interface the change in refractive index at the boundary is not abrupt. The form of the refractive profile can be well approximated by subdividing the continuous function into a series of discrete layers The measurements were performed at the small angle neutron scattering facility SANS-2 at the FRG-1 research reactor in Geesthacht, Germany with a wavelength resolution Δλ/λ of 10%.The monochromatic neutron beam is polarized in a Schärpf-type supermirror polarizer ). The measurements were performed with a variable collimation, because with this technique the signal-to-background ratio can be improved by two orders of magnitude. Therefore at the end of the collimation path five diaphragms with different widths can be moved into the beam. The change of the collimation was carried out, when the projection of the sample surface to the neutron beam due to the sample rotation reached the next diaphragm width.The sample was magnetized in a superconducting magnet with a field strength of 4.0 T, and with the magnetic field direction being parallel to the flight direction of the incident neutron beam. During the measurements the sample was in plane magnetized, the incident neutron beam was polarized parallel ). A movable beam stop absorbs the transmitted beam.The experiment is equipped with a position sensitive counter, which registers the specularly and off-specularly reflected neutrons with an effective pixel size of about The finite efficiency of the polarizers and flippers requires some important correction of the measured data. This is especially the case if the reflected intensity of the two spin directions is different. Then small intensities of the depolarization influence significantly the reflectivity. The correction must be done for the efficiency of the flippers and for the polarization of the neutron beam.The experimental determination of the flipping efficiencies was performed at SANS-2 without sample at a wavelength of with Δλ/λ=10%. Four primary beam intensities corresponding to the four possible conditions of the flippers (i.e., flipper 1 on and off, flipper 2 on and off) are measured and the flipping efficiencies are calculated The polarization of a neutron beam is defined bywhere I+ indicates the percentage of the neutrons with one spin direction, I− with the opposite spin direction. Four distinct measurements are necessary to determine the polarization. The sample should be a good polarizer with, on one hand, large intensity difference between the two non-spin flip (NSF)-reflectivities (r++,r−−) and, on the other hand, with a zero spin flip (SF) – reflectivity (r+−,r−+). The theoretical intensity for every spin state can be calculated and compared with the measurements The substrate plays an important role in the production of thin film layers. Its temperature, structure and surface roughness have a crucial influence on the structure and quality of the thin film layers. Double-sided polished quartz glass plates were the substrate of choice in this case. The ripple was measured with the laser interferrometry and is better than 5×10−2 mrad. The roughness was determined experimentally with the unpolarized neutron reflectometry and typical is better than 1 nm. This agrees quite well with the data provided by the manufacturer.A Cu–Co-alloy sputter target was first produced via melt-metallurgy and subsequently solution annealed in vacuum at 900°C for 10 days consisting of a solid solution with 2 at.% Co. The alloy target and the pure Co target were mounted in the ultra-high vacuum chamber of the sputtering facility. The layers of the sample were produced by triode-magnetron sputtering.For calibration of the rate monitors test layers were produced, the thickness and chemical composition were characterized with totalreflection X-ray fluorescence analysis (TXRF) 20 nm CuCo was prepared on a quartz glass substrate.For complementary transmission electron studies the same three-layer system was prepared on a fine grid with a mesh size of 200 mesh shows schematically the structure of the layered system as determined from TEM.Spin flip (SF) as well as non-spin flip (NSF) reflectivities of the as-received sample is shown in . By means of model calculation described in , scattering length density profiles parallel (η↑↑) and antiparallel (η↑↓) to the polarization with respect to the sample magnetization could be determined (). The accuracy of the determined layer thickness was computed within an error of ±0.5 shows a good agreement between the calculated reflectivities and experimentally determined ones. Apparent spin flip reflectivities are observed and calculated from the as-received sample due to incomplete polarization and flipping efficiencies.The scattering length density distribution η is given by the concentration distribution of the elements perpendicular to the surface of the layer system. The scattering length density is composed of a nuclear and a magnetic part and the sum depends on the magnetization direction of the Co layer in relation to an external magnetic field (see ) is nearly identical to the geometry of the layer system, whereas η↑↓ clearly reveals the profile of the Co layer. Thus, the r++ reflectivity is very sensitive to the total thickness of the layer system, whereas r−− is very sensitive to the thickness and position of the cobalt layer in the system shows the nomenclature of the parameters.For the evaluation of the accuracy of the fit, each fit parameter has to be changed slightly. The resulting calculated reflectivity curve has to be compared with the reflectivity curve based on the unchanged fit parameter. Then it becomes possible to estimate the accuracy of this fit parameter . Other parameters are less sensitive to minor variations and can be estimated within an accuracy of 5%.This full characterized sample was annealed in a pre-evacuated quartz tube subsequently filled with 300 mbar argon gas for a better heat transfer. After the heat treatment the sample was cooled in a cold argon gas stream.Isothermal heating at 125°C for either 4 h or 6 h did not lead to any changes in the reflection profiles, the same holds true for a heat treatment 1 h/150°C. Minor changes in the reflection profiles became discernible after 1 h at 200°C () as a result of interdiffusion at interface boundaries of the layered system. After 1 h at 300°C () this effect was more pronounced. After 1 h at 400°C no further changes can be detected and evidently the process was terminated.A comparison of the parallel spin direction of samples given different heat treatments () shows no changes in the reflection profile. In contrast in the antiparallel spin direction the reflected intensity significantly decreases with increasing aging temperature ( shows the parameters of the three-layer system after the heat treatment. The Co-interlayer becomes thinner during annealing (see also Due to the low value of the diffusion constant of Co at 300°C this surprising effect cannot be attributed to bulk diffusion of Co into the surrounding Cu-matrix. It is rather attributed to grain boundary diffusion of Co into the Cu-layers. As the solubility of Co in Cu is only 0.2%, the Co atoms cannot be homogeneously dissolved in the Cu matrix, in particular, as the Cu–1.74%Co are already supersaturated with Co at aging temperatures below 500°C. Thus, during the heat treatment small Co clusters must be formed in the Co-enriched interfaces between the Co and the Cu layers (). The layer system contains many grain boundaries (), so that the clusters originate preferentially at such grain boundaries This hypothesis is supported by the appearance of SF reflectivities (, marked regions). The weak SF-reflectivities also result from these not fully magnetized cobalt clusters which act as magnetic inhomogeneties. Such clusters can only arise along grain boundaries by grain boundary diffusion. The expected increase in the Co-concentration of the Cu-layers cannot be measured due to the insufficient contrast.The concentration profile of a ferromagnetic Cu–Co–Cu three-layer system, in as-received and annealed conditions, was analyzed using spin polarized neutron reflectometry with spin analysis of the reflected neutrons. The Co layer was found to be thinner during annealing at moderate temperatures (⩽300°C). This process has been attributed to grain boundary diffusion of Co into the Cu layer and inhomogeneous nucleation of cobalt clusters along grain boundaries in the Cu layers.Flexible metal–insulator–metal (MIM) devices for plastic film AM-LCDWe developed new flexible metal–insulator–metal (MIM) devices subject to plastic film substrate. The structure of the MIM device is that a Ta2O5 insulator is covered with two flexible Al electrodes on both sides. The flexible structure of the MIM device was successfully fabricated applying our own etch-free process.In recent, wireless mobile communication such as IMT-2000 has been progressed remarkably. In the IMT-2000, high performance display with high quality, high speed, low weight, and low power consumption is required. The switching device satisfying these criteria is metal–insulator–metal (MIM) on plastic film substrate. The merit of MIM device is that high quality moving pictures can be displayed with low power consumption So, in this work, we attempted to solve these problems by developing new design and process. In the new design, the MIM device is composed of a Ta2O5 insulator covered with two flexible Al electrodes on both sides. Thus, current etching technology cannot be applied because bottom Al electrode is damaged during the etching procedure for patterning of top Al electrode. So, we developed our own etch-free process, and fabrication of the flexible MIM device was tried using the newly developed process. The fabricated MIM device was analyzed and characterized to optimize the performance.As a substrate, flexible polycarbonate film was used. The size and thickness were 7×7 cm and 0.1 mm, respectively. Both bottom electrode and top electrode were made of ductile Al. After formation of bottom electrode on the film substrate, TaOx was formed by rf sputtering at room temperature using the Ta2O5 target. The TaOx layer was patterned with our own etch-free process in order to avoid damage of film surface by etching solution including HF. Then, Al top electrode was formed on top of the oxide layer with sputtering and etch-free process, too. Finally, post annealing treatment was done below 150 °C under the various conditions The electrode for the flexible MIM device should be subject to polymeric film substrate. For subjecting to the substrate, difference in both thermal and mechanical properties between the two materials should be low because failure can arise from the difference in thermal expansion and mechanical stress between the film and the substrate.Based on these parameters, we calculated to optimize the flexible design of new MIM structure using Stoney's equation . Therefore, in order to minimize the stress, we selected Al thin film as electrode. The Al material has low specific-resistance satisfying requirements to MIM-LCD as well as the flexible properties Using the flexible electrodes, a new structure was designed. depicts a cross-sectional view of the structure of MIM device. The bottom electrode and top electrode are the same flexible Al material, and they are located on both side of insulation layer. Since the electrode material is very ductile, the MIM device is expected to be subject to the flexible substrate.Flexible MIM device was fabricated using the newly designed structure. First, Al bottom electrode was patterned on the polycarbonate film. For observing the surface state of the bottom electrode by the material, we deposited Al and Ta on the polycarbonate film with sputtering. shows the surface states of the bottom electrodes. As shown in (a), many cracks were generated at Ta bottom electrode. It is considered to be owing to the thermal and mechanical mismatch between the flexible film substrate and the hard metallic film layer. That is, high temperature induced by collision of the Ta particles with polycarbonate film generates considerable thermal expansion of the polymer substrate during the sputtering deposition. During cooling, the difference in shrinkage between the two materials generates the distortion of the film substrate. Accordingly, driving force acts to relieve the distortion and cracks occur on rigid Ta film layer at the expense of the relief of the distortion. On the contrary, the deposition of Al thin film was resulted in the surface without crack and distortion. (b) represents the surface of the Al layer on the polycarbonate film substrate. The surface was very clean and no defect was found. It is assumed that relatively lower value of Young's modulus and coefficient in thermal expansion induced only small amount of stress and distortion of the polymeric film substrate. In addition, no crack was generated at the Al layer because it has mechanically ductile properties. Accordingly, the Al bottom electrode was turned out to be the very appropriate for the flexible film substrate.After the formation of Al bottom electrode, the TaOx insulation layer and Al top electrode were deposited in sequence. The etch-free process was developed and applied to the fabrication of these two layers. The current process cannot be applied since it is based on etching method. If applied, the Al bottom electrode will be damaged severely because the material is the same that of top electrode. For certifying this assumption, the current etching process and newly developed process were applied, respectively. As expected, extremely different results were appeared. In (a), Al bottom electrode of MIM device fabricated by current process was seriously etched during the patterning of Al top electrode. The surface observed with SEM was very rough and partially sunk as shown in (b). The rough surface of the electrode should be improved because it induces the locally concentrated electron flow leading to the acceleration of failure (a), the morphology of the top electrode was improved. The SEM observation of the surface also revealed that the Al bottom electrode was not damaged during the top electrode formation of the same material.The MIM device fabricated with the newly developed process is represented in . As shown in this figure, the MIM device with the Al flexible electrodes on both sides was fabricated very well.For investigating the electrical properties, I–V characteristics of the flexible MIM device was measured using HP 4145B analyzer. Before characterization, post-annealing treatment was performed in order to stabilize the characteristic of the MIM device. Four types of annealing treatment were done according to the atmosphere environment and annealing step, respectively. That is, the annealing was done after fabrication of TaOx or whole device under the O2 gas or vacuum environments, respectively.The I–V characteristic curves of the MIM devices are represented in . In case of the annealing treatment after fabrication of whole devices, the curve shows nearly perfect symmetric properties. It is considered to due to the symmetrical electrode structure of the MIM device because the symmetric properties are strongly dependent upon the difference in work function value between the two materials In this work, the MIM devices subject to flexible substrate were attempted to fabricate by applying the newly developed flexible structure. As a result, flexible Al/Ta2O5/Al MIM structure could be fabricated successfully by our own etch-free process, and the electrical properties were very excellent satisfying the requirements of device for active matrix display. It is expected that the flexible MIM device will be applied to a switching device for plastic film AM-LCD.Vacuum heat treatment effect on the thermophysical properties of BSCCO systemVacuum heat treatment effects at 850 °C for (3, 6, 9 and 12 h) on the crystalline structure and superconducting, as well as on the mechanical properties of Bi4Sr3Ca3Cu6Ox (BSCCO) have been studied. The highest critical transition temperature Tc 102 K was achieved at annealing time tann |
= 9 h. This maximum value is attributed to movement of excess oxygen from annealed sample. Phase examination by X-ray diffraction (XRD) revealed that the gradual formation of the high-Tc phase (2 2 2 3) due to the prolongation of annealing time up to 9 h. Thereafter, the high-Tc phase starts degrading and Tc reduces. The distortion of the 2 2 2 3 phase is suggested by the broadening of the different XRD peaks. Also, at the same above annealing time (9 h), SEM observation shows good quality plate-like crystal with relatively oriented. On the other hand, the Vickers microhardness (VHN) measurements was found to be dependent on the annealing conditions, a tremendously increase of VHN value 1.92 GPa (0.25 N) at tann |
= 9 h. This substantial increase is ascribed to improvement of the thermal conduction and stabilization of temperature distribution between superconducting grains.The annealing process is one of the important methods to accelerate the formation of the 2 2 2 3 phase On the other hand, since the BSCCO phases have a plate-like morphology with a high current carrying capacity in the a–b plane and also owing to the importance of this system, researchers around the world have been trying to develop the practical application potential of it. Most applications require both desirable superconducting properties (Tc, upper critical magnetic field HC2 and critical current density Jc) and mechanical integrity (fracture toughness KIC, ductility, and Vickers hardness VHN). Previous authors have observed that the BSCCO systems are generally brittle with unacceptably low strength In view of this problem, efforts have been made to improve Jc and the mechanical properties of the BSCCO system. Khalil For commercially useful applications, several processing techniques are under consideration. These includes doping Bulk superconducting materials with nominal composition Bi4Sr3Ca3Cu6Ox were fabricated by means of a ceramic solid state–state diffusion reaction. Fine powders of Bi2O3, SrCO3, CaCO3 and CuO with purities of more than (99.999%) with appropriate cationic ratio were weighed, thoroughly mixed and ground, and calcined in air at 800 °C for 18 h. The reacted material was then reground and cold pressed into disks or pellets by applying a pressure of 5 ton/cm2 using a cylindrical die made of stainless steel. Afterwards, the pellets were additionally sintered in air at 850 °C for 100 h. At the end of the sintering process the pellets were allowed to furnace-cool to room temperature. The vacuum annealing conditions of the samples were set at a constant temperature of 850 °C under 0.01 Torr for 3, 6, 9 and 12 h. At the end of each step of the vacuum annealing process the samples were studied for their electrical and structural properties. Identification of various phases present in the samples was done on the basis of the characteristics of their XRD peaks. X-ray diffraction patterns were recorded with CuKα radiation on a Schamdzu X-ray diffractometer.The temperature dependence of the electrical resistivity was measured by the standard four-point probe technique in the temperature range 77–300 K. Electrical contacts were made with silver paint curable at room temperature. The microstructure was investigated with a JEOL scanning electron microscope (SEM) Model JSM-5300. The samples were coated with a thin layer of gold to avoid the charging effect. Physical density measurements were performed by Archimedes method.To study the mechanical properties of vacuum annealed BSCCO material, Vickers microhardness was measured using an E. Ltd. Wetzlar microhardness tester with pyramid indenter. A pyramid indenter was employed in all the microhardness tests because it does not penetrate as deeply into the surface, thereby causing less cracking around the indentations. The contact loads (P) ranged from 300 gf (2.490 N) to 25 gf (0.245 N) for the examined samples. A loading time of 15 s was used to measure the diagonal of indentation with an accuracy of ±0.1 μm. An average of 10 indentations at different locations on the specimen surface was done to obtain representative mean values for each load.The Vickers microhardness values (VHN) were determined using the following equation:where “P” is the applied load in N and “d” is the diagonal length of the indentation mark in mm.The formation of different phases at various stages of vacuum annealing time has been made evident by the investigations of the XRD patterns. These studies also bring about the mechanism of growth of the high-Tc phase in the annealed sample. (a)–(d) illustrates the variation of X-ray patterns for the Bi4Sr3Ca3Cu6Ox sample as-sintered and vacuum annealed at 850 °C for (6, 9, and 12 h, respectively). H an L in this figure indicate the high-Tc (2 2 2 3) and low-Tc (2 2 1 2) phase, respectively. Other identified phases are written in terms of their chemical formula and denoted directly on the diffractograms above their corresponding peaks. For the as-sintered sample in (a), it can be seen that the majority of the peaks are due to the high-Tc (2 2 2 3) phase, while a minority of the peaks belong to low-Tc (2 2 1 2) phase. Some indications of formation of CuO with characteristic diffraction peak at 2Θ |
= 38.4° is present. (b), at relatively short time of vacuum annealing, e.g., 6 h, one can easily observe that the diffraction peaks corresponding to the high-Tc phase and the low-Tc phase have been found to develop with gradually increasing intensity. The impurity peak due to secondary phase CuO remains nearly the same. For 9 h of annealing time with the maximum value of Tc |
= 102 K, the peaks corresponding to the low-Tc phase have slightly decreased at 2Θ |
= 23.2° and 29.1°, as can be seen in the XRD patterns of (c). Most of peaks could be assigned to the high-Tc phase, i.e., the high-Tc phase dominates the phase maximum. This affirmative observation may be due to enhancement of outgoing oxygen from the annealed sample. Hence, this loss of oxygen favors the formation of the high-Tc (2 2 2 3) phase. Otherwise, one can infer that at an annealing time equal to 9 h, the extent of outgoing oxygen from this sample is greater than for samples annealed at any other time. When annealing is continued further, i.e., 12 h in (d), it can be noticed that the intensity of the diffraction peaks of the high-Tc phase start degrading, decreasing the value of Tc, while the peak of the low-Tc phase slightly increases at 2Θ |
= 29.2°. It is peculiar also to note the existence of the insulating phase Ca2PbO4 at 2Θ |
= 18.06° though in small quantities.A closer examination of the XRD pattern also reveals another important feature of the two superconducting phases. Firstly, it is observed that the peaks of the high-Tc phase are sharpest at the stage when Tc has its maximum value of 102 K. At annealing periods less and also greater than this optimized time the peaks are relatively broad. Secondly, the peaks due to the high-Tc phase are relatively sharper than are the peaks due to the low-Tc phase. These observations indicate the distorted nature of the low-Tc phase and show the best crystallinity when maximum Tc is attained for the high-Tc phase.The electrical features of the samples were examined by resistivity versus temperature measurements (ρ–T). depicts typical electrical resistivity as a function of temperature for the Bi4Sr3Ca3Cu6Ox samples (as-sintered and annealed at 850 °C with evacuation at 10−2 |
Torr for 3, 6, 9, and 12 h). In , the as-sintered sample exhibits a semiconducting character in the 300–138 K region. Below 138 K, the resistivity decreases gradually with decreasing temperature and reaches a zero-value at a critical transition temperature Tc |
≈ 77 K. After vacuum annealed for 3 h, the electrical behavior started to change. As seen in , a metallic behavior was obtained from 300 K up 123 K, After Tons |
≈ 123 K, the resistivity–temperature plot deviates from linearity and exhibits a noticeable resistivity drop with a small tail extending to the zero-resistivity state at Tc |
≈ 83 K. With prolongation of annealing time to 6 h, similar metallic behavior was observed, the Tons value was found to be 116 K which is lower than for the above annealing period and Tc was found to be 90 K, which is higher than for tann |
= 3 h. With more prolongation of annealing time to 9 h, the annealed sample exhibits a remarkable resistive behavior compared with tann |
= 6 h. The resistivity curve shows an almost complete transition at Tons |
≈ 110 K and a Tc value of 102 K, i.e., the prolonged vacuum annealing time up to 9 h appears to enhance the superconducting property. This behavior suggests that the 9 h annealed sample has a high fraction of the high-Tc phase (2 2 2 3). This result is also supported by XRD analysis. For the longest period of vacuum annealing, 12 h, the temperature behavior of ρ is qualitatively similar to the former annealing time 9 h but with higher values of ρ and a lower value of Tc |
≈ 89 K. This trend may be ascribed to structural defects, like variable metallic ordering patterns. Such effects could be related to the lowering of the transition temperature Tc. In general, we can infer that there are many oxygen vacancies in the as-sintered sample, which causes a structural disorder in the Bi–O sheets, by which many electrons are localized and the superconductivity is partially degraded. As a result Tc was depressed. However, it can be decided that vacuum heat treatment for narrow steps of time (3–9 h) leads to excess oxygen vacancies moving out the sample, and the moving of oxygen enhances the arrangement of the lattice and produces the carriers which are responsible for better superconducting properties. Under our experimental conditions, it can be concluded that as the excess of oxygen moves out the metallicity of the Bi–O layers increases and consequently the values of Tc is enhanced.It is worth mentioning that, we have made two important points here; firstly, by prolongation of vacuum annealing time, the superconducting phase coherence is enhanced. In particular, growth of superconducting phases, such as 2 2 2 3 and 2 2 1 2 phase is believed to be responsible for enhancement of superconducting properties. Secondly, excess oxygen in the Bi–O layers may enhance the square planar like arrangement around the copper ions and in turn the superconducting properties are enhanced. From the above results, one can easily deduce that the optimal annealing time = 9 h exhibits the best superconducting properties. Therefore, the in order to test the linear behavior of ρ(T) at normal state, an expression of the form:has been applied to the annealed samples in the temperature range between 130 and 300 K. The parameter “A” can be considered as the temperature coefficient of resistivity in the normal state with a value in the range 10−5–10−6 |
Ω cm k−1 which is usually obtained for high-Tc superconductors . It can be observed that for all times of annealing, the range above Tons is characterized by metallic behavior as usual for high-Tc superconducting materials. The transition to the superconducting state could be held at all times of annealing. The shortest period of vacuum annealing (3 h) was terminated by a short tail, confirming the role of impurities as has been confirmed by the XRD analysis, while, the most convenient time of vacuum annealing (9 h) was characterized by no tail. The variation of values of the parameters of Eq. , “A and B”, with the time of vacuum annealing is given in . It is found that “A” increases from 0.61 × 10−5 to 2.27 × 10−5 with prolongation of annealing time (3–12 h), which is in good agreement with results obtained for similar high-Tc superconductors materials To look into the morphology, we have performed scanning electron microscopy on the vacuum annealed sample. shows a micrograph of a pellet with a maximum value of Tc |
= 102 K. A plate-like phase seems to dominate the microstructure at this stage of annealing (9 h). Good quality plate-like crystals, quite moderate in size and relatively oriented, constitute the major phase giving rise to superconductivity The effect of vacuum annealing time at (850 °C) for periods (3, 6, 9, and 12 h) on the Vickers microhardness (VHN) of the Bi4Sr3Ca3Cu6Ox sample was examined. depicts the variation of measured microhardness (VHN) against the vacuum annealing time at certain applied loads (P |
= 0.25, 0.49, 0.98, 1.96 and 2.94 N). It can be observed that at all applied loads the variation of VHN exhibits an ellipse-shape type behavior with an apex at annealing time (tann) = 9 h and a trough at tann |
= 12 h. The tremendous increase of VHN values from 1.09 GPa (0.245 N) at tann |
= 3–1.92 GPa (0.245 N) at tann |
= 9 h is ascribed to the formation of good conducting channels between superconducting grains which can enhances the bridging between them. This is most probably the reason for the high values of VHN. Furthermore, the annealing process improves the thermal conduction and, hence, stabilizes the temperature distribution between superconducting grains. The monotonic decrease of VHN values 1.00 GPa (0.245 N) at tann |
= 12 h can be a consequence of the partial dissolution of superconducting (metallic) phases. Also, the prolongation of annealing time (12 h) may generate circumferential cracks. Therefore, the contact between grains weakens, consequently lowering the values of VHN. displays the change of VHN values with the applied load at tann |
= 0, 3, 6, 9, and 12 h. The curves in the graph show that disregarding the time of annealing the microhardness values decrease non-linearly as the applied load increases up to 0.980 N, thereafter, the VHN values tend to attain saturation. This type of non-linear behavior has also been observed in the literature for several kinds of materials and is termed an indentation size effect (ISE) In order to describe the ISE behavior of materials, several relationships between the applied test load “P” and indentation diagonal length “d” have been given in the literature. The simplest way to describe the ISE is Meyer’s Law (1908) where “n” is the Meyer number (or index) and “K” is the standard hardness constant. The values of Meyer’s index n are used as a measure of ISE. When n |
< 2, the hardness increases with decreasing applied load, which is termed normal ISE. When n |
> 2 the hardness values increase with increasing applied load which is called reverse ISE. When n |
= 2, the hardness is independent of applied load. However, the exponent n has values greater or lower than 2 in most cases. The values of n and K in Eq. may be obtained from the plots of Ln |
P against Ln |
d. Typical plots of the dependence of Ln |
P and Ln |
d for BSCCO samples at tann |
= 0, 3, 6, 9, and 12 h are shown in . The determined n values obtained from the slopes of the best-fit lines in this figure are tabulated in , it can be noticed that n values vary from 1.44 to 1.58. These values of n indicate that Meyer’s Law is obeyed in our results. On the other hand, the obtained results from indicate that the load independent hardness values are lower than those of the conventionally estimated ones (). This observation is in agreement with experimental results in the literature shows the apparent density change with increasing vacuum annealing time for the Bi4Sr3Ca3Cu6Ox samples. One can easily observed that the measured density increases with increasing annealing time and has a peak at tann |
= 9 h. This observation can be attributed to the fact that the prolongation of vacuum annealing time increases the ability of the 2 2 2 3 lattice for a favorable rearrangement, which leads to the increased density value In conclusion we formulate the following points as below:X-ray diffraction analysis suggests that the domination of the high-Tc phase (2 2 2 3) increases as vacuum annealing time increases and that the optimal annealing time with tann |
= 9 h produces a nearly pure Bi-2 2 2 3 phase.Vacuum annealing treatment causes an increase in the metallicity of the Bi–O layers and in turn the values of Tc are enhanced.SEM observations reveal that the sample with optimal vacuum annealing time tann |
= 9 h has good quality plate-like crystals that are quite moderate in size and relatively oriented.Vacuum annealing treatment has a significant effect on increasing the hardness. This high hardness can be attributed to improvement of the thermal conduction and hence stabilization of the temperature distribution between superconducting grains.The measured microhardness values of the BSCCO materials are load dependent. The variation of microhardness with load is non-linear. This observation is related to the indentation size effect (ISE).Vacuum annealing conditions lead to a substantial increase of the sample density. This increase is owing to long-range ordering of the sample’s crystal structure.Bainite/martensite duplex-phase high strength steelFatigue behavior of 1500 MPa bainite/martensite duplex-phase high strength steelThe fatigue properties of C–Si–Mn–Cr low alloy steel can be improved at the tensile strength of 1500 MPa or higher level by the following heat treatment: austenitizing at 900 °C for 20 min followed by air cooling and then tempering at 280 and 370 °C for 2 h. The steel had a duplex-phase of carbide-free bainite and martensite containing retained austenite of 8 vol% and higher (designated as CFB/M steel). The results obtained for the CFB/M steel were compared with those of the fully martensitic steel which was oil-quenched and then tempered at 280 and 370 °C for 2 h after austenitization at 900 °C for 20 min (designated as FM steel). The fatigue strength was determined by a method of axial fatigue test, and the fatigue crack propagation was studied by using compact-tension specimens. Compared to the FM steel, the CFB/M steel increased the fatigue strength and fatigue crack threshold (ΔKth), and then lowered crack propagation rate. In particular, the CFB/M steel had a better combination of strength, toughness and fatigue properties when tempered at 370 °C for 2 h.Bainite/martensite duplex-phase high strength steelThe fatigue strength would increase with the increase of their ultimate tensile strength for low or medium carbon steels with lower strength level, but for high strength materials, especially, the high strength steels whose ultimate tensile strength exceeded 1200 MPa, the fatigue strength would directly depend on both strength and toughness. The higher strength and toughness of low alloy high strength steels might be obtained through quenching and tempering at lower temperatures The steel used in the present investigation was 0.22C–1.8Si–2.3Mn–Cr steel which had been melted in a vacuum induction furnace. A 25kg ingot was forged at 1200 °C, then kept at 700 °C for 2 h and furnace cooled to room temperature. The martensite and bainite starting temperatures (Ms and Bs) of the steel measured by dilatometry were 295 and 330 °C, respectively. Some specimens were quenched into oil at room temperature for obtaining fully martensite steel (FM steel), and other specimens were air cooled for obtaining carbide-free bainite/martensite duplex-phase steel (CFB/M steel). Both the specimens were tempered at 280 or 370 °C for 2 h. The short tensile specimen having a gauge length of 40 mm and a diameter of 8 mm was utilized. The 10×10 mm2 U-notch standard impact specimens were prepared from the square bar. The fatigue specimens used were smooth hourglass-shaped specimens, as shown in . The gauge part was mechanically polished. The fatigue crack propagation tests were carried out by using compact-tension specimens shown in . Before the test, the fatigue crack was prepared at the root of the notch.The tensile test was conducted with an Instron AG-75TA universal electron testing machine at a crosshead speed of 0.5 mm/min. The tension–compression fatigue test was preformed on the servo-electro-hydraulic Instron-1603 testing machine, the test frequency was 165 Hz. The fatigue strength was measured with the stress ratio R=−1, and the cyclic load was applied up to a cycle number of Nf=1×107. The fatigue crack propagation rate was determined under a stress ratio R=0.1. First, the relationship of the fatigue crack length a to the cycle numbers Nf was measured, then the data were treated to obtain the da/dN–ΔK curve.The total retained austenite amount of the CFB/M steel and the FM steel was measured on a LDJ9600 vibration sample magnetometer. The microstructure was characterized using a scanning electron microscope (SEM) and transmission electron microscope (TEM). Fractographs were taken of the fresh fracture surface after the fatigue test using an SEM.The mechanical properties of the CFB/M and FM steel are shown in . The tensile and yield strength of the CFB/M steel slightly decrease, and its elongation and reduction of area are slightly improved with increasing tempering temperature, but the impact energy (aK) of the specimen tempered at 370 °C increases by about 10 J from that tempered at 280 °C. The toughness is improved visibly, which shows that the CFB/M steel has better tempering stability. For FM steel, the changes of its properties with the tempering temperature are similar to CFB/M steel, but the impact absorbed energy is reduced after tempering at 370 °C.The fatigue test results of the CFB/M and FM steels are shown in , it can be observed that the fatigue strength of the CFB/M steel is greater than 700 MPa, whereas the fatigue strength of the FM steel is greater than 635 MPa under both different tempering temperature. In addition, from the number of cycles to failure, it is suggested that the real value of the fatigue strength of the CFB/M steel tempered at 370 °C might be superior to that of tempered at 280 °C, but for the FM steel, the result is opposite to the above.The fatigue crack threshold stress intensity range (ΔKth) is a parameter which characterizes the beginning of fatigue crack propagation. The da/dN–ΔK curves of the CFB/M steel and the FM steel at stress ratio R=0.1 are shown in shows that the ΔKth values of the CFB/M steel after tempering at 280 and 370 °C are about 12.7 and 12.5 MPa m1/2, respectively; those of the FM steel at 280 and 370 °C are about 11.3 and 10.6 MPa m1/2, respectively. The results show that the ΔKth of the CFB/M steel is a little higher than that of the FM steel, and the effect of tempering temperature on the ΔKth of the CFB/M and FM steels is not apparent.From the fatigue crack propagation curves of two high strength steels, it can be seen that the fatigue crack propagation rate da/dN of the FM steel is obviously higher than that of the CFB/M steel. The effect of tempering temperature on the da/dN of the CFB/M steel is lower, and for the FM steel, the da/dN increases for the case of tempering at 370 °C.The microstructure morphology of two high strength steels tempered at 280 °C is shown in . The SEM observation revealed that for the CFB/M steel bainite, associated with martensite in acicular form, partitioned prior austenite grains and consequently, caused refinement to martensitic plate (). The SEM observation also revealed that for the CFB/M steel large angle boundaries were often found between neighbor interlath packets (); whereas for the FM steel, the large angle boundaries were hardly observed between the neighbor interlath packets (The micrograph of the CFB/M steel tempered at 280 °C observed by a transmission electron microscope is shown in shows that the retained austenite thin films between the interlath packets of lower bainite ferrite and in the lower bainite ferrite lath packets and martensite packets, also shows principally the retained austenite thin films between sub-lath in the lower bainite lath packets. , respectively. The microstructure characteristics of the CFB/M steel tempered at 370 °C resemble those tempered at 280 °C. The total retained austenite amount of the CFB/M steel tempered at 280 °C is about 8.7 vol%, that of tempered at 370 °C is about 8.1 vol%, but total retained austenite amount of the FM steel tempered at 280 °C or at 370 °C is far less than 1 vol%.The fatigue fracture photographs of the FM and the CFB/M steels tempered at 280 °C are shown in , where (a)–(c) are fatigue fractographies of the FM steel under the conditions of the stress amplitude of 650 MPa and the cycle number of 3.271×106. shows an enlarged fatigue origin region, the fatigue cracks initiate from the surface and sub-surface defect of the specimen then grow along radial direction, and the tearing ridges exhibit radiating distribution. The boxed area in . The fatigue fracture mode is transgranular quasi-cleavage failure, there are a few secondary cracks and rub marks, but the fatigue striation cannot be seen. shows the fatigue fractographs of CFB/M steel under the conditions of cycle stress amplitude of 750 MPa and cycle number of 3.739×106. For the CFB/M steel, it is observed from that the shear lip is wider and the fatigue crack propagation region is smaller. The arc in is the fatigue crack origin region. The fatigue crack initiates from the specimen’s surface. The boxed area in , and the morphology is similar to that of . The fatigue fracture morphologies of the FM and the CFB/M steels tempered at 370 °C resemble those tempered at 280 °C.For higher strength steels whose strength was above 1200 MPa, the literature At present, the measured fatigue strength of the CFB/M steel is greater than 700 MPa and it is about 46.7% of its tensile strength. It increases by 10.2% more than that of the FM steel. The ΔKth value of the CFB/M steel is about 12.7 MPa m1/2 and it increases by 20% from that of the FM steel, which means that the ΔKth is significantly improved, in comparison with the ΔKth of AISI4340, 300M and 50CrV as mentioned above. The fatigue crack propagation rate da/dN of the CFB/M steel is also lower than that of the AISI4340 and 300M and 50CrV steel The CFB/M steel tempered at 370 °C for 2 h has a better combination of strength, toughness and fatigue properties. It is therefore possible that the relaxation of residual strain produced by martensitic transformation occurs with an increase in tempering temperature from 280 to 370 °C, and the temperature range of the low-temperature temper embrittlement of the steel shifts following silicon addition and formation of the duplex-phase. Therefore, we can suggest that CFB/M steel exhibits greater fatigue crack propagation threshold resistance than FM steels.The CFB/M steel has higher σ−1, ΔKth and lower da/dN, which is closely related to its microstructure. In the CFB/M steel, the bainite associated with martensite in acicular form partitioned prior to austenite grains and caused refinement to martensitic plate. By the way, as a result of separated effect of the lower bainite lath on the prior austenite grain, martensite plate packet boundary neighbored with the bainite plate is large angle boundary, which makes the effectual grain size as the fatigue fracture unit reduce and lead to the change of the crack orientation and crack branching. Therefore, the fatigue crack propagation threshold resistance increases and the fatigue crack propagation rate decreases. In addition, the certain content of silicon is able to inhibit the carbide precipitation during the bainite phase transformation, and the retained austenite thin films are formed at the prior austenite boundaries, the lath bundle boundaries of the lower bainite and martensite and sub-lath boundaries The fatigue properties of C–Si–Mn–Cr low alloy steel can be improved at the ultimate tensile strength of 1500 MPa or higher by the following heat treatments: austenitizing at 900 °C for 20 min followed by air cooling and then tempering at 280 or 370 °C for 2 h. The steels had a duplex-phase of carbide-free bainite and martensite containing retained austenite of 8 vol% and higher (designated as CFB/M steel).The fatigue strength of the CFB/M steel is greater than 700 MPa independent of tempering temperature; whereas, the fatigue strength of the fully martensitic steel (FM), which was oil quenched and then tempered at 280 or 370 °C for 2 h after austenitization at 900 °C for 2 min, is greater than 635 MPa.The fatigue crack threshold ΔKth of the CFB/M steel is about 12.7 MPa m1/2, and increased 20% than the ΔKth of the FM steel under the stress ratio of R=0.1. The fatigue crack propagation rate da/dN of the CFB/M steel is lower than the FM steel under the different tempering temperature. Especially under tempering at 370 °C, it is the lowest, but the da/dN of the FM steel under tempering at 370 °C is the highest.The effect of tempering temperature at 280 and 370 °C on the ΔKth of the CFB/M steel is not obvious, but that on the da/dN of the CFB/M steel is evident. The da/dN of the CFB/M steel tempered at 370 °C is slower in comparison with that tempered at 280 °C.From the present investigation, it is suggested that improvement of the fatigue properties of the CFB/M steel depends on its microstructure and the CFB/M steel has a better combination of the strength, toughness and fatigue properties when tempered at 370 °C for 2 h.On the influence of grain size on the TWIP/TRIP-effect and texture development in high-manganese steelsThe impact of grain size on the work-hardening rate of two high-manganese steels with nearly identical stacking fault energy values of approximately 16.6 mJ m−2 (X60Mn17 alloy) and 16.5 mJ m−2 (X30MnAl17-1 alloy) was analyzed via macro-texture and micro-texture measurements as well as quasistatic tensile tests and phase analysis. Both alloys exhibited differences in twin volume fraction and volume fraction of the martensitic phase. The twin volume fraction increased with grain size. The X30MnAl17-1 alloy with Al and lower C content showed a lower overall twin volume fraction than the X60Mn17 alloy without Al. Additionally, the phase fractions of ε-martensite and α′-martensite were very low in the X30MnAl17-1 alloy. The increased grain size and the subsequent increase in twin volume fraction led to a relative decrease of the volume fraction of grains with <111>-orientation parallel to the tensile axis for both alloys.The remarkable mechanical properties of fully austenitic high-manganese TWIP (TWinning-Induced Plasticity) and TRIP (TRansformation-Induced Plasticity) steels with 12–30 wt% Mn and stacking fault energy (SFE) values in the range of 12–55 mJ m−2 can be attributed to high dislocation densities and the activation of additional deformation mechanisms such as deformation twinning []. Pronounced planar dislocation glide leads to a lower probability of dislocation annihilation and the combination with gradual grain refinement via deformation twinning - the so-called dynamic Hall-Petch effect – results in high work-hardening rates (WHR) []. In high-manganese steels (HMnS) that exhibit the TRIP effect, the strain-induced transformation of austenite to ε-martensite (hexagonal close packed) or α′-martensite (body centered cubic) increases the WHR and can retard local necking []. Accordingly, the product of ultimate tensile strength and total elongation for HMnS in some cases exceeds 50,000 MPa% []. Those properties make high-manganese TWIP/TRIP steels a point of interest for crash-relevant car parts []. However, a low yield strength (YS), as is common for HMnS, is detrimental to the performance for conventionally-designed crash-relevant parts because the available deformation volume of such parts is low compared to the total elongation of HMnS []. Thus, increasing the YS is of high significance for the application potential of HMnS.Conventional approaches to increase the YS in HMnS include the reduction of grain size, the use of alloying (e.g. C, Si) and microalloying elements (e.g. V, Nb) or the application of recovery annealing heat treatments to retain the heavily-twinned, fine-grained microstructure after cold-rolling []. While the addition of alloying elements can increase the YS, it also changes the SFE and the WHR []. Although the increase of the YS via grain refinement is well documented in HMnS, the magnitude of impact on the evolution of WHR during the uniaxial tensile test in TWIP/TRIP steels is still not fully understood, although there are several studies on the impact of grain size on deformation twinning [In the past, attempts have been made to determine the solid solution strengthening contribution of various alloying elements, most often based on experimental data and linear regression models. Those models often imply negative solid-solution strengthening parameters for Mn, which is usually thought of as an element with a positive impact on solid solution strengthening []. This led to the development of new models, which perceived the concept of short-range ordering (SRO) as an additional contribution to the YS of HMnS []. SRO is the concept of interactions of carbon-metal pairs which result in local differences of ordering energies and make certain cell configurations more favorable to inhibit a central C atom []. Ab-initio calculations employing density-functional theory (DFT) have been utilized to determine the differences in ordering energies []. Results from those approaches have shown good agreement with experimental data, e.g. from Mössbauer spectroscopy, atom-probe tomography (APT), small-angle neutron scattering (SANS) and in-situ synchrotron X-ray diffraction []. While the influence of SRO on the WHR in HMnS has been described as enhancing planar glide, its impact on secondary deformation mechanisms such as deformation twinning is still a point of debate [Analyzing the texture of a material is a powerful tool to investigate the material's response to deformation or heat treatment. The dependence of the activation of deformation twinning on the Schmid factor along with strong rotations of the grains leads to a texture evolution during uniaxial tensile tests that is dominated by grains with orientations close to the <111>||TA- (TA – tensile axis) or <100>||TA-fiber []. The typical recrystallization texture of high-manganese TWIP steels contains similar texture components as the deformation texture but is strongly weakened in intensity [The aim of this study is to analyze the impact of grain size on the activation of the secondary deformation mechanisms in two TWIP/TRIP HMnS (X60Mn17 and X30MnAl17-1) and the effect on mechanical properties. To this end, phase analyses and macro-texture measurements were performed by X-ray diffraction (XRD) to analyze the evolution of phase fractions of ε- and α′-martensite during uniaxial tensile tests in a fine-grained and a coarse-grained state. Quasistatic, uniaxial tensile tests were employed to evaluate the mechanical properties.The X60Mn17 alloy was produced via strip-casting with an as-cast thickness of approximately 3 mm, using a double-roller device, followed by in-line hot-rolling with a thickness reduction of roughly 20–30% to a strip thickness of 2.2 mm. A subsequent cold-rolling process with a thickness reduction of 50% led to a final strip thickness of 1.1 mm. The X30MnAl17-1 alloy was ingot-cast with ingot dimensions of 140 mm × 140 mm x 500 mm. Subsequently, a homogenization annealing at 1150 °C for 5 h and hot forging at 1150 °C from 140 mm × 140 mm–55 mm × 55 mm was employed, followed by an additional homogenization annealing at 1150 °C for 2 h and hot-rolling to a sheet thickness to 2.2 mm. The following cold-rolling thickness reduction of 50% resulted in a final thickness of 1.1 mm.Tensile test specimens with 30 mm initial gauge length and 6 mm gauge width were machined using water jet cutting. The TA was perpendicular to the rolling direction (RD). Heat treatments of the tensile test specimens to achieve the fine-grained (designation FG) and coarse-grained (designation CG) microstructures were performed in a salt bath furnace. Average grain sizes davg and the grain size distribution were determined with the line-intercept method using optical microscopy images of the microstructure after etching with a Klemm-solution, taking annealing twins into account. The chemical composition regarding the alloying elements with relevant alloying contents of both HMnS in wt%, the respective SFE values that were calculated using a subregular solution thermodynamic model [] and the heat treatments for the respective material states are given in Metallographic preparation for Electron Backscatter Diffraction (EBSD), XRD and macro-texture measurements was performed by means of mechanical grinding with SiC paper up to 4000 grit with subsequent mechanical polishing using 3 μm and 1 μm diamond suspension. Electropolishing was then performed at 22 V using an electrolyte containing 700 ml ethanol (C2H5OH), 100 ml butyl glycol (C6H14O2) and 78 ml perchloric acid (60%) (HClO4). The duration of electropolishing was 20 s for EBSD measurements and 2 min for the XRD measurements, respectively.EBSD analysis was performed on the sheet surface of tensile test specimen with 12 mm gauge length and 2 mm gauge width, using a Zeiss Sigma scanning electron microscope (SEM) at an acceleration voltage of 15 kV and a working distance between 15 mm and 20 mm, a step size of 70 nm and utilizing a post-processing routine employing the HKL Channel 5 software, as well as the MTEX toolbox. A noise reduction was employed by removing wild spikes and considering at least 5 neighboring data points. Phase analysis and XRD measurements were conducted on the mid-layer of the sheet thickness using a Bruker D8 Advance X-ray diffractometer equipped with a HI-STAR area detector operating at 30 kV and 25 mA to acquire three incomplete (0–85°) {111}-, {200}- and {220}-pole figures, which were used to calculate the corresponding orientation distribution function (ODF) via the MATLAB-based MTEX toolbox []. Volume fractions of the texture components were calculated by using a radius of 15° around the ideal position.Uniaxial, quasistatic tensile tests were conducted using a Z250 tensile testing machine by ZwickRoell. For each alloy state in the same condition, at least 3 tensile tests were conducted. Strains were measured with the contact-type extensometer MultiXtens by ZwickRoell. The strain rate from the beginning up to 0.2% yielding was kept constant at 0.00025 1/s, after which the strain rate was increased to 0.001 1/s over the duration of 10 s and then kept constant until fracture. The WHR throughout this manuscript is defined as dσtruedεtrue and as such, is the slope of the flow curve []. To describe the evolution of the WHR with increasing true strain, the strain hardening exponent nH from Holloman's equation has been calculated for any given true strain, using the relationship nH=dlnσtruedlnεtrue. shows the grain size distribution and the average grain size davg for all material states as determined by using the line-intercept method.The selected heat treatments led to a logarithmic distribution of grain diameters with significant differences in the grain size of each state of alloy. At the same time, davg of the fine-grained and coarse-grained material states are similar. Both fine-grained material states show average grain sizes of approximately 4 μm, although the grain size distribution of the X30MnAl17-1_FG alloy is narrower but shows one distinct runaway value with a grain diameter of 24 μm. The average grain size of the X60Mn17_CG alloy (16 μm) was slightly lower than that of the X30MnAl17-1_CG alloy (22 μm). Overall, the increase in annealing temperature and annealing duration resulted in significant differences between the average grain size for both alloys.The results of the quasistatic tensile tests are shown in The results for each material state were reproducible. (a) shows that the coarse-grained X60Mn17 alloy had a significantly reduced elongation to fracture (εf) of 20.2% compared to the X60Mn17_FG alloy with 52.9%, while εf for the X30MnAl17-1_CG condition was slightly increased to 69.2% from 62.7% for the X30MnAl17-1_FG condition. In both cases, the grain refinement led to a significantly higher yield strength (YS) and ultimate tensile strength (UTS). For the X60Mn17 alloy, the YS and UTS increased by ΔσYS,X60Mn17=184MPa and ΔσUTS,X60Mn17=306MPa respectively. The impact of the reduction in grain size was less pronounced for the X30MnAl17-1 alloy at ΔσYS,X30MnAl17−1=84MPa and ΔσUTS,X30MnAl17−1=138MPa. Additionally, the fine-grained states displayed a higher initial WHR, as can be seen in (d). This is also highlighted by the evolution of nH with increasing strain, as shown in (c). At early stages of deformation (εtrue<0.05), nH is higher for the fine-grained material states. During further plastic deformation, nH increases for both material states, but more significantly for the coarse-grained material states. As a result, nH is higher for the coarse-grained material states in both alloys at approximately εtrue>0.05. Both states of the X60Mn17 alloy showed serrated flow curves, which is common for high-manganese TWIP steels that do not contain Al []. None of the tested alloy states exhibited a significant necking behavior before fracture. An overview of the mechanical properties of the tested material states is given in The macro-texture evolution of the four material states is shown in as inverse pole figures (IPF) relative to the tensile axis (TA) for the fully recrystallized states (0%), after 20% elongation and after fracture (εf). Additionally, the texture indices T and the respective elongation to fracture εf for the specific samples are given.The IPFs indicate that while the recrystallization texture (0%) was very weak, the macro-texture evolution during tensile straining (20% and εf) was dominated by a stronger <111>||TA- and a softer <100>||TA-fiber. This is also visible in the corresponding orientation distribution function (ODF) sections at ϕ2 = 45°, shown in The definition of texture components used in While all tested alloys showed a weak but complete α-fiber in the recrystallized state, the {110}<100>Goss (G) texture component softened during elongation, leading to an incomplete α-fiber at fracture that was dominated by the {110}<112>Brass (B) texture component. Additionally, the intensity of the {112}<110>Rotated Copper (RtCu) texture component increased, as well as the intensity of the {001}<100>Cube (C) and {110}<110>Rotated Goss (RtG) texture components. The X60Mn17_CG alloy was an exception as the texture remained very weak even after fracture. Since the total elongation of the X60Mn17_CG alloy was only slightly over 20%, differences between the texture at 20% elongation and εf were marginal (cf. ), as is also indicated by the low increase of the texture index from T = 1.9 at 20% elongation to T = 2.0 at εf. The evolution of the volume fractions f of the main texture components is illustrated in , revealing results corresponding to the ODF sections shown in With increasing elongation, the volume fractions of the B and RtCu texture component significantly increased for all alloys and conditions except X60Mn17_CG, in which the B texture component slightly decreased. The volume fractions of C and RtG texture components slightly increased or remained at a similar level. In contrast to that, G and Cu texture components decreased in volume fraction. are excerpts of larger EBSD maps with the focus of comparing the twinning behavior of the (a) X30MnAl17-1_FG and (c) X60Mn17_FG alloys with a coarse-grained (b) X30MnAl17-1 and (d) X60Mn17 after 20% uniaxial deformation. In addition, for the X60Mn17_CG alloy, the microstructure after fracture is shown in Comparing the quantity of detected boundaries between grains with Σ3 CSL (Coincidence Site Lattice) orientation relationship in (a)–(e), the EBSD measurements indicate that the coarse-grained alloys included more deformation twins, while the number of Σ3-grain boundaries was the lowest for the X30MnAl17-1_FG alloy. Although Σ3-grain boundaries could also be an indication of annealing twins, the shape and size of the majority of twins in (a)–(d) is indicative of deformation twins. As annealing twins develop during heat treatments at high temperatures, they typically have straight boundaries and are larger than deformation twins, which are usually very thin and tapered [(f) for the X60Mn17_CG alloy shows that a significant fraction of ε-martensite and α′-martensite is present after elongation to fracture. As highlighted in (f), microcracks close to the fracture surface were formed at phase boundaries of ε- and α′ during deformation.The results of the phase analyses conducted from X-ray diffraction are summarized in The XRD results show that (i) the phase fractions of ε-martensite and α′-martensite were higher in the X60Mn17 alloy than in the X30MnAl17-1 alloy and (ii) that those phase fractions were higher for the X60Mn17_CG condition than in the X60Mn17_FG condition.As stated in the introduction, the low YS of many HMnS is problematic in cases of crash-relevant applications []. Accordingly, the impact of a conventional method to increase the YS like grain refinement on the mechanical properties of TWIP/TRIP HMnS is of special interest. In the following passages, first, the presented results of the tensile tests and EBSD measurements will be used to discuss the influence of grain size on the WHR and YS for the X60Mn17 and X30MnAl17-1 alloys. Afterwards, the macro-texture measurements will be analyzed with respect to the observed differences in twin volume fraction.(a) shows that the decrease in davg from 16 μm in the X60Mn17_CG alloy to 4 μm in the X60Mn17_FG alloy led to a significant increase in YS. The same observation was made for the Al-alloyed HMnS, where the higher annealing temperature and annealing time increased davg from 4 μm to 22 μm. The differences in grain size resulted in different initial levels of the WHR and nH after yielding (cf. ), which can be explained by a reduced initial mean free path (MFP) of dislocation movement after the elastic-plastic transition. This induced different dislocation-generation rates and a higher dislocation density in the fine-grained alloys at the start of plastic deformation []. On the other hand, the strain hardening exponent nH of both coarse-grained alloy states is higher than for the fine-grained alloys at εtrue>0.05 (cf. (c)). A rapid increase in WHR is commonly explained in terms of the activation of a secondary hardening mechanism, such as the TWIP or TRIP effect, which induces a reduction of the MFP [, where the EBSD IPF-maps indicate that after 20% uniaxial deformation, the twin density of the coarse-grained conditions (cf. ) was higher, although all alloys showed the TWIP effect. Accordingly, the nH curves (cf. (c)) indicated that the initially higher WHR in the fine-grained alloys, which is due to grain refinement, is compensated by an additional hardening mechanism in the coarse-grained alloys. For the X30MnAl17-1_CG alloy, the higher strain hardening exponent nH at εtrue>0.05 (cf. (c)) compared to the fine-grained material state resulted in a very similar WHR at εtrue>0.4 (cf. (d)). Since the coarse-grained alloy state exhibited a lower flow stress (cf. (b)), this resulted in a higher elongation to fracture compared to the fine-grained alloy state. Additionally, the XRD measurements (cf. ) indicated that the ε-martensite and α′-martensite phase fractions were significantly higher for the X60Mn17_CG condition compared to the fine-grained material. Accordingly, the TWIP and TRIP effect were more active in the coarse-grained material in the sense that the volume fraction of twins and the phase fraction of martensite were substantially higher. This led to higher values of nH. In contrast to the X60Mn17 alloy and despite similar SFE values (cf. ), the phase fractions of ε-martensite and α′-martensite did not increase significantly during plastic deformation in the X30MnAl17-1 alloy. Accordingly, the X30MnAl17-1 alloy could also be classified as a TWIP alloy to clearly distinguish the deformation behavior from the TWIP/TRIP X60Mn17 alloy. The decrease in εf for the X60Mn17_CG alloy can be explained in terms of the high phase fractions of ε-martensite and α′-martensite at a relatively low elongation to fracture. Especially α′-martensite is significantly harder than austenite, which is due to the tetragonal distortion of the body centered cubic cell of α′-martensite [], and influences the strain partitioning behavior, leading to the preferential accommodation of plastic deformation in austenite []. While the TRIP effect can have a positive impact on ductility, rapidly occurring γ→ε and γ→ε→α' transformations lead to an early saturation or conclusion of the austenite transformation. This can lead to crack initiation at phase boundaries (cf. ], which in turn leads to preliminary fracture []. This is also reflected by the sharp drop in the WHR curves prior to fracture for the X60Mn17 alloys in The macro-texture evolution shown by means of ODF sections in is in accordance with the frequently reported development of a stronger <111>||TA- and a softer <100>||TA-fiber with the main texture components being B and RtCu (<111>||TA) as well as RtG and C (<100>||TA). Dislocation glide leads to a rotation of grains to orientations on the <111>||TA- and <100>||TA fibers. Subsequently, the Schmid factor for twinning is high in <111>||TA oriented grains. Twinning in grains with orientations close to <111>||TA results in twin orientations close to <100>||TA []. Comparing the volume fractions of the texture components after deformation of the X30MnAl17-1_FG and X30MnAl17-1_CG conditions ((a)), the coarse-grained material shows a stronger increase of C and RtG texture components, indicating an increase in twin volume fraction, as both C and RtG are components of the <100>||TA-fiber.The relatively lower volume fraction of B in the X30MnAl17-1_CG alloy compared the fine-grained condition (cf. ), can be explained in terms of a higher fraction of grains of this orientation that exhibit deformation twinning. show the calculated positions of first-order deformation twins in the ODF section at φ2 = 45° for the B and RtCu texture components. Due to the use of a 15° radius around the ideal position to calculate the volume fraction of each texture component, the volume fraction of RtG is increased by a fraction of deformation twins formed in B-oriented grains, which appear in the φ2 = 45° ODF section at Φ = 90° and slightly higher φ1-values then the RtG texture component. Similarly, this is also the case for deformation twins in RtCu-oriented grains, which appear in the φ2 = 45° ODF section at φ1 = 0° and slightly lower Φ-values then the RtG texture component. Comparing the ODF sections at φ2 = 45° after fracture for the X30MnAl17-1 alloy (cf. ), the X30MnAl17-1_CG alloy revealed a slight peak in intensity at the position of deformation twins in B-oriented grains while the X30MnAl17-1_FG alloy did not. Accordingly, this also indicates that the slightly increased volume fraction of RtG and slightly lower volume fraction of B in the X30MnAl17-1_CG compared to the fine-grained condition was due to an increase in twin volume fraction. Comparing the X30MnAl17-1 alloys after 20% elongation, the increased twin volume fraction in the coarse-grained alloy results in a higher volume fraction of C (cf. ) and a stronger <100>||TA-fiber in general, as shown in the IPFs in Comparing the X60Mn17 alloy conditions, the volume fraction of the B and RtCu texture components (<111>||TA-fiber) in the coarse-grained material state increased less than in the X30MnAl17-1 alloy. This can be explained by an increase in twin volume fraction for the coarse-grained alloy (similar as in the X30MnAl17-1 alloy) on the one hand and the partial transformation to ε-martensite and α′-martensite on the other hand. The nucleation of martensite is promoted in areas of high dislocation density, such as twin-twin intersections and ε-martensite-twin intersections []. Since <111>||TA is a favorable orientation for deformation twinning, more nucleation sites for ε-martensite and α′-martensite are available in grains of such orientation []. With increasing deformation, the phase fraction of ε-martensite and α′-martensite increased (cf. ), leading to a reduced phase fraction of austenite and a reduced volume fraction of the <111>||TA-fiber texture components B and RtCu.The differences in twin volume fraction between the X60Mn17 alloy and the X30MnAl17-1 alloy in the EBSD maps in can also be discussed in context of the texture evolution in both alloys. The lower relative increase of the <111>||TA-fiber texture components B and RtCu in the X60Mn17 alloy after 20% elongation shown in the corresponding ODFs (cf. ) compared to those of the X30MnAl17-1 alloy indicate a higher twin volume fraction in the Al-free alloy. This is reflected by two findings: Firstly, the shift in intensity in the ODF sections at φ2 = 45° towards the position of first-order deformation twins of RtCu indicates that 20% elongation led to a higher volume fraction of first-order deformation twins in the Al-free HMnS. Secondly, the observation that the intensities around the B texture component increased only slightly suggests that a higher volume fraction of grains exhibited deformation twinning or underwent a martensitic transformation instead of rotating towards the <111>||TA-fiber. Because the nucleation of martensite is promoted at twin-twin intersections, the lower twin volume fraction in the X30MnAl17-1 alloy resulted in lower overall phase fractions of martensite (cf. ). Accordingly, the ductility is not negatively impacted by coarser grains in the X30MnAl17-1 alloy, as can be seen in . The difference in twin volume fraction for the X60Mn17 and X30MnAl17-1 alloys can be explained in terms of the impact of SRO []. Several authors point to a pinning effect of C and/or SRO clusters on the trailing partial dislocation of a stacking fault, increasing the stacking fault width and facilitating deformation twinning []. Al reduces the diffusion rate of C atoms in austenitic steels and increases the lattice parameter []. As a result, the distance of short-range jumps of C into preferential lattice positions, the likelihood of short-range jumps and the dislocation-pinning effect of SRO are reduced in Al-alloyed HMnS [To summarize, EBSD and macro-texture analysis showed that the coarse-grained condition of both alloys had a higher twin volume fraction, which in the case of the X60Mn17 alloy conditions resulted in a subsequently increased phase fractions of ε-martensite and α′-martensite. Due to the high hardness of α′-martensite, the elongation to fracture of the coarse-grained X60Mn17 condition was reduced.The impact of grain size on the mechanical properties, deformation mechanisms and texture development of two high-manganese steels (HMnS) was investigated in this study by means of mechanical testing, microstructural and texture analysis. Overall, the following conclusions can be drawn:A decrease in average grain size of the tested alloys from 16 μm (X60Mn17_CG) and 22 μm (X30MnAl17-1_CG) to 4 μm (X60Mn17_FG and X30MnAl17-1_FG) led to a considerable increase in yield strength (YS) and ultimate tensile strength (UTS). In the case of the X60Mn17 alloy, the YS increased from 275 MPa to 381 MPa while the elongation to fracture also increased significantly from 22% to 53%. For the X30MnAl17-1 alloy, the decrease in average grain size led to an increase in YS from 245 MPa to 338 MPa, which was accompanied by a small decrease in total elongation from 69% to 63%.The work-hardening rate (WHR) after the onset of plastic deformation is higher for the fine-grained material states, which is due to the initially smaller mean free path of dislocation movement. However, the strain hardening exponent nH is higher in the coarse-grained material states at εtrue>0.05. For the X30MnAl17-1 alloy, this difference in work-hardening behavior results in a higher elongation to fracture in the coarse-grained alloy.The higher values of nH at εtrue>0.05 for the coarse-grained states of both tested HMnS is due to a higher twin volume fraction in the coarse-grained condition, which effectively reduces the mean free path of dislocation movement during plastic deformation. Additionally, the higher twin volume fraction in the X60Mn17_CG condition facilitates the γ→ε- and/or γ→ε→α′-transformation, leading to higher martensitic phase fractions. In the X60Mn17_CG alloy, the high volume fraction of hard α′-martensite leads to early fracture.The impact of the influence of grain size on the activation of secondary deformation mechanisms is accompanied by changes in the texture development. The more pronounced TWIP- and/or TRIP-effect in the coarse-grained conditions lead to a relative softening of the <111>||TA-fiber in favor of grains with orientations close to the <100>||TA-fiber.This can be explained by a preferred activation of the TRIP and TWIP effect in favorably oriented grains close to the <111>||TA-fiber in austenite. For the TWIP effect, this results in first-order deformation twins close to the <100>||TA-fiber. Accordingly, both secondary deformation mechanisms lower the volume fraction of austenite grains rotating towards the <111>||TA-fiber during deformation.The authors declare no conflict of interest.Simon Sevsek conducted the mechanical testing and performed the EBSD measurements. Frederike Brasche conducted the phase analysis and XRD measurements. All authors participated in the analysis of experimental data and contributed in the writing process.The raw data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study. The processed data required to reproduce these findings cannot be shared at this time due to legal or ethical reasons.Optical Fiber Technology 12 (2006) 170–184 www.elsevier.com/locate/yofte Fiber misalignment in silicon V-groove based optical modules A. Priyadarshi a,b,∗ ,L.H.Fen a , S.G. Mhaisalkar a ,V.Kripesh b , A.K. Asundi c a School of Materials Science and Engineering, NTU, Singapore 639798 b Institute of Microelectronics, 11-Science Park Road, Science Park 2, Singapore 117685 c School of Mechanical and Aerospace Engineering, NTU, Singapore 639798 Received 17 May 2005 Available online 26 August 2005 Abstract The use of anisotropically etched silicon optical benches (SiOB) and optical adhesives have at- tracted much attention for the low cost assembly of optical modules. In this paper, the effect of attachment process parameters (such as the amount of bulk adhesive, and the thickness/symmetry of the adhesive layer between the fiber and V-groove walls) on the alignment of the fiber with the laser diode assembled on SiOB, is studied. Finite element analysis (FEA) on the fiber displacement within the V-groove at a given temperature, increases significantly for different thicknesses of adhesive lay- ers, present on either side of the fiber but does not change much with the total amount of adhesive in the module. Validation of the FEA data was accomplished by the use of in situ micro moiré in- terferometry, involving determination of the displacement of the centre of the fiber in the V-groove exposed to a temperature excursion ranging from room temperature to 125 ◦ C. Maximum displace- ments of the fiber centre in the vertical and horizontal directions relative to the SiOB were found to range between 0.69–1.21 and 0.36–0.42 µm, respectively at 100 ◦ C. The misalignment results ob- tained by moiré interferometry matched favorably with the finite element data. The simulation and experimental data in this study indicates that fiber misalignment due to temperature cycling can be reduced significantly if the fiber is located equidistant from the two inclined walls of the Si V-groove. © 2005 Elsevier Inc. All rights reserved. * Corresponding author. Fax: +65 67745747. E-mail address: [email protected] (A. Priyadarshi). 1068-5200/$ – see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.yofte.2005.07.003 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 171 Keywords: Silicon optical bench; V-groove; Single fiber; Adhesive joint; FEA; Micro moiré interferometry; Thermal stress; Misalignment 1. Introduction Assembly and packaging of multiple discrete optical components such as single mode fibers, lenses, laser diodes, and photo diodes, constitute a large fraction of the cost associ- ated with an optical module because of the precision requirements on the alignment [1–4]. To reduce the cost of the optical modules, it is important to simplify the package structures and to reduce the time for optical alignment. The silicon optical bench, with V-grooves for fiber placement, has become a popular technique for the assembly of optical modules due to its lower manufacturing cost and faster processing time [5–9]. MEMS devices, such as optical switches, actuator, sensors, and microsystem testing devices, use V-groove for the fiber alignment. The geometry of the V-groove facilitates defining the position of optical fiber, which simplifies the assembly process. To reduce the assembly cost further, the op- tical components are bonded with UV–curable adhesive, which offers more advantages in terms of mass production and economy than other joining techniques, such as laser weld- ing and soldering [10,11]. Apart from the low processing cost of the adhesive bonding, the technique is attractive for its ability to join a variety of materials, for the more homoge- neous stress distribution in joints, and the most important of all, for its transparent nature [12,13]. The surface and bulk properties of polymeric adhesives determine the quality of their preformance for a given application. Performance of the adhesive joints in the optical module is dependent on the properties of the optical adhesive, which, in turn is affected by environmental conditions (such as temperature and humidity). The performance of adhesive joints needs to be studied in detail for different systems in various testing conditions. It has been found in general that the stability and the reliability of the fiber-adhesive bonds are a major concern in optical modules [14–16]. Mechanical reliability problems are mainly induced due to the large shrinkage during the curing process [16,17] and the high coefficient of thermal expansion (CTE) of the adhesives. Adhesives, which are made of organic materials, have a coefficient of expansion that are an order of magnitude greater than the solder material. Large amount of elongation and contraction can take place in the adhesive joints due to the variation in the temperature. Because the CTE of the silicon substrate and the fiber is very low, the fiber movement would be mostly governed by the expansion and contraction of the adhesive. Due to this, misalignment is more probable in the optical fiber attached with adhesive. In a typical single-mode fiber application, a sub micrometer fiber shift induced by the temperature cycling results in significant loss in the coupled power, and performance degradation of the optical modules [3,4]. There are many factors involved in the adhesive- fiber joints such as design of the assembly platform, the amount of the adhesive and the adhesive material properties, which may be responsible for the fiber shift under temperature cycling. Furthermore, the order of the fiber shift is quite small, in the submicrometer range. It is difficult to measure a submicron level shift in the fiber during thermal loading. Thus, it is very hard to identify the factor which contributes most to the fiber shift during the thermal loading. 172 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 Fig. 1. (a) Schematic of laser diode to single mode fiber alignment on silicon optical bench. Different possible situation of fiber-adhesive joint in V-groove (b) fiber in direct contact with inclined walls, (c) fiber not in contact with wall but in symmetry position, and (d) unsymmetrical position of fiber. (e) Different forces acting on the fiber at the time of curing process of adhesive. Note the adhesive layer thickness between the fiber and V-grooves wall gap as shown in (c) and (d). The thermally induced fiber shift of fiber-solder-ferrule (FSF) joints in dual-in-line package (DIP) have been studied experimentally and numerically [18]. It is suggested to place the fiber in the center of ferrule to reduce the fiber shift of FSF joint in laser diode packaging based on the study. In this work, single mode laser module packaging on the SiOB with the V-groove groove for the fiber alignment is considered (Fig. 1). This model was selected because it represents the most important and crucial alignment in an opti- cal transceiver module for the communication network. In the paper, the fiber shift in the fiber-adhesive joint of the optical module is reported. Finite element modeling and experi- ments are carried out to understand the effect of adhesive thickness, between the fiber and inclined V-groove wall (discussed in Section 2), and the adhesive amount on the thermal induced displacement of the fiber. The sensitivity of the resulting alignment shift in the fiber because of the adhesive thickness variation between the fibers and V-groove walls is of primary interest. The temperature is varied from −40 to 125 ◦ C (JEDEC standard: JESD22-A104-B), and the fiber displacement induced by the thermal cycle is determined. Micro moiré interferometry (MMI) is used for in situ measurements of the fiber center displacement due to the temperature cycling. MMI has potential of measuring in plane dis- placement components with submicrometer sensitivity. The change in displacement field can be viewed in real time as the temperature is changed. The remaining sections of this paper are organized as follows. Section 2 describes the fiber-adhesive joint in the laser module and the positions of the fiber in the V-groove. The measurement method and the FEM method are presented in Sections 3 and 4, respectively. In Section 5 results of numerical simulations and measurements are discussed, and com- pared. A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 173 2. Fiber-adhesive joint in V-groove The fiber-adhesive joint in the silicon V-groove of the laser module is shown in Fig. 1a. The final position of the fiber depends on the how the fiber is held, the curing conditions, and the adhesive materials properties. Three different possible positions of the fiber in the V-groove after the adhesive curing are shown in Fig. 1: (A) the optical fiber is in contact with both the inclined walls of the V-groove (Fig. 1b), (B) the optical fiber does not touch either of the inclined walls (Fig. 1c), and (C) the fiber touches one of the inclined walls (Fig. 1d). The fiber can remain in contact with the V-groove walls if high pressure is applied through out the adhesive curing process. An intentional spacing, between the fiber and the walls, can be introduced for better alignment and to minimize the residual stress around the fiber, or for physical reasons. For example: (1) Generally, the laser diode chips have a dimensional uncertainty of more than ±10 µm in height and the location of the active area with respect to the upper surface has ±2 µm offset. This is one of the main barriers in adopting passive alignment in op- tical module assembly. In some modules during active alignment, to compensate for laser dimension uncertainty, the fiber is placed at a fixed distance from the V-groove walls. (2) Internal stresses exist within the optical adhesives wherever chain segments are un- der tension or compression. Mainly, the stresses arise during the curing process of the adhesives. They are often located at regions where parts of molecules are unequally attracted by the neighbors. Such unequal attraction can take place at the point of direct contact of the fiber and the V-groove wall in the fiber-adhesive joint. Resid- ual stresses can occur when the crosslink density changes rapidly from one location to another within the adhesive. These stresses which arise during curing or in use, can lead to failure of an adhesive because they may act as crack initiators. Stress distribution studies for solder and fiber joint on silicon substrate shows that a very high stress is developed underneath the fiber in the solder compared to the outer re- gion [22]. It was also found that shear strain increases significantly in the adhesive by decreasing the spacing between the fiber and substrate. Thus, a continuous adhesive layer between the fiber and silicon wall is important for reason of mechanical stabil- ity. (3) Optical adhesives may become entrapped between the V-groove walls and the fiber due to lack of appropriate pressure during placement, excess volume, and material shrinkage during the attachment process. In Fig. 1e different forces acting during the attachment process are shown. The fiber size is quite small (diameter 125 µm, and length few millimeters), thus buoyancy forces dominate over body forces before the curing. Throughout the curing process shrinkage of the adhesive can compel the fiber to separate from the inclined walls. Another factor is irregular surface (roughness) of V-groove wall and the low viscosity of the optical adhesives, which causes sliding of the adhesive during the dispensing. Based on the three-point contact principal, by applying pressure on the top of the fiber, misalignment due to curing shrinkage can be ruled out. But, this will introduce inhomogeneous stress in the joint, as discussed earlier. 174 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 The gap between the fiber and the inclined V-groove wall is quantified with the help of adhesive thicknesses. The adhesive thickness between the fiber and V-groove wall, shown in Fig. 1c, is defined as the shortest gap perpendicular to the inclined wall. 3. Sample preparation and experimental method 3.1. SiOB assembly A SiOB, with a U-groove for the 1.3 or 1.55 µm laser diode (LD) and a V-groove for the single mode fiber (Fig. 1a), was fabricated by standard KOH anisotropic etching [21]. The geometry of the grooves was determined based on the dimensions of bare single mode fiber (cladding diameter = 125 µm) and the LD chip (250 × 250 × 100 µm). To minimize the coupling loss due to the dimensional uncertainty of the chip, the width of U-groove is varied (320 ± 10 µm). After optimizing the SiOB dimensions and fabrication process, the attachment of laser diodes using solder and fibers using UV–curable adhesive, was com- pleted. The laser diode was placed on the solder preform in the U-groove, and the SiOB was heated above the solder melting point (290 ◦ C) to form the metallurgical bonds. A fi- nite pressure is applied, throughout the heating process, on the chip to make the molten solder flow faster and keep the LD in contact with the silicon walls. Next a bare fiber, stripped of its protective acrylic coating and cleaned using acetone, was placed onto the V-groove. To maintain the alignment of the fiber to the groove a small weight was used to hold down the fiber. In some of the modules, for compensating the diode chip dimension uncertainty, the fibers were placed at a fixed distance from the V-groove walls. The UV curing optical adhesive was dispensed into the V-groove for the fiber attachment. The ad- hesive was spread evenly throughout the substrate so as to ensure 100% uniform adhesion of the fiber. An UV light source, with 110 mW/cm 2 power at 365 nm, was used for 40 s to cure the adhesive and to bond the fiber onto the V-groove. 3.2. Sample preparation for micro moiré interferometry (MMI) For MMI, the specimens were prepared to reveal the silicon substrate, optical fiber, and adhesive bond interfaces. The laser diode module was carefully ground and polished to reveal the area of interest. The SEM picture of specimens, after grinding and polishing, shows the initial position of the fiber in the V-groove for two samples (Figs. 2a and 2b). It should be noted that the adhesive layer thicknesses between fiber and inclined walls cannot be controlled and are not the same for the samples. A crossed line diffraction grat- ing was transferred to the interface by the replication technique [20]. A thin layer of a transparent adhesive (BA-F114) was used for sticking the grating on the samples. There is no initial temperature history in the fiber-adhesive joint because of the room tempera- ture curing adhesive used for the grating transfer. The diffraction grating has a frequency of 1200 lines/mm and is cross-line grating with an orthogonality to within (π/2±0.5×10 −4 ) radians. A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 175 (a) (b) Fig. 2. SEM picture of fiber-adhesive joint in V-groove for moiré interferometry joint of (a) sample 1 and (b) sam- ple 2. Fig. 3. Schematic diagram of micro moiré interferometric setup. 3.3. Fiber-shift measurement setup A schematic of the micro moiré interferometer is shown in Fig. 3. Micro moiré inter- ferometry (MMI) is an optical technique that can determine in-plane displacement com- ponents with nanometer sensitivity [20]. The specimen grating interacts with the virtual reference grating, formed by two-beam interference, to generate moiré fringe patterns. The virtual grating, also known as the reference grating, controls the sensitivity of the mea- surement. A 30 mW Helium Neon laser (wavelength 633 nm) with a split fiber coupler is used to produce a virtual grating of 2400 lines/mm in the plane of the fiber adhesive joint. The SiOB is mounted on a three-axis translation aluminum stage with heater. The temper- 176 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 ature controller, which consists of a solid-state relay and a proportional integral differential controller, is used to control the temperature of the specimen on the heating block. 3.4. Moiré pattern capturing The samples were heated using the hot plate from room temperature (23 ◦ C) to 100 ◦ C as shown in Fig. 3 to simulate the temperature cycling profile. The thermocouple at- tached directly to the sample provided feed back to the temperature controller. This special arrangement helped in maintaining the quality of fringe pattern by avoiding any thermal drift in the environment near to the specimen surface. The temperature of the sample was allowed to stabilize for 10 min before the moiré patterns were captured. The displacement components in the X- and Y -directions were recorded at temperatures of 40, 60, 80, and 100 ◦ C. The number of moiré fringes in the region of interest (optical fiber-adhesive joint, approximately 200 µm in width, and height each) were insufficient for analysis. Thus, car- rier fringes were added before thermal loading, to realize the displacement components in the X- and Y -direction with better accuracy and precision. 4. Finite element analysis A steady-state thermo-elastic finite element analysis is carried out for analyzing the thermal effect on the shift of fiber center of the fiber-adhesive joints. The temperature T at a given point gives rise to thermally induced strains ε th given in a general three dimension (3-D) case as ε th = braceleftbig ε th x ε th y ε th z bracerightbig T = braceleftbigslurabove T slurabove T slurabove T 000 bracerightbig T α, slurabove T = T − T0, (1) where α is the matrix containing thermal expansion coefficients, T is the temperature at the given point, and T 0 is the temperature at which the structure is free of thermally induced strains (typically room temperature). ANSYS [19], a commercial finite element code, is employed for calculating the thermal strain in the elements of the module. Figure 4 shows a configuration of the 2-D mechanical model of optical module established by the ANSYS software. The four node thermal-mechanical coupled element (PLANE13) is used in this direct couple field analysis. The FEA model consists of more than 7000 nodes and the mesh in the neighborhood of the fiber-adhesive joint is refined in order to achieve suffi- cient accuracy in the calculation. The material properties for different components of the module are listed in Table 1. The diameter of single mode fiber core and cladding are 8.5 and 125 µm, respectively, and identical material properties have been considered for both. Temperature dependent coefficient of thermal expansion (CTE) and modulus of adhesive was determined experimentally by a thermal mechanical analyzer (TMA, TA Instrument Model 2940) and a dynamic mechanical analyzer (DMA, TA Instrument Model 2980). It is assumed that, materials in the package have isotropic material properties and thermal expansion coefficient of silicon does not change during the course of temperature cycling. Note that the presence of oxide on the silicon substrate is ignored. During the thermal loading, the silicon substrate was partially constrained with zero translation of the lower surface in Y -direction. A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 177 Fig. 4. A representative FEA model of a fiber-adhesive joint with the fiber in the V-groove center. Model is meshed with plane 13 element of ANSYS and region around the silicon-adhesive-fiber attachment is very fine meshed. Table 1 Thermo-mechanical properties of fiber-adhesive joint materials Young modulus (N/µm 2 )CTE× 10 −6 ( ◦ C −1 ) Single mode fiber (SMF) 0.80.5 Silicon (Si) 0.137 3.8 Adhesive (−40 ◦ C) 0.0012 60 Adhesive (100 ◦ C) 0.0010 120 Adhesive (150 ◦ C) 0.0007 120 Note. SMF and Si properties are as reported in Ref. [4]. 5. Results and discussion 5.1. FEA simulation results Figure 5a shows the Y -displacement of the fiber center with respect to the temperature for different adhesive thicknesses. It has been assumed that the fiber center is perfectly aligned with respect to the laser diode at room temperature (23 ◦ C) and the adhesive layer thickness on both sides of the inclined walls is equal. The X-displacement is found to be zero because of the symmetrical position of the fiber in the V-groove. The displacement of the fiber center is due to the adhesive expansion during a change in the temperature. The effect of silicon and fiber expansion is negligible because of the low CTE of these materials (Table 1). In Fig. 5b, the Y -displacement of the fiber center for various adhesive thicknesses is plotted as a function of the temperature. The displacement of the fiber varies linearly with respect to the layer thickness. For an adhesive thickness of 8 µm the fiber center is displaced by 1.1 µm, sufficient to cause significant coupling loss for a single mode optical module [3,4]. 178 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 (a) (b) Fig. 5. (a) Fiber displacements (node at fiber center) calculated with respect to temperature for different adhesive thickness. (b) At temperature 125 ◦ C, fiber shift with respect to the adhesive thickness. The thickness of optical adhesive, between fiber and the inclined groove walls, may be non-uniform due to various factors: mechanical problems, manual attachment process and adhesive shrinkage. The effect of non-uniformity of the adhesive thickness on the fiber displacement is also simulated by means of the same FEA model. Figure 6 shows the X- and Y -displacements of the fiber center, when the fiber is not equidistant from the inclined V-groove walls. In this case, the X-displacement of the fiber is non-zero due to the unsymmetrical position of the fiber in the groove. The X-displacement of the fiber is found to be less than 0.2 µm and the slope of the variation is also small (Fig. 6). Thus, the main reason for the optical loss in V-groove type optical module is the fiber shift in the Y -direction while the silicon V-groove helps in constraining the displacement in the X-direction. The displacement of the fiber in the asymmetric condition is more than in the symmetric case for the same increase in temperature, as shown in Fig. 7. Thus it is better to place the fiber symmetrically with respect to the inclined walls to avoid optical coupling loss due to the misalignment. Next the effect of bulk adhesive on the fiber displacement is studied. Figure 8 shows the displacement of the fiber center with respect to the adhesive area. A four-fold increase in adhesive area causes a shift of less than 0.1 µm in the fiber center. Thus as increase in the amount of adhesive does not affect the displacement of fiber significantly. A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 179 Fig. 6. Calculated fiber displacement shifts as a function of adhesive thickness between the fiber and the left wall of V-groove. Fiber is assumed to be in unsymmetrical position in V-groove. Fiber distance from both the walls is considered same for the gap of 10 µm. Fig. 7. Comparison of Y -displacement for asymmetric and symmetric position of the fiber in V-groove. 5.2. Displacement measurement results Figure 9 shows the fringe patterns representing the displacement fields in the X- and Y -direction, respectively for a fiber-adhesive bond at room temperature, 60 and 100 ◦ C respectively. From the fringe patterns it can be seen that the fringe density increases as the modules are subject to higher temperatures. In the same figure, it can be seen that the number of fringes in the epoxy area exceeds the number of fringes in the silicon substrate because of the high CTE of the epoxy compared to the silicon substrate (Table 1). The recorded fringe patterns obtained by MMI were used to find the displacements U and V in the X- and Y -directions, respectively of the center of single mode fiber with respect to the silicon substrate via the equations U = ∇Nx f ,V= ∇Ny f , (2) where ∇Nx and ∇Ny are change in the fringe orders in the X- and Y -directions, respec- tively, and f (2400 lines/mm) is the spatial frequency of the reference grating. From the 180 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 Fig. 8. Displacement of fiber center with adhesive area. Fiber is placed at a distance of 10 µm from both inclined walls of V-groove. Fig. 9. Displacement field (a) X-direction and (b) Y -direction for the fiber-adhesive joint sample 1 (Fig. 2a). X-displacement field patterns, the number of fringes from one edge of the V-groove to the center of the fiber is calculated while from the Y -displacement field pattern, the number of fringes from the lower vertex of the V-groove to the center of the fiber is calculated. 5.3. Comparison and discussion between FEA, and experimental results Figures 10 and 11 compare the displacement components of the fiber center for the two modules determined by MMI and FEA. The FEA simulation and the MMI experimental A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 181 Fig. 10. Measured and calculated displacement of fiber center of sample 1 (Fig. 2a). Fig. 11. Measured and calculated displacement of fiber center of sample 1 (Fig. 2b). results for the fiber shift of the two modules at different temperatures agree well with each other. This displacement was induced by the interaction of expansions in fiber, epoxy and silicon substrate, with the contribution from the epoxy being the largest due to its high CTE value (Table 1). From the graphs it can be seen that the magnitude of the vertical displacement for the fibers is larger than that of the horizontal displacement. This is due to the confinement of the optical fiber and the epoxy within the V-groove in the horizontal direction (Fig. 2). On the other hand, the vertical expansion of epoxy, being unrestricted, allows the optical fiber to move in the vertical direction (Fig. 2). Similar results for the X- and Y -displacements were found in the FEA simulation also. Although the modules showed similar trend in the displacements, the amount of vertical displacements are differ- ent. At 100 ◦ C, the maximum vertical displacement in module 1 is about 0.7 µm while for the module 2 it is 1.2 µm. The difference is caused by the difference in initial positions of the fibers with respect to V-groove walls (Fig. 2). In other words, the amount of adhesive between outer surface of the fiber and the wall affects the vertical displacement. A small horizontal displacement was also observed in the module, which was due to the off-center placement of the fiber in the V-groove. 6. Conclusions In this paper, the fiber center shift is examined numerically and experimentally in the fiber-adhesive joint on SiOB subjected to the thermal loads. These results can be used for 182 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 fabricating reliable and high-performance optical module packages with the V-groove for use in optical communication system applications. The following conclusion can be drawn from the performed study. • The effect of the adhesive layer thickness and the bulk adhesive amount on the fiber shift has been shown by numerical analysis. As a result, it is seen that the Y -shift increases with as increase in the adhesive thickness and the X-shift decreases as the fiber is placed symmetrically in the V-groove. It was also found that the effect of bulk adhesive on the fiber shift is negligible for the fiber-adhesive joint in the V-groove. • For a given adhesive thickness between the fiber and the inclined walls, the fiber dis- placement was found to change linearly with the temperature. The major cause of fiber shift comes from the adhesive expansion because of its high CTE value compare to the silicon and the optical fiber. The fiber displacement according to FEA calculation is 1.0 µm, for fiber-adhesive joints on SiOB, with an adhesive layer thickness of 10 µm at 125 ◦ C. • Micro moiré interferometry was used to understand the in situ submicron level dis- placement of the single mode fiber center during a temperature cycle. Maximum displacement of the fiber center in the vertical and horizontal directions with respect to the SiOB was measured to be 0.69–1.21 and 0.36–0.42 µm, respectively at 100 ◦ C. • Experimental results were compared with finite element analysis and were found to be in close agreement. Thus, it is shown that MMI is an effective tool to measure the misalignment of the optical components and its application can be further extended to study the fatigue failure of the optical modules. References [1] J.M. Verdiell, G.H. Wiseman, New approach to component-level packaging for 10 Gbps optical technolo- gies, Intel Develop. Update Mag. (2002) 1–6. [2] D. Ruxton, B. Ashby, Package deals, SPIE OE Mag. (2003) 26–28. [3] A.R. Mickelson, R.B. Nagesh, Y.-C. Lee, Optoelectronic Packaging, Wiley, New York, 1997. [4] A.R. Faidz, H. Ghafouri-Shiraz, K. Takahashi, H.T. Chuah, Analysis of combined ball lens and conically lensed fiber scheme to improve the coupling efficiency and misalignment tolerance between laser diodes and single mode fibers, J. Opt. Commun. 22 (3) (2001) 82–86. [5] M. Takahshi, T. Ido, T. Nagara, A polymer PLC platform with a fiber-alignment V-groove for a low-cost 10-GbE WWDM transmitter, IEEE Photon. Technol. Lett. 16 (1) (2004) 266–268. [6] T. Bricheno, H.F.M. Priddle, R.G. Peall, D. Spear, Packaging of optoelectronic devices with high coupling efficiency using silicon V-groove technology, in: Opt. Fiber Commun. Conf. Ser. Tech. Dig., vol. 2, 1998, paper ThS4. [7] D.C. Abeysinghe, V. Ranatunga, A. Balagopal, H. Mu, K. Ye, D. Klotzkin, A novel technique for high- strength direct fiber-to-Si submount attachment using field-assisted anodic bonding for optoelectronics packaging, IEEE Photon. Technol. Lett. 16 (9) (2004) 2150–2152. [8] D. Shimura, M. Uekawa, K. Kotani, Y. Maeno, H. Sasaki, T. Takamori, Silicon microlens on V-groove platform for low-cost and high-performance optical modules, in: IEEE Proceedings of 53rd Electronic Com- ponents and Technology Conference, 2003, pp. 1318–1322. [9] K. Kurata, K. Yamauchi, A. Kawatani, A. Goto, N. Kimura, K. Higashikawa, S. Dohmae, H. Tanaka, S. Ishi- kawa, Low cost optical module packaging techniques for optical access network systems, IEICE Trans. Electron. E80-C (1) (1997) 98–106. A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 183 [10] T. Qing, Y.C. Lee, Solder technology for optoelectronic packaging, in: IEEE Proceedings of 46th Electronic Components and Technology Conference, 1996, pp. 26–36. [11] N. Murata, Adhesive for optical devices, in: IEEE Proceedings of 48th Electronic Components and Tech- nology Conference, 1998, pp. 1178–1185. [12] Y. Yamamada, F. Hanawa, T. Kitoh, T. Maruno, Low-loss and stable fiber-to-waveguide connection utilizing UV curable adhesive, IEEE Photon. Technol. Lett. 4 (1992) 906–908. [13] Y. Hibino, F. Hanawa, H. Nakagome, M. Ishii, N. Takato, High reliability optical splitters composed of silica-based planar lightwave circuits, J. Lightwave Technol. 13 (1995) 1728–1735. [14] D. Wu, Y.C. Lee, Reliability study of an epoxy-bonded laser-to-fiber assembly, in: IEEE Proceedings of 48th Electronic Components and Technology Conference, 1998, pp. 1186–1191. [15] M. Derosa, S. Logunov, Photothermal behavior of an optical path adhesive used for photonics applications at 1550 nm, Appl. Opt. 40 (2001) 6611–6617. [16] M.A. Uddin, H.P. Chan, K.W. Lam, Y.C. Chan, P.L. Chu, K.C. Hung, T.O. Tsun, Delamination problems of UV-cured adhesive bonded optical fiber in V-groove for photonic packaging, IEEE Photon. Technol. Lett. 16 (2004) 1113–1115. [17] Y. Lin, W. Liu, F.G. Shi, Adhesive for optical devices, in: IEEE Proceedings of 52nd Electronic Components and Technology Conference, 2002, pp. 1178–1185. [18] W.H. Cheng, M.T. Sheen, C.P. Chien, H.L. Chang, J.H. Kuang, Reduction of Fiber Alignment Shifts in Semiconductor Laser Module Packaging, J. Lightwave Technol. 18 (6) (2000) 842–848. [19] ANSYS modeling and meshing guide: ANSYS release 6.0, ANSYS Inc., 2001. [20] D. Post, B. Han, P. Ifju, High Sensitivity Moire: Experimental Analysis for Mechanics and Materials, Springer-Verlag, New York, 1994. [21] E.J. Murphy, Fiber Attachment for guided wave device, J. Lightwave Technol. 6 (1988) 862–871. [22] M. Rassaian, M.W. Beranek, Quantitative characterization of 96.5Sn3.5Ag and 80Au20Sn optical fiber sol- der bond joints on silicon micro-optical bench substrates, IEEE Trans. Adv. Pack. 22 (1999) 86–93. Anish Priyadarshi graduated from the Materials Science Department, Indian Institute of Tech- nology, Kanpur, with a Bachelors’s degree in May 2001. He is presently pursuing a Ph.D. degree in the School of Materials Engineering, Nanyang Technological University (NTU). His research project on Optoelectronic Transceiver Modules is funded by A*STAR as part of the Optical Network Focus Interest Group (ONFIG) and represents a collaborative effort between Institute of Microelectron- ics (IME) and NTU. His research interests include optoelectronics packaging and reliability testing, optical metrology, and modeling of optical interconnects as well as thermomechanical behavior of packages and materials. Dr. Subodh Mhaisalkar is an Associate Professor at the School of Materials Engineering, Nanyang Technological University (NTU), Singapore. Prior to NTU, Subodh has over ten years experience in senior R&D and Process Engineering Positions in the field of microelectronics. He has held positions of Director of Engineering in ST Assembly&Test Services and senior managerial po- sitions in National Semiconductor and Singapore Institute of Manufacturing Technology (SIMTech). In his career in the industry, he has established design and advanced process development for pack- aging of microprocessors, flip chip assemblies, chip scale packages, and ball grid arrays. Subodh received his Bachelor’s degree from IIT Bombay, and Masters and Ph.D. degrees from The Ohio State University. His current research interests and areas of expertise include microelectronics and optoelectronics packaging, advanced silicon process technologies, nanostructured interconnect ma- terials, and reliability. Dr. Vaidyanathan Kripesh carried out his doctoral degree at Max Planck Institute for Metal- forschung, Stuttgart, Germany, in area of microelectronics. His research interests are 3D-stacked modules, electro-optical integration, and wafer level packaging technologies. He worked as visit- ing scientist at Infineon Technologies, Munich in the 3D-integrated ICs. He is presently working 184 A. Priyadarshi et al. / Optical Fiber Technology 12 (2006) 170–184 at Institute of Microelectronics, Singapore in area of 3D-stacked silicon micro modules and wafer level process. He has authored more than 25 publications and he holds 12 patents to his credit. Dr. Vaidyanathan Kripesh is a IEEE&IMAPS Member and he is presently the President of IMAPS, Singapore Chapter. Anand K. Asundi graduated with a B.Tech (Civil Eng.) followed by a M.Tech (Aero. Structures) from the Indian Institute of Technology, Bombay. Subsequently he received his Ph.D. degree from the State University of New York at Stony Brook. Following a brief tenure Virginia Tech., he joined the University of Hong Kong in 1983. He was with the University till 1996 as Senior Lecturer, Reader, and Professor. He is currently Professor and Director of the Sensors and Actuators Programme at the Nanyang Technological University in Singapore. His teaching area is in solid mechanics and his research interests are in optical methods in micromechanics and biomechanics, on-line structural health monitoring and fiber optic bio-chemical sensors. He has published extensively and presented seminars/talks at various institutions and at international conferences. He is Editor of Optics and Lasers in Engineering and was Associate Editor with Experimental Mechanics. He is a Chartered Engineer, Fellow of SPIE, and Member of I.MechE and OSA. He is also R&D Chair of the Photonics Association (Singapore) and Chair of SPIE Singapore Chapter. Fiber misalignment in silicon V-groove based optical modulesThe use of anisotropically etched silicon optical benches (SiOB) and optical adhesives have attracted much attention for the low cost assembly of optical modules. In this paper, the effect of attachment process parameters (such as the amount of bulk adhesive, and the thickness/symmetry of the adhesive layer between the fiber and V-groove walls) on the alignment of the fiber with the laser diode assembled on SiOB, is studied. Finite element analysis (FEA) on the fiber displacement within the V-groove at a given temperature, increases significantly for different thicknesses of adhesive layers, present on either side of the fiber but does not change much with the total amount of adhesive in the module. Validation of the FEA data was accomplished by the use of in situ micro moiré interferometry, involving determination of the displacement of the centre of the fiber in the V-groove exposed to a temperature excursion ranging from room temperature to 125 °C. Maximum displacements of the fiber centre in the vertical and horizontal directions relative to the SiOB were found to range between 0.69–1.21 and 0.36–0.42 μm, respectively at 100 °C. The misalignment results obtained by moiré interferometry matched favorably with the finite element data. The simulation and experimental data in this study indicates that fiber misalignment due to temperature cycling can be reduced significantly if the fiber is located equidistant from the two inclined walls of the Si V-groove.Anish Priyadarshi graduated from the Materials Science Department, Indian Institute of Technology, Kanpur, with a Bachelors's degree in May 2001. He is presently pursuing a Ph.D. degree in the School of Materials Engineering, Nanyang Technological University (NTU). His research project on Optoelectronic Transceiver Modules is funded by A*STAR as part of the Optical Network Focus Interest Group (ONFIG) and represents a collaborative effort between Institute of Microelectronics (IME) and NTU. His research interests include optoelectronics packaging and reliability testing, optical metrology, and modeling of optical interconnects as well as thermomechanical behavior of packages and materials.Dr. Subodh Mhaisalkar is an Associate Professor at the School of Materials Engineering, Nanyang Technological University (NTU), Singapore. Prior to NTU, Subodh has over ten years experience in senior R&D and Process Engineering Positions in the field of microelectronics. He has held positions of Director of Engineering in ST Assembly&Test Services and senior managerial positions in National Semiconductor and Singapore Institute of Manufacturing Technology (SIMTech). In his career in the industry, he has established design and advanced process development for packaging of microprocessors, flip chip assemblies, chip scale packages, and ball grid arrays. Subodh received his Bachelor's degree from IIT Bombay, and Masters and Ph.D. degrees from The Ohio State University. His current research interests and areas of expertise include microelectronics and optoelectronics packaging, advanced silicon process technologies, nanostructured interconnect materials, and reliability.Dr. Vaidyanathan Kripesh carried out his doctoral degree at Max Planck Institute for Metalforschung, Stuttgart, Germany, in area of microelectronics. His research interests are 3D-stacked modules, electro-optical integration, and wafer level packaging technologies. He worked as visiting scientist at Infineon Technologies, Munich in the 3D-integrated ICs. He is presently working at Institute of Microelectronics, Singapore in area of 3D-stacked silicon micro modules and wafer level process. He has authored more than 25 publications and he holds 12 patents to his credit. Dr. Vaidyanathan Kripesh is a IEEE&IMAPS Member and he is presently the President of IMAPS, Singapore Chapter.Anand K. Asundi graduated with a B.Tech (Civil Eng.) followed by a M.Tech (Aero. Structures) from the Indian Institute of Technology, Bombay. Subsequently he received his Ph.D. degree from the State University of New York at Stony Brook. Following a brief tenure Virginia Tech., he joined the University of Hong Kong in 1983. He was with the University till 1996 as Senior Lecturer, Reader, and Professor. He is currently Professor and Director of the Sensors and Actuators Programme at the Nanyang Technological University in Singapore. His teaching area is in solid mechanics and his research interests are in optical methods in micromechanics and biomechanics, on-line structural health monitoring and fiber optic bio-chemical sensors. He has published extensively and presented seminars/talks at various institutions and at international conferences. He is Editor of Optics and Lasers in Engineering and was Associate Editor with Experimental Mechanics. He is a Chartered Engineer, Fellow of SPIE, and Member of I.MechE and OSA. He is also R&D Chair of the Photonics Association (Singapore) and Chair of SPIE Singapore Chapter.Structure and electrochemical properties of composite electrodes synthesized by mechanical milling Ni-free TiMn2-based alloy with La-based alloysIn the present study, hydrogen storage composite electrodes were prepared by mechanical milling the powder mixtures of Ni-free Laves phase alloy Ti0.9Zr0.2Mn1.5Cr0.3V0.3 (AB2) with LaNi3.8Mn0.3Al0.4Co0.5 (AB5) and La0.7Mg0.25Zr0.05Ni2.975Co0.525 (AB3.5), respectively. X-ray diffraction (XRD) measurements found that the basic phase structure (hexagonal C14) was still maintained in TiMn2-based alloy after short-time mechanical milling with additional La-based alloys. The fine particles of La-based alloy were found dispersing over the bulk particle of TiMn2-based alloy by observations of scanning electron microscopy (SEM) with energy dispersive spectrometer (EDS). The electrochemical studies showed that the additional La-based alloys greatly improved the discharge capacity of the composite electrode. The maximum discharge capacity reached 310.4 mAh/g and 314.0 mAh/g for AB2–10 wt.% AB5 and AB2–10 wt.% AB3.5 electrodes, respectively, which was much higher than the maximum 48.6 mAh/g of original Ti0.9Zr0.2Mn1.5Cr0.3V0.3 alloy electrode. Electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) measurements suggests that AB3.5-type alloy as a surface-modifier is beneficial to the decrease of the charge-transfer reaction resistance. The mechanical milling with AB5-type alloy was found improving the hydrogen diffusion in the bulk of the alloy from the results of anodic polarization measurement.Among the hydrogen storage alloys that have been extensively investigated, Ti-based AB2-type Laves phase hydrogen storage alloys are potential candidates for negative electrode materials in Ni-MH batteries due to their larger hydrogen absorption capacities at ambient temperature In recent years, the mechanical milling treatment is regarded as an effective method in surface modification and improving some electrochemical properties of the hydrogen storage alloys such as discharge capacity, cyclic stability or kinetics Up to now, many studies showed that some of new series of R–Mg–Ni-based (where R is a rare earth or Y, Ca) alloys with higher hydrogen storage capacity were developed. Pan et al. As mentioned above, the structure of La–Mg-based alloy is different from that of AB5-type alloy and its electrochemical properties are superior to those of AB5-type alloy. Therefore, in this study, AB5-type alloy LaNi3.8Mn0.3Al0.4Co0.5 and La–Mg-based AB3.5-type alloy La0.7Mg0.25Zr0.05Ni2.975Co0.525 were selected as surface modifiers for Ni-free Laves phase Ti0.9Zr0.2Mn1.5Cr0.3V0.3 alloy. The AB2–AB5 and AB2–AB3.5 composites were prepared by mechanical milling method. Their structures and respective effects on the electrochemical performances were investigated systematically in the present work.Nonstoichiometric Ni-free AB2-type Laves phase alloy Ti0.9Zr0.2Mn1.5Cr0.3V0.3, AB5-type alloy LaNi3.8Mn0.3Al0.4Co0.5 and AB3.5-type alloy La0.7Mg0.25Zr0.05Ni2.975Co0.525 were prepared by vacuum magnetic levitation melting under argon atmosphere. The ingots were turned over and re-melted three times for homogeneity. The purity of starting elemental metals was higher than 99%. Then the ingots were mechanically crushed and ground into the powder of 200-mesh size for mechanical milling treatment. These alloys were represented as AB2 alloy, AB5 alloy and AB3.5 alloy hereafter, respectively.The alloy powder of AB2 was mixed homogenously in the weight ratio of 9.0:1.0 with AB5 and AB3.5 alloy powder, respectively, and ground by QM-1SP planetary ball miller under 0.2–0.3 MPa argon atmosphere for 2 h. In each stainless milling pot, the ball-to-powder weight ratio was 20:1. As compared, the AB2 alloy was ball milled under the same condition. The crystal structure of the alloys was characterized by XRD (Rigaku D/max-2500, Cu Kα, 40 kV, 250 mA). The surface morphologies of the alloys were observed using scanning electron microscopy (SEM, JSM6360LV) linked with energy dispersive X-ray spectrometer (EDS, Oxford INCA). A N4 plus laser scattering particle meter was used to measure particle size distribution of alloy powder.The tested electrodes for electrochemical measurements were fabricated by mixing 100 mg the prepared alloy powder with 300 mg electrolytic Ni powder. The mixture was then pressed into a pellet of 10 mm in diameter under a pressure of 30 MPa. Both sides of the electrode pellet were coated with two foam nickel sheets, then pressed at 6 MPa and tightly spot-welded. A nickel lead wire was attached to this pressed foam nickel sheet by spot welding. Electrochemical measurements were performed at 303 K in a standard open tri-electrode electrolysis cell consisting of a working electrode (the MH pellet electrode for studying), a sintered Ni(OH)2/NiOOH counter electrode, and a Hg/HgO reference electrode immersed in the 6 M KOH electrolyte. Charge–discharge cycles of alloy electrodes were conducted by an automatic LAND battery test instrument. The electrodes were charged for 2 h at a current density of 300 mA/g, rested for 5 min and then discharged to the cut-off potential of −0.6 V versus Hg/HgO reference electrode at a current density of 100 mA/g.The electrochemical impedance spectroscopy (EIS), cyclic voltammograms (CV) and anodic polarization (AP) measurements were conducted on Zahner Elektrik IM6e electrochemical workstation at 50% depth of discharge (DOD) at 303 K. The scan rate of CV measurements was 50 mV/s within the scan potential range from −1.2 V to 0 V. The frequency range of EIS was from 10 kHz to 5 mHz with an amplitude of 5 mV versus open circuit potential. Before the EIS and CV measurements, the alloy electrodes were first activated by charge–discharge for three cycles. The anodic polarization curves were measured by scanning the electrode potential with the rate of 5 mV/s from 0 mV to 600 mV (versus open circuit potential). The hydrogen diffusion coefficient was determined by potential step method, which was performed under the constant potential of +500 mV for 3600 s at the fully charged state of the fourth charge–discharge cycle. shows that X-ray diffraction patterns of ball-milled AB2 alloy, AB2–AB5 and AB2–AB3.5 composite alloys. It was found that the Ti–Mn-based AB2 alloy could be indexed as the hexagonal C14 Laves phase and some impurity of metallic vanadium. It was also found that mechanical milling with 10 wt.% rare-earth alloys did not change the basic structure (hexagonal C14 Laves phase) of the original AB2 alloy. illustrates that SEM images and EDS patterns of ball-milled AB2 alloy and AB2–AB3.5 composites. It was found that both AB2 alloy particles and rare-earth alloy particles were reduced in size. The surface of the milled alloy particles became much rougher after mechanical milling. This means that some fresh surfaces have been generated during the process of mechanical milling. In the case of ball-milled AB2–10 wt.% AB3.5 composite alloy, as seen from (b), the differences of EDS patterns between micro-area (I) and (II) revealed that the small and white particles on the surface of the bulk alloy were the La0.7Mg0.25Zr0.05Ni2.975Co0.525 alloy particles which adhered to the bulk of Ti0.9Zr0.2Mn1.5Cr0.3V0.3 alloy particles. This phenomenon resulted from the different hardness and ductility between the two alloys. The rare-earth alloy particles might be pulverized more easily and adhered to the surface of the larger Ti0.9Zr0.2Mn1.5Cr0.3V0.3 alloy particles according to the detection of EDS. presents the discharge curves of ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes at the largest discharge capacity and some electrochemical properties are summarized in , a discharge capacity of 48.6 mAh/g without an apparent discharge plateau was observed for ball-milled AB2 alloy electrode, which was ascribed to the absence of Ni element that had a catalytic effect on the charge transfer reaction illustrates the discharge capacity versus cycle number of the ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes. We noticed that, although the ball-milled AB2 alloy electrode had an improved cycling stability (see ), it still could not meet the requirements of practical application due to its lower discharge capacity. For ball-milled AB2–AB5 and AB2–AB3.5 composite electrodes, it needs only two charge–discharge cycles (see ) to be fully activated. However, the discharge capacity rapidly decreased to 113.2 mAh/g and 123.8 mAh/g after 20 charge–discharge cycles for AB2–AB5 and AB2–AB3.5 composites, respectively. Further investigations on improving the cycling stability of the alloy electrodes are underway in our laboratory. illustrates the electrochemical impedance spectra of ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes at 50% DOD at 303 K. It was found that all the EIS spectra of these alloy electrodes consisted of a small semicircle in the high-frequency region and a large semicircle in the low-frequency region. Kuriyama et al. , the radius of large semicircle in the low-frequency region of ball-milled AB2 alloy electrode was greatly larger than that of AB2–AB5 and AB2–AB3.5 composite alloy electrodes, which indicated that the charge-transfer resistance decreased to a great extent after mechanical milling AB2 alloy with rare-earth alloys AB5 and AB3.5, respectively. Thus, it was believed that the rare-earth alloys AB5 and AB3.5 played an important role in improving the hydrogen reaction kinetics, especially in decreasing the charge-transfer resistance on the surface of alloy electrodes. Furthermore, the radius of large semicircle in the low-frequency region of AB2–AB3.5 composite alloy electrode was smaller than that of AB2–AB5 (see the inset in ), which indicated that mechanical milling with AB3.5 alloy was better than mechanical milling with AB5 alloy in decreasing the charge-transfer resistance. illustrates cyclic voltammograms of the ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes at the 50% DOD at 303 K. It was shown that for each alloy sample, the oxidation peak appeared at the potential of around −500 mV (versus Hg/HgO). However, in the investigated potential range the reduction peak did not appear, which might be due to the rapid and hardly detective hydrogen absorption during charge process as reported in Ref. , the height and area of oxidation peak increased to a certain extent after mechanical milling AB2 alloy with rare-earth alloy AB5 and AB3.5, which indicated that the electrochemical kinetics and the discharge capacity had been improved. Moreover, it was also found that the effect of AB3.5 alloy on improving the discharge capacity and the reaction kinetics was better than that of AB5 alloy. This result was in good agreement with that obtained from the maximum discharge capacity and EIS measurements. shows the anodic polarization curves of the ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes at the 50% DOD and 303 K. It can be seen that, for all the anodic polarization curves, the anodic current density increased first with increasing overpotential and finally reached a maximum defined as the limiting current density IL. The limiting current density IL indicated that an oxidation reaction took place on the surface of the negative electrode and the general oxidation product prevented further penetration of the hydrogen atoms . The limiting current density IL was 889.6 mA/g, 1423.0 mA/g and 1412.2 mA/g for AB2, AB2–AB5 and AB2–AB3.5 alloy electrode, respectively. The variation of the IL values indicated that the hydrogen diffusivity in the alloy electrodes increased after mechanical milling with rare-earth alloys. Moreover, the effect of AB5 alloy on improving the behavior of hydrogen diffusion was better than that of AB3.5 alloy. shows the semilogarithmic curves of the anodic current versus time responses of the ball-milled AB2, AB2–AB5 and AB2–AB3.5 alloy electrodes. According to the model proposed by Nishina at al. where i (A/g) is the diffusion current density, D (cm2/s) the hydrogen diffusion coefficient, C0 (mol/cm3) the initial hydrogen concentration in the bulk of the alloy, Cs (mol/cm3) the hydrogen concentration on the surface of the alloy particles, a (cm) the alloy particle radius, d (g/cm3) the density of the hydrogen storage alloy and t (s) is the discharge time. From the particle size measurement as mentioned in Section shows that the average particle radius of Ti0.9Zr0.2Mn1.5Cr0.3V0.3 alloy is 266 nm after mechanical milling treatment for 2 h. According to our careful experiments, the particle size of the alloy after the addition of La–Mg-based alloy is almost the same as that of the sample without the La–Mg-based additive (see ). The hydrogen diffusion coefficient D values calculated by Eq. . It was found that the hydrogen diffusion coefficient D increased about one order of magnitude after mechanical milling with La-based alloys in which the maximum of the hydrogen diffusion coefficient D reached 3.19 × 10−14 |
cm2/s for the ball-milled AB2–AB5 alloy electrode. This result was consistent with that of anodic polarization curves in . It can be suggested that the additional La-based alloy particles with better electrochemical hydrogen storage capacity provide hydrogen diffusion pathways in the present composite electrode as mentioned above, which is helpful to improve both the hydrogen reaction kinetics and the discharge capacity.In order to improve the electrochemical characteristic of AB2 Laves phase alloy, AB2–10 wt.% AB5 and AB2–10 wt.% AB3.5 composite samples were prepared by mechanical milling Ti0.9Zr0.2Mn1.5Cr0.3V0.3 with small amount of rare-earth-based AB5-type LaNi3.8Mn0.3Al0.4Co0.5 alloy and AB3.5-type La0.7Mg0.25Zr0.05Ni2.975Co0.525 alloy, respectively. X-ray diffraction analysis showed that the addition of rare-earth-based alloys did not change the basic hexagonal C14 Laves phase structure of the original AB2 alloy. SEM images and EDS patterns indicated that rare-earth alloy particles adhered to the bulk of AB2 alloy. Compared with ball-milled AB2 Laves phase alloy, the dramatically improved electrochemical discharge capacity was obtained after mechanical milling AB2 with rare-earth-based alloys, which amounted to 310.4 mAh/g and 314.0 mAh/g for AB2–10 wt.% AB5 and AB2–10 wt.% AB3.5, respectively. However, the cycle life of these composite electrodes needed to be further investigations. The results of EIS and CV indicated that the mechanical milling with AB3.5-type La0.7Mg0.25Zr0.05Ni2.975Co0.525 alloy helped to decrease the charge-transfer resistance at the electrode surface. However, mechanical milling with AB5-type LaNi3.8Mn0.3Al0.4Co0.5 alloy was beneficial to improving the hydrogen diffusion behavior in the bulk of the composites.Microstructure of 24-1928 Ma concordant monazites; implications for geochronology and nuclear waste depositsThe microstructure of monazite was studied using scanning electron microscopy (SEM), electron microprobe analysis (EMP), X-ray diffraction patterns (XRD), and transmission electron microscopy (TEM). Four well-characterized monazites were investigated, having very different concordant U-Pb ages (24 to 1928 Ma), and up to ∼15 wt.% ThO2, and ∼0.94 wt. % UO2. The SEM and EMP analyses of polished single crystal fragments reveal the absence of significant chemical zoning. XRD and TEM investigations show that the monazites are not metamict, despite their old ages, very high abundances of radionuclides, and hence, high time-integrated radiation doses. Except for the youngest one, the monazite crystals are composed of a mosaic of crystalline but slightly distorted domains. This structure is responsible for the presence of (1) mottled diffraction contrasts on the TEM, and (2) a second structural phase (B), with very broad reflections in the XRD patterns. Older monazites receive higher self-irradiation doses, and hence, they contain higher amounts of this B-phase. For the 1928 Ma monazite, XRD reveals only the broad reflections of phase B, implying that the whole monazite was affected by radiation damage that resulted in total distortion of the lattice. It is concluded that radiation damage in the form of amorphous domains does not accumulate in monazite because self-annealing heals the defects as they are produced by α-decay damage. The only memory of irradiation-induced defects is the presence of distorted domains. As the diffusion rate of Pb in an undisturbed monazite lattice is extremely low, Pb loss due to volume diffusion out of the monazite lattice is virtually impossible. This is considered as one reason why almost all monazites have concordant U-Th-Pb ages. Moreover, as long-term self-irradiation effects are limited in monazite, we consider this phase as a good candidate for the storage of high-level nuclear waste under the aspect of its high resistance to irradiation.The accessory mineral monazite is a natural light rare earth orthophosphate (APO4: A = LREE, Th, U, Ca, Pb) that contains high concentrations of U and Th (up to 6 wt.% UO2 and 20 wt.% ThO2; e.g., ). Monazites are widely used in U-Pb geochronology because of their high actinide and low common Pb contents (). In contrast to zircon, monazite mostly gives concordant U-Pb ages (e.g., ), indicating that the U-Pb systems of the monazites are either completely reset or remain totally unaffected during most geological events. Nevertheless, some discordant U-Pb ages have also been reported. Most of these may be explained by the analysis mixtures of newly grown rims and inherited cores (e.g., ) and diffusive Pb loss is assumed only in rare cases (). To understand the geological significance of monazite ages measured, one needs to understand the mechanism of resetting the U-Pb system in this phase.Two processes are commonly considered to explain resetting of the U-Pb isotopic system: (1) Pb loss by volume diffusion out of the grain or (2) dissolution of the crystal by a fluid, followed by precipitation of a newly formed Pb-free crystal (). With respect to process (1), there have been several attempts to interpret discordant ages of monazites in terms of closure temperature model. In this model, resetting results from the diffusion of daughter isotopes out of the crystal. The assumed closure temperature depends on the size of the crystal, its shape, the cooling rate, and the diffusion coefficient of the daughter elements. Experiments on Pb diffusion by , however, showed that this diffusion is very slow in monazite, even at high temperatures. A different approach is to determine an empirical “geological closure temperature” using isotopic ages of different minerals combined with petrologically derived temperatures for metamorphic events. This method yielded temperature estimates as low as 530 ± 25°C (). It has been shown, however, that monazite inclusions that were shielded by host minerals such as quartz or garnet still retain “old” ages, despite annealing under granulite-facies conditions at temperatures above 800°C over longtime scales (). Such observations raise the question of whether diffusion is the main resetting mechanism. An alternative mechanism is dissolution/precipitation. This is the only resetting mechanism for the U-Pb system in monazites detected in the experiments of . The extent of the dissolution/precipitation process depends on the fluid composition () and is a more efficient mechanism for resetting than lead loss by volume diffusion.In minerals relevant for U-Pb dating, radioactive decay produces radiation damage that may partially or totally destroy the crystal lattice, thus producing so-called metamict domains. Whatever the mechanism of resetting is, the kinetics of resetting are strongly influenced by the degree of metamictization of the crystal. This structural state is essential, as physical properties of an amorphous phase differ significantly from those of its crystalline counterpart (e.g., ). It was suggested that Pb diffusion is enhanced in a metamict crystal () and radiogenic Pb diffuses much faster within “channels” that correspond to the percolating interface between amorphous and crystalline domains (e.g., ). Moreover, radiogenic Pb can be leached more easily from a damaged lattice (). Consequently, a damaged lattice will retain radiogenic Pb to a lesser degree than a perfect one, resulting in discordant ages for phases that exist in such a structural state. As pointed out by , microstructural investigations are, therefore, of fundamental importance to better understand isotopic ages of minerals.Characterising the structural state of U- and Th-rich phases is essential when considering their potential use in nuclear waste deposit strategies. Monazite-based ceramics have been first proposed by as a crystalline matrix for immobilizing radionuclides, especially the actinides. This concept is currently the focus of research on nuclear waste deposit strategies (e.g., ). The advantage of studying natural monazites in this context is that this mineral provides data on the response of the lattice to irradiation over geologic timescales at very low dose rates (<10−17 dpa/s; ). In contrast, studies using actinide-doped phases (10−10–10−8 dpa/s) or charged-particle irradiation (10−5–10−2 dpa/s; ) simulate very high dose rates, but only over short time periods.Monazite receives intense self-irradiation doses during its geologic history because of its generally high U and Th contents. During an α-decay event, a radionuclide liberates its energy by ejecting an α-particle while the remaining nucleus is recoiled in the opposite direction. Most of the atomic displacements that result in amorphization of a crystal lattice are caused by α-recoil nuclei (e.g., ). Up to now, monazite was rarely found in the metamict state despite the fact that it generally experienced intensive radiation doses (), and clear evidence of radiation damage was limited to isolated nm-sized domains within the crystal ( were the first to study a monazite in detail by using various complementary analytical methods, including X-ray diffraction (XRD), transmission electron microscopy (TEM), Raman, and cathodoluminescence. The studied monazite was a chemically homogeneous Brazilian monazite (called Moacir) that yielded a concordant U-Pb age of 474 Ma and a U-Th/He age of 479 Ma. This monazite also showed nm-scale defects induced by α-decay. The new result of this study, revealed by the XRD technique, is the presence of two distinct monazite “phases,” A and B, that have different lattice parameters. Because monazite A shows sharp reflections of high amplitude and larger lattice parameters, it was interpreted to be well-crystallized monazite that had an expanded lattice as a result of Helium accumulation. As monazite B exhibits very broad, low amplitude reflections, it was interpreted to be a Helium-free distorted monazite crystal lattice, which can be attributed to old α-recoil tracks. prompted new questions: Will this phenomenon be visible in other monazites? Is there a correlation between the B domains, the U, Th contents of the monazites, and hence, the self-irradiation dose received by monazite?To address these questions, we performed a comparative study of four well-dated monazites ( They were selected for the following reasons: (1) all have concordant U-Pb ages (). Hence, we can exclude recent alteration events that would complicate data interpretation; (2) their ages range from 24 to 1928 Ma; (3) the contents of radionuclides vary widely, reaching up to ∼15 wt.% ThO2 and up to ∼0.94 wt.% UO2. It follows from (2) and (3) that: (4) these monazites received quite different self-irradiation doses, ranging from 0.13 to 17 × 1019 α-decay/g. Using essentially XRD and TEM techniques, we investigated the micro-nanostructure of these monazites. The final aim of this study is to compare the effects of long-term self-irradiation in these monazites over a wide range of α-decay doses to specify implications for U-Pb geochronology and nuclear waste storage.The youngest monazite YS35 comes from a garnet-biotite leucogranite layer in the Diancang Shan Gneisses from the Red River Belt, China. The monazites YS35 have concordant U-Pb ages of 24.2 Ma (). A large series of U-Pb dating results from various accessory minerals (), some 40Ar/39Ar data from micas, and a modelled Ar degassing pattern of K-Fsp from the Ailao-Shan-Red River shear zone () show that this belt cooled rapidly after granitic-alkaline magmatism and high-strain ductile deformation, reaching the 300°C isograd at 20–19 Ma. This fast cooling was followed by a very inactive period of ∼15 m.y. during which the gneisses remained at ∼300°C. Final cooling occurred after ∼4 Ma during exhumation and erosion of the gneiss that was associated with dextral strike-slip shear along what is the presently active Red River fault. These monazites are ∼200 × 400 μm large and transparent yellow grains.The second monazite, Moacir, is a large (centimeter-sized) yellow-orange single crystal from a pegmatite in the Itambé Province, Brazil (). It has a concordant U-Pb age of 474 Ma () and a U-Th/He age of ∼ 479 Ma (R. Pik, private communication). Moacir is the reference monazite for this study as it has been previously characterized using SEM, EMP, XRD, TEM, Raman microprobe, and cathodoluminescence techniques (The third monazite, Madagascar, is a yellow, centimeter-sized, single crystal that has given concordant U-Pb ages of 545 Ma (). A U-Th/He analysis yielded an identical age (R. Pik, private communication) indicating that, like for Moacir, all helium produced by α-decay during the existence of the crystal is essentially still present in the monazite lattice. This monazite was separated from an apatitite from the Anosyan granites, Madagascar.The oldest monazite, DIG19, comes from exceptionally fresh, felsic high-grade metamorphic gneiss from Sverdrup Inlet, Devon Island, N.W.T., Canada. It gave concordant U-Pb ages of 1928 Ma (). A biotite-feldspar-whole rock Rb-Sr isochron of this gneiss corresponds to an age of 1814 ± 8 (). The high-grade crystalline basement is covered by Proterozoic metasediments, upon which undeformed and unmetamorphosed Cambrian to Devonian sediments rest unconformably. This simple geology indicates that the DIG19 monazites remained in the upper crust at low temperature for at least 540 Ma. The monazite crystals are yellow and typically 100 × 100 μm in size.All monazite grains were individually mounted in epoxy, polished, and carbon coated in preparation for imaging by back-scattered electron (BSE) and secondary electron (SE) techniques. Images were acquired using the JEOL JSM-840A SEM at the Institut für Planetologie, WWU Münster. Acceleration voltage was typically 20 kV.The EMP analyses were obtained using the CAMECA SX-50 electron microprobe at the GFZ-Potsdam, equipped with a wavelength dispersive spectrometer. The operating conditions were: accelerating potential 20 kV, beam current 40–60 nA, and 1–2 μm beam diameter. For details, see About 1 mg of powdered monazite was measured in transmission mode using the fully automated STOE STADI P diffractometer (Cu-Kα1 radiation) equipped with a primary monochromator and a 7°-position sensitive detector (PSD) at the GFZ-Potsdam. Details are given in . Two kinds of diffraction patterns were recorded. One pattern spanned from 5 to 125° (2θ) with a step size of 0.01°. The intensity was counted for 135 s at each position. The second pattern covered the 2θ range between 26 and 28° with the same step size and a counting time of 1000 s per detector step. The unit-cell refinements were performed using the Rietveld-refinement program of the GSAS software package (). The XRD patterns in the range of 26–28° (2θ) were fitted by a Gauss + Lorentz area function to determine the full width at half maximum (FWHM), areas, and the maximum position of the (200) reflection.Monazite YS35 was not studied by XRD due to the insufficient amount of sample material.The YS35, Moacir, and Madagascar monazites were cut in random orientation, and TEM foils were prepared by hand polishing and argon ion milling at 5 kV (Institut für Planetologie, WWU Münster). Due to its very small size, the DIG19 monazite was prepared using the Focused Ion Beam technique FEI FIB200 at the GFZ-Potsdam. For both thinning techniques, particular care was taken to avoid artificial defect production by the ion beam during milling. For the conventional Ar-milling technique, demonstrated by preparing both untreated and annealed (at 1000°C) samples, that the observed defects are not artefacts from the milling process. demonstrated early on that TEM analysis of crushed monazite fragments and ion-thinned foils give the same results. For the FIB technique, results are the same. The gallium beam used in the FEI FIB200 is responsible for implantation of gallium ions during milling, causing gallium peaks in Energy Dispersive Spectroscopy (EDS) analyses, and creates a ∼20 nm amorphous layer near the surface. However, the latter phenomenon does not adversely affect the observations.The TEM studies were carried out using the Philips CM 200 TEM at the GFZ-Potsdam operated at 200 kV and the JEOL 3010 TEM at the Institut für Planetologie, WWU Münster operated at 300 kV. Both instruments are equipped with a LaB6 electron source. High-resolution TEM (HRTEM) images were acquired as energy filtered images applying a 10 eV window to the zero-loss peak using a Gatan GIF™ system. Bright Field images (BF) of all samples were done close to a zone axis, i.e., where a maximum contrast is present, allowing direct comparison of the results.. Data for the reference monazite Moacir are from The BSE micrographs show that the Moacir and Madagascar monazites are chemically homogeneous ( whereas, monazites YS35 and DIG19 display a vague zoning (The EMP analytical results are compiled in The four monazites have different chemical compositions. Importantly, they show large variations in their actinide and lead contents: The ThO2 content ranges from ∼6 to 15 wt.%, the UO2 content from ∼0.13 to ∼0.94 wt. %, and the PbO content amounts up to ∼0.95 wt. % (). The Moacir and the Madagascar monazites have approximately the same age, but very different U and Th contents; hence, they have received different irradiation doses. The chemical homogeneity of the Moacir and Madagascar monazites was investigated by ; therefore, only representative analyses of these monazites is given in Using actinide contents and U-Pb ages, we calculated the theoretical self-irradiation doses of the monazites (e.g., , correspond to the number of α-decay events that have occurred during the geologic history of the respective monazite. The self-irradiation dose increases from 0.13 × 1019 α-decay/g for the youngest monazite YS35, over 2.77 × 1019 α-decay/g for the Moacir monazite, 6.12 × 1019 α-decay/g for the Madagascar monazite, to 17 × 1019 α-decay/g for DIG19, the oldest monazite. For comparison, estimated doses for some well-studied zircon samples amount to only 1 × 1019 α-decay/g (e.g., . Ceramics, doped with 1 wt.% 239Pu, will receive a dose of ∼10 × 1019 α-decay/g in 100 000 yrs (The X-ray diffraction patterns of the monazites Moacir, Madagascar, and DIG 19 are characterized by relatively sharp reflections of relatively high intensities. As shown in a). The number of these shoulders (arrows in a) decreases with higher self-irradiation doses due to the increased broadening of the peaks. In their study of the Moacir monazite, pointed out that this broadening is most pronounced along (200) reflection, i.e., at 2θ diffraction angles of 26 to 28°. Therefore, we obtained high-resolution X-ray diffraction patterns in this range for the two other monazites as well. (A typical diffraction pattern in the 2θ range of 26–28° is given for the Moacir monazite in b. It shows one characteristic reflection (A) of high amplitude that is very sharp (FWHM = 0.091°) and, additionally, a broad (FWHM = 0.425°) shoulder (B) of low amplitude ( The B-reflection maximum is located at the 2θ value of 27.04° (3.295Å) whereas the maximum of the A-reflection is located at 26.88° (3.314 Å; ). The area percentage of A and B reflections are 47 and 53%, respectively (). Rietveld-refinement of the XRD data for the unheated monazite, assuming only a single-phase (A) with a monazite structure, failed. Introducing a second phase (B, monazite structure assumed), however, resulted in a successful Rietveld refinement. Consequently the XRD patterns are interpreted to consist of two phases, both having monazite structure but with different lattice parameters (). Monazite (A) shows larger lattice parameters (1% in volume) than the synthetic reference crystal CePO4 from . The lattice parameters of monazite (B), however, correspond well to those of this reference crystal.Like the Moacir monazite, the X-ray pattern of the Madagascar monazite consists of the two structurally distinct monazite phases A and B with slightly different lattice parameters (). Phase A shows sharp reflections with FWHM(200) of 0.088°. Phase B, in contrast, exhibits very broad reflections with FWHM(200) of 0.336° and an amplitude lower than that of phase A (). In the case of the Madagascar monazite, the integrated area under the B-reflection, ∼80% for the (200) peak, is larger than for the Moacir monazite, where we observe only ∼53% for (200). Thus, the contribution of the B-reflections is larger in the Madagascar compared to the Moacir monazite. In addition, the A-reflection maximum for the Madagascar monazite is shifted towards lower 2θ values compared to the Moacir monazite.In contrast to the other monazites, monazite DIG19 shows only one reflection. This reflection is very broad (FWHM(200) = 0.276°) and has its maximum at a 2θ value close to the maximum reflection of phase B in the Moacir and Madagascar monazites.All four monazites show mottled diffraction contrasts in their BF images, but with significant differences among them ( The young YS35 monazite displays just a few, spot-like contrasts (a), indicating that monazite YS35 is an almost perfect crystal. The Moacir monazite has many mottled diffraction contrasts. The areas showing these contrasts are larger (∼5 nm-sized domains) and appear to be more frequent compared to those of the YS35 monazite (b). This effect is even more pronounced in the Madagascar monazite (c) where the domains are again larger (∼10 nm across) than in the Moacir monazite. Finally, monazite DIG19 (d) shows very large domains (∼80 nm across) of mottled diffraction contrasts. These patterns are very similar to the TEM dark field images published by . All these mottled diffraction contrasts are the result of a mosaic structure of the crystal, i.e., a lattice comprising many smaller domains that have slightly different orientations. The dark areas represent domains oriented close to a zone axis, and the bright ones are regions that are not diffracting.The SAD patterns of all studied monazites ( inserts a1, b1, c1, and d1) show sharp reflections. Evidence for amorphous rings or distorted reflections is lacking. The HRTEM image of monazite YS35 shows neither defects, nor distorted domains. The lattice is almost perfect (a). The Moacir and Madagascar monazites revealed some isolated ca. 5 nm2-sized areas where the lattice fringes were blurred (b and c). The distorted areas impose a local strain on the lattice, causing an inhomogeneous contrast distribution. As illustrated in the Inverse Fourier Transform images (IFFT) of b2 and c2, these domains correspond to regions where the lattice is distorted. Finally, the HRTEM image of monazite DIG19 (d) displays contrasts comparable to those of the Moacir and Madagascar monazites, but the IFFT image shows edge dislocations (d2; see arrow). In this monazite, the domains seem to be more distorted than in the other monazites. have demonstrated that the Moacir monazite is not a true “single crystal,” as shown by the simple SAD pattern but consists of a mosaic of tiny (∼5 nm2), slightly distorted, domains that correspond to old defects induced by radioactive decay. These differences in orientation are responsible for the presence of (1) the mottled diffraction contrasts in the BF images, and (2) a second structural phase (B), which produces very broad reflections in the XRD patterns. Broadening of the B reflections is due to different degrees of distortion. In addition, the XRD pattern revealed that the rest of the monazite is a perfect “Phase A” monazite crystal but with expanded lattice caused by the accumulation of Helium.Based on these observations on the Moacir monazite, suggested that HRTEM images are not appropriate to highlight variable defect concentrations in monazites because the 25 nm2 HRTEM images () might not provide a representative sampling of the small number of distorted areas. In contrast, the XRD technique () and TEM-BF images cover μm2-sized areas (), allowing a clear distinction among various defect concentrations due to the larger amount of sample studied (i.e., a few mg for XRD). For setting up a classification scheme of self-irradiation-induced defect concentrations in monazites, we therefore recommend using XRD patterns and BF images. The HRTEM images, however, still provide two important pieces of information, namely, that the monazites are not amorphous (a1 to d1),, and that the defects are only nm-sized, localized, distorted areas (Both BF and HRTEM images revealed the almost perfectly crystalline lattice of monazite YS35. We infer that the corresponding XRD pattern should have very sharp reflections with a high intensity without shoulders, similar to a synthetic monazite (). As previously noted, the A-reflection maximum for the Madagascar monazite is shifted toward lower 2θ values as compared to that of the Moacir monazite. This can have either (1) a chemical or (2) a structural reason. In the first case, the different LREE, Th, U contents in the two monazites () result in different lattice parameters, as demonstrated in many previous studies (e.g., for the Moacir monazite, is that the lattice expansion is due to the presence of helium. As the Madagascar monazite has much higher Th and U contents than the Moacir monazite at approximately the same age, it contains more helium, probably causing the significant lattice expansion. Moreover, the Madagascar monazite was more affected by radiation damage, causing lattice distortion, which, in turn, results in a higher volume percentage of phase B (80 vol.%) as compared to the Moacir monazite (53 vol.%). On BF images, this is visualized by the increasing size of the mottled diffraction contrasts (b and c). In the case of monazite that received the highest dose, DIG19, XRD yields only the broad reflection B (). This implies that the complete monazite was affected by radiation damage, resulting in the total distortion of the lattice. If undistorted, perfectly crystalline domains A are still present in monazite DIG19, they were not resolved by the XRD technique. The BF image of monazite DIG 19 (d) confirms this result: the mottled diffraction contrasts are very large and seem to affect the whole lattice.Helium concentration data are only available for the Moacir and Madagascar monazites (; R. Pik, private communication): In both samples, the U-Th-He and U-Pb ages are concordant. This fact indicates that the rocks containing both these different monazites were never heated above the closure temperature for He in monazite, i.e., ∼220°C (). Surprisingly, although both monazites must have stayed at such low temperatures for a long time, they contain only quite low defect concentrations induced by self-irradiation. Amorphous domains are absent, only some misoriented regions interpreted as old α-recoil tracks, have been found. Two explanations are possible: (a) either the lattice can heal at such low temperatures, in accordance with experimental data (), or (b) ionization created by α-particles, or electrons from α-decay in the U and Th decay chains partially anneal structural damages created by the recoil of nuclei (). Future experimental studies should focus on the healing mechanisms of monazite.To conclude, this study demonstrates that despite the high self-irradiation doses of up to 17 × 1019 α-decay/g, none of the monazite samples is metamict. Lattice distortion only occurred in nm-sized isolated domains. Even the high defect concentration featured in the 1.9 Ga monazite DIG19, in which the lattice is totally distorted, was insufficient to open and disturb the U-Th-Pb decay systems in these crystals. Although this defect concentration seems to be very high for a monazite, much higher defect concentrations occur in zircon, in which amorphous domains are very common. Finally, to answer our questions, it was demonstrated that the phenomenon described by , i.e., the presence of two different structural phases in monazite, of which one is distorted, is also observed in other monazites. The correlation between these “B domains” and the self-irradiation dose received by a monazite is a major outcome of this study.The present study confirms that radiation damages, in the form of amorphous domains, do not accumulate in monazite because defects heal faster by self-annealing than the lattice is damaged. The only memory of self-irradiation induced defects in the monazites is the presence of the distorted domains. These domains are visible using TEM and correspond to the “B-phase” observed in XRD. Taking this observation together with results from diffusion experiments (e.g., ), and dating results showing that the U-Pb clock is resistant to temperatures exceeding 800–850°C (e.g., ), we conclude that radiogenic Pb does not diffuse out of monazite grains at geologically measurable rates. Moreover, have shown that Pb fits very well in the monazite lattice, occupying probably the same structural site as the parent elements U and Th. This argument helps to understand why lead is obviously fixed quite well in the monazite lattice. Therefore, it is quite improbable that the diffusivity of Pb is enhanced due to the presence of radiation damage (see also ). The combination of all these arguments explains why isotopic dating of monazite mostly yields concordant U-Pb ages. Furthermore, it is highly unlikely that monazite U-Pb ages represent “cooling ages.”The results of this study have certain implications for the storage of nuclear waste. Natural minerals such as monazite are a great tool to test the long-term behaviour of waste-form phases in specific geological environments (e.g., ). Monazite has already been proposed for use in single-phase ceramics to immobilize actinides (): natural monazite contains large amounts of LREE3+, U4+, and Th4+ (), for which the actinides Am3+, Cm3+, Np4+, Pu4+ are good substitutes. Another reason concerns a property of monazite featured in this study, i.e., their mostly concordant U-Pb ages despite high self-irradiation doses over a long period of time. In the case of monazite DIG 19, for example, the U-Pb system remained closed for nearly 2 billion yrs. Therefore, monazite ceramics seem to be good candidates for the immobilization of high-level nuclear waste, as it may be able to retain actinides within its lattice for very long periods of time. It is, however, essential to point out the limitations of the current study. Precise data on the temperature and duration of the annealing of the investigated monazites are not available. Therefore, the timing of the development of the observed domain structures remains unconstrained. Moreover, our observations provide no insight into the annealing mechanisms (temperature, ionization?) of these defect structures. Future experiments should focus on determining possible annealing mechanisms in both doped and natural monazites. Other aspects that are of interest for nuclear waste deposit strategies are solubility, dissolution, doping, and charge balancing of the ceramics. Further studies under a variety of physical conditions are necessary to better understand the behaviour of such monazite-ceramics in aqueous solutions (). It should be the final aim of all these studies to construct models that allow the prediction of the structural state of monazite-type phases at a given time. A study on the microstructure and tribological behavior of cold-sprayed metal matrix composites reinforced by particulate quasicrystalIn the present study, mechanical blends of AlCuFeB quasicrystal and tin bronze powders were deposited by cold spray process to obtain metal matrix composites (MMCs) reinforced with quasicrystalline particulates. The influences of the incorporation of quasicrystal particles on the particle deposition behavior, microstructure and microhardness of the composite coatings were investigated. In order to evaluate the influence of reinforcing quasicrystal phase on the tribological behavior of the coatings, ball-on-disc sliding tribological tests were conducted in an ambient condition. The results showed that the incorporation of quasicrystal particles reduced the porosity and increased the microhardness of the composite coatings. At the same time, a reduction of the friction coefficient and an increase of wear rate were found. Wear mechanisms were discussed and correlated to the microstructure and microhardness of coatings.Quasicrystal materials have unique atomic structures and consequently well-determined physical, chemical and mechanical properties, such as low surface energy, high hardness, low coefficient of friction (COF) and good wear and corrosion resistance Cold spray is a low temperature deposition process in which particles are accelerated through a De-Laval type nozzle, and the coating in cold spray process is obtained solely as a result of the accumulation of plastic deformation of solid particles impinging upon a substrate. As a competitive candidate for preparing high quality coatings, cold spray technique has attracted much interest in fabrication of metallic coatings since its emergence. Cold spray can also be used to prepare MMC coatings, which usually exhibit improved physical or mechanical properties. This provides a new means for the preparation of desired MMCs, and the properties of cold-sprayed MMCs may be conveniently tailored by a proper addition of reinforcing particles. In this study, mechanically blended mixtures of AlCuFeB quasicrystal and bronze powders were deposited by cold spray process, aiming at tailoring quasicrystal-reinforced MMC coatings. Powder mixtures with different AlCuFeB contents were used to prepare composite coatings for evaluating the influence of quasicrystal fraction on the characteristics of deposited coatings. The structure–property relationships of the composite coatings were investigated. As one of the major objectives of this study, the tribological behavior of the quasicrystal particles reinforced coatings was investigated and correlated to their microstructures and mechanical properties.Inert gas atomized AlCuFeB quasicrystal (referred as QC hereinafter) and Cu-8wt.%Sn (CuSn8) powders were used as the feedstocks in this work. The as-atomized QC powder is in a nominal atomic composition of Al59.1Cu25.6Fe12.1B3.2. Both QC and CuSn8 particles have a spherical morphology. Particle size distribution were measured by Mastersizer 2000 (Malvern, UK). The size of the both particles follows a typical Gaussian distribution feature. For the CuSn8 powder, the d (10), d (50) and d (90) were 8.4 μm, 17.1 μm and 33.8 μm, respectively. As to the QC powder, the d (10), d (50) and d (90) were 7.9 μm, 17.1 μm and 32.2 μm, respectively. It is clear that the sizes of CuSn8 and QC powders were in a similar distribution range. Both of the powders were in-house made and more technical details regarding manufacture of these powders are available in our previous publication A commercially available Kinetic 3000 cold spray system (CGT GmbH, Germany) coupled with a self-designed nozzle was used for coating deposition. Compressed air and high pressure argon were used as the main gas and the powder carrier gas, respectively. Pressure and temperature of the main gas were maintained at about 27 bars and 540 °C, respectively. The pressure of the carrier argon gas was 30 bars. The standoff distance between the nozzle exit and the sample surface was kept constant at 30 mm for all cold spray processes. The moving speed of spray gun manipulated by an ABB robot was maintained at 50 mm/s. Four passes of the spray gun were made for each sample, and the thickness of obtained coatings was about 350 μm. In this work, mild steel disks with a dimension of Ø43 × 4 mm were chosen as substrates, and the disks were degreased and grit blasted following usual procedures prior to coating deposition.The cross-sectional microstructure of the prepared coatings was examined by optical microscope (Nikon, ECLIPSE ME600) and scanning electron microscope (JEOL, JSM-5800LV, Japan). The porosity level of coatings was evaluated based on polished cross-sections by using an image processing method. Ten optical images were used to calculate the average porosity. It should be noted here that only the deeper zone was taken into account for analyzing the porosity level because the top layer was relatively porous than the deeper zone in the coatings. Energy disperse x-ray analysis have been conducted for at least three times over different areas of each coating to determine the fraction of QC within the coatings. Microhardness of coatings was measured by a Vickers hardness tester (Leica VMHT30A, Germany) at 2.94 N load with 15 s dwelling time, and the corresponding sizes of the indentation marks were larger than a single splat during the hardness measurement. The microhardness given in this study is an average of at least 10 measurements randomly taken from the cross-sections of coatings.Tribological tests were conducted in a dry sliding condition through a re-constructed ball-on-disc tribometer (CSM, Switzerland) to evaluate the tribological performances of the prepared coatings. All the coatings were ground using diamond sandpapers (following 300 mesh, 600 mesh, 800 mesh, 1200 mesh and finally 4000 mesh), then they were polished using diamond slurries down to an average surface roughness of about 0.04 μm. All the coatings were treated in the same procedure to assure that they could have a same initial surface topography. WC-Co balls with a 6-mm-diameter and a mirror finished surface were used as the counterpart. The polished coatings slide against the WC-Co ball in a linear oscillation mode at a mean sliding velocity of 70 mm/s. The length of each stroke was 11.6 mm, and the frequency of the oscillation movement was kept at 7 Hz. The normal load was in the range of 2 N and the total sliding distance was 280 mm. The positive peaks of the COF curves were extracted and averaged as the mean COF for each tribotest. Morphologies of the wear traces were examined by optical microscope and scanning electron microscope. In this work, the width of at least four locations along the wear trace were measured by microscope and averaged to compare coatings' wear rates. The wear rate in this work is defined as the width of wear trace divided by the normal load (N) and total sliding distance (m).(a) and (b). It seems that both the pure CuSn8 and the composite coatings present compact structures with few defects on the interface of splats. Compared to pure copper, the deformation ability of bronze particles is usually insufficient for the closure of the interfacial gaps between deposited particles due to a relatively lower deformability. For the composite coatings, QC particles are dispersed uniformly in the bronze matrix. It seems that most of the embedded QC particles have a dimension less than 10 μm, and only the small QC particles (approximately below 5 μm) have maintained their original spherical morphology, as shown in (c). Statistically, the average size of QC particles deposited into the coating is obviously smaller than that in the starting powder. Presumably, part of the larger QC particles in the starting powder has rebounded off during the impacting process and consequently were not deposited into the CuSn8 matrix. Fragmentation of QC particles during the deposition process may also be responsible for this decrease, i.e., part of the larger QC particles have fragmented into smaller ones. SEM observation shown in (d) clearly indicates that some of deposited QC particles have broken into smaller pieces during the deposition process. The small fragments presented in (b), as marked by circles, are considered to be residuals of the crushed QC particles, while the rest of the broken particles could have been rebounded off during the impacting process. The fragmentation of brittle AlN particles has been also observed in the cold-sprayed Al/AlN composite shows the porosity level of the prepared coatings. Overall, the porosities of composite coatings are much lower than that of the pure CuSn8 coating, and a slight decrease of porosity level with the increase of the QC fraction in starting powder is observed for the three composite coatings. This indicates an important influence of the incorporation of QC particles in the starting powder on the densification of coating structures. Similar results can be found in the literature, e.g., Irissou et al. (b) shows that the volume fractions of QC phase are 7.5%, 11.5% and 20.5% for the QC19, QC36 and QC57 coatings respectively. These values are lower than the starting mixed volume fractions. This behavior has been also observed in previous work, when the starting volume fraction (75 wt.%) of Al2O3 particles was reduced to 26 wt.% after cold-spraying of MMCs The microhardness of the coatings is illustrated in shows the COFs and the wear rates of the CuSn8 and the composite coatings tested in a dry sliding condition. The COFs of the coatings are in the same range, i.e., from 0.67 to 0.72, and only a slight decrease in the COF with the increasing QC fraction is observed. The slight reduction in COF can be attributed to the decreased adhesion between the sliding pairs with enhancing QC particle fraction. The wear rates of CuSn8 and QC19 coatings are similar, but they are lower than that of QC36 and QC57 coatings. A previous study (b) indicates that different wear behaviors have taken place for the coatings. shows that a lot of grooves and craters were generated in the wear trace of the CuSn8 coating, as indicated by circles, and the wear of the CuSn8 coating is mainly characterized by both micro-ploughing and particle delamination. The fatigue wear of the CuSn8 coating is mainly attributed to the delamination and rupture of coating layer as a result of the initiation and propagation of micro cracks. For the QC19 coating, it seems that the micro-ploughing is no longer the most important factor contributing to the wear. However, more and even tinier craters were generated in the worn surface, as indicated by single arrows in (b). The formation of these tiny craters was assumed to be resulted from the delamination of the surface layer. Due to the reinforcing effect of QC particles, a lower in-depth deformation and thus a thinner strain layer involved in the frictional process were expected for the QC19 than those of the CuSn8 coating. As a result, the micro-ploughing of CuSn8 coating can be alleviated due to the load-bearing effect of QC particles, as demonstrated by the presence of many scratches in (b). Moreover, the initiation and propagation of micro-cracks can only occur within a relatively thinner sub-surface layer. This is assumed to be the main reason for the generation of the tinier craters. The QC36 and QC57 coatings have a higher wear rate than the QC19 coating. It is observed that numerous deep worn areas were generated in the wear trace and most of them were filled with small wear particles, as indicated by the single arrows in (c). For particulate reinforced MMCs, the interfaces between reinforcing particles and the matrix are weakly bonded regions Composite coatings present much denser microstructures than the pure CuSn8 coating, and the porosity level of composite coatings decreases with increasing the content of incorporated QC phase. It is believed that the hammering effect on the coating exerted by the QC particles enhances the coating porosity.The ratio of the deposited QC particles to its fraction in the initial powders is about 30%–40%. The low ratio is ascribed to the rebounding of QC particles upon impacting and thereby the low deposition efficiency.The coating hardness increased with the increase of the incorporation of QC phase. The reinforcing effect of QC particles is evident.The COF decreases slightly with the increase of the QC content in coatings. Incorporation of low-loading QC particles improves the wear resistance of CuSn8 coating. The improved wear resistance can be mainly attributed to the reinforcing effect of the QC particles, which reduce ploughing on the coating and also reduce the depth of frictional layer. However, a higher concentration of QC phase adversely aggravated the wear resistance of QC36 and QC57 coatings possibly due to the fact that debonded QC particles can cause severe abrasion.Fine-grain-embedded dislocation-cell structures for high strength and ductility in additively manufactured steelsMetals manufactured with either a fine-grained structure or a nano-grained structure possess the ultrahigh strength at the expense of ductility. Therefore, materials that combine high strength and ductility should be immediately manufactured for a wide variety of practical applications. In this study, we report a bulk stainless steel with fine grains embedded by dislocation cells using additive manufacturing (AM) methods, which can efficiently and economically manufacture complex structures to achieve the desired mechanical properties. Compared with conventionally cast and forged 316L steels, the average yield strength and tensile elongation of AM 316L steels are 170% and 45% higher, respectively, attributed to a synergistic deformation of the fine-grain-embedded dislocation-cell structures. High strength is expected from a block of dislocation cells and solute atoms to the planar slip of dislocations, whereas high ductility depends upon the slip of fine-grain boundaries associated with the altered shape of an embedded dislocation cell. Moreover, as deformation proceeds, the complex interlaced dislocation cell network structure and deformation twinning can further enhance work hardening as well as tensile elongation. This study proposes a fine-grain-embedded dislocation-cell structure to achieve high-strength-and-ductility materials.Material scientists and engineers have long been devoted their efforts to finding and creating new structures to eliminate the trade-off between strength and ductility []. For example, an austenitic 316L stainless steel (SS) is one of the most widely-used steels for structural applications, owing to its valuable combination of high corrosion resistance, oxidation resistance, and welding performance. However, a fine-grained 316L SS suffers from a major drawback, namely its limited yield strength (250–300 MPa) []. For several years, the application of classical strengthening mechanisms, such as, dislocation strengthening, solute strengthening, grain boundary (GB) strengthening, and precipitation strengthening has enabled the development of a series of engineered steels with the improved strength []. Unfortunately, improving the strength of 316L SS often reduces its ductility. Thus, to eliminate or alleviate the trade-off between strength and ductility, the efficient and economical methods should be employed to meet large-scale industrial needs.In this study, we consider the traditional processing methods that result in a single outstanding performance (high strength or high ductility). We determine whether we can effectively combine these multiple mechanisms to achieve the metallic materials with high strength and ductility. It is well known that the fine-grained material and nanograined material have high ductility and high strength, respectively []. The coarse/fine grains are refined by the formation of dislocation cells during severe deformation, resulting in nanograins []. The size of the dislocation substructure, such as Taylor lattice, cell block, and dislocation cell, is used to evaluate the strain hardening in an Fe30.5Mn2.1Al1.2C steel during tensile deformation []. The increased carbon content reduces the stacking fault energy in high-entropy alloys, and leads to the non-cell-forming structure []. However, with respect to regarding steels, regulations of various mechanisms in terms of how dislocation cells control the strength and ductility, as well as how to prepare the dislocation cell to obtain the excellent and comprehensive performance, are not fully understood.Therefore, the coordination of dislocation cells plays a critical role in guiding the mechanical properties of metals and alloys during the large-deformation process. Here, based on the “multiscale nested-grain group” view, the multiscale fine-grain-embedded dislocation-cell structure is proposed to realize the high strength and ductility materials. Most importantly, this structure can be prepared using three-dimensional (3D) printing technology []. In the current study, additive manufacturing (AM) is used to manufacture SS 316L. The purpose of this study is to clarify the dislocation cell evolution responsible for the deformation mechanism in AM 316L. Specifically, the goal of this investigation is to understand the critical role of dislocation cell in the development of high-strength-and-ductility AM 316L.An FS271 M machine (Farsoon, Inc, China) was used to perform the selective laser-melting (SLM) processes of 316L SS under the high purity nitrogen gas for avoiding the oxygen contamination. The spot diameter of the laser beam was 80 μm. The 316L SS powders were prepared by the Ar gas atomization, and their average particle sizes measured using a laser particle size analyzer were 32 μm. The 304 steel plate was used as the building substrate. The SLM parameters were listed in . During the AM process, the laser was rotated to an angle of 67° for each layer to ensure material compactness, and nitrogen atmosphere was used to prevent the oxidation. Specimens with a size of 100 × 40 × 40 mm3 were prepared for microstructural observation and mechanical tests. After AM process, the samples were annealed at 400 °C for 3 h to release the residual stresses. The chemical composition of the as-built 316L SS is listed in The solidification structure of AM 316L depends on the effects of solute and thermal gradients, based on the relationships between the ratio of the temperature gradient (Tg) to the growth rate (R), and the ratio of the solidification undercooling (ΔT) to the diffusion coefficient (DL) []. The plane-equiaxed solidification is formed when TgR<ΔTDL, and the cellular and columnar solidification occurs when TgR>ΔTDL. In particular, this method can provide the effective and economical control over the microstructures and mechanical properties, which are essentially determined by the powder size, cooling rate, and laser energy density.AM 316L specimens for microstructure characterization were cut, using electron-discharge machining. The samples were ground by SiC papers with different granulometries and then mechanically polished. To reveal the microstructure, AM 316L specimens were measured by a scanning electron microscope (SEM) equipped with an electron backscatter diffraction (EBSD) analyzer and transmission electron microscope (TEM). The TEM samples of AM 316L were obtained by a focused ion beam (FIB).To evaluate the mechanical properties of AM 316L, a tensile test was conducted at a constant strain rate of 1.0 × 10−3 s−1 at room temperature. The tensile experiments were performed on a material test machine (MTS 810), following the ASTM: standard (gauge dimension is φ6 × 40 mm2). Moreover, to validate the effect of microstructure on the mechanical property, a Vickers micro-hardness test was conducted with a load of 4.9 N for a loading time of 15 s for each indentation using a micro Vickers machine.To obtain a fundamental understanding of how the multiscale fine-grain-embedded dislocation-cell structure affects the mechanical properties of AM 316L, the microstructure of AM 316L should be characterized in detail. The grain morphology and orientation measured using high-resolution EBSD reveal that the dislocation cell structures exist inside the grains as a result of minor and continuous orientation changes within a single grain (a, d). Overall, a highly unconventional grain shape, distribution, and continuous orientation change dominates the microstructure of AM 316L, thus forming a fine-grain-embedded dislocation cell structure. b and d shows that AM 316L possesses a large fraction of low-angle grain boundaries (LAGBs, <10°, approximately 31% of the total GBs). The average size of grain with high-angle grain boundary (HAGB) is 19.7 ± 17.8, and the average size of grain with LAGB 19.6 ± 17.5 μm. Here, the average grain size is determined by the software of Image-Pro Plus 5.0 from 10 samples along the tensile direction. The average dislocation cell size is 500 nm in the interior of the micron grain (g, h, j), which are measured using the software of Image-Pro Plus 5.0 with the dislocation cell more than 600 from 10 samples along the tensile direction. Thus, AM 316L is an unconventional material with a multiscale fine-grain-embedded dislocation-cell structure, a large fraction of LAGBs, and a band-like grain shape.The SEM micrographs of AM 316L illustrate the melt pool and individually parallel melt traces generated by the scanning laser beam (e and f), where the elongated grain is observed to be mainly parallel to the building direction over several adjacent melt pools. This trend agrees with the previous result [], where the solidification occurs by the process of epitaxial growth. The melt pool is surrounded by a fusion line, and shaped like a fish scale. GB across the fusion line is clearly found (c, e, f). The fine columnar structure of the multiscale fine-grain embedded dislocation cell structure is observed in each large grain (g–j). The crystal anisotropy and heterogeneity are attributed to various solidification rates and very high temperature gradients caused by the rapid movement of the laser beam. The dislocation cell structures exhibit a bar-like shape in the longitudinal direction and a hexagon, pentagon, or square shape in the transverse direction. This trend is because of the difference in the growth direction of the large elongated grains containing them (Zones Ⅰ and Ⅱ in g). In addition, the dislocation cell structures are more inclined to grow along the building direction (see, e–g). This phenomenon conforms the constitutional supercooling theory and compositional fluctuation, as proven in recent studies [a shows an optical micrograph, following the hardness test and the relative position of the test points. The test points are evenly distributed along the horizontal and vertical directions to determine whether the hardness points are on the fish-scale scan-fused line. b presents the relationship between the microstructure and hardness in the build-up and trace directions, respectively. The average hardnesses counted from 10 independent samples are 286 ± 10 HV and 291 ± 14 HV in the build-up and trace directions, respectively. The hardness in the build-up direction is more uniform than that in the trace direction. The average hardness in AM 316L is significantly higher than that of conventionally produced AISI 316L with approximately 195 HV. This trend is due to the embedded dislocation cell strengthening in AM 316L. In recent studies [], many dislocations are concentrated at the boundaries of the dislocation cells. Moreover, the test points on the fusion line are indicated by the larger semi-filled points, as shown in b. A comparison of the hardness obtained on and outside the fused line reveals no significant differences due to the homogeneous and fully dense fusion line.To investigate the mechanical response of the AM 316L, an engineering stress-strain curve is derived, as shown in a. The yield strength and elongation of the AM 316L are 600 MPa and 45%, respectively. Compared with the tensile properties of traditional casted 316L with the same chemical composition [], the yield strength of AM 316L is over 1.7 times greater, whereas the elongation is nearly the same. As previously mentioned, AM 316L has a multiscale-nested-grain structure (), whereas the conventionally casted 316L structure consists of only single-scale grains. In i and j, the synergistic interactions between the fine grains and dislocation cells occur during the plastic deformation, where the dislocation cell structure is retained and clearly elongated. In addition, to assess the structural homogeneity on the mechanical properties of the AM 316L, the hardness measurements in the build-up and trace directions are performed (Multiscale-nested-grain-size strengthening can be explained according to the theory of the grain-boundary strengthening and dislocation-cell strengthening []. Thus, the yield stress of the multiscale-nested-grain material can be expressed as [where σ0 is the friction stress.fn is the ratio of the embedded dislocation cell volume to the total volume, and fm is the ratio between the fine grain volume and the total volume. The multiscale fine-grain embedded dislocation cell framework consists of large and small grains, namely, fn+fm = 1. kH-P is the Hall-Petch coefficients of fine grains. K is a constant, μ is the shear modulus, b is the Burgers vector, and M is the Taylor factor. dn and dm are the grain sizes of fine-grains and embedded dislocation cells, respectively. According to existing studies [], σ0=157MPa, kH-P=0.29MPa⋅m1/2, K=3.7, μ=65GPa, b=2.5×10-10m, and M=2.44 can be obtained. fn and fm are 0.8 and 0.2, respectively. dn and dm are 19 μm and 0.5 μm, respectively. Thus, based on Eq. , the yield stress of AM 316L is derived from 10 independent tension results. The yield strength as measured by the experiment has a higher fitting value because of the collaborative strengthening mechanism in the multiscale nested-grain material. Thus, the yield strength of AM 316L is considerably affected by the grain size due to the presence of several GBs to impede the dislocation movement. The high strength of the multiscale nested-grain material is mainly derived from the embedded dislocation cell.b compares the elongation to the fracture and yield strength of 316L prepared by various processes []. The 316L steels with nanostructures have high strength (yield strength is approximately 0.8–1.6 GPa) but moderate ductility (elongation is approximately 5–27%). The 316L steels with annealed fine grains exhibit superb ductility (elongation is approximately 30–85%) but moderate strength (yield strength is approximately 0.1–0.4 GPa).To reveal the effect of the multiscale nested-grain structure on the ductility, the critical uniform plastic strain is investigated. For the stress–strain curve of the metal material that obeys a power law of σ(ε)=σy+l1εn, the uniform plastic strain, ε∗, is derived from the following equationwhere σy is the yield stress, and sn is the strain-hardening coefficient. Considering the parabolic stress–strain curve of a multiscale nested-grain material (n = 1/2), the uniform plastic strain can be expressed based on Eq. The stress-strain curves for the micro- and nanocrystalline components are described by the following relationswhere σm0 and σn0 are the yield stresses of microcrystalline and nanocrystalline, respectively, and lm and ln are the strain-hardening coefficients of macrograins and nanograins, respectively.For a multiscale nested-grain material, the stress–strain curve can be expressed as:In a general case, using conditions for the loss of stability in plastic yielding based on Eqs. –(6), the corresponding critical uniform strain can be expressed as:where the equivalent strain-hardening coefficient λ=(fnln+cfmlm)/fn+cfm, in which c is the non-uniformity of the plastic-strain distribution between multiscale-nested-grain structural components.Thus, to increase the ductility of multiscale nested-grain structural materials, increasing the effective coefficient of strain hardening and the coefficient of plastic strain distribution between structural components is necessary (). The AM 316L from our work exhibits an exceptional combination of high strength and ductility, which is ascribed to the synergistic interactions in the multiscale nested-grain structure (The detailed interaction between the dislocation and GB is shown in a. The dislocation patterns of AM 316L consist of discrete dislocations and dislocation pile-ups. The dislocation pile-up is formed by a block of the dislocation cell and solute atoms to the planar slip of dislocations, which contributes to the high strength. The partial dislocations could be nucleated in dislocation cells at high strains, resulting in the great ductility. This is because more partial dislocations are allowed to be stored in dislocation cells [b shows the deformation of dislocation cell, which contributes to the plastic deformation of AM 316L. As the plastic deformation proceeds, the complex and interlacing dislocation-network structure and deformation twinning are clearly observed in the TEM image, as given in b–d. Furthermore, the deformation twinning occurs in the fine grain (d) to enhance the plasticity of AM 316L.To understand the deformation mechanism in detail, the schematic in shows the coordination of the dislocation cell embedded in the fine grain. High strength is generally expected from the nested ultrafine grains or the embedded dislocation cells, which serve as obstacles to the moving dislocations in the fine grain interior. Good ductility depends on the deformation of GBs associated with the nucleation and gliding of dislocations. In the initial stage of deformation, through the movement of dislocations along GBs, the slip in the fine grain adapts to the strain until this process becomes unsustainable owing to the limited deformation of GBs. During a large deformation, the internal stresses under external loading are produced by the deformation of GBs, the embedded dislocation cells start to deform through mechanisms of shape change. The embedded dislocation cells are then elongated, and finally, the GB and boundary between dislocation cells crack or fracture.We have developed a multiscale fine-grain embedded dislocation cell structure to achieve excellent overall performance of strength and ductility in AM 316L. The microstructures, which contain the dislocation cell embedded in the fine grain, lead to high strength and ductility. The excellent overall performance is attributed mainly to the composite structure and coordinated deformation of the multiscale nested intergrain, based on the multiscale fine-grain embedded dislocation cell group. The local strain-hardening results in an overall yield strength owing to the multiscale strain coordination, particularly in the dislocation cell. The multiscale fine-grain embedded dislocation-cell structure can be applied to other metal materials and provides a useful way to design the high-strength materials with great ductility.Jia Li: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing, Software. Ming Yi: Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing. Hongyu Wu: Investigation, Methodology, Validation, Visualization. Qihong Fang: Conceptualization, Funding acquisition, Writing - review & editing, Project administration, Resources, Supervision. Yong Liu: Funding acquisition, Writing - review & editing, Project administration, Resources, Supervision. Bin Liu: Funding acquisition, Writing - review & editing, Project administration, Resources, Supervision. Kun Zhou: Funding acquisition, Writing - review & editing, Resources, Supervision. Peter K. Liaw: Investigation, Methodology, Funding acquisition, Writing - review & editing, Project administration, Resources, Supervision.The authors declare no conflict of interest.Blade fracture analysis of a motor cooling fan in a high-speed reciprocating compressor packageThis paper presents the root cause analysis of the blade fracture of the motor cooling fan in a reciprocating compressor package. The dynamic analyses results showed that the first-order torsional natural frequency of the fan was within ±5% of the fifth-order excitation frequency. In addition, the maximum dynamic stress of the fan operating in conditions of torsional resonance was 40.88 MPa, and exceeded the allowable stress of 36.12 MPa. Furthermore, the maximum dynamic stress occurred at the same locations where the cracks initially emerged. It was concluded that the excessive dynamic stress caused by the torsional resonance was the root cause of the fatigue fracture of the fan blade. Design improvements were implemented to adjust the natural frequencies of the fan, and included reducing the height of the blades, increasing the number of blades, and changing the weld locations. After the implementation of these modifications, the first-order torsional natural frequency of the new fan avoided the resonance regions, and the maximum dynamic stress of the new fan was 0.69 MPa, which was considerably lower than the allowable stress. No fracture problems occurred on the new fan.Motor-driven reciprocating compressors are extensively used in the petrochemical fields. The oscillatory torques on the compressor crankshaft will induce the torsional vibration of the shaft system. The shaft system of the reciprocating compressor package generally consists of the compressor crankshaft, coupling, motor shaft, and other components mounted on the shafts. The motor cooling fan—one of the auxiliary components on the motor shaft—will be subjected to the centrifugal force and the oscillatory torques transmitted from the compressor crankshaft. If the fan exhibits mechanical natural frequencies (MNFs) that are coincident with the compressor's excitation frequencies, or if the deformation or stress amplitudes off-resonance are too severe, premature failure of the fan may occur, and the reliability of the compressor system will diminish. For instance, premature fracture occurred on some blades of the motor cooling fan in a natural gas compressor package on an offshore platform. This compressor was a four-throw and balanced-opposed reciprocating compressor, and was driven by a motor with a rated speed of 995 r·min−1, as shown in shows some cracks on the blades of the motor cooling fan. The cracks initiated at the root of the blades and propagated to the middle of the blades. To ensure production, an undamaged fan was used to replace the damaged one, but the cracks occurred again within a period of one month of continuous operation. Therefore, it was necessary to identify the root causes of the fracture problem of the fan blades.The fractures of the blades in the turbo machines have drawn increased attention during the past few decades. The fatigue failure of the blades can be caused by many reasons, such as vibration, corrosion, fretting, thermal stress, and others []. Vibration is a main cause of the fatigue failure of the impellers and blades []. Modal and harmonic response analyses are extensively used to predict the vibration characteristics of the structures, and identify the root causes of premature failures [] conducted modal and harmonic response analyses for the blades of a gas turbine generator cooling fan, and found that the first-order MNF of the blades was close to the excitation frequency, and that the dynamic stress at the root of the blades at resonance exceeded the endurance limit of the material. Poursaeidi et al. [] investigated the failures that occurred at the first-stage compressor blades of a gas turbine by conducting metallographic observations, computational fluid dynamics (CFD) simulations, and static and dynamic analyses. The results indicated that the fracture of the blades may be attributed to the structural resonance. Kou et al. [] performed the modal, aerodynamic, and dynamic response analyses on a compressor blade, and found that the first-order resonance that occurred at the critical speeds may have contributed to the failure of the blade. However, few studies have been published on the fatigue fractures of the motor cooling fans in the reciprocating compressor packages. In addition, in this study, the alternating torque acting on the motor cooling fan was calculated based on a torsional analysis of the shaft system, and was used to predict the dynamic responses of this fan.The root causes of the premature fractures that occurred on the blades of the motor cooling fan were studied herein. The static stress caused by the centrifugal force was calculated based on a static analysis. The low-order vibration modes of the fan were estimated based on numerical and experimental modal analyses. Furthermore, the dynamic stress induced by the alternating torques was predicted using a harmonic response analysis. In addition, some design improvements were proposed to solve the blade fracture problem.Both the static and dynamic characteristics of the motor cooling fan were investigated to determine the root causes of the fractures of the blades. Firstly, the centrifugal force was applied to obtain the static stress distribution on the blades. Secondly, a numerical modal analysis was conducted to evaluate the risk of structural resonance. An experimental modal test then was conducted to verify the effectiveness of the numerical model, and estimate the damping ratio of the fan. Moreover, the working torque of the fan was calculated, and was subsequently applied to predict the dynamic stress responses of the fan.The motor cooling fan discussed in this study consists of eleven blades, a base plate, a top plate, and a hub. Different parts of the fan were connected by welding. The outer diameter of the fan was 800 mm, the height of the blades was 270 mm, the thickness of the blades was 5 mm, and the mass of the fan was 19.22 kg. The motor shaft and the shaft hole on the fan hub were interference fitted, and the fan hub was connected with a plate fixed at the end of the motor shaft using four bolts. The material of the fan was aluminium alloy AL6082 with a density of 2700 kg·m−3, Young's modulus of 7.1 × 104 MPa, Poisson's ratio of 0.33, ultimate tensile strength of 245 MPa, and yield strength of 140 MPa. The conditional fatigue limit of AL6082 obtained by the high-cycle fatigue test was 86 MPa []. Taking into account a surface factor of 0.9, a stress concentration factor of 0.7, and a safety factor of 1.5, the allowable stress of the material was 36.12 MPa [A finite element model of the fan was established to calculate the static stress distribution on the fan induced by the centrifugal force. The blades, base plate, and top plate were meshed with hexahedral elements using the sweeping method comprising three elements along their thickness directions. The hub and weld seams were meshed with tetrahedral elements. The type of contact pairs between different parts was set as bonded. Since the distance between the fan and the closest bearing was 80 mm, the motor shaft was not included in the numerical model. A total of 756,003 elements were generated in the numerical model of the fan. Further increments in the number of elements can lead to minor improvements in the numerical accuracy. The cylindrical constraint was applied to the shaft hole on the fan hub, which means that the shaft hole on the hub was constrained in the radial and axial directions, but was free in the tangential direction. The centrifugal force—a typical source of static stress for rotating blades—was exerted on the numerical model by applying a rotating speed of 995 r·min−1.After constructing the numerical model and applying the boundary conditions and loads, the displacements of all nodes can be computed by solving the static equilibrium equations:where K, x, and P are the stiffness matrix, displacement vector, and static load vector of the structure, respectively. The stress vector of each element σε can be calculated based on the following relationship [where xε is the displacement vector of the element, matrix Dε is related to the material properties, and matrix Bε is determined by the element's shape function. The stress at each node can then be obtained. The von Mises stress, also known as the equivalent stress, was adopted to evaluate the stress state of the motor cooling fan. According to the fourth strength theory [where σ1, σ2, and σ3 are the maximum, middle, and minimum principle stresses, respectively. show the deformation and stress distributions on the fan calculated by the numerical static analysis. It can be seen that the blades were extended outwards under the action of the centrifugal force, and the stress concentration occurred at the root of the blades. The maximum static stress on the fan was 2.12 MPa, which was much smaller than the allowable stress of 36.12 MPa. This indicated that the centrifugal force was not the main cause of the fracture failure.The equation of motion for a viscous damped structure can be written aswhere M and C are the mass and damping matrices of the structure, respectively, ẍ and ẋ are the acceleration and velocity vectors of the structure, respectively, and F(t) is the dynamic load vector. Numerical modal analysis is a useful method employed for the calculation of the structural modal parameters. For the small damping structures, the damping matrix is usually neglected in the numerical modal analysis []. The undamped free vibration equation can be expressed asSubstituting the special solution x=φejwt into Eq. This is a generalised eigenvalue problem. The eigenvalues ω012, ω022 … ω0n2, can be calculated by solving the following equation, we obtain the eigenvectors φ1, φ2 … φn, where ω0i and φi are the MNF and mode shape of the structure, respectively.A numerical modal analysis was carried out to evaluate the possibility of resonance. Damping was not taken into account in this analysis. The rated speed of the compressor was 995 r·min−1. Thus, the first-order excitation frequency was 16.58 Hz, and its integer multiples were the excitation frequencies. According to the API 618 standard [], to minimise the risk of torsional resonance, the torsional natural frequencies of the compressor shaft system should not lie within the margin of ±10% for the first-order excitation frequency, and in the margins of ±5% for the second- to the tenth-order excitation frequencies. The first six orders of the MNF predicted by the numerical modal analysis are shown in . We can see that the first-order natural frequency of the motor cooling fan was 78.83 Hz, which was within ±5% of the fifth-order excitation frequency, namely within the range of 78.77–87.06 Hz. In addition, the first-order mode shape was the torsional vibration of the blades, as shown in . Moreover, the maximum stress of the first-order mode shape occurred at the same locations as the crack locations, as shown in . It was indicated that the fatigue failure of the fan may be caused by the torsional resonance. From , we can also notice that the second-order MNF of the fan was 241.58 Hz, which exceeded by far the tenth-order excitation frequency of 165.83 Hz. Since the high-order harmonics of the torque fluctuation are generally too weak to cause harmful vibration responses, the second-order and high-order vibration modes of the fan were not discussed in detail here.A forced modal test was conducted to identify the MNFs, damping ratios, and mode shapes of the fan, and to verify the effectiveness of the numerical model. A motor cooling fan without the cracks was fixed on a shaft, and the shaft was constrained on a skid in our laboratory. The measuring points were distributed on one of the blades because of structural symmetry. Thirty-six measuring points were selected on the blade, and the twentieth point was chosen as the excitation point, as shown in . The single input and multiple output (SIMO) method was adopted in this experiment to obtain a column of elements in the frequency response function (FRF) matrix of the fan []. During the experiment, the excitation was always applied to the twentieth point, and the acceleration sensor was moved through all the measuring points.. An impact hammer (MSC-3), with a piezoelectric force sensor integrated in it, was used to excite the fan. The sensitivity of the impact hammer was 0.4 mV·N−1, and its non-linearity was no more than 2%. To obtain a good coherence in the frequency range of interest, a nylon hammer tip was selected because it generated a stationary force spectrum in the frequency range of 0–400 Hz, which contains the first six orders of the MNF of the fan. The responses were acquired by a triaxial piezoelectric acceleration sensor (PCB 356B18). The sensitivities of the acceleration sensor in the X, Y, and Z directions were 1030 mV·g−1, 1048 mV·g−1, and 992 mV·g−1, respectively, where “g” represents the gravity acceleration. The measuring range of the acceleration sensor was ±5 g, its non-linearity was no more than 1%, and its transverse sensitivity was no more than 5%. A four-channel data acquisition (DAQ) module (NI-9234) was used to simultaneously record the force and acceleration signals. The DAQ module was plugged into a four-slot NI compact DAQ chassis (cDAQ-9174). The chassis was connected to a laptop using a USB cable. Signals obtained from the modal test were processed by a data acquisition and signal processing software (Coinv DASP V10). The sampling frequency fs was set to 1280 Hz, and the number of sampling points N was set to 4096. Thus, the frequency resolution Δf = fs/N was 0.3125 Hz. To reduce the power-line interference at 50 Hz and its multipliers, the experimental system was grounded. To reduce the effects of random noise, each point was measured five times, and the measured data were linearly averaged. Moreover, the force and exponential windows were applied to the force and acceleration signals, respectively, to decrease the effects of noise. The FRFs between the response and excitation points were calculated using the H1 estimator []. The impulse response functions (IRFs) can then be calculated by performing the inverse Fourier transforms to the FRFs. The eigen-system realisation algorithm (ERA) method, one of the advanced time domain methods, was adopted to estimate the modal parameters using the IRFs [ shows the stabilisation diagram of the ERA method, which was used to distinguish the physical (structural) poles from the mathematical ones []. To construct a stabilisation diagram, the poles were estimated at increasing model orders. Physical poles should readily align, while the mathematical poles are scattered. In , the horizontal axis represents the frequency, the vertical axis denotes the modal order, and the letter “s” highlighted in red indicates the stable poles. The poles identified in two consecutive modal orders will be regarded as stable poles if the differences of the modal parameters between these two orders are less than certain threshold values. The threshold values for the frequency, damping ratio, and eigenvector used in were 0.5%, 10%, and 5%, respectively. In addition, other letters in , such as “o”, “v”, and “f”, indicate that only one or two modal parameters stabilise with an increasing order number. Furthermore, the green curve in is the complex modal indicator function (CMIF). With the help of the stabilisation diagram, the first six orders of the vibration mode of the fan were identified. The mode shapes were normalised by the mass normalisation method.The first six orders of the MNF and damping ratio estimated by the modal test are listed in , and the first-order mode shape is shown in . The first six orders of the MNF predicted by the numerical modal analysis showed good agreements with the experimental results with relative errors less than 3%. Furthermore, the first-order mode shape obtained by the modal test was the torsional vibration of the blade, consistent with the numerical result. The comparison of the modal parameters between the numerical and experimental results indicated that the numerical model of the fan was reliable. The first-order damping ratio of 0.0013 was applied in the subsequent dynamic response analysis of the fan. We can also notice that the first-order MNF of the fan obtained by the modal test was 78.76 Hz, which was 0.01 Hz smaller than the lower limit of the resonance region (78.77–87.06 Hz). It should be noted that the mass of the acceleration sensor will slightly decrease the MNFs of the fan. To evaluate the mass loading effect of the acceleration sensor, we added a point mass element weighing 50 g at different positions on a blade in the finite element model. Numerical modal analysis results showed that the decrease of the first-order MNF of the fan after adding the point mass element was within the range of 0.27%–0.73%. Therefore, the actual value of the first-order MNF of the fan shall be within the resonance region, and torsional resonance may occur on the fan during operation.Firstly, a torsional vibration analysis of the shaft system of the reciprocating compressor package was conducted to obtain the working torques of the fan. A harmonic response analysis was then performed to predict the dynamic stress of the fan under the action of the working torques.To obtain the working torques of the fan, a torsional vibration analysis of the shaft system of the reciprocating compressor package was conducted. The rotating components of the shaft system mainly included the compressor crankshaft, the large end of the connecting rod, coupling, motor shaft, motor rotor, and the motor cooling fan. The reciprocating components, such as the piston, piston rod, cross-head, and the small end of the connecting rod, also affect the torsional characteristics of the shaft system. The mass-elastic model of the shaft system was established using the lumped-mass method. The rotating components with large inertia were simplified as the lumped-mass elements, and the rotating components with small inertia, or the dispersed components, were simplified as the elastic elements. The equivalent inertia values of the reciprocating components were added to the inertia of the crank pins. The shaft system of the reciprocating compressor package was simplified into twenty-nine lumped-mass elements and twenty-eight elastic elements. The inertia and stiffness of each section in the shaft system were provided by the manufacturers of the compressor, coupling, and motor.Torsional modal analysis was also conducted to obtain the torsional natural frequencies of the shaft system without damping considerations. The first-order torsional natural frequency of the shaft system was 91.44 Hz, which was outside the ±5% margins of the fifth- and sixth-order excitation frequencies. Thus, torsional resonance of the shaft system would not occur. Forced response analysis was subsequently conducted to calculate the dynamic torsional responses of the shaft system to the compressor's excitation torques. The compressor's excitation torques were applied to the lumped-mass elements of the four throws of the crank pins, and a damping ratio of 0.01 was utilised. The working torques of all locations in the shaft system were then obtained. The working torque of the fan was estimated, and the Fourier decomposition was performed to obtain the working torque harmonics at each order of the excitation frequency, as shown in Since the first-order MNF of the fan was within ±5% of the fifth-order excitation frequency, namely within the range of 78.77–87.06 Hz, structural resonance was likely triggered during operation. A harmonic response analysis was carried out to study the dynamic stress response of the fan in the resonance condition. The fifth-order working torque of 195.18 N·m was applied to the outer surface of the fan hub, the mode superposition method was utilised, and the damping ratio of 0.0013 that was obtained from the modal test was then used. After conducting the harmonic response analysis, the deformation and stress distributions in the frequency range of 78.77–87.06 Hz were obtained. The deformation and stress curves for some of the monitored points were obtained as a function of frequency. It can be observed that the maximum deformation and stress responses of the fan occurred at 78.82 Hz owing to torsional resonance. The deformation and stress contours of the fan in the case of torsional resonance are shown in . It can be observed that the working torque caused the torsion of the blades that resulted in a stress concentration at locations near the weld seams between the blades and the hub. The maximum dynamic stress of the fan in the case of torsional resonance was 40.88 MPa, which exceeded the allowable stress of 36.12 MPa. The location of the maximum dynamic stress matched the locations where the cracks were initiated. It can be indicated that the excessive dynamic stress at the root of the blades caused by torsional resonance was the main cause of the fatigue fracture of the fan blades.Based on the above analyses, it was determined that the fatigue fracture of the fan was caused by the torsional resonance that occurred on the fan blades. The fracture problem of the fan can be solved using the following approaches: (1) changing the stiffness or mass parameters to adjust the MNFs of the fan and avoid resonance, (2) increasing the damping of the fan to reduce the dynamic responses, or (3) changing the torsional vibration characteristics of the shaft system of the reciprocating compressor package to reduce the working torque of the fan. Considering the repair cost and time, the following design improvements were adopted to adjust the MNFs of the fan: (1) the height of the blades were reduced by 50 mm, the number of blades was increased by four, and the area of the top plate was increased to improve the torsional stiffness of the fan, (2) the blades were welded with the base and top plates but were not welded with the hub to reduce the stress concentration zones on the blades. The fan hub is in common to ensure full interchangeability. The material of the new fan was still the aluminium alloy AL6082. The weight of the new fan was 22.01 kg. The comparison of the geometrical models before and after the improvements is shown in The numerical model of the new fan was established with 768,384 elements. The cylindrical constraint was exerted on the shaft hole on the hub. After the modal analysis, the first three orders of the vibration mode of the new fan were obtained, as shown in . The first- and second-order MNFs of the new fan were 73.04 Hz and 73.17 Hz, respectively, and the corresponding mode shapes were bending vibrations of the fan. The third-order MNF of the new fan was 90.53 Hz, and the corresponding mode shape was the torsional vibration of the fan. shows the comparison of the Campbell diagrams before and after the improvements. Before the improvements, the first-order MNF of the original fan was within ±5% of the fifth-order excitation frequency, which indicated that the fan may resonate during operation. After the improvements, the first two orders of the MNF of the new fan were outside the ±5% margins of the fourth- and fifth-order excitation frequencies, and the bending resonance would thus not occur. In addition, the third-order MNF of the new fan was also outside the ±5% margins of the fifth- and sixth-order excitation frequencies, which indicated that the torsional resonance was prevented after the improvements.To further evaluate the dynamic stress response of the new fan, two harmonic response analyses were conducted considering the fifth and sixth orders of the working torque, respectively. The frequency ranges of these two analyses were set at ±5% of the fifth- and sixth-order excitation frequencies, namely, within 78.77–87.06 Hz and 94.52–104.47 Hz, respectively. Some monitored points on the new fan were selected, and the deformation and stress curves varying with the frequency of these points were estimated. It was found that the maximum dynamic response of the new fan occurred at the frequency of 87.06 Hz. The stress distribution at 87.06 Hz is shown in . It can be seen that the maximum dynamic stress was 0.69 MPa, and occurred at the root of the blade, which was much smaller than the allowable stress of 36.12 MPa. As a result, the fatigue fracture would not occur on the new fan owing to the avoidance of torsional resonance. No fracture problem has occurred within the one-year period that followed the replacement of the fan that verifies the effectiveness of the design improvements.The most important conclusions of the present study are as follows.The root cause of the blade fracture of a motor cooling fan in a high-speed reciprocating compressor package was investigated using static and dynamic analyses. The modal analysis results showed that the first-order MNF of the fan was close to the fifth-order excitation frequency, and the first-order mode shape was the torsional vibration of the blades. The harmonic response analysis results showed that the maximum dynamic stress of the fan in the resonance condition was 40.88 MPa, which exceeded the allowable stress of 36.12 MPa, and occurred at the same locations where the cracks initially emerged. It was concluded that the excessive dynamic stress at the root of the blades caused by torsional resonance was attributed to the fatigue fracture of the fan.To adjust the MNFs of the fan and avoid the structural resonance, some design improvements were implemented, including reducing the height of the blades, increasing the number of blades, and changing the weld locations. The MNFs of the new fan avoided the resonance regions, and the dynamic stress response of the new fan was considerably lower than the allowable stress. No fracture problem has occurred on the new fan since the replacement of the fan.At the design stage of the reciprocating compressor package, it is necessary to conduct modal and harmonic response analyses on the components mounted on the shafts, such as the motor cooling fan. The torsional natural frequencies of these components should be well separated from the excitation frequencies to avoid the torsional resonance and insure the reliability of the compressor system.Machine learned feature identification for predicting phase and Young's modulus of low-, medium- and high-entropy alloysThe growth in the interest and research on high-entropy alloys (HEAs) over the last decade is due to their unique material phases responsible for their remarkable structural properties. A conventional approach to discovering new HEAs requires scavenging an enormous search space consisting of over half a trillion new material compositions comprising of three to six principal elements. Machine learning has emerged as a potential tool to rapidly accelerate the search for and design of new materials, due to its rapidity, scalability, and now, reasonably accurate material property predictions. Here, we implement machine learning tools, to predict the crystallographic phase and Young's modulus of low-, medium- and high-entropy alloys composed of a family of 5 refractory elements. Our results, in conjunction with experimental validation, reveal that the mean melting point and electronegativity difference exert the strongest contributions to the phase formation in these alloys, while the melting temperature and the enthalpy of mixing are the key features impacting the Young's modulus of these materials. Additionally, and more importantly, we find that the entropy of mixing only negligibly influences the phase or the Young's modulus, reigniting the issue of its actual impact on the material phase and properties of HEAs.High-entropy alloys (HEAs) are compositionally complex alloys containing multiple principal elements in near equiatomic proportions, and have received widespread attention due to their promising structural properties Fundamentally, lowering the Gibbs free energy of mixing (ΔGmix) in ΔGmix=ΔHmix−TΔSmix stabilizes the alloy. Here, T is the absolute temperature, ΔHmix is the enthalpy of mixing, ΔSmix=−R∑i=1n(CilnCi) is the entropy of mixing Recent efforts utilizing machine learning Data assemblage and feature selection: Available experimental data is collected from existing literature . The data for crystallographic phase prediction (‘phases dataset’) consists of 329 entries where 159 are BCC HEAs, 111 are FCC HEAs and 59 are multiphase. Since there are relatively smaller number of HCP HEAs discussed in the literature, we do not include them in ‘phases dataset’ to avoid creating a biased dataset that reduces the accuracy of the predictive model . Tm is an indirect measure of strength of interatomic interactions (b) it is seen that one of the features, am (lattice constant calculated by the rule of mixtures), has a negligible impact in determining the Young's modulus. Based on this observation, it is redundant to include ∆a (difference in lattice constants) as an additional feature. Moreover, the ‘phases dataset’ does contain both the ΔTm and Δa as features. This concept stems from the fundamental principle of single phase solid solution formation that emphasizes on calculating the ‘concentration averaged difference’ in parameters like electronegativity or atomic size when describing HEAs because the latter do not have a single dominant solvent or a solute Data analysis is initially performed by calculating the Pearson correlation coefficient P. Here, P = 1 denotes strong positive correlation and P = –1 denotes a strong negative correlation. The absence of any significant correlation amongst any pair of features indicates that all metrics should be considered in the model.Model construction: After assembling the data and evaluating the feature values, we employ the datasets to construct two machine learning models, one for predicting the Young's modulus and the other for the crystallographic phase. In each case, the dataset is classified into a training set (90% of the data) and a test set (10% of the data). For example, in our ‘phases dataset’, we use around 296 (90%) data points for training the model and around 33 (10%) data points for testing the model. This process is essential to build a robust model with minimum training error (discussed in following sections) before the model can be adopted for actual usage. The errors generated in the predictions of the test dataset are quantified and minimized by altering the parameters of the models, which are retrained iteratively on the training data followed by re-predictions on the test subset until the error is minimized below a threshold. The model is finally used for predicting crystallographic phases and Young's moduli of alloys in . The final model predictions are validated with experimental measurements.Model for predicting crystallographic phase: The phase prediction is performed by employing the Gradient Boost Classification outputfromtree=∑residual∑[(previousprobability)i×(1−previousprobability)i]Subsequently, the log of the odds is updated aslog(odds)=previousprediction+∑i=1numberoftrees(learningrate)×(outputfromtree)Once the new log(odds) value is obtained, we again convert it to a probability (P) as discussed previously. The process is repeated by calculating the new residuals and at every successive step the residual value is expected to decrease until it attains a minimum at the last tree. The final probabilities are used to classify the data into favorable or unfavorable outputs.Model for predicting Young’s modulus: Gradient boost for regression E=Emean+[(learningrate)×residual]1+[(learningrate)×residual]2+...+[(learningrate)×residual]nwhere the subscripts 1, 2, …, n represent the number of the trees. At each successive tree, the residual value decreases and reduces to a minimum at the final tree that provides the closest prediction to the measurements.Experimental validation: We synthesize 26 alloys composed Mo-Ta-Ti-W-Zr elements, as listed in , by arc-melting compressed pellets of elemental powder mixes (Sigma-Aldrich, purity ≥ 99.9%) in Argon atmosphere (pressure 30 psi) on a water-cooled copper hearth. Powdered metals are used to minimize the occurrences of elemental macro-segregation and achieve homogeneity in the alloys. Powders are mixed thoroughly in a laboratory jar mill (Thomas Scientific Series 8000), at an optimum rpm to ensure a balance between the centrifugal and gravitational forces. Next, the powder mixes are compressed to a pressure of 5000 psi on cylindrical pellets of diameter 20 mm using a Carver hydraulic press. The pellets are then arc-melted (in Edmund Bühler GmbH Mini Arc Melting System), cooled and re-melted a total of four times to ensure improved homogeneity. An X-ray diffractometer (Panalytical Empyrean vs. 7.9f 20170530 X-ray Diffraction Unit) is employed for characterization of the crystal structure, with the 2θ scan ranging from 10 to 90°, with the radiation from a 45 kV, 40 mA copper target. Subsequently, the data is analyzed with Malvern Panalytical HighScore software package serves as the validation set for both classification and regression models. Nanoindentation is performed on mounted and polished samples using a Hysitron TI900 nanoindenter with a Berkovich tip. Indents are performed using load control with a maximum load of 5 mN and a 5–2–5 s load-hold-unload profile. An array of 25 indents (5 by 5 pattern) with a spacing of 10 µm in the x and y was performed on each sample. Modulus and hardness are determined using the Oliver and Pharr method Crystallographic phase predictions: The tunable parameters in a machine learning algorithm (gradient boost classifier in this case) are known as the hyperparameters. The hyperparameters for gradient boost algorithm are predominantly “learning rate” and “n-estimators” . These hyperparameters can be set to an optimum value while training the algorithm on the dataset, to eventually produce minimum training error. Here, 90% of the data from training set is employed to train the model and the remaining 10% is used as a test set to check for error in prediction. We define error aserror=numberofincorrectpredictionstotalnumberofdatapointsThe error is calculated for all possible combinations of 15 “learning rate” values ranging from 0.001 to 0.5 and 15 “n-estimators” ranging from 50 to 1000. illustrates the hyperparameters’ optimization (the process of the guiding the hyperparameters to an optimum value) to minimize error. A learning rate of 0.05–0.2 with 450–600 estimators produces a minimum error of 0.3077 as identified in region R in (a). Thus, any set of hyperparameter values within that range can be considered for the final model. Here, the learning rate is set to 0.1 and n-estimators is set to 500. The output produced by the model after the experimental dataset for the 26 alloys, is listed in . On comparison between the measured and predicted values for crystallographic phases, our model is able to accurately identify the lattice structure of 14 out of the possible 23 alloys (the crystal structure test data contains 3 alloys that form HCP lattices, and our existing model can predict only FCC, BCC or a mix of FCC and BCC). The empirical model ) which categorizes them all as BCC phase. Hence, the empirical model cannot be guaranteed to yield correct results when applied to a refractory system such as the one under investigation.Young’s moduli predictions: We define the mean absolute error as meanabsoluteerror=∑i=1n|yi−xi|n, where yi is the prediction and xi is the measured Young's modulus from experiments, while n is the total number of data points available. From the hyperparameter optimization illustrated in (b), we find that a learning rate of 0.0031 and n-estimators of 1000 produces the minimum mean absolute error of 23.59 GPa, and consequently adopt that for the final model. The model output reveals that 19 of the 26 predictions are within an error of 20%, while 14 are within a 12% margin. We strongly assert that the accuracy is limited by the sparse data available to us, and will improve as the additional simulation and experimental results are incorporated into training the model. In the current state of the model, the root-mean square error (RMSE) for this regression model was calculated to be 87.76%.The relative importance of the different features in the structure and modulus predictions are illustratively compared in . For crystallographic phase prediction, the order of feature importance is noted to be Tm > Δχ > ΔTm > am > Ω > ΔHmix > Δa > λ > ΔSmix > δ, while for Young's modulus, we find the order to be Tm > ΔHmix > λ > Ω > Δχ > δ > ΔSmix > am. These results evince that certain descriptors like Tm and ΔHmix are more crucial than the Hume-Rothery parameters (Δχ and δ) in determining the Young's modulus, while Tm , Δχ , ΔTm are key in estimating the phases that form in these multicomponent alloys.Tm and Δχ are predicted to be the key parameters in determining the alloy phases. Since a high Tm of the alloying elements is analogous to their high bonding energies, the constituent metal with the highest melting temperature would preferably not allow the inclusion of another metal atom with a lower bonding energy and hence the system would form multiple phases. On the other hand, the difference in electronegativity is directly responsible for the probability of intermetallic compound (secondary phases) formation. ΔTm plays a vital role in determining phases as it conveys information on relative bond strengths of the metals in the alloys; the lower the ΔTm, the higher is the probability of the formation of single phase as the metallic bond strengths in each metal will be identical. Δa and am exhibit strong contribution in determining the phases, relative to ΔSmix, as similar lattice parameters enable the formation of single phase to be more probable. Interestingly, contrary to the Hume-Rothery guidelines and an earlier report Our results conclusively suggest that melting point of alloys, which is an indirect metric of bond strength In summary, we employ available data and experimental measurements to construct machine-learned correlations to predict the Young's modulus and crystallographic phase of low-, medium- and high-entropy alloys. Thermodynamic, geometric and phenomenological parameters are chosen as features and the final models are built to enable predictions for 26 equiatomic alloys in Mo-Ta-Ti-W-Zr elemental family. There is good agreement between the model predictions and the experimental results, which corroborate that additional material features, apart from the Hume-Rothery parameters, contribute significantly in predicting the crystallographic phase and Young's modulus of low-, medium- and high-entropy alloys. A key insight lies in the insignificant effect of the entropy of mixing as a feature for structure and property predictions, even for the HEAs.The data and methods reported in this paper are available from the corresponding author upon reasonable request.Correlation of the tribological behaviors with the mechanical properties of poly-ether-ether-ketones (PEEKs) with different molecular weights and their fiber filled compositesThe tribological behaviors of three poly-ether-ether-ketones (PEEKs) with different molecular weights and their SCF (short carbon fiber)/graphite/PTFE (polytetrafluoroethylene) filled composites were examined using a block-on-ring apparatus under dry sliding conditions. Tensile tests, hardness measurements and dynamic mechanical thermal analysis (DMTA) of the PEEK based materials were also performed. The tribological behaviors of PEEK based materials were correlated with their mechanical properties and the tribological mechanisms were discussed based on scanning electron microscope (SEM) inspections of worn surfaces and wear debris. Under a low apparent pressure, a high material ductility seems to reduce the wear rate of pure PEEK through alleviating the microcutting effect exerted by the protruding regions of the counterpart. Under a high pressure, however, a high stiffness seems to improve the wear resistance of pure PEEK by reducing the plastic flow occurring in the PEEK surface layer. After incorporating SCF/graphite/PTFE fillers, the wear rate of PEEK was decreased significantly. Thinning and cracking of SCF are supposed to be the important factors determining the tribological behaviors of the composites.Poly-ether-ether-ketone (PEEK) presents the combination of a good tribological performance and a high strength and therefore, it is being widely used as tribomaterial Generally, for a single-phase material, the combination of a low surface energy, a high stiffness, and a high toughness results in a good tribological performance, i.e. low friction coefficient and wear rate. The structure of polymeric materials, e.g. molecular weight and crystalline structure, exerts important roles on their properties including tribological performances Compared with a single-phase polymer, the polymer composite presents a more complicated structure–tribology relationship. Generally, the roles of fillers can be summarized into the three following aspects: lubricating effect, improving mechanical properties, e.g. compressive strength and stiffness, and promoting the formation of a homogeneous transfer film. The internal lubricants refer to the materials with a low surface energy, e.g. polytetrafluoroethylene (PTFE), and the layer-structural materials in which the layers are linked by weak Van der Waals bonds, e.g. graphite, MoS2, etc. A pioneering work was carried out by Voss and Friedrich Effort on fundamental understanding of tribology is always crucial for the formulation of tribomaterials. Surely, it is of interest to understand how PEEK's mechanical properties affect its tribological behaviors. In this work, the mechanical properties of three pure PEEKs and two SCF/graphite/PTFE filled PEEK composites with different matrix molecular weights were characterized. The tribological behaviors of the PEEK materials were examined and correlated with their mechanical properties. The objective of this paper is to describe a comprehensive effort to correlate the tribological behaviors of PEEK materials with their mechanical properties.Three pure PEEKs referenced in this paper as MA (material A), MB (material B), and MC (material C) and two SCF/graphite/PTFE filled PEEK composites referenced as MB FC30 and MC FC30 were used in this work. MB FC30 and MC FC30 were compounded, respectively from MB and MC with each 10 wt% SCF (9.1 vol%), graphite and PTFE. The molecular weights of the three PEEKs follow the order: MA < MB < MC. Plates were compression molded at 400 °C and slowly cooled to room temperature in the mold. All samples were prepared under the same compression and cooling conditions. Two kinds of plates with dimensions of 100 mm × 80 mm × 4 mm and 100 mm × 80 mm × 2 mm were prepared. The thick plates are employed for tensile and tribological tests while the thin ones are used for dynamic mechanical thermal analysis (DMTA) and hardness measurements. It should be noted that using the same processing parameters, the PEEKs with different molecular weights have different crystallinities: a high molecular weight corresponding to a lower crystallinity All tests were performed at room temperature (23 °C) on a Zwick 1474 universal testing machine at a constant crosshead speed of 1 mm/min. The measurements followed DIN EN ISO 527 using dumbbell shaped specimens. The specimens with a 4 mm thickness were machined from the compression molded plates. The displacement of each specimen during tension was accurately measured by an extensometer. All presented data correspond to the averages of five measurements.The measurements of material universal hardness (HU) were performed on a Shimadu DUH-202 dynamic ultramicrohardness tester using a Vickers hardness indenter. The maximum load was 100 mN. The depth of penetration of the indenter tip was measured dynamically when applying and releasing the load and the hardness data were determined by the final depth after releasing the load completely. All presented data are the mean values of 10 measurements.DMTA tests were performed using a Gabo Qualimeter Explexor with tension configuration. The complex modulus and loss factor of each specimen with a dimension of 50 mm × 12 mm × 2 mm were determined at a constant frequency of 10 Hz. The samples were heated from 20 °C to 200 °C at a rate of 1 °C/min.The specimens for tribological tests, having a dimension of 4 mm × 4 mm × 12 mm, were cut from compression molded plates. The tribological tests were performed using a block-on-ring apparatus. The counterpart was a 100Cr6 steel ring with a 60 mm diameter and a mean roughness, Ra, 0.2 μm. In order to reduce the running-in process, the specimens were “pre-worn” with grinding paper before the test (firstly P 800 and then P 1200) to an arc outline to match the configuration of the counterpart. During the test friction force was measured and dynamically recorded into a computer. The ratio between the friction force and load equals friction coefficient. For each test the friction coefficient refers to the mean value of the data recorded after running-in process. All the tests in this work were conducted for 20 h under dry conditions at room temperature. Specimen's mass loss, Δm, was measured after frictional test and the specific wear rate wS of the material was calculated using the equation:where ρ is the density of the specimen, F the normal load applied on the specimen during sliding, and L is the total sliding distance. The inverse of wear rate is usually considered as material's wear resistance. The sliding velocity was controlled by the rotating speed of the ring and was fixed at 1 m/s in this work. As suggested in a previous work shows respectively the Young's modulus and elongations at break of the five studied materials. Compared with MA, MB presents a slightly lower Young's modulus but a slightly higher elongation rate. MC presents the lowest Young's modulus but the highest elongation rate. c shows the hardness of the three pure PEEKs. As is seen, MC exhibits the lowest hardness and MA presents the highest hardness. shows the DMTA results. Being consistent with the above results of tensile tests and hardness measurements, MC exhibits the lowest storage modulus but the highest loss factor near its glass transition temperature (Tg). Summarized, an increase in PEEK molecular weight corresponds to a decrease in material stiffness but an increase in material ductility. shows the overview of the tensile fracture surface of MA. Two distinct zones, indicated respectively as zone I and zone II, were noticed on the fracture surface. b and c shows respectively the zones I and II with higher resolutions. Clearly, zone I corresponds to fracture initiation and zone II corresponds to fracture propagation. The failures initiate from randomly distributed impurities acting as stress concentration sites at the early stage of tensile tests. The drawn nature of the morphology in zone I suggests ductile deformation rather than stable crack growth. Once fracture is initiated, a fast interlayer crack takes place in zone II. d shows the overview of the fracture surface of MC and e illustrates the zone corresponding to fracture initiation. Compared with MA, MC exhibits a higher ductility and therefore a longer fibration process occurs.Being filled with SCF/graphite/PTFE, MB FC30 and MC FC30 present a significantly improved tensile modulus. However, the ductility of PEEK is much decreased. f shows the overview of the fracture surface of MC FC30. Unlike pure PEEK, the fracture initiation zone cannot obviously be distinguished from the fracture surface of the composite. The DMTA results shown in verify that the composites present a higher tensile modulus than pure PEEKs. g and h shows representative PEEK/SCF interfaces observed on the fracture surfaces of MB FC30 and MC FC30, respectively. The arrows in the images indicate the carbon fibers. It is clearly seen from the fracture surface of MC FC30 that some PEEK material sticks on the fiber. This permits to conclude that the matrix/fiber adhesion is stronger in MC FC30 than in MB FC30.Friction coefficients and wear rates were summarized in . For three pure PEEKs, in studied range, the increase in apparent pressure leads to slightly higher friction coefficients. The friction coefficient seems to increase slightly, if any, with increasing molecular weight. Under low apparent pressures (1 MPa and 2 MPa), the increase in molecular weight corresponds to a lower wear rate. Under 4 MPa, however, the increase in molecular weight corresponds to a higher wear rate. After incorporating SCF/graphite/PTFE, the wear rates of the PEEKs are significantly decreased by 93.0–96.3%. Under 1 MPa and 2 MPa, incorporating the multiple fillers does not change significantly the friction coefficients of PEEKs. Under 4 MPa, however, especially for MC FC30, the incorporation of the fillers distinctly decreases the friction coefficient. The wear rates of the composites are increased by raising the apparent pressure from 1 MPa to 4 MPa. Since the apparent pressure exerts important influences on the tribological performances of the materials, the following discussions on tribological mechanisms were split into two sections according to the pressure. respectively shows the worn surface and wear debris of MA obtained under 1 MPa. The sliding direction is indicated in the worn surface by an arrow. The keen-edged grooves and particle-like wear debris suggest that a microcutting effect constitutes the main wear mechanism. The sliding is essentially governed by the dynamic process occurring in the surface layer involved in the friction process c shows the worn surface of MC. Compared with MA, the grooves on the surface of MC are alleviated. This can be the reason why the wear rate of MC is lower than that of MA. However, besides the grooves, short ripple-like deformations perpendicular to the sliding direction are noticed on the surface of MC. This morphological feature is much more obvious on the worn surface of MC produced under 4 MPa (). These ripple-like deformations are assumed to be caused by a stick-slip motion of the counterpart. Due to the adhesion between the two sliding pairs, the PEEK surface presents a larger tangential deformation along the sliding direction than the sub-surface material. Accordingly, plastic flow of PEEK surface occurs.d shows the worn surface of MC FC30. During the frictional process, carbon fibers support most of the load and internal lubricants, i.e. graphite and PTFE, reduce the adhesion between the sliding pairs. Under a low apparent pressure, the fibers were ground progressively and therefore the thinning of fibers mainly determines the wear rate.With increasing pressure, the thickness of the frictional layer can be larger. a–c respectively shows the worn surfaces of MA, MB, and MC produced under 4 MPa. Interestingly, compared with the worn surface produced under 1 MPa, the grooves on the worn surface of MA produced under 4 MPa are less severe. This suggests that the microcutting effect can no longer be a dominant factor determining PEEK's wear rate. d shows the wear debris of MA produced under 4 MPa. Bamboo-raft-like, film-like, and rod-like debris, which are indicated as I, II, and III in the figure, are noticed. The rod-like debris is assumed to derive from fractured bamboo-raft-like debris. The film-like and bamboo-raft-like debris suggests that viscous flow With comparing the worn surfaces of the three pure PEEKs produced under 4 MPa, one can find that the plastic flow occurring in the frictional layer is more evident for materials with a lower stiffness and a higher ductility. As is seen from b, short ripple-like deformations are noticeable on the worn surface of MB produced under 4 MPa, while periodic and long ripple-like deformations are clearly observed on the worn surface of MC (c). Due to the adhesion force, the PEEK surface presents a larger tangential deformation than the sub-surface material. Once the stress applied on the polymer surface exceeds the critical stress , for the two composites the increase in pressure leads to a monotonic increase of wear rates. Under a high pressure cracking and debonding of fibers become important factors contributing to material loss. shows the worn surface of MB FC30 produced under 4 MPa with different magnifications. Compared with the worn surface produced under 1 MPa (d) more scratch traces are noticed on the worn surface. The scratch of the surface is caused by the cracked fibers (b). Under a high pressure, fractures occur in fibers when they are impacted by the protruding regions of the counterpart. c–e respectively shows the squared zones indicated as I, II, and III in b. These three figures clearly show the behavior of a broken fiber after being removed from the matrix. When the broken parts are removed from the matrix, they scratch the matrix in the following sliding process. Moreover, when the broken parts collide with other fibers, they can cause further fractures of fibers. This scratch effect seems to be important in the zone where carbon fibers are normal to the contact surface.f shows the worn surface of MC FC30. Compared with MB FC30, removal of fibers is reduced and less scratch traces are observed on the worn surface of MC FC30. Two factors can contribute to the improvement in the material's wear resistance. Firstly, the less stiff matrix can absorb more impact energy on fibers through elastic deformation. Secondly, the better matrix/SCF adhesion surely reduces the removal of fibers from the matrix.In the present work the mechanical and tribological properties of three PEEKs with different molecular weights were examined. The influence of apparent pressure on the tribological behavior was investigated. The tribological mechanisms were discussed and correlated with the mechanical properties. The following conclusions can be drawn:Under the same compression molding and cooling conditions, the increase in molecular weight corresponds to a higher material ductility and a lower material stiffness.The tribological mechanism of PEEK is closely related to its mechanical properties. Under a low pressure, the microcutting effect exerted by the protruding region of the counterpart constitutes the main wear mechanism. In this case, an increase in the material's ductility decreases the wear rate. Under a high pressure, however, plastic flow occurring in PEEK surface layer and material transferring to the counterpart become important factors contributing to material loss. In this case, a high stiffness seems to benefit the material's wear resistance.Being filled with SCF/graphite/PTFE, PEEK exhibits a much improved wear resistance. The increase in apparent pressure increases the wear rates of the composites. Under a low pressure, thinning of fibers dominates the tribological behaviors of the composites. Under a high pressure, cracking and debonding of fibers are important factors determining the tribological behaviors of the composites. MC FC30, in which the matrix has a higher ductility, exhibits a lower wear rate than MB FC30.Fiber-based computational modeling of rectangular double-skin concrete-filled steel tubular short columns including local bucklingBoth the external and internal thin-walled steel sections of a rectangular double-skin concrete-filled steel tubular (RDCFST) short column loaded axially may be susceptible to progressive local buckling, which is rarely included in mathematical modeling programs employing the fiber discretization scheme for such composite members. This paper provides a description of a new computational simulation technology, which is developed for the nonlinear fiber analysis of short RDCFST columns axially loaded to failure. The progressive localized-buckling failure of thin-walled steel sections are included in the formulation of the computational simulation method. Experimental measurements and finite element analysis results obtained by ABAQUS software by other researchers are employed to evaluate the accuracy of the computer algorithms developed. Numerical studies are undertaken to ascertain the responses of RDCFST columns to the change of design parameters. Proposed is a mathematical expression for the determination of the axial resistances of RDCFST columns considering the post-localized buckling strength of thin-walled steel sections. It is demonstrated that the proposed computational modeling and design technologies yield good predictions of the responses of RDCFST columns and can be used to undertake the nonlinear simulation of RDCFST columns with any class of steel sections.Rectangular double-skin concrete-filled steel tubular (RDCFST) columns as shown in have been developed by modifying conventional concrete-filled steel tubular (CFST) columns Experimental research on the behavior of short RDCFST and SDCFST columns has been reported by investigators. Zhao and Grzebieta Most of the numerical studies on circular DCFST columns where both tubes were circular and square DCFST columns composed of an internal circular tube were performed by means of employing the commercial software ABAQUS. Huang et al. Rectangular CFST and DCFST short columns whose steel sections are identified as slender or non-compact will undergo localized buckling when they are loaded axially. Uy Computational models have been presented by researchers for the analysis of short SDCFST and RDCFST columns. The analysis of SDCFST short columns was undertaken by Zhao et al. More recently, numerical and experimental investigations into the behavior of CFST and DCFST columns filled with high-strength or ultra-high-strength concrete have been undertaken by various researchers The computational modeling method is developed by means of utilizing the fiber discretization technique and has been implemented in a computer program. The fiber scheme discretizes the cross-section of a RDCFST column into small elements as shown in . The fiber method does not employ contact elements to model the interface between steel and concrete and does not divide the column into elements along the length of the column so that it significantly saves the model development and computational time when compared to the traditional finite element method are employed to calculate fiber stresses from given incremental axial strains. The assumptions made in the formulation of the fiber model are: (a) all fiber elements are subjected to the same axial strain under a given axial compression; (b) the localized buckling of both inner and outer tubes is considered; (c) the improved ductility of concrete due to steel encasement is included in the material laws for concrete; (d) failure occurs if the extreme compression fiber strain of concrete reaches the specified maximum strain; (e) the time-dependent behavior of concrete due to shrinkage and creep is ignored.The idealized material model of stress-and strain for steel in compression and tension illustrated in is used for cold-formed steels having yield stress below 460 MPa. The tri-linear curve is employed to model the response of high-strength steel material. The rounded part (0.9εsy<εs⩽εst) is represented by using the expression of Liang in which σs represents steel stress, εs denotes steel strain, fsy stands for the steel yield stress, εsy corresponds to the steel yield strain, εst is assigned to 0.05 for cold formed and high-strength steels. The ultimate strain εsu is chosen as 0.2 for mild structural steels and 0.1 for high-strength steels to consider the steel ductility.The confinement of concrete in a rectangular CFST column is limited to the corners of the column and does not increases the compressive strength of the filled concrete schematically illustrates the stress–strain relations of concrete in RDCFST columns. The stress–strain relationship for concrete is described by Parts OA, AB and BC. The material laws for the filled concrete reported by Mander et al. in which σc stands for the concrete compressive stress in longitudinal direction, fcc' represents the concrete strength in compression, εc denotes the concrete strain, εcc' is the strain that corresponds to fcc' and Ec is concrete Young’s modulus, which can be computed by the expression suggested by ACI Committee 363 The strain ε'cc depends on the concrete strength and is calculated by the following equations given by Liang ε'cc=0.002forf'cc≤28(MPa)(fcc'-28)/54000+0.002for28<f'cc≤82(MPa)0.003forf'cc>82(MPa)The Parts AB, BC, and CD of the concrete stress–strain response shown in are described by the mathematical expressions provided by Liang σc=f'ccforε'cc<εc≤0.005100f'cc(0.015-εc)(1-βc)+βcf'ccfor0.005<εc≤0.015f'ccβcforεc>0.015where βc represents the section effect on the ductility and varies with the width-to-thickness ratio (Bs/t) of the column cross-section, where Bs is the larger ofBo and Do. The coefficient βc was proposed by Liang βc=1.0forBs/t⩽241.5-Bs/(48t)for 24<Bs/t⩽480.5 forBs/t>48The effective strength (fcc') of concrete in compression is affected by the concrete quality, column size and loading methods. It is determined as γcf'c , where fc' denotes the concrete cylindrical compressive strength, and γc is the coefficient given by Liang where Dc is selected as the greater value of (Do-2to) and (Bo-2to) for a rectangular cross-section, Do represents the depth of the cross-section, Bo denotes the width of the cross-section and to stands for the thickness of the outer steel tube.The internal and external tubes in a RDCFST column that are axially loaded are vulnerable to local buckling. When the in-plane compressive stress applied to the steel tube wall reaches its critical buckling stress, the initial localized instability occurs. Liang et al. σcr=fsy1.198×10-7bt3-9.869×10-5bt2+0.005132bt+0.5507where σcr is the initial localized buckling stress of the imperfected steel tubular wall, b stands for the clear width of the wall and t is the tube thickness. Eq. can be applied to rectangular RDCFST columns with clear b/t ratios ranging from 30 to 100.After the initial localized buckling, the post-buckling of steel tubes in a RDCFST column under increasing compressive load takes place progressively. The post-buckling strength of steel tube walls of a rectangular CFST column was ascertained by Liang et al. . These effective-width expression of Liang et al. beb=1.921×10-6bt3-3.994×10-4bt2+0.02038bt+0.5554in which be stands for the plate’s effective width. Eq. can be used to compute the effective widths of cross-sections that have the ratios of b/t varying from 30 to 100 The effective width of a steel plate decreases with increasing the applied compressive stress. In contrast, its ineffective width increases with increasing the applied compressive load. The maximum ineffective width (bmne) of the steel plate at the ultimate strength state is determined as bmne=b-be. Liang The fiber-based computational method for simulating the progressive post-buckling of steel plates developed by Liang . The fiber mesh of the cross-section is maintained in the nonlinear analysis of the column under consideration. In this scheme, steel fibers lied in the ineffective width (bne) of a steel tube wall under a given strain increment are given zero stress value as shown in while the stresses of steel fibers in the effective width are maintained and this process is repeated until bmne is attained. When the maximum ineffective width bmneof the tube wall is reached, the fibers in the effective width of the steel tube wall are assigned to the yield stress. This implies that the progressive post-buckling of the steel tube under increasing axial compression is characterized by the in-plane stress redistribution within the steel tube wall. Further details on the simulation of the gradual localized buckling can be found in the work of Liang The ductility indicator, which represents the axial strain ductility of a RDCFST column, is determined bywhere εu.col represents the axial strain that corresponds to the applied axial load falling to 90% of the column ultimate load or the ultimate strain in the post-yield range of strain-hardening. The column yield strain εy.col is determined as ε0.75/0.75. The strain ε0.75 is computed using the strain corresponding to the applied axial load that attains 75% of the column ultimate load.To verify the accuracy of the computational modeling and simulation algorithms, the predictions are compared against the test results provided by Tao and Han . The ratio of the strength of concrete cylinder f'c to the concrete cube strength was taken as 0.84 in the computer simulation as suggested by Oehlers and Bradford , the computer simulation program yields good results. The average value of Pu,num/Pu,exp ratios is 0.96, which is sufficient for engineering design purpose. However, the computed resistance of Specimens FR100 × 4-NS20 × 2.5-C40 is less than 90% of the experimental value. The cause for this is probably because the actual strength and quality of concrete in these specimens are unknown. As depicted in , the simulated load–strain relations of RDCFST short columns are generally in good agreement with those determined from experiments. Moreover, the computer modeling technique simulates well the residual strengths of the tested DCFST columns. However, there is a discrepancy between numerical simulations and experimental measurements because the idealized material laws and material properties of steel and concrete materials were used in the simulations, which might not represent the actual responses of the steel and concrete materials used in the tested column specimens. The Specimen Reference Sample shown in (b) exhibited very brittle failure, which might be caused by the poor quality of concrete and the premature fracture of the steel tubes in the tested specimen. Consequently, the post-peak experimental responses of the column deviated considerably from the simulation.The fiber-based computational model is further verified by comparing the computational results of Specimens D-SS-a, X-0.49 and X-0.64 with those obtained by using the nonlinear finite element analysis software ABAQUS by Ding et al. that the axial load–strain responses predicted by the developed fiber-based model are in good agreement with those obtained by the sophistical finite element software ABAQUS. Both computer programs produce almost the same initial axial stiffness and ultimate axial load for each column, and very close post-peak behavior. However, it should be noted that the fiber-based computational model is significantly more computationally efficient than the finite element model created by ABAQUS as pointed out by Liang The influences of geometric and material design parameters on the structural responses of RDCFST short columns were investigated by using the computational algorithms proposed. A total of 24 RDCFST columns shown in were analyzed by taking into account the experimentally identified failure mode of local buckling. It was assumed that the internal and external steel tubes had the same material properties. A value of 200 GPa was used for Young’s modulus of steel material.The computational algorithms written were used to determine the effects of the Di/Do ratio on the structural responses of RDCFST columns. The Di/Do ratios were determined to be 0.4, 0.5, 0.6 and 0.7 by means of altering the depth of the inner tube only as given in Group 1 in gives the simulated responses of RDCFST columns having different Di/Do ratios. The figure indicates that changing the Di/Do ratio from 0.4 to 0.5, 0.6 and 0.7 causes a reduction in strength by 4.9%, 9.5%, 11.2% and 17.5%, respectively. This is because of the reduction in the concrete area which shares a large part of the applied load. As presented in , the column’s ductility improves with an increase in the Di/Do ratio.Two cases were investigated to evaluate the significance of Do/to ratio on the responses of RDCFST columns. The first case was to vary the depth of the outer steel box while other parameters were not changed as shown in Group 2 in presents the axial load–strain relations of RDCFST columns, which are a function of the Do/to ratio. A marked increase in the column’s initial stiffness and axial resistance is obtained by means of using a larger Do/to ratio. Changing the Do/to ratio from 50 to 60, 65 and 70 results in the increases of 16.7%, 32.9% and 46.7% in the column’s capacity, respectively. It is illustrated in that the ratio of Do/to notably influences the post-peak performance of RDCFST columns. The relationship between the Do/to ratio and the column ductility is shown in , which indicates that a decrease in the ductility index is caused by increasing the Do/to ratio. When altering the ratio of Do/to from 50 to 60, 65 and 70, the ductility indicator varies from 5.71 to 5.40, 5.30 and 5.20, respectively.The second situation was to change the thickness of the external tube only as demonstrated in Group 3 in . This means that the width B0=375mmand depth D0=600mm of the external tube were unchanged but its thickness was varied to determine the Do/to ratios of 50, 60, 65 and 70. shows that the larger the Do/to ratio, the lower the initial stiffness. A reduction in the axial resistance of RDCFST short columns is expected when the Do/to ratio increases. As shown in , the use of a larger Do/to ratio causes a slight reduction in the column ductility. The ductility indicators of the RDCFST columns having Do/to ratios of 50, 60, 65 and 70 are 5.50, 5.40, 5.36 and 5.30, respectively.The Di/ti ratios of 25, 30, 35 and 40 of the columns in Group 4 shown in were determined by altering the inner tube thickness from 10 mm to 6.25 mm. The load-axial strain responses of RDCFST columns that have different Di/ti ratios are given in . Increasing the Di/ti ratio slightly decreases the initial-stiffness of the columns. In addition, the use of a larger Di/ti ratio considerably reduces the capacity of RDCFST columns. The relationship between the Di/ti ratio and the ductility is shown in . The column’s ductility indicator reduces slightly from 6.15 to 5.91.The responses of RDCFST short columns in Group 5 given in have been determined by the developed computer program to assess the significance of steel yield stress on their performance of ductility and strength. The predicted nonlinear responses of RDCFST columns to various steel yield stresses are presented in . The steel yield stress does not affect the column’s initial stiffness. However, the use of steel tubes with higher yield stress improves the axial resistance of the RDCFST column. The column’s capacity could be improved by 8.33%, 16.6% and 33.3% respectively by means of changing the yield stress from 250 to 300, 350 and 450 MPa. shows the ductility indicators of these RDCFST columns designed by using steel tubes that have different yield stresses. The ductility indices of DCFST columns that have the yield stresses of 250, 300, 350 and 450 MPa have been determined as 6.03, 5.81, 5.44 and 4.9, respectively. It can be concluded that the steel yield stress does not have a significant influence on the column ductility.The RDCFST short columns in Group 6 were designed with different concrete compressive strengths varying from 40 MPa to 100 MPa. The predicted relationships between the applied load and strain for RDCFST short columns with concrete strength as a design variable are given in . It is discovered that the axial capacity of RDCFST columns is remarkably improved by filling higher strength concrete into double-skin tubes. The axial load-carrying capacity of the composite column of double skins would have an increase of 13.8%, 24.1% and 63.7%, respectively if the concrete strength of 40 MPa is replaced by 50, 70 and 100 MPa. This implies that higher strength concrete can be used to improve the axial capacity of short RDCFST columns. presents the relationship between the ductility indicator and concrete design strength. The use of higher concrete strength in RDCFST columns results in lower ductility. The ductility index of the RDCFST column filled with 40 MPa concrete is 6.15, which is reduced to 3.80 when 100 MP high-strength concrete is employed. This is attributed to the brittle property of high strength concrete.The load distribution in Specimen C19 shown in was ascertained by means of conducting a nonlinear inelastic analysis on this column using the computer program. shows the computed load–strain responses of the filled concrete and steel sections in the composite column. The figure indicates that the filled-concrete, outer steel section and inner steel section share 56.1%, 37.2% and 16.9% of the ultimate axial load of the column, respectively. A large portion of the applied load is resisted by the filled concrete.The developed computational modeling and simulation program can be used directly in the design of RDCFST shot columns. However, a simple expression for design purpose is needed in practice. As discussed previously, the use of non-compact or slender steel section in RDCST columns leads to economical designs, but localized buckling, which should be explicitly included in design. A new design expression for determining the section axial capacity of RDCFST short columns including post-local buckling effects is proposed here aswhere Pu represents the ultimate axial load of the column; Aseo and Asei represent the effective areas of the outer and inner steel sections, respectively, and are calculated by using the effective width expressed by Eq. ; fsyo and fsyi stand for yield stresses of the external and internal steel sections, respectively; Ac represents the concrete cross-sectional area.To verify the proposed design expression, the computed strengths of RDCFST columns by employing Eq. , where Pu,cal is the axial capacity calculated by using Eq. . The agreement between the calculations and experimentally measured values is shown to be good. The mean of the computed to the experimental loads is 0.959. The statistical analysis indicates that the standard deviation of Pu,cal/Pu,exp is 5% and the coefficient of variation is 5.3%. This suggests that the design expression proposed can be utilized in designing RDCFST short columns.This paper has described a computer simulation and modeling method incorporating an efficient fiber discretization scheme for the modeling of RDCFST and SDCFST short columns loaded concentrically to the localized buckling failure. The computational modeling technique proposed can detect the onset of initial localized buckling and simulate the progressive post-buckling of both outer and inner thin-walled steel sections in the inelastic nonlinear response analysis of RDCFST columns. The comparison of fiber-based calculations with experimental measurements as well as the nonlinear finite element analysis results produced by ABAQUS documented elsewhere has verified the accuracy of the computational technique. The computational results obtained in the parametric studies provide a new insight into the behavior of RDCFST short columns where the progressive localized instability of steel tubes takes place under increasing compression. A simple expression has been proposed and can be used to design RDCFST columns. The developed computational modeling and simulation technology provide structural designers with efficient tools that can be employed to predict the responses of high-strength RDCFST columns constructed by rectangular steel sections of any class.Muhammad Rizwan: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing–original draft, Visualization. Qing Quan Liang: Conceptualization, Methodology, Software, Validation, Writing – review & editing, Supervision, Project administration. Muhammad N.S. Hadi: Validation, Writing – review & editing, Supervision.The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Optimisation of the superplastic forming of aluminium alloysSuperplastic forming involves the shaping of metal sheets by gas pressure at elevated temperatures. It relies on the fact that fine-grained metals can exhibit high sensitivities of flow stress to strain rate, but this usually only occurs at quite slow strain rates. As a result, the process is much slower than conventional pressing operations and this is a major factor in process costs. Although there have been several attempts at optimising the process to minimise the forming time, most do not achieve a true optimum. In the present work, the non-linear system is optimised within the constraints of product thickness and damage (cavitation) level by searching a control space of enclosed volume history. The finite element method is used in that process, and good constitutive models and damage evolution models are required. These models are described, along with the optimisation strategy, and results are given for the forming of commercial aluminium alloys.At elevated temperatures, certain fine-grained alloys can exhibit very large ductilities when deformed at relatively low strain rates. This ductility, coupled with low flow stresses, allows the forming of complex shapes in sheet metal using relatively modest gas pressures The underlying material behaviour – superplasticity – is characterised by a high sensitivity of stress to strain rate. The usual measure of this, the rate sensitivity index, m, is given bywhere σ is the stress and ε˙p the (plastic) strain rate, with the partial derivative defined at constant temperature, T, and microstructural state, {Λ}. Superplasticity usually involves m |
> 0.4 and this leads to large tensile ductilities, typically >200%, and counteracts strain localisation in forming. In aluminium alloys, especially, the ductility is often not limited by macroscopic strain localisation but by the formation of internal cavities. This cavitation can reach levels of several vol.% without fracture, and this will clearly have an influence on service performance of the formed material, in addition to the thickness of the formed sheet.A significant issue with superplasticity is that the rate sensitivity, m, varies not only with temperature and microstructure but also with strain rate. It decreases at rates faster than some optimum value at a given temperature, and so attempting to form at too fast a rate will mean a reduction in the property responsible for counteracting strain localisation. This is the fundamental issue in forming cycle optimisation.There have been several attempts to optimise superplastic forming. Most have involved the finite element method to simulate the process. Many early approaches The complexity of the material behaviour, as well as other aspects of the process such as geometry and friction, mean that superplastic forming is a non-linear process, and finding a true optimum is rather more difficult than maintaining some simple strain rate sequence. Bate et al. A very simple model was used in this “proof of concept” work. This was an axisymmetric rigid-plastic model, based on the formalism given by Wang The constitutive model of the material is of paramount importance. The type of model used has been dealt with in some detail by Ridley et al. where Q is the activation energy, R the gas constant and T the absolute temperature. The equation for flow stress is then:where C, B0, α and β are material constants, and the evolution of grain size, D, is given byIsotropic visco-plasticity was assumed, although ultimately this may not be adequate and there are indications that the grain growth – and so the hardening – may be strain-state dependant The forming operation was defined as the time to give a certain volume enclosed by the deforming sheet. This was described by a weighted sum of orthogonal polynomials:where t is the time at enclosed volume V (which is zero at t |
= 0), pi are Legendre polynomials and v is a variable, ranging from −1 to +1, which is a linear function of V. The coefficients, ci, then define the forming sequence. Because orthogonal polynomials are involved, these coefficients define a point in a control space. For a given optimisation, the total volume and the total forming time are set. This reduces the degrees of freedom from N |
+ 1 to N, so the space of coefficients to be searched becomes N-dimensional.The result of each finite element run, corresponding to a point in the control space, was parameterised as either the minimum product thickness in the formed part or the maximum level of cavitation. Only the former will be considered here. The optimisation is then rationalised as a search of the control space to find the maximum value of the minimum thickness in a component.In the early stages, a simple line search of the control space was used. This involved examining the neighbourhood of a trial point in discrete steps, and either moving the search to the appropriate trial point if it gave a better value, or reducing the search step size if it did not. While this worked, it was relatively inefficient, and a Nelder–Mead simplex algorithm was adopted for N |
= 2. This type of algorithm can fail if local optima occur, but these have not been detected in the shapes studied. provides a simple shape for demonstrating the results of the optimisation process. The predicted optimal volume and pressure histories for forming a 0.7 l conical dome in 1000 s, using data for AA5083 Al–4.5%Mg alloy at 530 °C, with a N |
= 5 control space are shown in . Each finite element simulation took about 1 s on a standard personal computer, and about 50 simulations were required to locate the optimum. There is, in fact, little benefit in using N |
> 3 and this reduces the number of simulations required. Note that volume rate is treated as the control, and pressure as a response. The reduction in pressure occurring for the optimal sequence in may cause problems with very simple pressure-controlled systems.Trials were run for a variety of forming times, and the results in terms of minimum thickness are shown in . Even with this rather undemanding shape, considerable time saving is predicted by using optimised forming sequences. For a given minimum thickness, the optimal forming time is always less than half of that required at constant volume rate.A further issue addressed is that of robustness. In the actual forming process, there will be variability in the material and in the forming conditions. Mechanical testing has shown that commercial materials give quite reproducible results, but there may be differences in forming temperature, for example. Introducing a variation in temperature of 10 °C between optimisations gives the results shown in . The difference in material behaviour due to the temperature difference has made only a small change to the optimal volume history, and if the optimum path calculated for 530 °C is used for simulated forming at 520 °C, the resulting thickness is within 0.2% of that from the optimal path calculated for 520 °C. For 540 °C, the difference is <0.1%.Although the use of an optimal volume path is robust, this is not the case if pressure were used to control the process. This is clear from where, although the volume histories are similar, the pressure histories are markedly different. Attempting to control the forming process via pressure could lead to large deviations from the predicted result unless very close control of material and temperature were possible.The use of multiple finite element simulations within a control-space optimisation strategy has the potential of giving truly optimal forming conditions, which other methods are unlikely to achieve. Using enclosed volume rate gives robust control, such that the optimum is relatively insensitive to variation in material behaviour. The method depends on having sufficiently accurate constitutive models of the material behaviour.Development of processing maps for Al/SiCp composite using fuzzy logicIn the present study the aluminium reinforced with 10% SiCp particulate composites were produced through powder metallurgy route and processing map was developed. The composite specimens were subjected to compression test at different strains, strain rates and temperatures. Using the load-stroke data, flow stress–strain were evaluated. Fuzzy logic has been implemented for the prediction of flow stress. Power dissipation efficiency and instability parameter were evaluated and plotted with temperature and stain rate using the predicted values of flow stress from fuzzy logic. The contour plots were superimposed to obtain the processing maps. The dynamic recrystallisation zones and instable zones were identified and validated through micrographs. The processing maps can be used to select optimum strain rates and temperatures for effective hot working of Al/10% SiC composites.The development of MMCs has been driven by the need for structural materials with high specific strength and stiffness. Reinforcements may be continuous in the form of fibres or discontinuous in the form of whiskers and particles or chopped fibres. While fibre reinforced composites offer the highest specific stiffness along reinforcement direction particulate MMCs are isotropic in their properties and are easier to process via powder metallurgy or casting route. The intrinsic workability of MMCs depends upon the initial microstructure as decided by the alloy chemistry and prior processing history and its response to the applied temperature, strain rate and strain in processing Modeling of process and system identification using input–output data has always attracted many research efforts. In fact, system identification techniques are applied in many fields in order to model and predict the behaviors of unknown and/or very complex systems based on given input–output data Aluminium Silicon carbide composite were made by a powder metallurgy route. Aluminium powder was blended with 10% volume of silicon carbide powder of average size 25 μm to make the SiCp/Al composite. The powder mixture was degassed at 450 °C for 1 hr under a vacuum pressure of 10−4 |
Torr. The powder mixture was vacuum hot pressed at 450 °C at a pressure of 75 MPa to a size of 20 mm diameter and 20 mm height. The sintering was done at 500 °C for 2 h. The Hot compression tests were performed on a 10T FIE servo controlled universal testing machine for different strains (0.1–0.5), strain rates (0.0087–2.7 s−1) and temperatures (300–500 °C). Using the test data, the processing maps were obtained for different strains. After compression testing, the specimens were immediately quenched in water and the cross-sections were examined for microstructure. The microstructure of the specimens was obtained through Versamet 2.0 Microscope with Clemex vision Image Analyser and domains with maximum efficiency and flow instability zones were validated.The conventional regression equation for description of the flow stress during hot deformation in which Z is the Zener–Holloman parameter, Q the activation energy for deformation, R the gas constant, A and α the constants related to materials and n is the stress exponent. Roberts and Ahlblom is a function of as an independent variable. The parameters calculated in this work for Al/10% SiC composites areThe comparison between the measured and predicted flow stress using the above method has been made. The mean error D is calculated aswhere pi is the predicted value and mi the measured value. It can be seen that the mean error of the predicted flow stress using regression method is up to 12.5%; its accuracy is thus rather low. Hence an attempt has been made to predict flow stress using fuzzy logic.A fuzzy logic unit comprises a fuzzifier, membership functions. A fuzzy rule base, an inference engine and a defuzifier. First, the fuzzifier uses membership functions to fuzzify the S/N ratios. Next, the inference engine performs a fuzzy reasoning of fuzzy rules to generate a fuzzy value. Finally, the defuzzifier converts the fuzzy value into a crisp value.The concept of fuzzy reasoning is described briefly based on the three-input–one-output fuzzy logic unit.The fuzzy rule base consists of a group of if-then control rules with the two inputs, x1, and x2 and one output y, i.e.Rule1:ifx1isA1andx2isB1andx3isC1thenyisD1elseRule2:ifx1isA2andx2isB2andx3isC2thenyisD2else⋮Rulen:ifx1isAnandx2isBnandx3isCnthenyisDnelseA1, B1, C1, D1 etc. are fuzzy subsets defined by the corresponding membership functions. In this paper, four fuzzy subsets are assigned for strain rate and five subsets each are assigned to temperature and strain. Ten fuzzy subsets are assigned to the flow stress. Various degree of membership to the fuzzy sets is calculated based on the values of x1, x2, x3 and y1. Sixty fuzzy rules are derived directly based on the experimental results. By taking the max–min compositional operation, the fuzzy reasoning of these rules yields a fuzzy output. The membership function of the output of fuzzy reasoning can be expressed asμCo(y)=(μA1(x1)∧μB1(X2)∧μC1(y)∨⋯μAn(X1)∧μB(X2)∧μc(y)where A is the minimum operation and V is the maximum operation Finally, a defuzzification method, called the center-of-gravity method The comparison between the measured and predicted flow stress using fuzzy logic has been made and the mean deviation is 7.2% which is significantly lower than that of regression analysis.Traditional curve fitting techniques have been in use for obtaining these maps. Processes such as hot working can encompass a range of complex non-linear and interaction effects. This is enhanced with the fact that strain rate values generally used for carrying out experiments cover a range of four to five orders of magnitude. Hence, the traditional curve fitting techniques carried out to compute the flow stress values at finer temperature and strain rate intervals, using the experimental data points, may not be appropriate for modeling these highly complex effects. On the other hand, fuzzy logic has been found capable of learning from a data set to describe the non-linear and interaction effects with great success. The advantages of fuzzy logics are that the functional relationship between various variables can be obtained even if the form of non-linear relationship is unknown and some of the experimental data are faulty. This makes the fuzzy logic techniques a robust technique for obtaining the functional relationship in any engineering problem. It has been applied in a number of areas of engineering. In this paper, fuzzy logic model is applied for predicting flow stress and obtaining the processing map. Hot working data reported for Al/10% SiC has been selected for illustrating the methodology. Using the flow stress data fuzzy rule has been developed and the predicted flow stress value using fuzzy logic has been used to estimate power dissipation efficiency and instability parameter for the development of processing maps.The stress–strain (σ versus ɛ) curve and the flow behavior of materials are strongly dependent on the strain rate and temperature. The general form of the flow behavior at constant strain and temperature is expressed aswhere K and m are considered as constants. From m can be obtained asFor a stable material flow, value of m lies between 0 and 1. A close observation of experimental data reveals that m is different at different strain rates. The power dissipated per unit volume P, is given byIn which G is the dissipated co-content given byand J is the dissipated co-co-content given byThe maximum dissipater co-content Jmax is given byThe efficiency of power dissipation, η is defined as corresponding to its value at limiting strain rate, J is obtained asHence, the efficiency of power dissipation, η, as proposed by Prasad and Sasidhara It is to be mentioned that efficiency can be directly computed by obtaining J from numerical integration procedure carried out by Narayanamurthy et al. The regime, where the metallurgical instability during plastic flow occurs, was obtained in which the following instability condition is satisfied:The parameter ξ(ɛ) may be evaluated as a function of temperature and strain rate to obtain an instability may, where metallurgical instability during plastic flow occurs in regimes where ξ(ɛ) is negative.Once the flow stress is obtained the efficiency of the power dissipation is obtained as. For evaluating the integral in the above equation, the integral in Eq. ∫0εσdε˙=∫00.0087σdε˙+∫0.0087εσdε˙=σε˙m+1ε=0.0087+∫0.0087εσdε˙The first integral on the right hand side of Eq. is evaluated by assuming the power law nature of the σ versus ɛ curve. The value of m at ɛ |
= 0.0087 s−1 is obtained by finding the slope of the logσε˙ curve close to the point ɛ |
= 0.0087 S−1. The second integral is evaluated using trapezoidal rule, the value of ɛ being obtained from the fuzzy model. The instability parameter ξ(ɛ) as defined in Eq. was obtained by using central difference method The compression tests on cylindrical Al/10% SiC composites were conducted at different temperatures ranging from 300 to 500 °C, strains (0.1–0.5) and strain rates (0.0087–2.7 s−1). Flow stress data has been obtained from the load-stroke data. The flow stress data was implemented through fuzzy logic for prediction. A comparison between prediction capability of regression analysis and fuzzy logic has been made. The prediction error was found to be less for fuzzy logic. shows the plot of experimental value of flow stress and fuzzy logic predicted flow stress. The average deviation of error is −0.3526% for 300 °C at 2.7 s−1 and the deviation is 13.32% for 500 °C at 0.0087 s−1. Larger error values are corresponding to low values of flow stress. The absolute error varies from −0.12 to 5.372 MPa for 500 °C at 0.0087 s−1. Thus even though the absolute error values are small, the percentage error is more due to low value of flow stress. The low values of stress are prone to be sensitive to errors. More uniformity can be achieved by taking more number of membership functions in the low value range. The low stress range happens to be the boundary of the input domain, i.e. low strain rates and high temperatures. Hence more number of membership functions should be taken in the boundary of the inputs and small values of the output.Using the flow stress data, the strain rate sensitivity, power dissipation efficiency and flow instability parameter were estimated and processing maps were constructed. Microscopic examination revealed the mechanism of deformation under different test conditions.The flow curves obtained for Al/10% SiC composites deformed in compression at 300 °C at different stain rates ranging from 0.0087 to 2.7 s−1 are presented in . The flow stress is significantly low at lower strain rates whereas the work hardening rate is relatively high. Hence the flow stress is found to increase with increase in strain. At lower strain rates, the deformation is isothermal but at high strain rates it is adiabatic. The increase in strength is attributed to dispersive hardening effect of SiC particles. The matrix around the SiC particles presents much higher dislocation density than that of normal alloy. The high dislocation density regions restrict the plastic flow and contribute to the strengthening and strain hardening The flow curves for different strain rates at constant temperature of 400 °C are shown in . The Al/10% SiC composites undergo greater strengthening and work hardening at lower temperature. At lower strain rates, the strain hardening is more and is compensated by softening due to higher temperature. Hence steady state flow curve is observed. As the temperature increases the strengthening effect of SiC particles is considerably diminished. The periodic decrease in flow stress at higher strain rates is due to dynamic recovery and dynamic recrystallisation. Dislocation removal during straining accomplished by the motion of high angle grain boundaries that lead to the annihilation of dislocations on more massive scale. DRX is limited to FCC metals that have relatively low yield strength. This in principle lead to decrease in flow stress by about 50% when it is periodic type or grain coarsening type. In composite materials however the relative flow softening is limited to 20% partially because of inhomogenities in the nature of flow generally induced under normal testing conditions, but also because of the grain boundary drag. shows the power dissipation efficiency contours plotted against the temperature and strain rate for 0.5 strain. The efficiency represents the relative rate of entropy production during hot deformation and characterize the dissipative microstructure under different temperatures and strain rates. The maximum efficiency of power dissipation in the domain of DRX is about 30–40% for low SFE materials and 50–55% for high SFE materials such as aluminium. The contours in this domain are generally widely spaced Dynamic recrystallisation (DRX) is a beneficial process in hot deformation since it not only gives stable flow and good workability to the material by simultaneously softening it but also reconstitutes the microstructure. The DRX globularises the acicular preform microstructure and redistribute the prior particle boundary defects in powder metallurgy compacts to facilitate further processing and eliminates discrete particle defects in compacts by transferring mechanical energy across the hard particle interfaces to refine them. The DRX domain is chosen for optimizing hot working and controlling the microstructure and is a safe domain for bulk metalworking . DRX domain generally occurs in the homologous temperature range of 0.7–0.8 and strain rate range of 0.1–1 s−1 for low stacking fault energy materials. The strain rate range is still lower for high stacking fault energy materials. The maximum efficiency of DRX domain is 30–40% for low stacking fault energy materials and 50–55% in the high stacking fault energy materials. The contour in DRX domain will be normally wide spread representing a steep hill and present a fairly wider window.The flow instability occurs in two different zones. At higher strain rate range of 0.523–2.7 s−1 between a temperature range of 300–320 °C the instability has occurred. The second zone has occurred at higher strain rate range of 1.7–2.7 s−1 between a temperature of 340–480 °C.Localization may occur in two modes. In the first mode of localization, discontinuous deformation often localizes in narrow zones and the neighboring regions remain intact. Factors such as material compositions, boundary conditions and type of loading may affect localization. A material can fail due to the formation and growth of micro cracking at randomly distributed locations. In the second mode localization may result in motions along preferential directions called shear bands and can lead to extensive deformation and failure At higher strain rates, heat generated due to local temperature rise by plastic deformation is not conducted away to the cooler regions of the body since the available time is too short. The flow stress in deformation will get lowered and further plastic flow will be localized. The band gets intensified and nearly satisfies adiabatic conditions. Such bands are called adiabatic shear bands, which exhibit cracking, recrystallisation along macroscopic shear planes represents the formation of shear bands at a high strain rate of 2.7 s−1 and a temperature of 300 °C for 0.5 strain.The presence of SiC particles in aluminium matrix when deformed causes the interface to crack and debond. Since the matrix undergoes plastic flow while the particles do not deform. When the accumulated stresses become large, the interface may separate or the particles itself crack at lower temperature and higher strain rates. This may lead to the creation of microstructural damage due to cavity formation ultimately contributing to ductile fracture. Voids generated by particle cracking and debonding causes macroscopic crack propagation from the surface to the interior linking the voids. Debonding of SiC particles observed at 2.7 s−1 and 400 °C is shown in Hot compression tests were performed on Al/10% SiC composites produced through powder metallurgy. The flow stress was evaluated for a temperature range of 300–500 °C and a strain rate range of 0.0087–2.7 s−1. The power dissipation efficiency and instability parameters were evaluated and processing maps were constructed for 0.1 and 0.5 strains. The optimum domains and instability zone were obtained for the material. The domains for hot working significantly differ from that of pure aluminium. The microstructure evaluation leads to particle fracture, debonding adiabatic shear band formation and matrix cracking which led to flow instability of the composites. The super plastic deformation and dynamic recrystallisation zones correspond to optimum working regions were identified. The fuzzy logic has been implemented in the development of processing maps successfully.Surface-hydrophilic and protein-resistant silicone elastomers prepared by hydrosilylation of vinyl poly(ethylene glycol) on hydrosilanes-poly(dimethylsiloxane) surfacesPoly(ethylene glycol) (PEG) was grafted on the hydrophobic poly(dimethylsiloxane) (PDMS) surface as a hydrophilic outmost monolayer for potential applications such as biomaterials, microfluidics, and anti-fouling coatings. A two component sylgard 184 PDMS (part A of poly(hydrogen methyl siloxane) (PMHS) and part B of poly(methyl vinyl siloxane) (PMVS)) was vulcanized in an off-ratio of, A/B > 1/10, where excess Si–H species exist in the elastomer. An allyl PEG took the hydrosilylation reaction with the surface Si–H groups, catalyzed by the Speier's or Karstedt's catalyst, and resulted a hydrophilic and anti-fouling surface. Atomic Force Microscopy (AFM) observations revealed a well uniform surface with a surface height variation at angstrom scale. Bovine Serum (BS), Bovine Serum Albumin (BSA), Lysozyme (Lys), and Rb IgG/FITC were used to evaluate the protein adsorption on PEG-coated PDMS surface by means of the well-established ATR-FTIR and fluorescence scanner. Results indicated that the PEG monolayer was efficient in reducing the adsorption of proteins.Silicone elastomers have been extensively used in the fields of biomaterials, micro- and nano-technologies such as microfluidic devices To solve the problem, some literatures have reported the grafting of PEG monolayer by silanization technique ). Since PEG is a water-soluble, very flexible, and highly mobile polymer PDMS (Sylgard 184) was from Dow Corning (Midland, MI, USA). Polyethylene glycol mono-methylether (M |
= 1900) was obtained from Alfa Aesar (A Johnson Matthey Company). Allyl bromide and Karstedt's Catalyst (1,3-divivinyl-1,1,3,3-tetramethyldisiloxane Pt(0) complex) were obtained from Aldrich. PEG-200 and PEG-2000 were obtained from China Pharmtec Union (Shanghai, China). Rb IgG/FITC (Peroxidase Conjugated Affinipure Goat Anti-Mouse IgG) was purchased from Beijing Zhongshan Company (Beijing, China), Lys and BSA were obtained from Shanghai Bio Life Science & Technology Compony (Shanghai, China), BS was obtained from the Institute of Medicine and Molecule of Nanjing University (Nanjing, China). Speier's Catalyst (Chloroplatinic Acid) and other reagents were reagent grade and were used without further purification.ATR-FTIR spectra of PDMS films on a wedged germanium crystal were recorded using a Bruker IFS66/S spectrometer with a DTGS detector. The germanium crystal has the configuration of a trapezoid with dimensions of 52.5 mm × 20 mm × 2 mm, which gives 25 reflections for one incidence. All spectra were obtained at 45° angle of incidence for 50 scans with a resolution of 4-cm−1 in the 4000–400-cm−1 region. A freshly cleaned germanium crystal surface was recorded as a reference. Information 1 μm below the PDMS surface can be obtained. Unless specifically stated, all spectra were transferred from ATR unit to absorbance unit. FTIR-TR spectra were recorded on a VECTOR 22 Bruker spectrophotometer with KBr pellets in the 4000–400-cm−1 regions. 1H NMR spectrum was recorded on a DRX500 Bruker spectrometer at 298 K with TMS as an internal reference. Contact Angles (CAs) of the droplets were measured with a Rame-Hart 100R Contact Angle Goniometer. The samples were imaged with a Multimode Nanoscope IIIa AFM (Veeco/Digital Instruments, Santa Barbara, CA) in air under tapping mode (Tip: Silicon MDT; Curvature Radius: <10 nm; Force Constant: 7.5 N/m).According to Williamson's synthesis, under the nitrogen atmosphere, to a solution of PEG-2000 (20 g, 0.010 mol) in 100 mL dry THF was added NaH (0.28 g, 0.012 mol) at 0 °C, the mixture was stired for 1 h. A solution of allyl bromide (1.48 g, 0.012 mol) in 50 mL dry THF was then dripped into the reaction mixture. The resulting mixture was further stirred at room temperature for 48 h. After reaction, centrifuge, the solution was removed and the residue was dissolved by chloroform. Centrifugal separation was run once again, and the solvent was evaporated under reduced pressure. The yield based on PEG-2000: 55%. FTIR-IR: (cm−1) 842, 962, 1066, 1120, 1281, 1343, 1468, 1647, 2884, 2965, 3097, and 3454. 1H NMR δ (CDCl3, ppm): 2.723(1H, HO–), 3.898(2H, HOCH2–), 3.995(2H, HOCH2CH2O–), 3.65[4(n |
− 1) H, (CH2CH2O–)n], 3.904(1H, CH2CHCompound 2 was prepared from Polyethylene glycol monomethylether according to the synthesis of 1. The yield based on polyethylene glycol monomethylether: 90%. FTIR-IR: (cm−1) 847, 963, 1060, 1113, 1280, 1343, 1467, 1619, 1668, 2746, 2889, 2962, 3074, and 3424. 1H NMR δ (CDCl3, ppm): 0.879(3H, CH3O–), 3.543(2H, CH3OCH2–), 3.551(2H, CH3OCH2CH2O–), 3.65[4(n |
− 1) H, (CH2CH2O–)n], 4.020(1H, CH2CHCompound 3 was prepared from the parent PEG-2000 according to the synthesis of 1, a more superfluous NaH was used to ensure complete sodium alcoholate resultant. The yield based on polyethylene glycol 2000: 95%. FTIR-IR: (cm−1) 842, 962, 1060, 1112, 1281, 1343, 1468, 1616, 1637, 1728, 2695, 2887, 2956, 3069, and 3445. 1H NMR δ (CDCl3, ppm): 3.65(CH2CH2O–), 3.949(CH2CHPart A and B of PDMS were carefully mixed in a polyethylene beaker in 1:1, 1:2, 1:3, 1:4, 1:5, 1:7, 1:10, 1:30, respectively, and degassed under vacuum until no visible bubbles were observed. They were then cast on a plastic disc and instantly put into a vacuum oven and vulcanized at 80 °C for 40 min. After they were removed and cooled to room temperature, the PDMS were carefully peeled off. A thin slice of the sample (52.5 mm × 20 mm × 2 mm) was cut to closely match the crystal area on the horizontal ATR accessory and measured with ATR-FTIR. Those results were used to quantify the Si–H concentrations of diverse elastomers.To a solution of 1 (0.4 g, 0.2 mmol) in dry PEG-200 (20 mL) was added Speier's Catalyst (8 mg, 2 × 10−5 |
mol) or Karstedt's catalyst (Pt concentration: 20 ppm), followed a piece of PDMS was immersed, the mixture was sonicated for 0.5 h at 80 °C under the assistant of Speier's catalyst or for 4 h at 45 °C under the assistant of Karstedt's catalyst. The samples were taken out, and sonicated in methanol and water three times for 30 s, respectively, then, dried with nitrogen. Reagents 2 and 3 were treated with the same process. The amount of Si–PEG concentration was quantified by ATR-FTIR.Prior to the adsorption, both modified and unmodified PDMS elastomers were gently rinsed with deionized water. Then, they were immersed in phosphate-buffered saline (PBS) buffer (pH 7.4) of BSA (10 mg/mL), Lys (10 mg/mL) and BS (5%, m/m), respectively. Adsorption was allowed to proceed in an incubator for 2 h. Upon completion of adsorption experiment, the samples were washed three times with PBS buffer and deionized water, respectively, in order to remove non-adsorbed protein and salt. After drying with nitrogen, each sample was measured by ATR analysis.The PDMS samples with or without PEG Monolayers coatings were dipped three drops (3 μL) of PBS buffer (pH 7.4) of Rb IgG/FITC (0.01 mg/mL), and incubated for 2 h in the dark at room temperature. The resulting samples were washed by methanol, PBS buffer and deionized water, respectively. Instantly, the samples were used for scanning fluorescence images. Fluorescence images acquired by a Laser Scanner (LS300, Tecan) were analyzed with software Image J1.30, for each fluorescent image, the average pixel intensity was determined. The net signal was determined by subtraction of the local background from the average intensity.According to the supplier (Dow Corning), Sylgard 184 is normally used at a weight ratio of A:B = 1:10. In that ratio, the two functional Si–H and vinyl groups have equal moles and the crosslinking reaction can be completely finished by vulcanization. When the ratio is more than 1:10, more hydride groups will be present in the mixture. After vulcanization, excess hydride groups remain intact in the elastomer. The Si–H residue can easily be detected by ATR spectroscopy for a sharp characteristic band at 2160 cm−1. To obtain a map of Si–H in the PDMS elastomers with different off-ratios, the ATR spectra of diverse off-ratio PDMS after pre-curing were shown in The strongest band is a double-peak at 1072 and 1027 cm−1, attributed to the stretching vibration peaks of Si–O–Si. The asymmetric and symmetric stretching vibrations of C–H3 occur at 2962 and 2906 cm−1, respectively. A sharp single peak is seen in every spectrum at 1260 cm−1, attributed to the deformation vibration of C–H3 in Si–Me2 group According to Beer's law, the amount of Si–H is proportional to its specific peak area. However, the absolute value of the peak area is affected by many parameters such as the instrument, the sample size, and the vulcanization degree. Since the concentration of SiMe2 in PDMS with different ratios remains nearly constant, its deformation band at 1260 cm−1 is taken as an internal reference. The ratio of the absorbance peak area of Si–H (2160 cm−1) versus SiMe2 (1260 cm−1) in ATR-FTIR spectra is used as a marker for the Si–H concentration (A2160/A1260). describs the relationship of A2160/A1260 versus the weight ratio of A/B., the Si–H concentrations increase nearly linearly with the concentration of part A after 1:10. For the elastomers vulcanized by diverse ratio, when the ratio is less than 1:3, the Si–H residue concentration (below 8.1%) is relatively low, therefore the coating PEG monolayers cannot exhibit good wettability and protein resistance. But when the ratio is more than 1:2, the elastomer becomes fragile. For example, the elastomer of 1:1 is very brittle and its elongation ratio and Young's modulus are also limited records the ATR-FTIR curves of the H-terminated and 1-grafted off-ratio elastomers. Obviously, after PEG was grafted, both Si–H vibration peaks (2160 and 910 cm−1) decrease distinctly. Meanwhile, the characteristic vibration peaks of PEG come up: a new vibration peak of 1335 cm−1 is attributed to the C–H2 bending vibration derived from PEG molecules; its symmetric stretching vibration at 2866 cm−1 shoulders the symmetric stretching vibration peak of C–H3 at 2909 cm−1; its asymmetric stretching vibration at 2953 cm−1 greatly strengthens the asymmetric stretching vibration of C–H3 at 2962 cm−1; a broad vibration peak of 3440 cm−1 appears due to the –OH group of PEG or its H-bonding forming.The integrated intensity of Si–H bands diminished following hydrosilylation, hinting the consumption of the surface Si–H bonds to afford the covalent attachment of PEG. The Si–H stretching band can therefore be used as an index for the reaction progress. Speculating that the consumed Si–H bonds completely convert to Si–C–PEG monolayer, the average conversion efficiency E (%) (E |
= (A0 |
− |
A1)/A0, where A0 and A1 are the integrated peak areas of the native H-terminated PDMS and PEG-modified PDMS in the Si–H region (2160 cm−1), respectively) was rather desired. Three E% values for 1, 2, and 3 are about 85%, 90% and 85%, respectively. Some extended experiments revealed, carrying out the reaction at a higher temperature or for a longer time would yield a greater E value. But those procedures can cause side-effects such as the apparent Pt(0) deposition, thus creating difficulties in cleaning samples.CA measurement is a simple and direct method to characterize the hydrophilicity of samples. From our determinations, the CA of non-coated PDMS elastomer is 104° for water and 73° and for PEG-200, consistent with its high hydrophobicity recorded the CA data of PEG grafting elastomers. Obviously, after PEG was grafted, those surfaces show a degree of hydrophilicity. Comparing the CA data from both catalysts, the monolayers from the Karstedt's catalyst seem relatively dense. Therefore, the corresponding CAs are small. An interesting phenomenon is that the CAs of PEG-grafted surfaces shrank distinctly within several minutes, and they became completely stable after tens of minutes. We propose that the varying CAs probably derive from the rearrangement of the flexible PEG molecules upon exposure to water. The good hydrophilicity of PEG monolayer is generally attributed to the formation of a dense pre-wetting water layer due to H-bonding plotts CA versus PEG density. Expectedly, as PEG density increases, the CA data decrease. It has been reported that the hydrophiliicity obtained by plasma and other modifications would soon diminish because of contamination by adsorbed airborne contaminants such as hydrocarbons or reorganization or the diffusion of the hydrophobic species. But in our experiments, the hydrophilicity can last for a long time.FTIR techniques can directly provide qualitative and quantitative information about deposited protein on polymer surface records the BSA adsorbed curves on 1-grafted surfaces. In , the vibration peaks of Amide I and II always appear at 1660 and 1540 cm−1 synchronously, and their peak areas decrease as PEG densities increase. To evaluate the relative amounts of the adsorbed proteins, the spectra data were normalized using the peak of 1260 cm−1. Those Amide I concentrations (A1660/A1260, the vibration of Amide I locates at 1660 cm−1) of BSA and Lys were shown in Previous literatures have investigated several theoretical approaches to understand protein–polymer interactions ), this datum corresponds to the Si–H concentration of 1:3 elastomer. After 1:3, the protein adsorption levels off. According to protein adsorption model in reference , because the molecular size of Lys (Lys has a molecule weight of 16 k Dalton) is smaller than that of BSA (BSA has a molecule weight of 67 k Dalton), the IR intensity of peptide bands for Lys is less than that for BSA. Comparing , as the wettability increases, the protein-deposited amount decreases. Therefore, the protein-resistant capability goes with their hydrophilicity. compares the protein resistance of 1-, 2-, and 3-grafted surfaces. From it, we can see that all the PEG coating surfaces show a desired protein-resistant capability. The proteins-deposited concentration on the bare elastomer surface was used as an index for evaluating protein-resistant capability, data for 1, 2, and 3 were 17%, 19%, and 28% to BSA, and 19%, 16%, and 35% to Lys, respectively. Clearly, the deposited proteins of 1-grafted surfaces are close to those of 2, and better than those of 3. evaluates the protein-resistance capability of PEG grafting elastomer in BS solution. Clearly, the protein deposited amount on 1-grafted elastomer surface is much less that on H-terminated non-coated elastomer surface. Counted by peak areas, the protein amount of 22% remains after PEG (1) was grafted. Protein resistance was also confirmed by fluorescence images. The images of the IgG/FITC adsorption on the non-coated and 1-grafted PDMS surfaces were recorded in . Counted by software of Image J 1.30, the corresponding fluorescence intensities were listed in . Clearly, IgG/FITC deposited strongly on non-coated PDMS surface, whereas the adsorption on the PEG (1)-grafted PDMS surfaces is nearly negligible, only 4∼10% remains. The protein adsorption experiments illuminate that those PEG monolayers are efficient in rejecting protein adsorption.To elucidate surface topographical change of the elastomer surfaces before and after PEG-coating, 1-, 2-, and 3-grafted surfaces were performed AFM analyses (AFM images were showed in Hydrosilane-PDMS was prepared by a convenient strategy of vulcanizing off-ratio PDMS mixtures, and the resulting Si–H residual of diverse off-ratios was determined by ATR-FTIR. Three PEG species, Polyethylene Glycol Monoallyl Ether, α-Allyl-ω-Methyl-Polyethylene Glycol, and Polyethylene Glycol Diallyl Ether, were successfully grafted on the surfaces of hydrosilane-PDMS through the hydrosilylation reaction under the assistant of Pt catalysts. PEG modifications and protein adsorptions were monitored by ATR-FTIR and Fluorescence Scanner, and the results revealed that the protein-deposited amounts greatly decreased after PEG was grafted. AFM observations revealed the uniform surfaces with the varying heights of 5.2, 1.7, and 4.8 Å after PEG grafting.Supplementary data associated with this article can be found, in the online version, at Design of mechanical properties of a Zn27Al alloy based on microstructure dendritic array spacingThe present work focuses on the influence of the as-cast dendritic microstructure of a ZA27 alloy (Zn–27 wt%Al) on tensile mechanical properties. A low carbon steel chill was used in a unidirectional solidification experimental set-up in order to permit a wide range of dendritic spacings to be obtained along the casting. Experimental results include transient metal/mold heat transfer coefficients (hi), tip growth rate (VL), secondary dendrite arm spacing (λ2), ultimate tensile strength (σU) and yield strength (σy) as a function of solidification conditions imposed by the metal/mold system. Experimental laws relating σU and σy with secondary dendrite arm spacing are proposed. It was found that both tensile properties increase with decreasing λ2. A predictive theoretical dendritic growth model has been compared with the present experimental observations. Expressions correlating tensile properties, dendritic spacing and solidification thermal variables have been established. Such expressions permit the control of as-cast microstructures by manipulating casting variables, such as the cooling rate and the tip growth rate and can be used as an alternative way to design mechanical properties.To address the increasing demand for high performance high quality die castings, a class of zinc based engineering cast alloys have been developed, in particular for applications in the automotive industry. Three members of this family of alloys, generally identified industry-wide as ZA-8, ZA-12 and ZA-27, have been shown to have good physical, mechanical and tribological properties and are commonly used as foundry alloys in a variety of applications The effect of microstructure on metallic alloys properties has been highlighted in various studies and particularly, the influence of grain size and dendrite arm spacing upon the mechanical properties and corrosion resistance has been reported For cast metals, however, it is not always true that the strength improves with decreasing grain size. Strength will increase with grain size reduction only if the production of small grains does not increase the amount of microporosity, the percentage volume of second phase or the dendrite spacing It is well known that there is a close correlation between thermal variables and the solidification structure and as a direct consequence, morphological structure parameters such as grain size and dendritic arm spacing also depend on heat transfer conditions imposed by the metal/mould system. Thus, the control of solidification thermal variables such as tip growth rate (VL), thermal gradient (GL), cooling rate (T˙), and local solidification time (tSL) permits a range of microstructures to be obtained The present work focuses on the influence of heat transfer solidification variables on the microstructural formation of Zn–27 wt%Al alloy castings and on the development of correlations between dendritic spacing and mechanical properties. Experimental results include transient metal/mould heat transfer coefficient (hi), secondary dendrite arm spacings (λ2), ultimate tensile strength (σU) and yield strength (σy).Heat flow across the metal casting/mold interface can be characterized by a macroscopic average metal/mold interfacial heat transfer coefficient (hi), given by:where q [W] is the average heat flux, A [m2] is the area and TIC and TIM are casting and mold surface temperatures [K] at the interface. It is well known, that during the solidification process, the mold gradually expands due to heat absorption, and the metal casting is subjected to shrinkage. As a result, a gap develops due to insufficient contact between metal and mold, and as a direct consequence, hi decreases rapidly. In previous articles the transient interfacial heat transfer coefficient has been successfully characterized by using an approach based on measured temperatures in casting and numerical simulations provided by a heat transfer solidification model where hi [Wm−2 |
K−1], t is the time [s] and Ci and n are constants which depend on alloy composition, chill material and melt superheat.Thermal solidification variables during solidification of binary alloys can be analytically described as a function of metal/mold parameters and casting operational conditions and consequently, as a function of metal/mold interfacial heat transfer coefficient (hi) gives a typical example of tip growth rate (VL), as a function of hi. However, other thermal variables such as temperature gradient (GL), tip cooling rate (T˙) |
and local solidification time (tSL) may also be expressed as a function of metal/mold solidification parameters, as described in previous articles VL=2αSLϕ222kSϕ2(TSol-T0)nπ(TLiq-T0)exp(ϕ12)[M+erf(ϕ1)]hi+SL,where αSL is the thermal diffusivity of mushy zone, ϕ1 and ϕ2 are solidification constants It is well established that under most conditions of solidification the dendritic morphology is the dominant characteristic of the microstructure of metallic alloys. The dendritic array is characterized by interdendritic spacings which are recognized to have a significant influence on tensile properties. Several theoretical models have been proposed in the literature to describe the dependence of primary and secondary dendrite arm spacings on solidification variables such as initial alloy composition, growth rate and thermal gradient. Bouchard and Kirkaldy have established a compendium of steady and unsteady state formulations for such spacings where a2 is the secondary dendrite arm-calibrating factor, which depends on the alloy composition, Γ is the Gibbs–Thomson coefficient, DL is the solute chemical diffusivity in the liquid, k0 is the partition coefficient, C0 is the alloy composition and TF is the solvent fusion temperature. The secondary spacing given by Eq. refers to that of the initial dendritic growth. The calibrating factor a2 is incorporated to take into account among other uncertainties, a ripening correction for secondary spacings.a shows the casting assembly used in solidification experiments. The main design criterion was to ensure a dominant horizontal and unidirectional heat flow during solidification. This objective was achieved by adequate insulation of the chill casting chamber. A low carbon steel chill was used at a normal environment temperature of about 25 °C (initial mold temperature), with the heat-extracting surface being polished. Experiments were performed with a Zn–27 wt%Al alloy which was melted in an electric resistance-type furnace, degassed and then poured into the casting chamber with a melt pouring temperature of about 10 pct above the liquidus temperature. This alloy was prepared by using commercially pure metals, as described in . The thermophysical properties of such alloy, which were used to run the necessary simulations, are shown in Temperatures in both metal and mold were monitored during solidification using a bank of six type J thermocouples accurately located with respect to the metal/mold interface (01 located inside the mold and 05 along the casting at different positions from the metal/mold interface). All the thermocouples were connected by coaxial cables to a data logger interfaced with a computer, and the temperature data were acquired automatically. b exhibits location of specimens that were taken for optical metallographic examination and tensile testing, the latter according to specifications of ASTM Standard E 8 M. Longitudinal specimens were obtained from the solidified casting with the selected section being polished and etched to reveal the microstructure. The etchant was a solution of 0.5% HF in water. An image processing system was then used to measure the secondary dendrite arm spacing (20 measurements for each selected position from the casting surface).Experimentally monitored temperatures at one position in metal and another in mold, were compared with theoretical predictions of a finite difference solidification model to determine the transient metal/mold heat transfer coefficient (hi) shows typical experimental thermal responses compared to those numerically simulated and the resulting hi coefficient profiles as a function of time.The results of experimental thermal analysis inside the casting have also been used to determine the tip growth rate (VL), as a function of time and/or position. shows a comparison between the experimental and calculated tip growth rate as a function of position from the metal/mold interface. The calculated tip growth rate was obtained by using the analytical expression given by Eq. . In this equation, the appropriate experimental values of hi, given in Discrepancies between experimental and calculated tip growth rate are mainly caused by uncertainties in the thermophysical properties and the presence of convection currents in the liquid metal induced by fluid motion during pouring which were not considered by the analytical model. exhibits typical Zn–27wt%Al microstructures at different locations from the metal/mold interface. The structure immediately after solidification will be formed by an Al-rich dendritic matrix involved by an interdendritic eutectic mixture. During subsequent cooling a eutectoid decomposition takes place shows the measured secondary dendrite arm spacing (λ2) expressed as a function of distance from the casting surface. It can be observed that, as expected, λ2 increases with distance from casting surface due to the corresponding decrease in cooling rate. illustrates the calculated and measured secondary dendrite arm spacing as a function of tip growth rate. The theoretical approach was that due to Bouchard and Kirkaldy . They suggest a calibrating factor a2, ranging between 4 and 11 for a number of metallic binary alloys, but they do not recommend a specific value for Zn–Al alloys. In our experimental investigation a calibrating factor of 11 appears to be appropriated for the alloy examined, as indicated by the good agreement observed in between measured and calculated secondary dendrite arm spacing. The analytical expression for VL, expressed by Eq. in order to establish a general formula permitting the secondary spacing to be expressed as a function of transient solidification variables. shows the experimental results of ultimate tensile strength (σU) and yield strength (0.2% proof stress) (σy) as a function of secondary dendrite arm spacing (λ2). It can be seen that both tensile properties increase with decreasing secondary dendritic arm spacing. The Zn–27wt%Al alloy exhibits good values of σU and σY for relatively high strenght aplications. For example, in a range of about λ2 |
= 60 μm, σU and σy values are 300 MPa and 260 MPa, respectively. can incorporate models expressing λ2 as a function of thermal solidification variables (such as Bouchard–Kirkaldy validated in ) and metal/mold heat transfer coefficient permitting expressions correlating mechanical properties with solidification conditions to be established. Additionally, if the analytical expression describing VL (Eq. ) is also inserted either into the theoretical (Eq. ) or the experimental law for λ2, both σU and σy can be correlated with solidification conditions. By using the experimental laws obtained, i.e., λ2=57VL-2/3, σU=114+1500λ2-0.5 and σy=65+1650λ2-0.5 and the analytical expression given by Eq. , the following expressions can be established:σU=114+198.72αSLϕ222kSϕ2(TSol-T0)nπ(TLiq-T0)exp(ϕ12)[M+erf(ϕ1)]hi+SL1/3σy=65+218.52αSLϕ222kSϕ2(TSol-T0)nπ(TLiq-T0)exp(ϕ12)[M+erf(ϕ1)]hi+SL1/3 permit the control of as-cast microstructures by manipulating casting variables, such as mold type (M), hi, initial melt temperature, metal and mold thermophysical properties, and can be used as an alternative way to design mechanical properties.In order to investigate the role of secondary dendrite arm spacing on mechanical properties of a Zn–27wt%Al alloy, solidification experiments and tensile tests were carried out. The following main conclusions can be drawn from the present experimental investigation:The experimental expressions correlating the ultimate tensile strength and yield strength with secondary dendrite arm spacing for a Zn–27wt%Al alloy have shown that both tensile properties increase with decreasing dendrite spacing.The control of as-cast microstructures, by manipulating casting processing variables, such as the cooling rate and the tip growth rate can be used as an alternative way to design mechanical properties.Equation of state for technetium from X‐ray diffraction and first-principle calculationsThe ambient temperature equation of state (EoS) of technetium metal has been measured by X-ray diffraction. The metal was compressed using a diamond anvil cell and using a 4:1 methanol-ethanol pressure transmitting medium. The maximum pressure achieved, as determined from the gold pressureEquation of state for technetium from X-ray diffraction and first-principle calculations scale, was 67 GPa. The compression data shows that the HCP phase of technetium is stable up to 67 GPa. The compression curve of technetium was also calculated using first-principles total-energy calculations. Utilizing a number of fitting strategies to compare the experimental and theoretical data it is determined that the Vinet equation of state with an ambient isothermal bulk modulus of B0T=288 GPa and a first pressure derivative of B′=5.9(2) best represent the compression behavior of technetium metal.Technetium, the lightest radioelement, exhibits a hexagonal-closed-packed (HCP) structure as a pure element. Nuclear fission reactors produce considerable quantities of 99Tc, which has one of the largest fission yields (~6%) as well as a long half-life (t1/2=213 000 years, β−=293 keV). Technetium chemistry has been explored for applications in nuclear waste and reprocessing since the 1960s, but a number of fundamental properties of the element remain poorly studied. The phase diagram of metallic technetium is one of the least explored for any transition metal with minimal investigations establishing the structure and properties of technetium metal at high pressures. Accordingly, understanding the pressure and temperature dependence of thermal, mechanical, and electronic properties of technetium is of significance to nuclear waste storage and condensed-matter physics.Owing to the challenges of studying radioactive elements, previous measurements of pressure-dependent properties of technetium are sparse and are summarized here. In 1955, using an opposed anvil geometry, Bridgman observed the resistance and shear characteristics of technetium up to 10 GPa and found the hcp phase stable with no irregularities , with no reference to how these values or other properties vary under high pressures. In 1989, Guillermet Here we report room-temperature synchrotron diffraction measurements up to 67 GPa and establish an EoS derived from non-hydrostatic high-pressure diffraction experiments. In addition, first-principles total-energy calculations have been carried out to determine the structural properties and compressibility of bulk technetium up to 273 GPa. Our objective is to determine an EoS of technetium and create a baseline for further high pressure studies of technetium metal and relevant systems.Technetium-99 is a weak beta emitter (Emax=293 keV). All manipulations were performed in a radiochemistry laboratory at UNLV designed for chemical synthesis using efficient HEPA-filtered fume hoods, and following locally approved radioisotope handling and monitoring procedures.Technetium-99 metal was synthesized as polycrystalline aggregates in the radiochemistry laboratory at UNLV as previously described Angular-dispersive X-ray powder diffraction data of technetium and gold in a diamond-anvil cell was obtained at beamlines16-ID-B and 16-BM-D of the High Pressure Collaborative Access Team (HPCAT) at the Advanced Photon Source (APS). In order to prevent dispersal of radioactive material in the case of diamond failure, mylar windows were added to the DAC containment. At each beamline, diffraction patterns were collected with a MAR345 image plate with incident monochromatic beams of 29.200 keV and 33.169 keV X-rays. The sample to detector distances were calibrated with CeO2 (SRM 674b), then the two-dimensional images were integrated and corrected for distortions using FIT2D First-principles total-energy calculations were performed using spin-polarized density functional theory, as implemented in the Vienna Ab initio Simulation Package (VASP) The interaction between valence electrons and ionic cores was described by the Projector Augmented Wave (PAW) method Structural optimization was carried out using the Monkhorst–Pack special k–point scheme Representative diffraction patterns measured over the pressure range of this work are shown in . The mylar windows contributed three reflections to each diffraction pattern at 2θ=6.4°, 8.6°, 10.9°; these reflections were excluded from the final refinements. Eight reflections belonging to the sample were identified in the region of 2θ=10–30°. At each pressure, seven peaks were observed which index to gold. The sample pressure was determined using the EoS parameters recommended by Takemura and Dewaele, A 4:1 mixture of methanol–ethanol is commonly used as a quasi-hydrostatic medium and is reported to provide hydrostatic conditions up to 10.5 GPa shows the evolution of the micro-stress as a function of pressure. Below 9 GPa, when the medium is a fluid, there is no significant change in micro-stress. Once the medium is solid, the stress profile displays three domains: strongly increasing up to 22 GPa, invariant from 22 to 40 GPa and then gradually increasing beyond 40 GPa. The considerable amount of micro-stress, at these pressures introduces uncertainty into lattice parameters and consequently the equation of state.In light of the variation of micro-stress with pressure, two pressure ranges were considered in determining the EoS, 0–10 GPa and 0–67 GPa. The first range is in the hydrostatic regime for this pressure medium and the second is the full pressure range of this study. Using the software package EoSfit7GUI The EoS parameters for the hydrostatic pressure range refined using the second order Vinet equation are V0=14.301(4) Å3/atom, B0=333(8) GPa and for Birch-Murnaghan equation are V0=14.305(4) Å3/atom, B0=321(8) GPa. The refined EoS parameters are V0=14.307(7) Å3/atom B0=310(20) GPa, B′0=6(6) for both the third order Vinet and Birch-Murnaghan equations. The data from the full pressure range was fit with all parameters free obtaining V0=14.310(4) Å3/atom B0=309(7) GPa and B′0=4.7(4) for the Vinet equation and V0=14.309(4) Å3/atom, B0=309(7) GPa and B′0=4.6(4) for the Birch–Murnaghan equation. For the combined X-ray and Ultrasonic analysis, the isothermal bulk modulus was fixed at B0=288 GPa, the converted value from the adiabatic bulk moduli determined from the ultrasonic measurement The calculated volumes of technetium in hcp packing show a continuous decrease up to 273 GPa. The choice in functional has been previously demonstrated to work reliably for technetium-containing structures EoS parameters are compared by pressure range, and EoS type in . There is no appreciable difference between the quality of the fit or the resulting parameters between the two EoS forms except in the case of the 2nd order equations in the hydrostatic domain. Hydrostatic data is preferred for EoS determination, but the total compression in this range is only ~3%, which is inadequate to define the parameters of a 3rd order EoS. The hydrostatic data fits an unreasonably stiff model at high pressures, b. Finally, the difference between observed and predicted pressure shows a random distribution as can be seen in the difference plot of c. As expected the spread in the data under hydrostatic conditions is considerably lower than that at higher pressures. At low pressures the difference is less than 0.4 GPa while over the entire pressure range the spread is four times greater.There are a number of factors that need to be taken into consideration when evaluating our equation of state for technetium. First, analysis of the two pressure ranges offer different interpretations for the compressibility of technetium with the hydrostatic range indicating that the bulk modulus is greater than 320 GPa while the nonhydrostatic data suggest that it is less than 310 GPa. While a larger pressure range alone does not automatically improve the equation of state, it is necessary to achieve a sufficiently large compression in order to estimate the bulk modulus and begin to refine higher order parameters. The pressure range achieved in the diffraction experiment under hydrostatic conditions resulted in a compression of ~3% while the maximum compression for the experiment was ~15%. The conflict between the need for larger compression and hydrostatic conditions is a challenge for incompressible material such as technetium.Over the years, bulk moduli of most elements, including technetium, have been repeatedly revised, reflecting the challenges associated with these measurements. The selection listed in is representative of recent work on technetium. The consensus is that the bulk modulus is just under 300 GPa. In this study a compression curve was produced from DFT methods, which when fit to an equation of state agrees well with the diffraction data. Previously, using the same GGA/PW91 methodology, the bulk modulus was reported to be 298 GPa At ambient conditions, rhenium and technetium are isostructural and rhenium is commonly used as a nonradioactive technetium homolog that can provide valuable insights on the behavior of technetium. The equation of state of rhenium has been extensively studied by X-ray diffraction up to the highest achievable static pressures An accurate determination of the bulk modulus of technetium will solidify our understanding of the trends describing the 4d transition metals. The bulk moduli of the transition metals are plotted against the number of d electrons in . Manganese, 3d5, represents a minimum in the bulk moduli. This minimum is not seen in rhenium for the 5d metals and is suggested to be at technetium for the 4d metals. Theoretical studies of the transition metals An analysis of the equation-of-state for technetium has been performed from X-ray diffraction data, ultrasonic measurements and DFT calculations. The study produced three equations of state to consider for technetium which agree very well in the experimental pressure range. The EoS refined from only the diffraction data yielded the best match to theory but the fit using the fixed bulk moduli give the best representation of the physical behavior of technetium metal, Bo=288 GPa and B′o=5.9(2). We have confirmed that the HCP phase is stable up to 67 GPa. Nonhydrostatic behavior is observed at high pressures but did not prevent a quality fit from being obtained for the equation of state as is evident from the low average difference in the calculated pressure. This equation of state is considered a reliable description of technetium compression behavior within the bounds reported here. To further our understanding of the pressure-dependent behavior of technetium and periodic trends of transition metals it would be beneficial to investigate the effects of temperature and pressure on the structural properties of technetium under hydrostatic conditions up to higher pressures.Development of polyoxadiazole nanocomposites for high temperature polymer electrolyte membrane fuel cellsNovel nanocomposite membranes were prepared with sulfonated polyoxadiazole and different amounts of sulfonated dense and mesoporous (MCM-41) silica particles. It has been shown that particle size and functionality of sulfonated silica particles play an important role when they are used as fillers for the development of polymer electrolyte nanocomposite membrane for fuel cells. No significant particle agglomerates were observed in all nanocomposite membranes prepared with sulfonated dense silica particles, as analyzed by SEM, AFM, TGA, DMTA and tensile tests. The Tg values of the composite membranes increased with addition of sulfonated silica, indicating an interaction between the sulfonic acid groups of the silica and the polyoxadiazole. Constrained polymer chains in the vicinity of the inorganic particles were confirmed by the reduction of the relative peak height of tan |
δ. A proton conductivity of 0.034 S cm−1 at 120 °C and 25% RH, which is around two-fold higher than the value of the pristine polymer membrane was obtained.Incorporation of inorganic particles or fillers into polymeric materials has been the subject of growing interest in different technological areas such as fuel cells Nanocomposites can offer advantages when dealing with the following fuel cell key issues: optimization of the membrane–electrode–catalyst interface, preparation of membranes able to effectively operate above 100 °C and under external low humidification in fuel cells fed with hydrogen and preparation of membranes with low alcohol crossover When the SO3H-MCM-41 particles were used as fillers in a polysiloxane-based membrane, similar to the pristine polymeric matrix, the composite membrane containing 9.6 wt.% SO3H-MCM-41 proton conductivity values significantly dropped at medium temperatures (110–130 °C) and low humidity (17%) showing very low proton conductivity (10−11 to 10−10 |
S cm−1) Dicarboxylic acid 4,4′-diphenylether (DPE, 99%, Aldrich), dimethyl sulfoxide (DMSO, >99%, Aldrich), hydrazine sulfate (HS, >99%, Aldrich), sodium hydroxide (NaOH, 99%, Vetec), poly(phosphoric acid) (PPA, 115% H3PO4, Aldrich), phosphoric acid (85%, Aldrich). All chemicals were used as received.Functionalized silica and the mesoporous silica were synthesized and characterized as described elsewhere The synthesis condition has been selected considering previously reported synthesis method for sulfonated polyoxadiazoles with high molecular weight and high sulfonation level (S/C 0.103) Homogeneous membranes were cast from solutions with a polymer concentration of 4 wt.% in DMSO. After casting, the DMSO was evaporated in a vacuum oven at 60 °C for 24 h. For further residual solvent removal, the membranes were immersed in water bath at 60 °C for 48 h and dried in a vacuum oven at 60 °C for 24 h. The final thickness of the membranes was about 50 μm. Nanocomposite membranes were prepared by adding 2.5–10 wt.% of functionalized filler (based on polymer content) into the 4 wt.% polymeric solution. The solution was stirred for 6 h and cast on a glass plate at 60 °C for solvent evaporation and dried following the same protocol described for the membranes prepared only with the polymer. The final thickness of the membranes was in the range 50–70 μm. The sulfonated membranes were converted into its acid form by immersing the cast membranes in 1.6 M H3PO4 at room temperature for 24 h, followed by immersion in water for 2 × 24 h to ensure total leaching of residual phosphoric acid.The membrane morphology was observed by scanning electron microscopy (SEM) type LEO 1550VP. The samples were previously coated with gold in a sputtering device. Atomic force microscopy (AFM) of samples was done by using a Multi Mode Scanning Probe Microscope Model with a nanoscope IV controller by Digital Instruments Inc. (Veeco Metrology Group). AFM observations were carried out in air at ambient conditions (25 °C) using tapping mode probes with constant amplitude. Micrographs in topographic and phase contrast modes were obtained.Thermogravimetric analysis (TGA) experiment was carried out in a Netzsch 209 TG, equipped with a TASC 414/3 thermal analysis controller. The film sample, under nitrogen atmosphere, was heated from 100 to 700 °C at 10 °C/min. Dynamic mechanical thermal analysis (DMTA) was used for determination of glass transition temperature (Tg), storage modulus (E′), loss modulus (E″) and loss tangent (tan |
δ). DMTA was performed using a TA instrument RSA 2 with a film tension mode at a frequency of 1 Hz and 0.1N initial static force. The temperature was varied from 25 to 500 °C at a heating rate of 2 °C/min and at a constant strain of 0.05%. Tensile testes were conducted on Zwick–Roell equipment according to the ASTM D882- 00 and operating at a cross-head speed of 5 mm min−1 at room temperature. The reported values correspond to an average of five specimens.Ionic conductivity was measured in the frequency range 1–106 |
Hz with an oscillating voltage of 100 mV. The pellets of 8 mm in diameter and 0.5–1 mm in thickness were prepared by pressing the sulfonated silicas. The pellets were placed between two thin graphite slices and then put into a PTFE specimen holder where the pellet is clamped between two sintered metal electrodes. Measurements were performed with a flow cell purged with wet gas; the relative humidity was controlled by setting the temperature of a water reservoir heated at suitable temperature between 45 and 140 °C.Proton conductivity was measured by the AC impedance spectroscopy in the frequency range 10–106 |
Hz at signal amplitude ≤100 mV and obtained from the impedance modulus at zero phase shift (high frequency side) with 5–25% of relative humidity. Measurements were performed with a flow cell purged with wet nitrogen; the relative humidity was controlled by bubbling nitrogen gas in water heated at a suitable temperature between 49 and 81 °C. The impedance measurements were carried out on stacks containing up to five membranes (cumulative thickness around 400 μm). The spectrometer used was a Zahner IM6 electrochemical workstation.For the preparation of the nanocomposite polyoxadiazole membranes four different fillers were used, being three of them constituted of dense sulfonated silica particles and one of mesoporous sulfonated silica particles. The structures of the sulfonated silicas are depicted in . Because of the size (1–5 μm) of the SO3H-MCM-41 particles, microcomposite polyoxadiazole membranes could be prepared only. An interesting aspect of the use of the SO3H-MCM-41 particle is its nanoporous structure and the presence of functionalized groups within the porous, as characterized elsewhere . A proposed attaching of the SiO2OHC3H6–SO3H group onto the pore wall of Si-MCM-41 is shown in The dense sulfonated silica particles have been used as fillers in two different polymeric matrices, sulfonated polyetherketone shows the ionic conductivities obtained for the pellets of sulfonated silicas measured under different relative humidity conditions. Data for the SO3H-MCM-41 As expected, the conductivity of the fillers has a strong dependence on the level of hydration and temperature. The high conductivity values at RH = 100% (up to 10−2 |
S cm−1) sharply decrease upon dehydration (up to 10−5 |
S cm−1). The higher conductivity obtained for the silica-fluoropropane in all conditions may be attributed to its acidity. The electronegative fluorine atoms probably have significant contribution to the formation of hydrogen bonds with water. The low proton conductivity values obtained for the SO3H-MCM-41 at low relative humidities probably results from the non-uniform distribution of sulfonated ions in the pores (The ionic conductivity data at 120 °C and under 5–25% RH of the composite membranes containing 5 wt.% of fillers are shown in . To analyze the effect of each filler without the influence of eventual residual phosphoric acid used for converting the sulfonated membrane into its acid form, the conductivity of the membranes in salt form was measured. All composite membranes had higher proton conductivity than the pristine polymer membrane in all range of relative humidity. The significant difference between conductivity data of composite membranes and the pristine polymer membrane at low relativity humidity (e.g. 15% RH) is probably a consequence of the better water retention capacity of the composite membranes conferred by the sulfonated silicas. Considering the error bar for the measurements, no significant difference among the fillers could be observed, except for the nanocomposite prepared with silica-telechelic. Similar to the systems based on sulfonated poly(ether ether ketone) ) before the incorporation in the membrane, this result supports once more that the amphoteric character of the sulfonated telechelic attached to the surface of the silica containing both basic nitrogen sites as well as acid sulfonic acids favor additional points for proton jumps.The lower influence of conductivity on the relative humidity for the nanocomposite membrane prepared with silica-telechelic could be attributed to a favored Grotthuss mechanism by the amphoteric telechelic structure. The high values obtained under low relative humidity suggest a close distance between the –SO3H groups as well as the presence of additional sites for protonation, strengthening a mechanism of proton transport by the diffusion of protons within hydrogen bonded structure All nanocomposite membranes prepared with sulfonated dense silica particles were transparent, indicating the absence of significant particles agglomerates. On the other hand, because of the large particle sizes of the sulfonated mesoporous silica particles, SO3H-MCM-41, which are around 1–5 μm, turbid microcomposite membranes could be prepared only. The homogeneity of all composite membranes was confirmed by the SEM images. As shown in , well homogeneous nanocomposite membranes without particle agglomerates were prepared with sulfonated dense silica particles. A homogeneous distribution of SO3H-MCM-41 can also be observed even though the particles are quite large.AFM observations were used to confirm that the sulfonated dense silica particles are homogeneously distributed. In the phase contrast mode, the effect of the topography is much smaller and the influence of other parameters like the difference in hardness of inorganic particle and organic polymer matrices can be clearly identified. In this case, the inorganic particles can be differentiated and assigned to the bright areas. The analysis of the AFM images () of the nanocomposite membranes containing sulfonated silicas confirms the particle nanosizes in the range 10–15 nm for the nanocomposite containing 5 wt.% silica-fluoropropane (a), 20–67 nm for the nanocomposite containing 5 wt.% silica-telechelic (b) and 20–75 nm for the nanocomposite containing 10 wt.% silica-telechelic (c), respectively. By increasing the filler concentration the occurrence of agglomerates increases.The particle size has an effect on the mechanical properties of the composite membranes. Mechanical properties of the nanocomposite membranes prepared with dense sulfonated silicas were kept unchanged, while for the microcomposite membrane prepared with 5–10 wt.% SO3H-MCM-41 the mechanical properties slightly decreased (). The Tg values of the composite membranes slightly increased with addition of sulfonated silica, which might be caused by the hydrogen bonding between the sulfonic acid groups of the silica and the polyoxadiazole. Interaction between the sulfonated silicas and the polyoxadiazole results in constrained polymer chains in the vicinity of the inorganic particles. The depression in tan |
δ indicates the reduction of chain mobility during the glass transition clearly shows the reduction of the relative peak height of tan |
δ and the increase of the Tg values with increasing amount of sulfonated silicas. This result supports once more the very well dispersed silica particles within the polymeric matrix. On the other hand, for 10 wt.% SO3H-MCM-41 aggregation of the large size particles leads to a decrease of mechanical properties as well as insignificant change in the Tg value (Thermal stability of composite membranes prepared with 2.5–10 wt.% of silica-telechelic and SO3H-MCM-41 particles was analyzed by TGA. As shown in , the pristine polyoxadiazole membrane has two distinct regions of weight loss. The first occurs between 250 and 370 °C with an approximate weight loss of 4% due to decomposition of sulfonic acid groups. The second region starting at 469 °C is associated with the loss of volatiles caused by the degradation of the polyoxadiazole. These two weight loss temperatures slightly shifted toward higher temperatures with addition up 5 wt.% loading of silica-telechelic, while for 10 wt.% loading of silica-telechelic they remained unchanged. The addition of 2.5 wt.% of SO3H-MCM-41 did not alter the degradation pattern of the pristine polyoxadiazole membrane. On the other hand, with 5 and 10 wt.% loading of SO3H-MCM-41 the first region of weight loss shifted toward lower temperature, probably because of the lower interaction between the sulfonic acid groups of the large mesoporous silica particles and of the polyoxadiazole chains. As the sulfonic acid groups are more located inside the pores Because of the better results observed for the nanocomposite membrane containing silica-telechelic, these membranes were converted into acid form and their proton conductivity values were measured as shown in . In the acid form, proton conductivity value of the nanocomposite membrane containing 5 wt.% silica-telechelic is only slightly higher than the observed for the pristine polymer membrane as a result of the already very high proton conductivity of the polymer (10−2 |
S cm−1), which exceeds by more than three orders of magnitude of the silica-telechelic (). No appreciable difference in proton conductivity values for the membrane containing 2.5–5 wt.% silica-telechelic was observed. Combined with the mechanical tests, this result indicates insignificant difference in the membrane properties for this loading range. On the other hand, for the membrane containing 10 wt.% silica-telechelic proton conductivity markedly increases. The proton conductivity was 34 mS cm−1, much higher than 17–21 mS cm−1, the value for the pristine polymer membrane. This increase should be expected taking into account that the number of hydrophilic sulfonic acid groups introduced along with the functionalized silica also increased, while simultaneously a good dispersion of silica-particles was also guaranteed. Based on this result, 10 wt.% of silica-telechelic is considered here within the studied concentration range the optimal loading for this system.Particle size and functionality of sulfonated silica particles play an important role when they are used as fillers for the development of polymer electrolyte nanocomposite membrane for fuel cells. Though high conductivity values are obtained for the sulfonated silica particles containing fluorine atoms, sulfonated silica particles containing both basic nitrogen sites as well as acid sulfonic acids lead to a significant improvement of nanocomposite polyoxadiazole conductivity when inserted into the polymeric matrix. All nanocomposite membranes prepared with sulfonated dense silica particles were transparent, indicating the absence of significant particles agglomerates. On the other hand, because of the particle sizes of the sulfonated mesoporous silica particles SO3H-MCM-41 particles (1–5 μm), turbid microcomposite membranes could be prepared only. Mechanical properties of the nanocomposite membranes prepared with dense sulfonated silicas were kept unchanged, while for the microcomposite membrane prepared with 5–10 wt.% SO3H-MCM-41 the mechanical properties slightly decreased. The Tg values of the composite membranes increased with addition of sulfonated silica, indicating interaction between the sulfonic acid groups of the silica and the polyoxadiazole. Constrained polymer chains in the vicinity of the inorganic particles were observed for both nanocomposite prepared with sulfonated silica particles and with mesoporous silica particles by the reduction of the relative peak height of tan |
δ. A proton conductivity of 0.034 S cm−1 at 120 °C and 25% RH, which is around two-fold higher than the value of the pristine polymer membrane was obtained.Gas detonation forming process and modeling for efficient spring-back predictionAluminum cylindrical cups are formed with gas detonation forming technology and finite element modeling of aluminum cylindrical cup production with the detonation forming technology is performed. The forming process simulation is carried out in two-step analysis. 2D and 3D computational models are constructed with both explicit and implicit dynamic analyses are performed. The effects of detonation pressure and die design parameters also are investigated. The thickness distribution and deformed geometry of the cups are compared with the experimentally determined values. The spring-back predictions based on the explicit and implicit methods are criticized in terms of deformed shape accuracy and elapsed time.The gas detonation forming (GDF) of cylindrical aluminum cups is a dynamic production process based on the pressure energy produced instantaneously by the shock waves inside a combustion chamber The FE computations were based on both two-dimensional (2D) implicit dynamics and three-dimensional (3D) explicit dynamics formulations In the experiments and FE simulations, the geometry of die and blank holder are kept fixed, and the detonation pressure is changed to deform the aluminum blank of 0.5 mm thickness and 50 mm diameter. The optimum detonation pressure is identified as the maximum process pressure by which the aluminum cup is produced free of defects.The calculated cup deformations, thickness distributions and radial total strains are compared with the experimentally obtained values. The relative differences found between the experiment and simulation results are discussed.The 2D implicit dynamic FE analyses are carried out using the Ansys software and the implementation of the Lemaitre–Chaboche material model is performed using the user-programmable material subroutine of the software The GDF system is specially manufactured for the present work. Briefly it consists of five main parts; explosion chamber, vacuum pumps, velocity and pressure measurement assembly, gas control system, and ignition system, as shown . The explosion chamber is made of special seamless steel pipes and has two parts The detonation forming is a transient dynamic deformation process and the process simulation is the finite element solution of dynamic equilibrium equations , is measured in the chamber using a quartz crystal sensor with a sampling frequency of 500 kHz for 10 ms and contains high frequencies whose effects may be negligible and requires extremely fine mesh for the blank to account the correct material response. To eliminate the high frequency-noise effects, the measured pressure signal is filtered with a low pass filter with a cut-off frequency of 5000 Hz.The other assumptions are related to modeling of the friction and structural damping in the simulation model. Due to the difficulties in describing friction forces between the blank and the die surfaces, no-stick condition is assumed and the Coulomb friction model is used in 2D and 3D computations. The proportional structural damping is used to damp out the vibrations for spring-back analyses and dissipates the kinetic energy due to the high frequency noise during deformation. The mass and stiffness proportional damping coefficients are calculated using fast Fourier transform (FFT) analysis of the measured pressure signal. The mass matrix multiplier is calculated using the lowest frequency, 98 Hz, in the FFT plot of pressure signal in . In the calculation of stiffness matrix multiplier, a 10% of critical damping is sought in the frequency band of 100–5000 Hz.The 2D FE model is constructed based on the axisymmetry assumption for the forming process. The schematic of the axisymmetric model is shown in . In the solution of dynamic equilibrium equations, the Newmark time integration method is used FE solution at successive discrete time points. An amplitude decay factor of 0.005 is applied for all computations.The aluminum blank is modeled with 900 axisymmetric linear isoparametric selective-reduced integration elements The isotropic linear work hardening and Lemaitre–Chaboche nonlinear work hardening models were used for the deformation of the aluminum blank during detonation forming and spring-back deformation. The strain rate effects were not included for both models. The mechanical properties of the blank, used for isotropic linear work hardening material model are, given in The three additional material properties for Lemaitre–Chaboche plasticity model are k, C and γ. For the identification of these parameters, the approximation is done by assumption in which the stabilized cyclic stress–strain curve was replaced with the equivalent stress-strain curve determined by monotonic uniaxial loading. The parameter identification procedure is given in A 3D simulation model including the aluminum blank, die and blank holder was constructed and the FE solution of dynamic equilibrium equations was performed using the explicit time integration method. The details of the explicit dynamic formulations used in the computations are given in about the axis of symmetry. The die and blank holder surfaces are modeled as rigid segments. To represent the die profile geometry accurately and to reduce the artificial friction force at the segment vertex, 15 equally spaced line segments are placed over the 90° section. The blank is modeled with four-noded selectively reduced integration co-rotational shell element with five through-thickness integration points. The location of through-thickness integration points and the weight factors are specified according to trapezodial integration rules given in The 3D finite element model is composed of 14 560 shell and 3432 contact segments. The mesh layout is given in . The only boundary condition is the detonation pressure history acting on to the aluminum blank. The contact between the blank and rigid die surfaces is defined with the penalty stiffness method. Mesh connectivity search algorithm is used for computational efficiency.The isotropic linear work hardening is used for the material description of aluminum blank with the specified parameters in . The strain rate effects are incorporated into the computations with the Johnson and Cook material model The detonation forming process produces high bending stresses on the aluminum blank after the time at which the peak detonation pressure is applied, the lower amplitude pressure loading together with this bending stresses relaxes the material by redistribution the internal forces and brings about a change in the deformed geometry at the peak pressure. The change in the cup profile is a predicted by performing a second-step analysis for both 2D and 3D FE models.For the implicit dynamic FE model, the spring-back analysis is performed as an unloading step in which the pressure loading is ramped to zero in 3 ms. To dissipate the kinetic energy of the deformed material, the proportional structural damping is applied.The spring-back analysis of the 3D explicit dynamic FE model uses the dynamic relaxation method and a constant relaxation factor of 0.995 is applied through the relaxation analysis. The convergence to a quasi-static state is controlled with ratio of the kinetic energy at time t to the kinetic energy at the start of relaxation analysis Deformations of the blank for three levels of the detonation pressure were analyzed using the isotropic linear work hardening (2D FE) model. For all pressure histories, a total duration of 3 ms is assumed and peak values of 2, 4 and 8 MPa are used for the computations. The variation of pressure load as a function of time is shown in . In the computations, both frictionless contact and contact with Coulomb friction is considered. For the contact friction, a factor of 0.12 was taken between blank and blank holder, where as for blank-die contact friction factor was taken as 0.08 shows the effective stress distribution at maximum pressure. The calculated deformations of the blank under contact with friction are shown in The calculated effective plastic strain values, for frictional contact condition, under three pressure levels are 0.18, 0.44 and 0.83, respectively and 8 MPa pressure level causes nearly complete filling of die cavity, for the 8 MPa pressure loading, the plastic deformation is well above the failure strain observed in tension tests. The cup forming experiment with a maximum pressure of 7.8 MPa was carried out which caused failure of the specimens as shown in and the . The rupture occurred at the section where the cup section is bent extreme level in which it was believed to be due to edge constraint of die.Using the 2D FE simulation results, five specimens were tested under different detonation conditions, where the maximum pressures were between 6.7 and 7.8 MPa. The measured and processed pressure histories were used as loading condition for the 3D explicit dynamic FE model. lists the maximum detonation pressures applied in the experiments and calculated averages for each specimen.The specimens A, B and C were observed to deform without a failure, whereas the specimens D and E were ruptured mainly located on the cup head section. This was attributed to the excessive straining together with the bending deformation caused by the die-wall constraint. The deformed shape of specimens A and C calculated with explicit dynamic FE analyses with the produced ones are given in From the experiments, the maximum pressure applied with the current die parameters and process conditions were limited to 7 MPa. The improvements that were applied to the contact interface conditions to reduce the friction force do not eliminate the rupture observed on the work piece. Similarly, reducing the blank holder contact area to the blank or additional lubrication did not eliminate the failure of the work piece. This condition was related to the extremely high-speed deformation of the core region. The velocity distribution for specimen E at time 0.24 ms is shown in The maximum accumulated effective plastic strain values calculated for specimens A, B, and C are 0.55, 0.61 and 0.76, respectively, using linear work hardening model in 3D FE analysis.The deformation of the aluminum blanks during the cup forming process can be described by dividing the section of blank into three regions. The region A is the portion of blank experiencing only a pure radial drawing between the die and blank holder. In all detonation pressures, this region is characterized by the increased thickness due to the circumferential membrane stress and folds forms during the process as in the conventional deep drawing process The blank material in region B experiences initially the plastic bending as shown in . As the process takes places, sliding and stretching are also produced, as illustrated in the superposed deformation plot of the 2D FE mesh in The deformation of blank material in region C is dominated throughout the process. As process takes the radius of the hypothetical punch, decreases from a very large value, theoretically infinity, to die opening distance. At this time the material, point 1, first rebound for some distance, then starts bending over the die bottom, occurring a small amount of sliding due to the continuing stretching at the same instant (). The similar deformation pattern is observed at the die side-wall; however the movement of point 2 is much slower. As the pressure level is increased, the material fills the die cavity. High bending stresses together with the thinning due to continuing stretching results in strain localization. shows the effective stress at the instant of rebound while shows the effective plastic strain distribution computed for the specimen C. The experiment results for the specimens D and E validated this condition because the cup failure due the rupture is identified at the same location.The total radial strain; calculated using the linear work hardening and Johnson and Cook material models are compared with experimentally determined values for specimen C. The maximum difference in the total strain in the radial direction is less than 8% based on these two material models, however the variation of strain as function of initial radial position shows large differences with the experimentally determined curve for specimen C, as given in . On the other hand, the total length of the cup section in the radial direction is predicted with an error of approximately 7 and 9% with the linear work hardening and Johnson and Cook material models, respectively. These simulation results are based on the 3D explicit FE model.In the case of the thickness strains, all material models predict the thinning for region C higher than the measured values for specimen C. The reverse situation exists for the region A. Compared to the other material models, the thickness strain variation predicted with the implicit solution using Lemaitre–Chaboche material model is the closest to the experimentally determined values and variation. gives the thickness strains calculated by using different material models and the experimentally determined curve.The cup profile in rz-plane for specimen C is measured at 16 points using NIKON Profile Projector machine. The final profile shape following the spring-back analyses of both 2D and 3D FE models are superposed for comparison purposes, as shown on . The maximum deviation of explicit 3D model (with linear work hardening material model) from the measured profile is observed approximately 4 mm at the die profile point. The deviation of explicit 3D model with Johnson and Cook material model is approximately 3 mm, and that of the implicit 2D FE result with the Lemaitre–Chaboche material model is approximately 2.5 mm at the cup head radius.All computations are performed on a Pentium III-866 processor PC running under Windows NT4.0 operating system. Comparison for the computational times for the process simulation with the FE models shows that the most efficient model is implicit 2D model with the linear work hardening model, as given in lists the computational times for the spring-back calculations and the same argument holds for elapsed times of the spring-back analyses of the FE models. The other point which needs attention is that the spring-back analyses of explicit 3D FE models take more computer time than the forming analyses using both linear work hardening and Johnson and Cook material models.The FE analysis of GDF of aluminum cylindrical cups is done. The dynamic forming process and spring-back deformations are calculated rate-independent plasticity models in 2D and 3D. The scaling of yield stress is incorporated into the computations with Johnson and Cook material model. The deformation shapes and maximum detonation pressure predicted for material failure limit are validated with experiment. Examining the deformations predicted with FE simulations, the characteristics of forming process are investigated and rupture formation is identified.The calculated thickness strain and radial strain distributions are compared with the experimental values. The maximum values computed show a fairly good agreement with experimental values, however the distributions are quite different than the experimental distribution. The cup profiles predicted with the different material models show moderate amount of shape difference with the measured profile by the 3D digitizing system.Optimum process parameters to produce green ceramic complex partsThe fragility of green ceramic compacts introduces considerable difficulties during green or bisque machining. This paper demonstrates methods developed to manufacture thin wall–thin floor, complex green ceramic parts to close tolerance. Hybrid finite element (FE)/mechanistic models were utilized in the development of the green machining process. An FE model was used to define cutting edge geometry and machining parameters that would reliably produce crack free parts. Mechanistic model was used to direct cutter path generation of a 5-axis milling machine having a large axial depth of cut, and to prevent edge chipping. The optimized cutter path eliminated any need for hand work before densifying the machined part.Engineering ceramics are used in a wide variety of structural applications. Most of these ceramic parts are formed from a powder, sintered at high temperature, and then machined to the required dimensions Machining processes for shaping ceramics are usually employed to make low volume, high precision, complex-shaped parts that are uneconomical for low yielding ceramic forming methods such as slip casting or injection molding. Ceramic machining can be divided into three categories as shown in , fired (dense diamond ground), bisque (partially fired), and green (shaped powder with or without binder). In , pre-formed ceramics are produced by slip casting, dry-pressing or injection molding New approaches are required to develop cost effective machining processes for ceramic components that will not induce surface cracks and maintain the desired surface roughness. One of these approaches is to identify the proper binders that make green ceramic more readily machinable, then utilize physics based modeling to identify the proper cutting geometry of the tool edge that will allow machining of green ceramics without the initiation of cracks on or below the surface This paper presents the approach used for cost effective green machining of complex ceramic components. As compared to conventional machining methods for ceramics, this approach can offer much higher material removal rates, especially for thin wall–thin floor complex parts made of ceramic materials. These complex-shaped parts are desirable for several different high temperature structural ceramic applications. shows a representative thin wall–thin floor part geometry similar to that was used for green machining development.The material selected in this study was silicon carbide (SiC). A CAD model for the selected geometry was developed to enable selection of the cutter path. The procedure that was used to optimize the green machining of the silicon carbide powder compact was as follows:define a binder system to achieve reasonable green part strengthdefine the optimum tool geometry for producing crack-free green machined surfacesperform experiments to define maximum axial depth with no edge chipping at exit and entranceutilize mechanistic modeling to direct the cutter path generation process to prevent chipping of the edge.The success in green machining of SiC depends on the introduction of a weak interlayer that deflects cracks during machining (binder) yet still provides adherence. A proper binder should make green silicon carbide less notch sensitive, enhance the material fracture toughness and provide sufficient green strength at a low content A combination of polyvinyl alcohol (PVA) and poly ethylene glycol (PEG) binder system was used. The binder system was ball-milled into the SiC powder using deionized water having pH adjusted to permit greater dispersion. The slurry was then spray dried to form dry, flowable granules. Several pellets were produced for various SiC/binder formulations and were each tested for biaxial flexure strength following ASTM C 1499-03. Post-baking with the PVA and PEG system brought the average strength to 7 MPa with a high point at 8 MPa, while the tensile strength was 5 MPa and yield stress was 2.1 MPa. The green density was 2.8 g/cm3, and Young's modulus was 1.2 GPa.Silicon carbide compacts were made using an isostatic, dry bag pressing process. The selected PVA/PEG system allowed increasing the unidirectional pressing pressure by 70% with negligible effect on the pellet strength or integrity. An isostatic press die was manufactured to produce blanks required to green machine the complex part (shown in ). The inner surface was pressed in its final form within the specified tolerances; this was able to reduce the total cost of machining by 20%.Orthogonal cutting FE models were used to simulate the green machining of silicon carbide. These models were developed using a commercial general purpose finite element solver, ABAQUS/Explicit™. The FE model assumed that the chip generation will be initiated in a brittle fracture mode based on a fracture strain criteria, which is a function of hydrostatic pressure and von Mises stresses. The model could simulate the effect of rake angle, cutting speed and undeformed chip thickness on crack initiation during cutting. A Drucker–Prager–Cap model (a materials option of the ABAQUS/Explicit™) was used for the description of the SiC flow behavior. The yield surface of this model consists of a linear slope describing failure under shear stresses with superimposed hydrostatic pressure, and an elliptical cap forming the boundary of the yield surface in the direction of the hydrostatic axis (where P is the hydrostatic stress and t is a combination of the second and the third stress invariant; tan |
β (β |
= 65°) is the slope of the shear failure line, d is particle cohesion strength and =0.01 MPa, R describes the shape of the cap =0.5, and α describes a transition surface between cap and failure line and is =0.1; Pb is a hardening parameter, and Pa is given byIn ABAQUS the hardening variable Pb is dependant on the volumetric strain, which is equivalent to 0.62 in this study based on the data available in Ref. , where ν is the cutting velocity, r the tooltip radius, f is the feed, and γ and ξ are the rake and clearance angles, respectively. The length of the workpiece is l and the height is h.The mechanical and physical properties of a typical tungsten carbide tool material as well as a commonly used diamond coating (for FE simulation) are shown in The analyses used were thermo-mechanically coupled, where heat is generated by material plastic deformation and friction between two contact surfaces. The coefficient of friction between SiC and the tool was assumed as 0.2. The gap conductance between two contact surfaces was assumed to be 500 kW m−2 |
°C−1.The primary objective of the numerical simulations is to study the effect of various process parameters on surface crack initiation during green machining of SiC. The parameters for the study include variable speed in the range of 50–150 m/min, and a feed range of 0.03–0.15 mm/tooth, as well as a rake angle range of −10° to 10° and the tooltip radius, range 0.02–0.05 mm. shows the effect of changing the rake angle from −10° to +5°. As it can be seen in -a, the positive rake angle led to large crack initiated on the machined surface, while the cracks are essentially non-existent when the rake angle was −10°. Speed was 150 m/min and feed was 0.06 mm/tooth., the predicted maximum depth of a crack induced in the surface varies when using different rake angles. Two different cutting speeds of 50 and 150 m/min were used. Chip thicknesses of 0.03 and 0.06 mm/tooth were used in this analysis. shows that the depth of the crack is insensitive to chip thickness at negative rake angles. The chip thickness effect is more pronounced as the rake angle increases. Hydrostatic pressure in the workpiece is mainly controlled by speed and tooltip radius. The increase in the hydrostatic pressure reduces the crack initiation, which indicates that higher speed and lower tooltip radius are recommended in green machining of SiC compacts. It is recommended, based on these analyses, that the speeds of 100–150 m/min, a tooltip radius of 0.02 mm, a feed of 0.03 mm/tooth, and a −5° rake angle be employed.Dry pressed silicon carbide test panels (53 mm × 53 mm) were used to evaluate the machining parameters defined from the FE analysis and to determine the maximum axial depth of cut (ADOC) that can be used without initiating edge chipping at tool exit. Diamond coated carbide ball end mills with negative rake angles were selected for a crack-free green machining process of ceramic parts and longer tool life. shows an example of some of the slot edges at tool entrance and exit. The maximum allowable ADOC for a ball end mill of 6.5 mm diameter was 22.5 mm. Up to this value no catastrophic breakage of the SiC tiles was observed.The silicon carbide panels withstood the cutting forces generated by the maximum ADOC of 22.5 mm. The feed was 0.06 mm/tooth and the speed was limited by the machine capabilities to 3000 rpm. These tests indicate that large ADOC is a feasible process in green machining of SiC. Using such high ADOC will reduce the production time and consequently the cost of green machining of SiC complex parts.In addition, during the generation of the tool cutter path, for the actual thin walled structure, extra care was taken in controlling tool motion at the lead and trailing edges of the thin wall structure as shown in . To eliminate edge chipping the tool was exited at an angle greater than 30° as illustrated in . Because green machining of ceramics generates dust like chips, very small values of cutting forces were measured; however, the tool helix angle was specified to direct dust away from the workpiece. Only tool wear and the tendency to encounter part edge fracture at the tool exit necessitate the control of tool motion and chip thickness.A 5-axis mechanistic cutting force model was developed utilizing the generalized approach proposed in some of our previous work . The mechanistic model determines the location and size of the contact area between the cutting tool and the workpiece. The cutter path generated for the milling process was analyzed to predict the instantaneous maximum chip thickness, feed rates, angle of contacts, width of cut, and axial depth of cut.These data are then used to optimize the cutting process to increase efficiency and eliminate hand finishing or rework. The feed rate was optimized utilizing the mechanistic model. shows the predicted and the recommended feed for roughing operations. The optimization constrain was to achieve constant maximum chip thickness during the roughing operations. The scheduled feed led to a 5% reduction in the machining time of the thin wall open cut operation. shows the optimized chip thickness for the finish path of the thin wall feature. An extra 8% reduction in machining time was obtained by optimizing the feed in the finishing operation.Further feed optimization of the cutter path indicated that a total of 60 min could be saved from the total machining time. shows the actual milling process of selected part configurations and -b shows crack/edge chipping free green SiC complex part.With reference to the results and analysis, the following conclusions are drawn:Proper finite element modeling of ceramic green machining allows prediction of optimum process parameters.Cutter path planning allowed the manufacturing of a thin wall–thin floor part free from cracking and edge chipping.Dust removal during high ADOC milling can be controlled by properly selecting the tool helix angle.The use of 5-axis machining, when properly applied, can produce very smooth surfaces and eliminate hand work.Effect of multiple secondary cracks on FRP debonding from the substrate of reinforced concrete beamsFor a FRP strengthened concrete beam, debonding is often initiated by the opening of a flexural or shear/flexural crack along the concrete beam. This debonding failure is often studied with the pull-off test. However, experimental results indicate that debonding on a concrete beam occurs at a significantly higher force than that in the pull-off tests, which can be ascribed to the presence of multiple cracks along the beam span. In the present investigation, an analytical model that accounts for the opening of shear/flexural cracks along the beam is developed. The proposed analytical model has also been verified with the experimental results.In the last two decades, the external bonding of fiber reinforced polymer (FRP) has become increasingly popular for structural strengthening and retrofitting of concrete structures due to the superior properties of FRP, such as high strength-to-weight ratio, high corrosion resistance, ease of application and good durability For many investigations, the pull-off test () is often employed to study debonding behavior between concrete and FRP plates. By pulling the FRP plate along the direction of its length, the maximum FRP stress/strain can be obtained for a given width and thickness of the FRP plate. For a concrete beam made with the same concrete materials as the pull-off test specimen, the maximum FRP stress/strain from the pull-off test is expected to be also applicable if the same FRP is used for strengthening the concrete beam. However, experimental results indicate that debonding models with parameters derived from the pull-off test tends to significantly underestimate the maximum FRP stress/strain in the strengthened beams Generally speaking, the debonding process can be divided into three stages as shown in a shows an early stage, when debonding is only occurring at the vicinity of the cracks. In this stage, relative shear sliding between concrete and FRP plate occurs in opposite directions at the two sides of the crack, as well as the interfacial shear stresses. b illustrates shear sliding and interfacial shear stresses when debonding has extended over two adjacent cracks. c shows further propagation of the debonded zone along the concrete-to-FRP interface. The shear stresses (and sliding) near the cracks become the same sign after the debonded zone has passed through the secondary cracks. In the above models, final failure is taken to occur when interfacial debonding has spread over a region between two adjacent cracks, which is corresponding to the condition in b. A recent set of experimental data from Khomwan et al. . When the load level is low, the relative displacement between concrete and the FRP plate fluctuates between positive and negative signs, which is consistent with the situation shown in a, where the relative displacement between concrete and FRP are of opposite signs on the right and left sides of the cracks. However, with increasing load, the relative displacement becomes positive along the whole member, meaning that the relative sliding become the same sign. In other words, unstable debonding does not occur when debonding spreads over the region between two adjacent cracks, as shown in b. Actually, unstable debonding occurs until the situation in c is attained. Therefore, existing models for debonding between two adjacent cracks are only focusing on an intermediate process rather than the ultimate failure state. To obtain the maximum FRP stress at debonding failure, the debonding has extended over two or more flexural or shear/flexural cracks should be considered.Following the above arguments, the main objective of this paper is to develop a new analytical model to describe FRP debonding under the presence of multiple cracks. The model is based on the concept of delayed softening behavior, which will be further discussed in the following section. To verify the model, experimental investigations are carried out using FRP strengthened beams with: (1) debonding failure under a major crack, and (2) debonding affected by opening of secondary cracks. From the former type of beams, material interfacial parameters are determined to provide input to the proposed model. The modeling results for debonding under the presence of multiple cracks are then compared to test results from the latter type of members.. With increasing load, a major flexural crack will form near the mid-span, and distributed secondary cracks will form in the flexural-shear span. For the sake of simplifying the explanation, only one secondary crack is assumed to affect the debonding behavior. Before debonding initiates from the bottom of the cracks, the relative displacement between concrete and FRP plate results from elastic shear deformation of the adhesive layer and the interfacial shear stresses are relatively small (a). With increasing loading, debonding occurs at the bottom of the cracks, and the relative displacements between concrete and FRP are in opposite directions on the two sides of a specific crack, as well as the interfacial shear stresses (b). With additional loading, interfacial debonding will eventually pass through two adjacent cracks, as shown in c. The secondary crack will open significantly when the two individual debonding zones are linked together. In this stage, the relative displacements between concrete and FRP become the same sign in the debonded zone. The opening of the secondary crack increases the displacement of concrete on the right hand side of the crack, resulting in decrease of the relative displacement between concrete and FRP. As shown in c, a “jump” in the relative displacement will occur at the secondary crack, which has also been verified by the test results of Khomwan et al. ). The reduced relative displacement is associated with an increase in shear stress at the crack, resulting in reduced shear softening rate along the debonded interface (c). The debonded zone can then grow to a larger size before final instability is reached. The maximum force that can be carried by the FRP under the major crack, which is an integral of the interfacial shear forces, is then also increased. (Note: if the secondary crack is absent in c, the shear stress will decrease monotonously towards the major crack rather then exhibiting a sudden jump. The instability condition, when the shear stress drops to zero, will then be reached much earlier.)To model the debonding behavior between concrete and FRP plates, an interfacial shear slip relation is needed. In the present work, the interfacial relation model is originally developed by our research group where τ0 and k are the interfacial material parameters defining the initial residual shear stress and shear softening rate after debonding. Based on Eq. , the distribution of tensile stress (σP) along the debonded part of the FRP can be calculated by the following differential equation:Also, in the elastic zone where debonding has not occurred, σP is given by:where β2=GhtEp1+EfAfEcAc, α2=hkβ2G-hk. In the equations, Ac and Af are the sectional area of the concrete beam and the FRP plate, while Ec and Ef are the respective Young’s modulus. h is the thickness of the adhesive and G is its corresponding shear modulus. Also, t is the thickness of the FRP plate.Solving the above equations with appropriate boundary conditions, the tensile stresses along the FRP plate and interfacial shear stresses are obtained as:σp=τstβtanh(β(L-L1))cos[α(x-L+L1)]+τotαsin[α(x-L+L1)]τp=-τsαβtanh(β(L-L1))sin[α(x-L+L1)]+τocos[α(x-L+L1)]The plate force F (acting below the major crack) is given by Apσp (L). To calculate the maximum value of σp(L), Eq. can be employed to find the value of L1 for the shear stress to drop to zero at the location of the debonding inducing crack (x |
= |
L). Putting this critical value of L1 into Eq. , the ultimate plate stress can be obtained.As a first attempt to model the effect of secondary cracks on FRP debonding, two major assumptions are taken in the model. Firstly, it is assumed the secondary crack to have an effect on the interfacial softening process only. If a secondary crack occurs within the elastic stress transfer zone, its effect will be neglected. Secondly, the effect of a secondary crack on debonding behavior will be considered only when it opens significantly. If the steel reinforcement has not yielded at the section of the secondary crack, the opening of the crack is taken to be too small to have an effect on debonding behavior. To see how one can model debonding under the presence of multiple cracks, the situation shown in c is considered. In the figure, the debonded zone is separated into two parts, with the first part (zone 1 of length L1) between the elastic zone and the secondary crack, and the second part (zone 2 of length L2) between the secondary crack and the major crack. L2 is the crack spacing along the beam, which can be calculated from various approaches. For zone 1, at the boundary between the elastic and debonded zone, the FRP stress is continuous and its value can be calculated from Eq. . Also, the shear stress is equal to the initial residual value τ0. With these conditions, the stress distribution within this zone can be obtained by solving Eqs. . In zone 2, at the secondary crack, the stress in FRP is continuous and can be obtained from the solution to the stress distribution in zone 1. The shear stress, on the other hand, will exhibit a sudden jump due to opening of the secondary crack. The value before the jump, τ1-, is obtained from the analysis for zone 1. The value after the jump (τ1+) should be smaller than initial residual value τ0 and is given by:where w is the width of the secondary crack and k is the shear softening rate after debonding.Knowing the FRP stress and interfacial shear stress at the beginning of zone 2, the stress distribution within the zone can again be calculated with Eq. . In reality, the crack opening and spacing depend on the applied moment as well as the steel and FRP reinforcement ratio. According to CEB-FIP where Srm is the average crack spacing. ρc,eff |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.