University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | In order to explore this phenomenon in more depth and to allow fair comparison
of these two flotation cells, a detailed pilot-scale study was initiated under controlled
experimental conditions. Figure 5.8 shows representative bubble images for both pilot-
scale flotation cells taken at four selected sampling locations. Both cells were operated at
7 m/s impeller tip speed, while the superficial gas velocity was fixed at 1.44 cm/s for the
Dorr-Oliver cell and was measured to be 1.41 cm/s for the WEMCO cell.
It can be readily noticed that, under analogous operating conditions, the WEMCO
cell tends to generate a larger number of smaller bubbles that can be detected at all
sampling locations. It can also be noted from the presented images that bubble
populations sampled below the discharge flow coming from the impeller contain an
insignificant number of big bubbles in comparison with locations 1 and 2. This indicates
that the process of bubble segregation, based on bubble size, occurs in the discharge
stream. Figure 5.8 also shows the assumed gas phase pathways within the cell.
Figure 5.9 and Figure 5.10 show bubble size distribution results obtained for the
0.8 m3 Dorr-Oliver flotation cell, at four different operating conditions.
162 |
Virginia Tech | Bubble diameters measured at different locations in the cell when it was operated
under low aeration rate (0.86 cm/s superficial gas velocity) and medium (5 m/s) and high
(7 m/s) agitation rates are presented in Figure 5.9. As can be seen from the figure, the
increase in agitation rate, under a low aeration rate, does not affect the characteristics of
the majority of bubbles present in the bubble population, which is well reflected through
the relatively constant number mean diameter (goes from 0.45 to 0.49 mm) obtained for
both conditions at locations 1, 2 and 3. On the other hand, Sauter mean bubble diameter
decreases from approximately 1.1 to 0.7 mm, indicating that the number of big bubbles in
the population drops significantly with the increase in agitation rate. This trend can be
clearly observed by looking at the area graph in the figure, indicating that the fraction of
total gas contained in large bubbles drops significantly when the agitation rate increases.
Results obtained for location 4 show an insignificant number of big bubbles carried into
the zone above the impeller, indicating that the machine dispersed gas effectively for both
operating conditions.
The effect of agitation rate increase on bubble size distribution under high
aeration rate (1.88 cm/s) is presented in Figure 5.10. In this case, a large fraction of gas
was detected at location 4 when the cell was operated under a 5 m/s impeller tip speed,
suggesting that the gas dispersion limit was exceeded at that operating condition and that
the cell was running under boiling conditions. On the other hand, a significantly large
fraction of gas was also noted at location 4 when the impeller tip speed was increased to
7 m/s. Even though this might not be a direct indication of boiling conditions in the cell,
it suggests that, under high aerating conditions in the forced aerated cell, the total volume
of gas introduced into the system cannot be completely dispersed into small bubbles in
166 |
Virginia Tech | the generator zone, which could potentially result in the escape of larger bubbles through
the impeller/stator gap. In addition, comparison of the number and Sauter mean bubble
diameters leads to the conclusion that, again, the majority of bubbles remain in the same
size scale (around 0.7 mm) but, on the other hand, the number of big bubbles in the
population decreases significantly when the impeller tip speed increases from 5 to 7 m/s.
This is also well reflected through the area graph presented in the Figure 5.10. Finally,
when comparing bubble sizes obtained at the same agitation rate and at two different
aeration rates, a significant increase of both number mean and Sauter mean diameters can
be noticed. This clearly indicates the strong impact of aeration rate on bubble generation
process, and therefore on bubble size, in forced-aerated flotation cells.
Figures 5.11 and 5.12 show bubble size distributions obtained during pilot-scale,
two-phase, gas dispersion testing of the WEMCO flotation cell. The operating conditions
were altered by increasing of the impeller tip speed, which resulted in the change in
aeration rate, which was accordingly measured. As impeller speed increased from 4 to 7
m/s, the aeration rate gradually increased from 0.5 to 1.4 cm/s. From the figures, the
measured number mean bubble diameter remains practically the same (0.45±0.04 mm)
under all operating conditions and at all sampling locations. This suggests that the basic
mechanism of bubble creation remains the same, since the majority of bubbles in the
population have the same characteristics for all tested conditions, which can easily be
noticed from the number frequency distribution graphs.
167 |
Virginia Tech | On the other hand, the Sauter mean bubble diameter and area graphs presented
show significant increase of a number of big bubbles in the cell with the increase in
agitation rate. At the location 2, the Sauter mean diameter increased from 1.1 mm at
lowest agitation rate to 1.6 mm at the highest aeration rate. Over the whole range of
agitation rates tested, bubble populations sampled at the location 1 contained significantly
larger number of big bubbles, which is seen in the strong peak over the large size
fractions in the area graphs.
For the range of operating conditions tested, a self-aerated flotation cell generates
a greater number of small bubbles (bubbles smaller than 0.5 mm) than a forced-aerated
cell. This can be clearly seen from the number frequency and cumulative frequency
distribution graphs (Figures 5.9 to 5.12) for both tested cells. This indicates that there are
some fundamental differences in the way bubbles are generated for these two flotation
cells.
5.5. DISCUSSION
The comparison between the gas dispersion results obtained for two flotation cells
reveals a relatively wide variation of bubble size distributions measured at different
locations in the turbulent zone of each cell. The data show that the characteristics of a
bubble population strongly depend on aeration rate for the forced–aerated cell and on
agitation rate for the self-aerated cell. Figures 5.13 and 5.14 provide a simplified graphic
representation of the gas phase pathways for each cell type under three extreme operating
conditions. The thickness of the arrows in the drawings is roughly proportional to the
fraction of the total gas volume that is transported in that direction.
170 |
Virginia Tech | Figure 5.13 shows gas phase pathways inside of the forced-aerated Dorr-Oliver
flotation cell at a constant high agitation rate and at three different aeration rates (low,
medium and high). The drawings suggest that the majority of the gas introduced into the
cell is carried by the discharge stream dispersed radially from the impeller. From this
stream, a fraction of the gas contained in small bubbles has a chance to be carried to the
bottom of the tank, where part is re-entrained into the generator zone. On the other hand,
based on the findings from this study, a larger fraction of the gas carried by the discharge
stream is moved toward the upper zone of the cell and eventually leaves the cell from the
top.
Depending on the operating conditions, one fraction of the gas contained in large
bubbles that are carried by the discharge stream coming from the impeller does not
follow the main fluid stream lines, but leaves the cell directly. This phenomenon is very
important for understanding the overall hydrodynamics inside of the flotation cell.
Generally, there are two main forces contributing to the hydrodynamics and,
consequently, to overall gas distribution pattern in the flotation cell: turbulent dispersion
forces, reflected directly through the drag force of the continuous phase (liquid)
(Simonnet et al., 2007), and buoyancy of the dispersed phase (gas) (Sokolichin et al.,
2004). For each location in the cell, the balance of these two forces defines the bubble
path and its terminal velocity. At the same agitation rate, an increase in aeration rate
increases both overall gas holdup in the system and volume of the recirculated gas, which
dampens the turbulent kinetic energy of the discharge stream. On the other hand, an
increase in aeration rate results in the increased production of large bubbles in the
generator zone. Both of these effects hinder the discharge stream capacity to transport
172 |
Virginia Tech | large bubbles and increase the probability of the of escape large bubbles from the radial
jet to the upper flow field. In this case, the probability of detecting large bubbles in the
discharge stream decreases as the bubble size increases. Moreover, the probability of
detecting large bubbles above the discharge stream, in this case, will be greater than in
the discharge stream.
At first, this mechanistic description might appear contradictory since the
discharge stream is the region of high energy dissipation rates and it is expected that the
large bubbles are broken-up here before they have a chance to escape, which should
result in the small bubbles in both regions. Even though, in general, this might be the case
for the majority of big bubbles, and it is well reflected through the similar shape of the
size frequency distribution curves (Figures 5.9 to 5.10), some bubbles will still have a
chance to escape. It is very important to take large bubbles into consideration since a
small number of large bubbles carry a significant fraction of total gas from the system
and in that way decrease process efficiency.
These results point to an important fact for modeling of the two phase flows in
flotation cells. Care must be taken when modeling break-up in the impeller discharge
stream since it is not the only phenomenon that defines final bubble size. Recirculation of
the primary bubbles and bubble buoyancy should also be incorporated into the modeling
codes for simulation of flotation systems.
As a result of the observed phenomenon represented in Figure 5.14, it can be seen
that the overall gas distribution in the cell goes from a uniform to non-uniform pattern as
the aeration rate is increased from low to high. This is understandable in a view of the
173 |
Virginia Tech | fact that significantly narrower bubble size distributions, which are shifted toward smaller
bubble sizes, are generated when the aeration rate is low, while broad bubble size
distributions, shifted toward larger bubble sizes, are generated during operation with the
high aeration rates and at constant agitation rate (right columns of Figures 5.9 and 5.10).
Generally, small bubbles tend to follow all major fluid streamlines in the turbulent zone
of the cell while larger bubbles have more chances to overcome local drag forces due to
their higher buoyancy, and are moving directly toward the cell surface.
This non-uniformity can be best observed by looking at the radial distribution of
the gas phase right below the pulp-froth interface. Here, uniformly distributed gas
entering the froth zone, while running at low aeration rates, becomes more concentrated
toward the cell wall as the aeration rate is increased. During operation with high aeration
rates, the gas distribution profile takes a saddle shape, with higher gas fractions leaving
the cell in the zone closer to the cell wall and around the impeller shaft. After this point,
if the aeration rate were increased, the gas dispersion capacity of the impeller would be
exceeded, which means the cell would run under boiling conditions, and the gas
distribution profile at the top of the pulp would be strongly skewed toward the center of
the cell.
Gas distributions for the three selected operating conditions for the self-aerated
WEMCO flotation cell are described in the Figure 5.14. The three limit operating
conditions reflect low, medium and high agitation rates. As can be observed in the figure,
the expected gas distribution pattern, in the turbulent zone of the self-aerated cell
resembles the pattern in the turbulent zone of the forced-aerated cell. Due to the shorter
radial distance from the disperser hood to the cell wall and shorter vertical distance from
174 |
Virginia Tech | the discharge stream to the cell surface, the gas distribution profile, close to the pulp-froth
interface, is somewhat different from that of the forced-aerated cell. For all selected
operating conditions, the fraction of the gas reporting to the froth will be slightly higher
close to the cell wall and gradually decreases toward the disperser hood. It is also found
that, for the self-aerated cell, the gas preferentially concentrates in the upper zone of the
cell with practically no bubbles present in the bottom half of the cell.
Schematic representations of bubble generation processes occurring in the high-
energy intensive (impeller) zone, for both mechanical flotation cells, are presented in
Figures 5.15 and 5.16. In both cases, gas cavities are formed at the low-pressure, trailing
edge of the impeller blades, which is the first stage of the process of bubble creation.
Thereafter, bubbles are shed from the tail of the cavity by the turbulent eddies. Small
bubbles are formed as the high circulating velocity in eddies dissipates through the radial
flow of the fluid. Therefore, energy dissipation occurring when turbulent, high-intensity
eddies disintegrate is one of the main factors important for the creation of small bubbles
(Stephenson et al., 1998). In the Dorr-Oliver cell (Figure 5.15), the gas is introduced
directly to the impeller through the six openings at the bottom of the impeller disc. The
introduced gas accumulates in the gas cavities where it is radially distributed toward the
cell body. Since cavities are, at the same time, receiving and releasing the gas, they are
also known as ventilated cavities.
175 |
Virginia Tech | The cavity profile, represented through the dashed lines in the figure, strongly
depends on the volumetric gas rate, the flow characteristics of the up-coming fluid stream
(gray bordered arrows in the figure), and the characteristics of the turbulent eddies
created at the impeller blade edge. Five different cavity profiles, as a function of different
aeration rates under a constant, high agitation rate, are shown in the figure. Numbers 1 to
5 reflect operating conditions with high to low aeration rates.
Profile 2 represents the operation just below the maximal gas dispersion limit,
which is typically well reflected through the minimal value of the gassed power to
ungassed power ratio for a certain agitation rate. This operating condition represents the
optimal operation point which typically results in the maximal performance. The upward
shift of the cavity profile (line 1 in the figure) represents process operation when the gas
dispersion limit is exceeded (boiling conditions). In this case, part of the gas introduced
into the impeller by-passes the generator zone and escapes directly through the gap
between the impeller and stator.
For the WEMCO cell, aeration rate is a function of the total fluid flow pumped
through the impeller. The tangential flow that is generated in the disperser region is
important for the creation of the forced vortex and surface aeration of bubbles, which
result in gas induction (Mundale and Joshi, 1995). The surface aeration of bubbles is the
main bubble generation mechanism in the self-aerated cells. During the operation, the
liquid level in the generator region and cell is directly affected by the impeller rotational
speed. Generally, as impeller speed increases the liquid level decreases. It is followed by
an increase in the volume of the introduced gas up to a certain point when it starts
decreasing. This reduction in the induced gas flow rate is a result of two phenomena
177 |
Virginia Tech | occurring in the generator cavity as suggested by Patil and Joshi (Patil and Joshi, 1999a;
Patil and Joshi, 1999b):
1. impeller increased exposure to the gas phase due to the very low liquid level,
which results in a sudden drop of the impeller pumping capacity, and
2. impeller drowning as a result of the flow reversal in the upper part of the
generator cavity, which reduces the impeller capability to induce the gas.
Figure 5.16 shows the estimated cavity profile formed at the low pressure side of
the WEMCO impeller when operated at optimal agitation rate. Profile lines presented in
the figure (left image) depict the gas cavity profile at different horizontal levels of the
impeller (right image). As can be seen from the figure, one side of the blade is almost
completely covered with the gas phase, while the other side of the blade is covered with
the liquid phase. Created long cavities and much longer impeller blade lengths, which are
capable of dispersing the gas by breaking the created free surface, support the creation of
the large numbers of small bubbles.
5.6. CONCLUSIONS
Operating conditions and created flow conditions in a mechanical flotation cell
have a considerable impact on the distribution of the gas phase throughout the cell.
Typically, large bubbles generated in the high-energy impeller-stator zone can be found
in the impeller discharge stream, suggesting that the large scale vortices present in the
discharge stream have the capability to capture and transport larger bubbles. On the other
hand, with the increase in aeration rate for forced-aerated cells or with an increase in
agitation rate of self-aerated cells, a significant number of large bubbles leaving the
178 |
Virginia Tech | generator zone can escape from the discharge stream. This phenomenon is strongly
affected by the balance of local drag force coming from the continuous phase and the
bubble buoyant force. However, when flotation cells are operated under optimal
conditions, the largest fraction of the total gas entering the cell is contained in small
bubbles, which are carried by the discharge stream to the tank wall. From there, one
fraction is transported to the bottom of the cell and reintroduced into the high-intensity
zone.
Findings from this study suggest that the bubble diameter in a flotation system is
not determined by a single phenomenon. Typically, bubble break-up due to the high
energy dissipation rates is the determining factor, but there are several other mechanisms
that have to be taken into account. Bubble buoyancy, recirculation of the primary bubbles
and trailing vortices generated behind the large bubbles and bubble swarms are some
mechanisms that should also be considered.
In summary, local bubble size distributions for two different types of mechanical
flotation cells and for different operating conditions have been reported. These results are
now available for the further development and refinement of existing flotation models
and for the validation of existing numerical simulations.
5.7. REFERENCES
Angeli, P. and Hewitt, G.F., 2000. Drop size distributions in horizontal oil-water
dispersed flows. Chemical Engineering Science, 55(16): 3133-3143.
Chesters, A.K., 1991. The modeling of coalescence processes in fluid-liquid dispersions:
A review of current understanding. Chemical Engineering Research and Design:
transactions of the Institution of Chemical Engineers, Part A(69): 259–270.
179 |
Virginia Tech | CHAPTER 6:
SUMMARY AND CONCLUSIONS
A fully-instrumented 0.8 m3 pilot-scale flotation circuit was developed for the
purpose of providing performance data that can be more readily utilized for the
engineering design, scale-up and optimization of industrial flotation circuits. This highly
flexible system enabled measurement, monitoring and control of a number of
hydrodynamic and metallurgical parameters over a wide range of operating conditions.
Additionally, a new, robust, in-situ bubble sampling apparatus was developed and
validated in both two- and three-phase conditions. The new apparatus allowed bubble
monitoring at different locations within the pilot-scale mechanical flotation cell, which
could not be performed in the past by utilizing some of the commercially available
bubble sampling systems. These newly developed state-of-the-art systems allowed for the
detailed investigation of flotation performance of different pilot-scale flotation machines
over a wide range of operating conditions. The flotation cell was operated as either batch
or continuous reactors in both two-phase (liquid/gas) and three-phase (liquid/gas/solid)
modes.
The main conclusions obtained from the comprehensive hydrodynamic and
metallurgical investigation are listed below:
The continuous flotation circuit was successfully and easily operated with slurry
containing more than 30% w/w of solids and allowed continuous and simultaneous
monitoring of multiple flotation parameters over a wide range of operating
conditions.
183 |
Virginia Tech | Two different bubble sampling methods (i.e, a new in-situ technique and a
commercially available ex-situ technique) were used simultaneously to investigate
local bubble size distributions in a pilot-scale machine. The measured bubble size
distributions and Sauter mean bubble diameters revealed significant differences
between the two sampling methods. Most importantly, the commonly used ex-situ
bubble sampling method failed to detect larger bubbles present in the flotation pulp.
Two types of image analysis software, the fully automated Northern Eclipse package
(edge detection image analysis technique) and semi-automated BubbleSEdit package
(cross-correlation image analysis technique), were used to analyze captured images
and to obtain bubble size distributions for each image set. The commonly used edge
detection image analysis technique generally failed to detect larger bubbles due to
their non-spherical shape and greater chance for bubble overlap.
The miscounting of large bubbles from the image sets can result in misleading
conclusions about the gas dispersion properties inside the flotation cell since large
bubbles represent a significant fraction of the total gas volume. Experimental data
collected in the current study indicated that the fraction of the total introduced gas
carried by larger bubbles (>1.5 mm) could exceed 80% of the total gas volume when
the cell was operated under high aeration rates and low agitation rates.
General simplicity and ease of use makes the ex-situ method useful whenever a large
number of tests have to be performed in a short timeframe. However, information
gained by the ex-situ sampling method gave local mean diameters of bubbles entering
the froth phase, which produced misleading results when the bubbles were sampled
from only one location.
184 |
Virginia Tech | Although the new in-situ sampling method and BubbleSEdit image analysis technique
was more demanding, this method provided a more realistic estimation of the true
bubble size distribution at all locations within a mechanical flotation cell.
Bubble populations were found to vary significantly at different vertical and radial
distances from the impeller/stator assembly, and the degree of the variation strongly
depended on the operating condition. Due to this variability, care must be taken when
performing bubble measurements in mechanical flotation cells. Bubbles found below
the froth-pulp interface contribute significantly to the processes occurring in the froth
zone, while bubbles sampled in the impeller discharge stream contribute to the overall
bubble-particle interaction dynamics occurring in the turbulent zone of the cell.
Therefore, in order to achieve better insight into the spatial gas distribution profile in
mechanical flotation cells, radial screening of bubble sizes needs to be performed.
The Sauter mean bubble diameters obtained in this study ranged from 0.5 to 3 mm,
which is wider than bubble size distributions previously reported in the literature. The
larger mean bubble diameters were obtained due to increased precision achieved with
the in-situ bubble sizing method by which up to 98% of all recorded bubbles in an
image were detected and included in the analysis.
The measurement of power consumption provided considerable insight into the gas
dispersion capabilities of the rotor/stator mechanism. Specifically, the ratio of
aerated-to-unaerated power plotted as a function of the aeration number provided
important information about the minimum agitation rate necessary for complete
dispersion of the gas introduced into a flotation machine. The first sign of the
185 |
Virginia Tech | transition of the overall flow pattern from “loaded” to “flooded” conditions can be
easily observed through power input monitoring.
The power input had a positive effect on the number of bubbles created in the cell,
which increased the probability of bubble-particle attachment, increased overall
carrying capacity, and therefore increases the flotation rate constant.
A decrease in mean bubble diameter was achieved by decreasing superficial gas
velocity and increasing specific power input.
Flattening of the D -P* trend was observed for all aeration rates tested, which
32
suggests that there is a minimum energy input needed to achieve an optimal bubble
size distribution in the cell for a given constant aeration rate.
For all particle sizes, high aeration and agitation rates resulted in higher material
recoveries. For fine and intermediate particle sizes, recovery increased as residence
time increased over all operating conditions.
For coarser particles, an increase in agitation rate increased off-the-bottom suspension
and increased particle concentration suspended in the pulp. This condition ultimately
created more favorable conditions for bubble-particle encounter in the high-turbulent
zone of the cell.
A correlation between the flotation rate constant and bubble surface area flux was
observed for all glass particles tested, while the nature of this correlation strongly
depended on the size of the particles.
As a result of the heterogeneous gas distribution within the cell, the bubble surface
area flux values estimated from global superficial gas velocities overestimated the
bubble surface area flux.
186 |
Virginia Tech | A significant number of large bubbles leaving the generator zone were found to
escape from the discharge stream with an increase in the aeration rate for forced-
aerated cells or with an increase in agitation rate for self-aerated cells. This
phenomenon is believed to be strongly affected by the balance of local drag force
coming from the continuous phase and bubble buoyant force.
When flotation cells are operated under optimal conditions, the largest fraction of the
total gas entering the cell was found to be contained in the small bubbles. These
smaller bubbles were carried by the rotor discharge stream to the tank wall. From
there, a fraction of these small bubbles was transported to the bottom of the cell and
reintroduced into the high-intensity zone created by the rotor.
The bubble diameter in a flotation system was not determined by a single
phenomenon. Typically, bubble break-up due to the high energy dissipation rates was
the determining factor, but there were several other mechanisms that must be taken
into account. Bubble buoyancy, recirculation of the primary bubbles, and trailing
vortices generated behind the large bubbles and bubble swarms are mechanisms that
should also be considered.
Local bubble size distributions were carefully measured for two different types of
mechanical flotation cells over a range of different operating conditions. These
experimental results are now available for the further development and refinement of
existing flotation models and for the validation of existing numerical simulations.
In summary, data obtained using the pilot-scale system can be used as a baseline for
advanced modeling, control and optimization of flotation processes. With its functional
versatility, the system can be easily adapted to almost any process condition and, in that
187 |
Virginia Tech | PROCESSING LOW RANK COAL AND ULTRA-FINE
MINERAL PARTICLES BY HYDROPHOBIC – HYDROPHILIC
SEPARATION
RIDDHIKA JAIN
ABSTRACT
This thesis pertains to the processing of ultra-fine mineral particles and low rank coal
using the hydrophobic–hydrophilic separation (HHS) method. Several explorative experimental
tests have been carried out to study the effect of the various physical and chemical parameters on
the HHS process.
In this study, the HHS process has been employed to upgrade a chalcopyrite ore. A
systematic experimental study on the effects of various physical and chemical parameters such as
particle size, reagent dosage and reaction time on the separation efficiencies have been
performed. For this, a copper rougher concentrate (assaying 15.9 %Cu) was wet ground and
treated with a reagent to selectively hydrophobize the copper-bearing mineral (chalcopyrite),
leaving the siliceous gangue minerals hydrophilic. The slurry was subjected to a high-shear
agitation to selectively agglomerate the chalcopyrite and to leave the siliceous gangue dispersed
in aqueous phase. The agglomerates were then separated from dispersed gangue minerals by
screening and the agglomerates dispersed in a hydrophobic liquid (n-pentane) to liberate the
water trapped in the agglomerates. The chalcopyrite dispersed in the hydrophobic liquid was
separated from the medium to obtain a concentrate substantially free of gangue minerals and
moisture. The copper recoveries were substantially higher than those obtained by flotation. The
HHS process was also tested on ultrafine mono-sized silica beads. The results were superior to
those obtained by flotation, particularly with ultrafine particles. The HHS process has also been
tested successfully for upgrading subbituminous coals. Low-rank coals are not as hydrophobic as
high-rank coals such as bituminous and anthracite coals. In the present work, a low-rank coal
from Wyoming was hydrophobized with appropriate |
Virginia Tech | Chapter 1: Introduction
1.1 Preamble
Chalcopyrite is the principal commercial source of copper. Presently, the mineral is
recovered by flotation after grinding the ore to fine sizes for liberation. Similarly, flotation is
used for the removal of silica from iron ore and phosphate concentrates. As ore grades becomes
lower, it is necessary to grind an ore finer to achieve liberation (Johnson, 2005). However,
flotation is inefficient for the recovery of fine particles below approximately 10 to 25 µm.
Spherical or oil agglomeration has been proposed as an excellent technique for
recovering fine particles that are difficult to recover by flotation. Oil agglomeration involves
addition of an immiscible liquid such as a hydrocarbon oil to an aqueous suspension of solids.
Upon agitation, there will be a distribution of the oil over the surface of the hydrophobic
surfaces, causing formation of liquid bridges or oil agglomerates. Addition of suitable collectors
to the process results in selective mineral recovery. Recovering coal fines by oil agglomeration is
a well-known process. However, recovering mineral particles with oil agglomeration is yet to be
tried outside the laboratory (House, C. I. and C. J. Veal, 1989). Some work on cassiterite,
ilmenite, hematite, barite and gold has been reported in the literature (House, C. I. and C. J. Veal,
1989).
Coal is used as a raw material for many chemical synthesis processes including
metallurgy and fuels for power plants due to its low cost. In 2011, 42% of the electricity in
United States used coal as its source of energy (EIA 2012). Over the coming years it is expected
that the demand for coal will rise. An estimated 52% of the world coal reserves consist of sub-
bituminous and brown coal also referred as low rank coal (LRC). With the impending energy
crisis, it will be increasingly necessary to use LRCs Although, low rank coal has low ash and
.
sulfur contents, its high moisture content, low calorific values and spontaneously combusting
characteristics makes it really difficult to utilize. So if it could be efficiently upgraded and
converted into high-grade, high-heating coal, it would greatly contribute not only to a stable
1 |
Virginia Tech | energy supply but also to environmental conservation as SO releases from the combustion plants
2
contributes largely towards acidic rain.
From the Energy crisis in 1970, numerous research and development projects of
conversion process for low rank coal started in several countries. Some of the conventional
processes developed to reduce moisture content of the low rank coal involve heating LRC above
the boiling point of water to vaporize the fluid. However, these processes have problems with the
large latent heat of water vaporization which increases its energy consumption and the product
becomes spontaneously combustible. As a result, the transportation and storage of processed low
rank coal become very difficult. For this purpose, a new and innovative process to reduce the
moisture content and increase the BTU value of low rank which can also solve the self-
combustion issue has to be developed having two main benefits: utilization of low rank coal
reserves around the globe to cater to energy needs and the environmentally friendly product with
low sulfur content.
1.2 Objectives
Virginia Tech developed a patented process known as dewatering by displacement
(DbD) in 1995 (Yoon and Luttrell, U.S. Patent No. 5,459,786) for dewatering of fine coal. The
main aim of this research work was to extend the application of dewatering by displacement
(DbD) process for processing of mineral particles in ultrafine size ranges as well as for up
gradation of LRCs. The research focused on studying the changes in the properties of low rank
coal and mineral (chalcopyrite and silica) with the test parameters.
For chalcopyrite and silica beads, the goal of this project was to develop a process to
facilitate the recovery of mineral particle in ultra-fine size range. The project was focused on
conducting laboratory-scale batch tests on different size fractions of both silica beads and
chalcopyrite. The surface of the ore particles was modified using hydrophobizing agent. The
modified ore was then subjected to oil agglomeration. The agglomerates were then dewatered
either dispersing them in hydrophobic liquid or filtered in presence of hydrophobic liquid so as
to facilitate the displacement of water by hydrophobic liquid.
2 |
Virginia Tech | For low rank coal, the goal of this project was to develop a process to remove the
excessive moisture content without making it susceptible to spontaneous combustion. The coal
surface was modified using different techniques to make it more hydrophobic. The modified coal
was then subjected to oil agglomeration. The coal particles were then dispersed in hydrophobic
liquid, which was recycled back into the process using double boiler and condenser unit. The
resulting clean coal should contain less than 12% moisture to achieve higher BTU values. In all
the experiments, pentane was used as hydrophobic liquid since it is affordable and can be easily
recycled using evaporation and condensation.
The project was focused on conducting laboratory-scale batch tests using three different
methods for making low rank coal surface hydrophobic in order to determine the most economic
method which will facilitate the scaling up of the process. During the experiment processes, the
low rank coal was made hydrophobic by surfactant coating, acid washing followed by
esterification or low temperature oxidation followed by esterification. The resulted hydrophobic
coal particles were subjected to oil agglomeration, which were then dispersed into hydrophobic
liquid using mechanical agitation/vibration.
1.3 Organization
The thesis has been broadly categorized into four main parts. The introductory part gives
an overview of the research objectives for both low rank coal processing as well as mineral
processing.
The literature review section of this thesis deals into prior work conducted by researchers
on the processing of low rank coal as well as oil agglomeration of mineral particles. It provides
an overview to familiarize the readers with basic concept and terminologies that have been used
in this research.
The experimental section is further divided into three parts. The first part describes the
chalcopyrite processing by hydrophobic- hydrophilic separation (HHS) technique and compares
with froth flotation. The second parts deals with processing of mono-sized silica by HHS and
compares with froth flotation. Finally low rank coal processing by HHS technique is discussed.
3 |
Virginia Tech | Chapter 2: Literature Review
2.1 Prior Work
2.1.1 Dewatering by Displacement
Virginia Tech developed a patented process known as dewatering by displacement (DbD)
in 1995 (Yoon and Luttrell, U.S. Patent No. 5,458,786), and has been making improvements. In
this process, a hydrophobic liquid is added to a coal slurry to displace the surface water. The
displacement occurs because the hydrophobic piqued has a higher affinity for coal surface than
water. The DbD process is depicted in Figure 2-1 (a), in which a hydrophobic solid particle 1
leaves an aqueous phase 3, and enters the hydrophobic liquid (oil) phase 2. This step is
spontaneous, when the change in Gibbs free energy (G) normalized by particle surface area (A)
is less than zero:
(1)
Combining Eq. (1) with Young’s equation, one obtains the following relation,
(2)
in which is the contact angle of oil on the hydrophobic surface as measured through water (3) .
According to Eq. (2), G < 0 when > 90o.
Figure 2-2 shows the contact angles of n-alkanes on a hydrophobic bituminous coal
surface. As shown, the contact angles are greater than 90o. Therefore, the hydrocarbon oils can
spontaneously displace water from the surface. Note here that G becomes more negative with
decreasing number of the carbons in the hydrocarbon chain. Therefore, a shorter hydrocarbon
chain would be a better hydrophobic liquid to displace water from the surface of hydrophobic
particles. One advantage of using a short chain hydrocarbon oil is that it can be readily recovered
and recycled after the DbD process.
5 |
Virginia Tech | typical sulfide flotation collector such as xanthate or thionocarbamate. Ultrafine silica particles
can also be separated from phosphate or iron oxides by selectively rendering the silica
hydrophobic. For coal, it is not necessary to use a hydrophobizing agent as higher rank coals
such as bituminous coal and anthracite are naturally hydrophobic. For subbituminous coals that
are naturally hydrophilic, one can use appropriate hydrophobizing agents and subject them to the
HHS process.
2.1.2 Low Rank Coal
Coal formation and reserve maturity determine the rank of coal. The rank of coal can be
seen to be the degree of maturation that occurs as coal metamorphoses from peat to anthracite
during the course of its formation process (Sondreal, E.et al, 1984)
Thus the sub-bituminous coal types and lignite can be seen to form low rank coal. This
type of coal has low carbon content and hence, the energy output achieved on combustion is also
lower. Thus, low rank coal refers to the coal type which has high moisture content and low
heating value. Consequently the price of low rank coal which amounts to 50% of total coal is
also lower compared to the higher ranks. It is imperative that low rank coal with its low pricing
and larger reserve needs ample attention so as to benefit from the same.
Figure 2-3 describes the types of inherent moisture present in low rank coal. In total we
can see five major types. Based on their characteristic properties, these can be classified as
interparticle water, adhesion water, capillary water, surface adsorption, organic water and
interior water (Kartikayen, M et al, 2009). These can be defined as follows:
1) Interior adsorption water is that which is deposited during formation and find itself
bound to the coal micro-particles and capillaries.
2) Surface adsorption water, as the name suggests finds itself on the outer surface of the
coal particles.
3) Capillary water is that which is found in the capillaries of coal particles.
4) Water found between two particles is termed as interparticle water.
5) Water which find itself adhering to the surface of the coal particles is adhesion water.
7 |
Virginia Tech | Organic water
Interior
adsorption
Surface adsorption
water
water
Capillary Water
Interparticle water
Coal particles
6) Organic water which is present as the carboxyl function group at the surface of coal
particles.
Figure 2-3: Inherent moisture structure in low rank coal
Traditionally this type of coal did not find preference for use in industries due to the
additional costs involved in the transportation of the fuel due to high moisture and the low output
in terms of energy content. The high moisture content of the low rank coal, or brown coal, makes
it heavy in weight and with the energy output being not nearly as comparable to the higher ranks,
it found its use near the mines for burning and heat generation processes.
However, over the years, low rank coal has been sought out for usage as fuel due to its
characteristic low sulfur content which makes it an environmentally friendly substitute compared
to its other sulfur-rich counterparts.
Another disadvantage of the low rank coal is the ease with which it combusts
spontaneously. Spontaneous combustion occurs when low rank coal is dewatered at high
temperature. Even at low temperatures, partial oxidation of the carbon takes place which releases
heat and cause a rise in the temperature of the coal. This temperature increases gradually
reaching a point at which spontaneous combustion takes place. Low rank coal has lower carbon
content which increases the possibility of an oxygen adsorbent functional group to exist in its
8 |
Virginia Tech | structure in comparison to higher ranking coal thus making it more susceptible to oxygen
adsorption both chemically and physically
In the techniques which have been employed so far to enrich the low rank coal, the
inherent moisture present as the organically bound water in the coal is also driven out due to the
drying techniques used. However, this causes the problem for the fuel composition as it makes
the fuel easily susceptible to the spontaneous combustion.
2.2 Fine Particle Processing: Flotation
Froth flotation (Figure 2-4) is the most vividly used process in mineral industry for
concentrating mineral in the size range 45µm to 150 µm. It is a physico-chemical separation
process based on the wettability of particles. In flotation cell, air bubbles are generated in the
pulp phase by agitator which selectively attached the hydrophobic minerals and carry them to the
froth phase, while hydrophilic gangue particles remain in the bottom of the cell. The froth is then
collected in the launder separating valuable mineral particles from gangue.
Figure 2-4: Chalcopyrite floatation using Lab scale Denver flotation cell
9 |
Virginia Tech | Flotation is controlled by both chemical and physical variables. Chemical Variables
include reagent dosage, pulp pH etc. Physical variables include particle size, feed rate, pulp
level, slurry density, aeration rate, impellor speed, conditioning time and froth height. The
separation efficiency of the process is also effected by type and configuration of flotation cell.
The efficiency of flotation process is highly dependent on the surface properties of the
mineral and gangue. The surface chemistry of the particle is controlled by addition of several
chemical regents like frother, collector, activator, depressant and pH modifier. Although
complex ore (like lead-Zinc sulfide ore) flotation process utilizes all of the reagents, single ore
(like chalcopyrite or silica) flotation often utilizes only collector and frother. Collector is the
organic surfactant which attaches on the surface of the mineral making it more hydrophobic and
hence facilitates the bubble – particle attachment (Wills, 2005). Frother are hetero-polar
surfactants consist of polar head and non-polar tail which are used to stabilize the froth and
hence inhibit the breaking of particles loaded bubbles in the froth phase (Wills, 2005).
Flotation is most efficient for particle size ranging from 45µm to 150µm. At finer particle
size, the bubble particle collision decreases, while for big particle, the probability of detachment
from the bubble surface increases. Apart from particle size, bubble size also affects the efficiency
of flotation process. Bubble size not only affects the recovery but it also affects the selectivity of
the process.
Although selective attachment of valuable mineral to the air bubble is the major
phenomena for recovery of valuable mineral, the separation efficiency of the flotation process is
also affected by degree of entrainment in the water phase which passes through froth phase and
physical entrapment between particles in the froth (Wills, 1995). The entrainment of unwanted
gangue in the water phase is very common in the industry and can be minimized by adding
several circuits to the flotation (rougher, scavenger and cleaner).
2.3 Oil Agglomeration
Selective oil agglomeration is commonly used to recover or separate fine particles
dispersed in water through addition of oil. On contrary, as the particle size decreases below
10 |
Virginia Tech | 50µm, the flotation process becomes less desirable due to decrease in selective recovery of
desirable mineral particles and increase in processing cost (Hazra, Rao et al. 1988). Oil
agglomeration is more economic and most effective in very fine particle size range as the particle
is fully liberated.
2.3.1 History
Oil agglomeration has been used in mineral processing industry since 1920’s for cleaning
of coal. The earlier process involved agitating a mixture of dry solids and oil in water to separate
oil wetted particles form water-wetted gangue minerals. The process has been subjected to
several modifications since then. The process, in its early years of development, could not
compete with the then existing processes due to the high cost of creating the agglomerates and
low oil recoveries. However, in 1970’s the development of oil agglomeration process was given
impetus because of the sharp increase in oil prices. Though the principle of oil agglomeration
process remained same, several patents were filed during this period. The process was even
tested on several pilot plant units (Mehrotra et al., 1983). Other possible uses of the process,
besides cleaning of coal like dewatering and waste water cleaning have also been explored.
Several methods to employ oil agglomeration on low rank coal were also researched in late
1980’s (Ikura and Capes, 1988). The oil glut in late 1980’s once again made the oil
agglomeration obsolete as an oil substitute was no longer needed (Mehrotra et al., 1983).
2.3.2 Factors affecting Oil agglomeration
Oil agglomeration is the selective wetting of hydrophobic particle, in aqueous
suspension, by oil. The process highly depends on the surface hydrophobicity of the particle and
that of the oil. The interaction between hydrophobic liquid and hydrophobic particle is controlled
by the surface free energies at the three interfaces, mixing intensity (high shear or Low shear),
mixing time and amount of the hydrophobic liquid used. For most effective process, from
thermodynamic standpoint, the surface free energies at solid/water interface and oil/water
interface should be high than surface free energies at solid/oil interface (Mehrotra et al., 1983).
From kinetics point of view, speed of agitation and total agitation time are the deciding factors.
High sheer quickly forms agglomerates but limit the size of agglomerates to small diameters,
whereas low sheer helps in growth of the agglomerates size. Optimum oil dosage and mechanical
11 |
Virginia Tech | mixing allows coal particles to collide and stick tighter by oil bridges and the growth of
agglomerates in incorporated by mixing time. Low oil dosage creates loose flocs structure whilst
very high dosage results in formation of emulsion rather than agglomerates.
Selecting right kind of oil is very important for successful agglomeration process both
form economic point of view and as well as selectivity of the process. Oils are conveniently
divided amongst heavy oil and light oil depending on the viscosity of oil. Some researchers have
shown that heavy oils are too viscous to be dispersed in the slurry while other researchers have
shown that with longer mixing time, highly viscous oils are also capable of high recoveries.
Some researchers have also concluded that heavy oils are less selective. It is also cited in the
literature that very light oil could not make the particles sufficiently hydrophobic. In summary,
the selection of oil is deeply affected by the surface properties of particles to be agglomerated
and the method of mixing of oil (Mehrotra et al., 1983)
Since oil agglomeration is based on differences in surface properties of organic and
inorganic matter, first being hydrophobic and later being hydrophilic, oil agglomeration is very
difficult for low rank coal. The low rank coal can be made hydrophobic either by addition of
hydrophobic collecting oil, and/or by adding surface active agents or electrolytes which adsorbs
on the coal surface and also modify the surface charges on both coal surface and oil droplets and
hence aid towards the agglomeration process (Gűrses et al., 1997). It is mentioned in literature
that agglomerate rates for low rank coal can be improved by adding ions to the water phase
or a small peroxide-group-containing chemical to oil phase. Using vacuum bottoms (highly
viscous IPPL/Cold Lake) at high temperature for oil agglomeration of lignite has also been
reported successful with 90% combustible recovery (Ikura et al., 1988)
2.3.3 Methods Employed for Dewatering Of Agglomerates
The moisture content in agglomerates mainly consist of the water droplets trapped inside
the agglomerate and surface moisture. A significant amount of research has been done at
Virginia Tech on reducing the moisture content of the agglomerates. Several technologies such
centrifuge, ultrasonic probe, dewatering by screen have been utilized for breaking and
dewatering of agglomerates. Centrifugation was able to produce moisture as low as 7.5% with
combustible recoveries in range of eighties to nineties (kara 2008). A proprietary method of
12 |
Virginia Tech | breaking the agglomerates has also been developed at Virginia Tech. While separation
efficiencies for both ultrasonic probe and the novel method were near 85% percentile when used
for breaking coal agglomerates, it was discovered that method developed at Virginia Tech is the
best process for breaking agglomerates with product moisture as low as 2.05% (Smith, Sarah
2012).
2.4 Esterification of Low Rank Coal
Esterification is the process of production of esters by heating carboxylic acid (-COOH)
and alcohols in presence of a catalyst. The double arrows in the reaction shown below signify the
reversible nature of the esterification reaction. The process of esterification, thus reaches
equilibrium easily and to be able to drive the reaction more to the right, the Le Chatelier’s
principle is commonly adopted. According to this principle, upon increase of any one of the
reactant concentrations, the reversible reaction under consideration can be driven forward to
favor product formation.
The use of catalysts can also cause the reaction to move forward. In most cases, sulfuric
acid is preferred for use as a catalyst favoring product formation. However, as part of this study,
hydrochloric acid was used, since it is found to be more suitable for demineralization processes.
A small quantity of catalyst is used, since catalyst consumption does not occur during the
progress of a reaction and the rate is not impeded by the lack of the same at any point. Another
method to increase the process rate is to supply heat to the esterification process. For the scope of
this project, the economically optimum temperature of fifty degrees was used and all
experiments for esterification were conducted at this temperature.
OH H+ O R'
+
R C R' OH R C + H O (3)
2
O O
The rate of the esterification reaction is highly dependent on the structure of acid and
alcohol, temperature, choice of catalyst, amount of the catalyst and amount ratio of acid to
alcohol. Since esterification is equilibrium reaction, the rate of the reaction can also be increased
by removing water from the final product. It has been cited in literature that methanol is the most
13 |
Virginia Tech | reactive and can esterify most of the carboxyl groups and the efficiency of the process decreases
with increase in carbon chain length in alcoholic group as the structural hindrance increases.
It has been reported in the literature that for coal, the internal surface area inclusive of all
capillaries and micro-pores is much larger than the external surface area. To make low rank coal
suitable for esterification, the low rank coal is either subjected to demineralization or low
temperature oxidation in order to create carboxyl acid group (-COOH) on the surface of low rank
coal. The esterification process not only takes place at the coal surface, but the alcohol penetrates
inside the pores as well. The shorter the alcohol chain length the more penetration into the pores
increasing the hydrophobicity of low rank coal surface. The product obtained by esterification of
low rank coal is hence suitable for oil agglomeration.
2.4.1 Pretreatment 1: Demineralization by HCl
Demineralization is the process of oxidizing and de-ashing low rank coal by treating it by
aqueous solution of acid. The process is alternatively termed as chemical leaching and is one of
the simplest ways to reduce inorganic impurities from brown coal (Wang et al., 1986, Yang et
al., 1985). Shamaras et al. (1996) were the first to report the use of acid for efficient
demineralization process. The most commonly used acids for demineralization are hydrochloric
acid or other suitable mineral acids. The process effects both physical and chemical properties of
coal and hence altering the suitability of coal for different process (Kister et al., 1988). It has
been proven successful in removing all the chemically bound Na, Ca, Mg, Ba ions from the
surface of low rank coal (Young and Niksa 1988), but quartz remain intact due to their
insolubility in HCl (Vamuka et al., 2006). The mechanism of ion exchange at the surface of coal
due to presence of HCl is explained by equation 3, which explains the formation of carboxyl acid
as the main product. The penetration of acid in the pores of coal supports the formation of
carbonyl or carboxylic groups and causes disappearance of aliphatic group (Kister et al., 1988).
The extent of demineralization depends on acid concentration, particle size of the coal,
temperature at which demineralization is carried out and total reaction time (Kister et al., 1988).
Ash content in the final product reduces with increase in acid concentration (Vaccaro, Salvatore
2010). It has been reported by Kister et al. (1988) that as particle size decreases, more
14 |
Virginia Tech | oxygenated groups are formed upon demineralization, hence, showing the inverse trend for
oxidation (surface reticulation effect).
(4)
In the past, the demineralization process is used as pretreatment to remove the mineral
matter present on the surface of the coal so as to make low lank coal more hydrophobic and
hence suitable for subsequent flotation process (Henkley, David W., 1974). The process has also
been reported as pretreatment for preparing coal as a feed stock for liquefaction process (Polat,
Chander et al. 1995). The process is used to produce ultra-clean coal (UCC) at elevated
temperatures (Vijaya, N. et al., 2011)
2.4.2 Pretreatment 2: Low Temperature Oxidation
Oxidation of coal can be seen to be the consumption of oxygen by the fuel which occurs
as adoption onto the surface of the fuel. However, oxygen consumption and oxygen adoption
cannot be regarded to be the same. This is because the consumption of oxygen by coal is the
result of oxygen adsorption as well as physical and chemical reaction at the surface pores and
particle voids. Interaction between coal and oxygen can be classified as either physical or
chemical adsorption process. While the physical adsorption is similar to condensation being that
it is a nonspecific type of adsorption, chemical adsorption on the other hand is surface specific
and involved forces which are stronger than those which results in physical adsorption. Thus,
while physical adsorption can form either mono or multi-layer at the surface, chemical
adsorption restricts itself to monolayer at the pores (Wang, Dlugogorski et al. 2003).
Oxidation of low rank coal at low temperature (< ) is described by consumption of
oxygen on both external surface and internal surface of the coal, resulting in formation of solid
oxygenated complex species or humic acids, such as carboxyl (-COOH), carbonyl (-CO) and
hydroxyl(-OH) groups in the coal matrix. Surface reaction between the coal and the oxygen was
15 |
Virginia Tech | found to be a first order reaction. This was because unlike first order reaction wherein the
oxygen concentration would be expected to decrease exponentially with time, in the case of this
particular reaction, the oxygen concentration declined linearly with time until the surface
concentration became zero (Rehman, Hasan et al. 2007).
The oxygen adsorbs both chemically and physically on the surface of the coal. Thus, the
oxidation process is effected by several factors including particle size, reaction temperature,
composition and physical properties of coal. It is observed that the rate of oxidation reaction is
inversely proportional to the particle size (Wang, Dlugogorski et al. 2003). However, the
reaction rate reaches the maximum at some point after which the further decrement in size has no
effect on the reaction rate (Rehman, Hasan et al. 2007). The effect of temperature on oxidation
rate usually follow the exponential trend given by Arrhenius equation. It is reported in the
literature that at 35ºC, the dominant product species includes phenolic groups while at 70ºC, the
formation of carboxylic groups is predominant (Rehman, Hasan et al. 2007). Inherent moisture
also plays important role in the oxidation of low rank coal as it act as catalyst in the oxidation
reaction (Rehman, Hasan et al. 2007).
16 |
Virginia Tech | Chapter 3: Experimental Results & Discussion
The research conducted as part of this thesis delves on using of HHS separation technique
for ore processing. The idea is novel, in that, not many researchers have explored the possibility
of using this technique as a means for processing of low rank coal or for mineral
chalcopyrite. The traditional flotation techniques exploit the hydrophilic hydrophobic properties
of the ore-gangue mixture, using the bubbles to separate the hydrophilic gangue from the
hydrophobic ore. This study builds upon the traditional techniques and takes into account the
thermodynamics as well as reaction kinematics, which could influence the separation process.
Batch tests were conducted using different methods of hydrophobic – hydrophilic
separation techniques on copper ore, mono-sized silica beads and low rank coal from Wyoming
basin. This chapter is divided into three sections, one each for copper ore, silica beads and low
rank coal respectively. The subsection for each experimental process is further divided to include
description of the apparatus used for the process the methods employed, followed by the results
and a discussion.
For HHS process, n-pentane by Alfa Aesar, was used to produce agglomerates. Pentane is
colorless, immiscible liquid with density of 0.631 . The liquid and vapours are highly
flammable. Pentane has a boiling point of C. Pentane creates an explosive environment on
reaching the concentration of 1.8% to 8% by volume are. The pentane used in the following
experiment was HPLC grade and contains of minimum of 99% pentane by volume (Alfa Aesar
2009). Pentane was also used to create hydrophobic liquid phase for breaking and dewatering of
agglomerates. To dewater agglomerates, 8 inch laboratory sieves of varying apertures were used
or Buckner funnel is used. The Additional supplies and apparatuses used for the individual
method will be discussed in the method’s individual section.
17 |
Virginia Tech | 3.1 Chalcopyrite Processing by HHS
Two methods of recovering Chalcopyrite from copper ore were employed: froth flotation
and selective oil agglomeration followed by HHS. The experimental setup employed for each
method is discussed in their respective sections.
3.1.1 Froth Flotation
A small scale Denver flotation cell was employed. A rougher concentrate from Utah
copper plant was used in the experiments. The slurry was grinded in laboratory scale ball mill for
varied time in order to study the effect of size on flotation recovery. Varied amount of potassium
amyl xanthate was used as collector. The conditioning time was 2 minutes after addition of
collector. MIBC (Methyl isobutyl Carbinol) frother was added to the floatation cell as was
necessary to maintain a solid layer of froth. The pulp pH was kept approximately at 9.5 and
calcium hydroxide used as was necessary to maintain the pH. Approximately 1 liter of slurry
was floated at a time and the froth was manually paddled off until the floatable solids were
depleted.
3.1.2 Hydrophobic – Hydrophilic Separation
To form agglomerates, a Ninja kitchen blender was used. A variable speed control drive
was employed in conjunction with the blender so that both high and low sheer mixing
environment could be created. A volume of copper ore slurry was poured into the blender and
mixed on high speed setting for 50 to 90 seconds after addition of pentane. The volume of the
pentane varied from 30% to 40% by weight of solid. Immediately after addition of pentane, an
obvious phase separation could be observed. Heavier gangue containing water remained in the
lower portion of the blender while less-dense golden chalcopyrite rested on the top of the
blender. The high sheer mixing was followed by low sheer mixing, in order to facilitate the
growth of the agglomerates in the size and hence enhance the dewatering stage. The blender was
set to run on low sheer for additional 5 to 8 minutes by lowering the speed of the blender using a
variable speed controller. The long time mixing allowed large agglomerates to form. The
agglomerates were poured across a small-mesh screen (ranging between 140 and 400 mesh
depending on agglomerate size) to dewater agglomerates.
18 |
Virginia Tech | The hydrophobic – hydrophilic separation took place in vibrating mesh or morganizer
(Smith, Sarah 2012). It consists of a custom made glass column 5 inches high and 1.5 inches in
diameter. A model 2007E electrodynamic shaker produced by Modal Shop INC. was used to
disperse mineral particles into the pentane phase. A shaft with 2 mesh discs extended from the
bottom of the shaker. The lower plate had an opening of 0.5 millimeters and the upper plate had
an opening of 80 millimeters. The mesh disc pulsated at 35 hertz with an amplitude of 0.5 inches.
The experimental setup is shown in Figure 3-1. Before the separation testing began, a water –
pentane interface is formed in the glass column by pouring water at the bottom and pentane into
the upper portion. The mesh disc were lowered into the column in such a way that the lower disc
sat at the pentane / water interface. The overflow of morganizer was sent to the
evaporation/condenser unit.
Figure 3-1 Experimental setup for Hydrophilic – Hydrophobic separation
The copper-pentane overflow form vibrating mesh was poured on the double boiler in
order to evaporate the pentane (Smith, Sarah 2012). As the pentane evaporated, it travelled
upward through the Teflon pipes and was condensed by two condensers employed in the circuit.
In order for pentane to evaporate and condense properly, a pump was used to pump the displaced
gas from the evaporation beaker into the reagent tank so as to minimize the pentane loses. The
condenser unit is shown in Figure 3-2.
19 |
Virginia Tech | Figure 3-2: Evaporation Unit used to recover pentane
500 millimeter of copper slurry was added to the blender. High sheer mixing was started
immediately after adding 20 ml of pentane to the slurry. The pentane and slurry were mixed for
50 to 90 seconds to form small powder like agglomerates. It was then subjected to low sheer
mixing for 5 minutes so that agglomerates can grow in size. The agglomerates were dewatered
on the screen.
The mechanical shaker was turned on and set to operate at 35 Hz. The agglomerates were
removed from the screen using laboratory spoon-spatula and dropped on the morganizer.
Immediately, the agglomerates dispersed into the pentane phase and gangue could be seen falling
into the water phase. Pentane was slowly poured into the top of the column to overflow the
chalcopyrite-pentane mixture. The process of adding more agglomerates and overflowing the
column was repeated until enough amount of the product for grade analysis has been collected. If
needed, ports at the bottom of the column were used to drain the gangue containing water so that
the interface could be maintained at the same level as the lower disc.
20 |
Virginia Tech | The chalcopyrite/pentane product was then poured into the evaporation unit. The pentane
was evaporated and condensed into the separate chamber, leaving behind dry chalcopyrite. The
dry chalcopyrite was removed from the system and the final product moisture was calculated.
Results from these experiments are discussed in section 3.1.3
3.1.3 Results: Chalcopyrite Processing by HHS
The rougher concentrator from Utah copper plant with copper grade 15.9% was grinded
in ball mill to obtain different size fractions. A Microtrac - X100 size analyzer was used to
measure the size distributions of each size fraction and the 80 percent passing size of the
fractions were found to be 100µm, 51µm, 40µm, 22µm and 20µm. The fractions were then
subjected to both flotation and HHS and the weight recovery, grade and copper recovery was
calculated at 17.6 lb. /ton dosage of potassium amyl xanthate, based on the grade analysis
obtained from FLSmidth analytical lab, Salt Lake City.
Table 3-1 shows the results obtained by flotation and HHS. As the particle size decreases
below 40µm , all the copper recovery, copper grade and weight recovery increases for HHS
while the trend is opposite for flotation . Figure 3-3(a) shows the ability of HHS process to
produce high recoveries even at fine size fractions which is in accordance with the findings cited
in literature for oil agglomeration (House, C. I. and C. J. Veal, 1989).Figure 3-3(b) shows the
decrement in copper grade, copper recovery and weight recovery for flotation process after
particle size decreases below 40µm. With decrement in particle size, the concentration of
surfactant per unit area decreases for both flotation and agglomeration but increasing the
agglomeration time and low sheer mixing increases the recovery of HHS process and hence
contribute towards the high recoveries of HHS process at lower size fractions. Table 3-2 focuses
on the results obtained by flotation and HHS for particle size of 22µm.
21 |
Virginia Tech | The agglomerates, formed using copper ore of 80% passing particle size of 22µm, were
dewatered by pouring them into the morganizer. The pentane overflow saturated with copper ore
particles was collected and poured into the condenser unit for pentane recovery. Moisture content
of the resultant product was calculated. Table 3-3 shows the results of the experiment conducted
on copper agglomerates to decrease their moisture content.
Table 3-3: Moisture content and solid recovery of final product from Morganizer
Agglomerate Product Moisture % Solid
Test No
Moisture (%) (%) Recovered
1 59.02 0.14 1.62
2 48.56 0.58 2.02
4 65.19 0.49 1.98
A second set of experiments were performed to study the effect of changing the
potassium amyl xanthate (KAX) dosage ranging from 2.2 lb. /ton to 26.4 lb. /ton on the weight
recovery of both flotation and HHS process. Table 3-4 shows the results obtained by flotation. It
is evident that as the particle size decreases beyond 40µm, flotation weight recovery decreases
irrespective of the collector dosage. Table 3-5 shows the results obtained for hydrophobic-
hydrophilic separation process. The weight recovery of the HHS process increases with increases
in hydrophobizing reagent dosage even at the finer size, showing that HHS process is suitable for
fine particle sizes as well. Figure 3-4 (a) and (b) shows the variation in weight recovery with
change in KAX dosage for different particles sizes.
Table 3-4: Effect of KAX dosage on weight recovery from Flotation
Amyl Flotation Wt. Recovery For Different D80 Particle Sizes (%)
Xanthate Particle Particle Particle Particle Particle
Dosage Size Size Size Size Size
(Lbs./ton) 110 µm 51 µm 40 µm 22 µm 20 µm
2.2 84.8 82.4 81.6
4.4 86.3 84.4 85.8 50.1 60.1
8.8 87.0 85.4 86.6 55.6 67.4
17.6 88.3 87.2 87.5 68.9 68.7
26.4 88.7 88.1 88.6 63.7 70.9
23 |
Virginia Tech | 3.2.1 Froth Flotation
The flotation tests were conducted in a one liter laboratory scale Denver floatation cell.
The material used in the flotation was technical quality glass spheres obtained (nominal size
35µm and 75µm) obtained from pottery industry. These samples were hydrophobized in a 4 x 10-
6 M solution of dodecylamine hydrochloride (DAH) collector, which was prepared in advance in
bulk by completely dissolving DAH in pure ethanol so as to ensure the uniformity in all tests.
The conditioning time was 2 minutes after addition of collector. MIBC (Methylisobutyl
Carbinol) frother was added to the floatation cell as was necessary to maintain a solid layer of
froth. Approximately 1 liter of slurry was floated at a time and the froth was manually paddled
off until the floatable solids were depleted.
3.2.2 Hydrophobic – Hydrophilic Separation
To form agglomerates, a Ninja kitchen blender was used. A variable speed control drive
was employed in conjunction with the blender so that both high and low sheer mixing
environment could be created. Silica along with desired volume of water was poured into the
blender and mixed on high speed setting for 50 to 90 seconds after addition of pentane. The
volume of the pentane varied from 15% to 20% by weight of solid. Immediately after addition of
pentane, an obvious phase separation could be observed. The agglomerated silica particles floats
on the top of the blender. The high sheer mixing was followed by low sheer mixing, in order to
facilitate the growth of the agglomerates in the size and hence enhance the dewatering stage. The
blender was set to run on low sheer for additional 5 to 8 minutes by lowering the speed of the
blender using a variable speed controller. The agglomerates were poured across a 140 mesh
screen to dewater agglomerate. The agglomerates were dewatered either by Morganizer or using
Buckner funnel.
100 grams of silica particles along with 500 millimeter of water were added to the
blender. The silica particles were hydrophobised either by 4 x 10-6 M solution of dodecyl amine
hydrochloride (DAH) collector, which was prepared in advance in bulk by completely dissolving
DAH in pure ethanol so as to ensure the uniformity in all tests, or by coating silica beads with
octadecyltrichlorosilane (OTS, 95% purity Alfa Aesar) followed by rinsing the hydrophobized
surface with toluene (99% purity, Fisher Chemical). High sheer mixing was started immediately
25 |
Virginia Tech | after adding 30 ml of pentane to the slurry. The pentane and slurry were mixed for 50 to 90
seconds so as to form small powder like agglomerates. It is then subjected to low sheer mixing
for 5 minutes so that agglomerates can grow in size. The agglomerates were dewatered on the
140 mesh screen.
For dewatering of agglomerates by Morganizer, the mechanical shaker was turned on and
set to operate at 35 Hz. The agglomerates were removed from the screen using laboratory spoon-
spatula and dropped on the morganizer. Immediately, the agglomerates dispersed into the
pentane phase (Figure 3-5) and the hydrophobic silica particles stay in the pentane phase.
Pentane was slowly poured into the top of the column to overflow the silica-pentane mixture.
The process of adding more agglomerates and overflowing the column was repeated until
enough product for moisture analysis had been collected. The silica/pentane product was then
poured into the evaporation unit. The pentane was evaporated and condensed into the separate
chamber, leaving behind dry silica beads. The dry silica was removed from the system and the
final product moisture was calculated.
Figure 3-5: Complete dispersion of silica particles in pentane phase (above pentane water interface)
in Morganizer
Buckner funnel was employed as another method for dewatering of agglomerates. The
funnel was connected to pump in order to pump out the water / pentane. The Buckner funnel was
lined with Whatman ashless filter paper and the agglomerates were poured over the filter paper.
Pentane was continuously poured over the agglomerates until the overflow from pump no more
contain water droplets. The filter cake is then subjected to moisture determination.
26 |
Virginia Tech | A different set of experiment was performed to study the effect of both hydrophobizing
agent (OTS or DAH) and dewatering technique on moisture of the final product from HHS
process. Table 3-7 shows the results of the experiments.
With OTS coated silica, moisture content less than 1% is achievable using Morganizer.
Whereas with DAH coating, the final product has moisture content varies from 5% to 6%. Same
trend is observed while using filtration for dewatering which confirms that OTS coated silica is
more hydrophobic than DAH coated silica. Results also show that dewatering process is more
efficient for particle size of 75µm than for particle size of 35µm.
Table 3-7: Moisture contents of the product of HHS process for OTS coated vs. DAH coated
Size (µm) Technique Moisture (%)
OTS Coated DAH
35 Morganizer .680 5.17
35 Morganizer 1.02 6.12
35 Morganizer 0.92 5.89
35 Filtration 9.15 12.42
35 Filtration 8.12 9.95
35 Filtration 7.89 15.12
75 Filtration 3.12 4.99
75 Filtration 3.89 5.82
75 Filtration 2.18 6.13
3.3 Low Rank Coal Processing by Hydrophobic-Hydrophilic Separation
The sample used for the testing in this section was obtained from Wyoming basin and
had an as received (AR) moisture content of 28%, 8398 AR BTU/lb value and 8.5% dry ash
content. The hydrophobization of low rank coal was achieved using three techniques. In the first
technique, the low rank coal was coated with reagent U (Span 80) and Diesel mixture (1:2 ratio)
or Span 20 and Diesel mixture. In second technique, the coal was first demineralized using HCl
acid and then subjected to esterification. In third method, the coal was first subjected to low
temperature oxidation followed by esterification. The hydrophobized coal was then
agglomerated in Ninja kitchen blender, which was used in conjunction with the variable speed
drive to control the speed of the blender. The hydrophobized coal was poured into the blender
and mixed on high speed setting for 50 to 70 seconds after addition of pentane. The volume of
the pentane varied from 20% to 30% by weight of solid. The agglomerated low rank coal
28 |
Virginia Tech | particles float on the top of the blender. The high sheer mixing was followed by low sheer
mixing, in order to facilitate the growth of the agglomerates in the size and hence enhance the
dewatering of agglomerates. The blender was set to run on low sheer for additional 5 to 30
minutes by lowering the speed of the blender using a variable speed controller. During low sheer
mixing, a small amount of pentane was also added to facilitate the bridging between the
agglomerates and also to incorporate for the pentane loss occur during long low sheer mixing.
The agglomerates were poured across a 140 mesh screen.
For dewatering, the agglomerates were dropped into the 250 ml glass beaker, filled with
100 ml of pentane, using spatula and then agitated by hands for 5 to 10 minutes to facilitate
breaking of agglomerates and dispersion of coal in pentane. The coal/pentane product was then
poured into the evaporation unit. The pentane was evaporated and condensed into the separate
chamber, leaving behind dry coal. The dry coal was removed from the system and moisture of
final product was calculated using ASTM method for moisture determination. The products are
then packed in sample bags and sent to Precision testing lab Inc, West Virginia for conducting
BTU analysis. The ash analyses was performed using LECO Model 601-400-600.
A set of experiments were conducted to study the effect on immersion of sample in
pentane. Theoretically pentane dislodges the water molecules from macro pores and capillaries
of the fuel sample in consideration. Initially the coal sample was immersed in pentane.
Continuous stirring using spatula was done to ensure that adequate molecular interaction took
place between the coal and pentane. The pentane/coal solution was then poured into the
evaporation unit. The dry coal was removed and analyzed for moisture content and BTU values.
3.3.1 Hydrophobization of Low Rank Coal Using Reagent Surfactant
In this technique, 150 ml of coal slurry with 20% solids was conditioned with desired
amount of surfactant, being Span 80 or Span 20 solution in diesel in 1:2 ratio, for 5 minutes to
ensure the maximum hydrophobization of the coal surface. Both Span 80 and Span 20 mixture is
prepared in large quantity in advance by mixing 1 amount of surfactant with 2 amounts of diesel
29 |
Virginia Tech | in order to ensure uniformity in all tests. The hydrophobized coal was then subjected to the
agglomeration followed by dewatering process.
Figure 3-7 shows the (a) agglomerates formed using this technique and (b) tailings
obtained upon screening the agglomerates.
Figure 3-7: (a) Agglomerates formed using Span 80 (b) Tailing obtained by screening the
agglomerate
3.3.2 Hydrophobization of Low Rank Coal by Demineralization Followed by Esterification
In this technique, 20 grams of coal was mixed with 100 ml of 0.1M HCl in a 250 ml
beaker. The beaker was then placed on Thermo Scientific stirring hot plate. The
demineralization/ acid washing process was carried out at 50ºC along with continuous stirring.
The processing time was varied from 1 to 4 hours. Figure 3-8 shows the experimental setup for
demineralization. The acid washed coal was then filtered using conical funnel lined with
Watman ashless filter paper. The filter cake was then poured in the 250ml glass beaker and to
that 100 ml of ethanol was added along with 10µl of 1M HCl. The esterification was also carried
out at 50ºC along with the continuous stirring for 3 hours. The ethylated coal was then filtered
using conical flask and the filter cake is then subjected to agglomeration followed by dewatering.
30 |
Virginia Tech | Figure 3-8: Experimental setup for demineralization
3.3.3 Hydrophobization of Low Rank Coal by Oxidation Followed by Esterification
In this technique, 20 grams of coal was placed in oven at 70ºC for time ranging from 1
hour to 8 hour. The oxidized coal was then poured in the 250ml glass beaker and to that 100 ml
of ethanol was added along with 10µl of 1M HCl. The esterification was carried out at 50ºC
along with the continuous stirring for 3 hours. The ethylated coal was then filtered using conical
flask and the filter cake is then subjected to agglomeration followed by dewatering.
3.3.4 Results: Low Rank Coal Processing by HHS
Hydrophobization of Low Rank Coal Using SurfactantThe method was successful in both
decreasing the moisture content as well as increasing the BTU/lb value of the final product.
Table 3-8 (a) and (b) shows the result of HHS testing on Wyoming coal using reagent U
and span 20. From the both tables, it can be seen that increasing the reagent dosage resulted in
increase of BTU values and decrease of moisture content for the Wyoming coal samples which
were tested. Additionally, increasing the low sheer mixing time also resulted in decrease of
moisture content of the product. Both the reagents were successful in hydrophobizing the coal
31 |
Virginia Tech | surface leading to the finding that sorbitan group was capable of adsorbing on the surface of low
rank coal thus altering the hydrophobicity of the surface.
Table 3-8 (a): Results of HHS Testing on Wyoming Coal Using Reagent U (Span 80)
Reagent U Shear Product Ash Product
Agglomerate Product Combustible
Dosage Time Moisture Rejection AR
Moisture% Ash% Recovery%
(lbs/ton) (min) % % BTU/lb
33.3 30 27.1 5.8 20.8 66.2 53.5 9814
33.3 15 44.6 6.2 38.2 95.9 26.7 7562
44,7 30 42.6 6.3 33.0 71.1 45.2 8194
50 30 28.1 6.0 4.1 95.5 29.4 11759
50 15 43.6 5.2 6.7 90.5 42.9 11543
50 5 46.2 5.8 6.0 87.1 38.8 11560
Table 3-8 (b): Results of HHS Testing on Wyoming Coal Using Span 20
Reagent U
Shear Agglomerate Product Product Product
Dosage
Time (min) Moisture% Ash% Moisture% AR BTU/lb
(lbs/ton)
55.5 30 62.9 5.9 5.6 11880
55.5 20 50.3 5.9 7.9 11147
55.5 5 63.6 6.6 17.7 9134
44.5 30 67.9 6.2 5.9 11739
3.3.4.1 Hydrophobization of Low Rank Coal by Demineralization Followed by
Esterification
The low rank coal obtained from the Wyoming basing was sieved into different size
fractions using sieves having openings of 75µm, 300µm, 600µm, 1.18mm, and 6.3mm. All the
size fractions are then subjected to acid washing for 4 hours followed by esterification for 3
hours. Table 3-9 shows the effect of particle size on moisture content and as received BTU/lb of
the final product. From the table, It can be seen that product BTU increases with increases in
32 |
Virginia Tech | particle size. It can also be deducted from the table that moisture content of the final product
decreases with increase in particle size.
Figure 3-9 shows the increasing trend for BTU values and decreasing trend for product
moisture.
Table 3-9: Effect of particle size on Moisture content and product AR BTU/lb.
Particle Particle Product % AR
Agglomerates Product Product
Size Range mean AR BTU
Moisture% Ash% Moisture%
(µm) Size(µm) BTU/lb. increase
-350+75 212.5 40.3 3.20 9.92 10827 28.92
-600+350 475 25.62 3.20 9.82 11019 31.21
-1180+600 890 28.34 2.87 8.4 11216 33.56
-
3740 37.63 2.30 6.27 11529 37.28
6300+1180
Effect of acid concentration was also studied in a different set of experiments. Low rank
coal in the size fraction 1.18mm to 600µn was subjected to demineralization using HCl with
concentrations 0.005M, 0.01M, 0.05M, 0.1M and 0.15M. Table 3-10 shows the results for this
experimental set. From Table, it can be seen that steady increasing of the acid concentration
results in the corresponding increase of product BTU values along with a trending decrease in
both product moisture content and ash values. However, these trends continue up to an optimum
value of 0.1M concentration of HCl, beyond which increasing concentration brought about
negligible changes in the characteristics. The trends can also be noted from the Figure 3-10.
33 |
Virginia Tech | Table 3-10: Effect of acid concentration on the product characteristics
Acid Agglomerates Product Product Product % AR BTU
Concentration(M) Moisture% Ash% Moisture% AR BTU/lb. increase
0.005 21.27 6.48 16.93 9448.38 12.51
0.01 29.06 2.86 15.24 9869.45 17.52
0.05 22.18 3.20 14.67 9888.04 17.74
0.1 20.34 2.87 8.40 11215.50 33.55
0.15 18.32 2.84 8.23 10801.33 28.62
A set of experiments were conducted to study the relation between acid wash time and
factors such as ash, BTU and moisture content of the final product. The trends can be seen in
Table 3-11. On increasing acid wash time, the ash content was found to lower significantly
however, on further increase of acid wash time no such marginal difference was seen as the
values were found to cluster around 3%. The values for moisture content was found to decrease
with decrease in acid was time, correspondingly the values for product BTU was found to
increase with increase in acid wash. Figure 3-11 also shows the continuous increase in BTU
values and decrement in product moisture in correspondence with the increase in
demineralization time.
11500 30
11000
25
P
d e v 10500 20 r o d
e u
ic 10000 c
e t
R 15 M
s 9500 o
A is
b 10 t u
l/ U 9000 r e
T (
B 5 %
8500 )
BTU/lb As Recieved
Product Moisture
8000 0
0 1 2 3 4
Acid Wash Time (Hour)
Figure 3-11: Effect of acid wash conditioning time on product BTU and moisture content
35 |
Virginia Tech | Table 3-11: Effect of acid wash conditioning time on the product characteristic
Acid Wash Agglomerates Product Product Product % AR BTU
Time Moisture% Ash% Moisture% AR BTU/lb. increase
0 - 8.5 28 8398 0.00
1 28.39 3.36 18.28 9680 15.27
2 25.62 3.40 10.64 10806 28.67
4 39.54 3.20 9.92 10826 28.91
Another set of experiments were conducted to find the optimum alcohol for esterification
process. It could be seen that on increasing the hydrocarbon chain length in alcohol, there was a
corresponding increase in product moisture and a decreasing trend was observed in product BTU
values signifying that methanol is the most appropriate alcohol for the esterification of low rank
coal followed by ethanol and higher chain length. The trends can be noticed in Table 3-12. The
same has been diagrammatically represented in the Figure 3-12. The findings can be justified by
the fact that the smaller the chain length of the alcohol, more easily it can penetrate inside the
pore structure and form ester at the capillaries and hence making coal more hydrophobic.
12000 25
11500
20
11000
b
L
/ U 10500 M
15
T o
B
t c
10000
is
t u
u r
d
o
9500 10 e
(
r %
P )
9000
R
A 5
AR BTU
8500
Moisture (%)
8000 0
1 2 3 4 5
No. of Carbon in Alcohol Chain
Figure 3-12: Effect of increasing alcohol chain length on product BTU and moisture content
36 |
Virginia Tech | decreases and correspondingly product moisture increases. The immersion time of particles in
pentane was increased to overnight in the second set of experiment and the final product
moisture is calculated. It can be seen from Table 3-14 that increasing the immersion time has
positive impact on moisture reduction which confirms the replacement of capillary moisture and
interparticle moisture by pentane. From Figure 3-13, it can be noticed that increases in
immersion time have positive impact on moisture reduction of middle range particle sizes
because the increment in immersion time helps the penetration of pentane into the pores and
hence cause the moisture reduction. However, the moisture content of very fine sized particles
and very coarse sized particles are unaffected by the increment in immersion time. It can be
justified by the fact that for very fine particle size, the pentane penetration into the pores is quick
due to large surface area and hence the increment in immersion time didn’t help much. For
coarse particles however, the penetration is very difficult due to minimum surface area.
25
20
) % 15
(
e
r
u
t s 10
i
o
M
5
15 mins
18 hours
0
0 2000 4000 6000 8000 10000 12000
Particle Size (µm)
Figure 3-13: Effect of immersion time in pentane on different particle sizes
3.3.4.2 Hydrophobization of Low Rank Coal by Oxidation Followed by Esterification
The low rank coal obtained from the Wyoming basing was sieved into different size
fractions using sieves having openings of 75µm, 300µm, 600µm, 1.18mm, and 6.3mm. All the
38 |
Virginia Tech | size fractions were then subjected to low temperature oxidation for 2 hours followed by
esterification for 3 hours.
Table 3-15 shows the effect of particle size on moisture content and as received BTU/lb
of the final product. It can be seen from the table that the process has minimal effect of ash
content of the feed. However, with decrease in particle size, the significant increase in BTU
values was noted along with significant decrease in moisture content of the feed coal which is in
accordance with the finding in literature (Rehman, M. et al, 2007) supporting the hypothesis that
smaller particles have large surface area for oxygen exposure. The trend in BTU values and
moisture content of the coal can be seen in Figure 3-14.
11000
25
10500
P
d e 20 r o
v d
e u
ic 10000 c
e 15 t
R M
s o
A 9500 is
b 10 t u
l/
U
r
e
T (
B 9000 5 %
BTU/lb As Recieved )
Product Moisture
8500 0
0 500 1000 1500 2000 2500 3000 3500 4000
Particle Size (m)
Figure 3-14: Effect on particle size product properties obtained from HHS using low temperature
oxidation
Table 3-15: Effect of particle size on product BTU and moisture content
Particle Size Particle mean Agglomerates Product Product Product % AR BTU
Range (µm) Size(µm) Moisture% Ash% Moisture% AR BTU/lb. increase
-350+75 212.5 27.3 7.2 9.3 10755 28.07
-600+350 475 15.2 7.4 7.6 10952 30.41
-1180+600 890 19.4 7.8 10.4 10225 21.80
3740 20.8 7.5 11.2 10029 19.4
-6300+1180
39 |
Virginia Tech | Chapter 4: Conclusion and Recommendation
A hydrophobic–hydrophilic separation (HHS) technique has been developed and tested
for processing of mineral particles in ultrafine size range as well for upgrading low rank coals. In
case of mineral processing in ultra-fine size range, the HHS process was found to be promising,
and can be used as an alternative to flotation for particles finer than 40 µm in size. In case of low
rank coal, the HHS process can be used to increase the heating (BTU) values decreasing the
moisture content.
The study showed that application of the HHS process for the processing of chalcopyrite
gave copper recoveries approximately 20% higher than flotation particularly at finer particle
sizes. Even at coarser particle sizes, the HHS process gave weight recoveries that are comparable
to flotation. However, the copper grades were lower.
It has been observed that the HHS tests conducted on a mono-sized silica beads gave a
minimum 25% higher recoveries than flotation at finer particle sizes. It has been found also that
the higher the hydrophobicity of silica particles, the higher the silica recoveries and the lower the
moisture of the products. The silica particles coated with OTS gave better results than the silica
particles coated with DAH, which can be attributed to the hydrophobicity difference.
It has been found that the HHS process is also useful for separating water from a low-
rank coal and hence increasing its heating value. The process requires that the low-rank coal be
rendered hydrophobic by using appropriate reagents. When using the Reagent U as
hydrophobizing agent, the moisture of the coal was reduced to less than 10% by weight from the
moisture content of 28% by weight in the raw coal. As a consequence, the heating value of the
processed coal was increased substantially. However, process required large amounts of the
hydrophobizing agents for the HHS process to be effective.
In the present work, a low rank coal was hydrophobized by esterification with short-chain
alcohols before subjecting the coal for the HHS process. Before the esterification, the low-rank
coal sample was treated by an acid treatment to increase the surface concentration of the
carboxylic acid groups. The acid treatment also helped ash rejection possibly by a dissolution
mechanism. Another methods employed to increase the surface concentration of carboxylic acid
41 |
Virginia Tech | Measurements, Modeling and Analysis of High Pressure Gas Sorption in Shale and Coal
for Unconventional Gas Recovery and Carbon Sequestration
Xu Tang
ABSTRACT
In order to exploit unconventional gas and estimate carbon dioxide storage potential in shale
formations and coal seams, two key questions need to be initially answered:
1) What is the total gas-in-place (GIP) in the subsurface reservoirs?
2) What is the exact ratio between bulk gas content and adsorbed gas content?
Both questions require precise estimation of adsorbed phase capacity of gases (methane and carbon
dioxide) and their adsorption behavior in shale and coal. This dissertation therefore analyzes
adsorption isotherms, thermodynamics, and kinetics properties of methane and carbon dioxide in
shale and coal based on experimental results to provide preliminary answers to both questions.
It was found that the dual-site Langmuir model can describe both methane and carbon dioxide
adsorption isotherms in shale and coal under high pressure and high temperature conditions (up to
27 MPa and 355.15K). This allows for accurate estimation of the true methane and carbon dioxide
GIP content and the relative quantity of adsorbed phases of gases at in situ temperatures and
pressures representative of deep shale formations and coal seams. The concept of a deep shale gas
reservoir is then proposed to optimize shale gas development methodology based on the successful
application of the model for methane adsorption in shale.
Based on the dual-site Langmuir model, the isosteric heat of adsorption is calculated analytically
by considering both the real gas behavior and the adsorbed phase under high pressure, both of
which are ignored in the classic Clausius–Clapeyron approximation. It was also found that the
isosteric heat of adsorption in Henry’s pressure region is independent of temperature and can serve
as a quantified index to evaluate the methane adsorption affinity on coal.
In order to understand the dynamic response of gas adsorption in coal for carbon sequestration,
both gas adsorption kinetics and pore structure of coal are investigated. The pseudo-second order
model is applied to simulate the adsorption kinetics of carbon dioxide in coals under different
pressures. Coal particle size effects on pore characterization of coal and carbon dioxide and
nitrogen ad/desorption behavior in coal was also investigated. |
Virginia Tech | Measurements, Modeling and Analysis of High Pressure Gas Sorption in Shale and Coal
for Unconventional Gas Recovery and Carbon Sequestration
Xu Tang
GENERAL AUDIENCE ABSTRACT
Shale gas is natural gas that is found trapped within subsurface shale formations, and the in-situ
pressure and temperature of shale formations can go up to 27MPa and 86℃. Shale gas, the main
component of which is methane, mainly consists of adsorbed phase and free compressed gas in
shale formations. The adsorbed phase accounts for 20-85% of the total gas-in-place resource. Thus,
the estimation of amount of methane adsorbed in shale under in-situ conditions are extremely
important for determining the total gas-in-place quantity and the working life of a shale gas
production well and its economic viability. This work provides a method for accurate estimation
of the shale gas-in-place resource under in-situ shale formation conditions. The method is based
on laboratory methane adsorption test data in shale at high pressure (up to 27MPa) and high
temperature (up to 82℃) conditions. According to this method, it was found that for depths greater
than 1000 m (> 15 MPa) in the subsurface, the shale gas resources have historically been
significantly overestimated. For Longmaxi shale (2500 – 3000 m in depth), classical approaches
overestimate the GIP by up to 35%. The ratio of the adsorbed phase compared to the free gas has
been significantly underestimated.
Shale gas production follows pressure depletion of shale formations. The pressure depletion
process allows methane in the adsorbed phase to become free gas, which is known as the physical
desorption process. Desorption is an endothermic process while adsorption is an exothermic
process, both of them are reversible. Thus, the heat transfer process during shale gas production
requires a thermodynamic analysis of methane adsorption in shale. This work investigates the
isosteric heat of adsorption for methane in shale by considering both the real gas behavior and the
volume effect of the adsorbed phase, not previously considered for methane in shale. The
temperature dependence as well as the uptake dependence of the isosteric heat can be readily
investigated by the applied method. This study lays the foundation for future investigations of the
thermodynamics and heat transfer characteristics of the interaction between high pressure methane
and shale.
This work also investigates gas adsorption kinetics properties in coal and the particle size effect
on pore characterization of coal using the gas adsorption approach. Results show that particle size
of coal samples can significantly influence the sorption behavior of gas in coal, which finally
affects pore characterization of coal. It is difficult to characterize the pore structure of coal using
only one coal particle size. Carbon dioxide adsorption kinetics in coal, which can be modelled by
the pseudo-second order model, is a combination of both bulk diffusion-controlled and surface
interaction-controlled processes; the former dominates the initial stage while the latter controls the
majority of the overall process. |
Virginia Tech | ACKNOWLEDGEMENT
First, I would like to thank my advisor, Dr. Nino S. Ripepi for giving me the opportunity to
complete this dissertation and for providing me the best study and research conditions in Virginia
Tech. Without his continuous encouragement and extensive discussion on this topic, this work
cannot be completed.
I am very thankful to Dr. Gerald H. Luttrell, Dr. Kray Luxbacher, Dr. Matthew Hall and Dr. Cheng
Chen for being the examiners of my dissertation. I would also like to give my special thanks to Dr.
Matthew Hall (University of Nottingham, UK) for supervising me when I was an exchange student
in the University of Nottingham. His tremendous knowledge and friendliness helped me to
understand the fundamental principle of gas adsorption.
Furthermore, I would like to thank my colleges in the mining department and VCCER (Virginia
Center for Coal & Energy Research) and for their help and support in the laboratory works: Charles
Schlosser, Kyle Louk, Ellen Gilliland, Scott Jeter, Cigdem Keles, Joseph Amante, Flora Lado,
Marina Rossi, Biao Li, Ming Fan, Kaiwu Huang. I gratefully acknowledge Dr. Alex O. Aning
(Materials Science and Engineering, Virginia Tech), Dr. Emily Sarver and Dr. Roe-Hoan Yoon
for their permission to use their laboratory instruments.
I would also like to thank several collaborators for their help in conducting the high pressure gas
adsorption tests in shale and coal and for their valuable discussions on this work: Dr. Zhaofeng
Wang, Mr Lingjie Yu and Dr. Nicholas P. Stadie.
Finally, I would like to thank and dedicate this dissertation to my family for their constant supports
throughout all those years. Special thanks go to my wife, Min Chu, for all her encouragement and
support all the time.
v |
Virginia Tech | PREFACE
This dissertation is submitted as a completion of the degree of Doctor of Philosophy at Virginia
Polytechnic Institute and State University. The research described here was conducted by the
author, Xu Tang, under the supervision of Dr. Nino S. Ripepi in the Department of Mining &
Minerals Engineering at Virginia Polytechnic Institute and State University.
This dissertation mainly comprises three fundamental works for high pressure methane adsorption
in shale for deep shale gas resource estimation (Chapter 2), thermodynamic analysis for high
pressure gas adsorption in shale and coal (Chapter 3), as well as gas adsorption kinetics analysis
and pore characterization of coal (Chapter 4).
In Chapter 1, the basic concepts for adsorption related phenomenon are briefly discussed for shale
gas development and geological sequestration of carbon dioxide in unconventional gas reservoirs.
The objective of this dissertation is also presented.
Chapter 2 represents a compilation of three separate manuscripts focusing on the methane
adsorption model in shale and its application for shale GIP resource estimation in deep formations.
First, analysis of laboratory data for methane adsorption in shale (303 - 355 K and up to 27 MPa)
proves the single-site Langmuir model becomes invalid under high pressure conditions. Thus, a
new concept, the deep shale gas reservoir, is introduced for the shale gas industry based on the
observed methane adsorption behavior in shale under high pressure conditions. The deep shale gas
reservoir study requires a new high pressure adsorption model. A dual-site Langmuir model is then
introduced to interpret observed methane adsorption behavior in shale. This model can not only
interpret all observed test phenomena but also is superior to available adsorption models in
literature. The proposed model herein allows accurate estimations of the true shale GIP resource
and the relative quantity of adsorbed methane at in situ temperatures and pressures representative
of deep shale formations.
Chapter 3 is composed of three manuscripts focusing on thermodynamic feature of methane
adsorption in shale and carbon dioxide adsorption in coal. On the one hand, the isosteric heat of
adsorption within Henry’s region is calculated for methane adsorption in coal, which can be used
to describe the adsorption affinity of different types of coal. On the other hand, the isosteric heat
of adsorption, considering both the real gas behavior and the contribution of adsorbed gas phase,
xiii |
Virginia Tech | is calculated analytically for high pressure methane adsorption in shale based on the dual-site
Langmuir adsorption model. Both the adsorption model and thermodynamic analysis for
supercritical carbon dioxide adsorption in coal are also explored in order to support an on-going
carbon dioxide sequestration field test in unminable coal seams.
Chapter 4 contains two manuscripts. The first one studies the carbon dioxide adsorption kinetics
properties of crushed coal using the pseudo-second order (PSO) model. Understanding both the
pore feature of coal and the dynamic response of coal to carbon dioxide sorption are important for
optimizing carbon dioxide injection methods in unconventional reservoirs such as coal seams and
shale formations to enhance natural gas production. The second exhibits how different coal particle
sizes used in the low pressure gas adsorption methods affects the pore characterization of coal
samples.
In Chapter 5, conclusions from this dissertation are summarized. Suggestions for future work, that
have not been covered in this work but deserve attention in future research, are presented.
The Appendix section contains both supplemental materials for this dissertation and the copyright
release documents from publishers for three published papers.
Part of this dissertation has been published in the following peer-reviewed journals:
Tang, X., Ripepi, N., Stadie, N. P., Yu, L., & Hall, M. R. (2016). A dual-site Langmuir
equation for accurate estimation of high pressure deep shale gas resources. Fuel, 185, 10-
17.
Tang, X., Wang, Z., Ripepi, N., Kang, B., & Yue, G. (2015). Adsorption affinity of
different types of coal: mean isosteric heat of adsorption. Energy & Fuels, 29(6), 3609-
3615.
Tang, X., Ripepi, N., & Gilliland, E. (2015). Isothermal adsorption kinetics properties of
carbon dioxide in crushed coal. Greenhouse Gases: Science and Technology. DOI:
10.1002/ghg.1562.
xiv |
Virginia Tech | Chapter 1 Introduction
1.1 Background
Unconventional gas now plays a significant role in the world energy profile because of the boom
in shale gas production over the past ten years. With the development of horizontal drilling
technology coupled with hydraulic fracturing, shale gas (primarily methane) has become the major
component of the total natural gas production in the United States [1-6]. Based on the successful
experience in the United States, different countries have launched a variety of projects to explore
their shale gas resource potential [7-8]. The principle of shale gas exploration and production has
followed the methodology developed for coalbed methane (CBM) because shale gas and coalbed
methane have some similar features. For example, gas in shale formations and coal seams under
reservoir conditions are mainly composed of adsorbed methane and bulk methane, which makes
them distinguishable from other gases like tight gas and conventional natural gas. Since the
adsorbed methane makes up a large portion of the total gas-in-place (GIP) resource for both shale
gas and CBM, it is imperative to understand the relationship between the adsorbed methane
quantity and the free methane quantity at reservoir conditions. Thus, the methane adsorption
behavior in shale and coal needs to be fully understood in order to accurately estimate the
CBM/shale gas resource.
In order to decrease greenhouse gases in the atmosphere like carbon dioxide, geological
sequestration of carbon dioxide in unconventional natural gas reservoirs like coal seams and shale
formations is likely a promising option [9-12]. The injected carbon dioxide can displace methane
in coal and shale and enhance natural gas recovery, which can help offset the cost of carbon capture
and storage. In order to initiate the carbon dioxide sequestration project in shale formations and
coal seams, the carbon dioxide storage capacity needs to be evaluated. Thus, the states of carbon
dioxide under reservoir conditions, such as adsorbed, bulk gas and dissolved phases, must be
investigated. Since the dissolved amount of carbon dioxide in reservoir water can usually be
neglected compared to the adsorbed and bulk phases, an accurate estimation of the adsorbed phase
becomes critical. This therefore requires a thorough understanding of carbon dioxide adsorption
behavior under reservoir conditions.
1 |
Virginia Tech | Based on the above discussions, it is noted that the adsorption phenomenon is extremely important
for the process of CBM/shale gas development and carbon dioxide sequestration. Thus, the basic
concepts for adsorption related phenomena are briefly reviewed in this section.
1.1.1 Gas adsorption phenomenon
Adsorption is a surface phenomenon where the density of a fluid near the surface of solid increases
as a condensed phase. The adsorption process is governed by not only the unique properties of the
solid (surface heterogeneities, etc.) but also the specific energy of the fluid (temperature, etc.).
Physical adsorption can be attributed to the weak van der Waals forces. Methane adsorption in
coal and shale belongs to physisorption.
In order to model gas adsorption behavior, different models have been proposed such as Henry’s
model [13], Langmuir’s model [14], BET (Brunauer–Emmett–Teller) model [15] and pore-filling
model [16-17]. Among these models, the Langmuir model is the most widely used one because of
its simplicity, effectiveness, and the reasonable explanation of its parameters. The Langmuir
equation was developed by Irving Langmuir in 1916, which is based on the following assumptions:
1) the adsorption sites are monolayer, independent, unique, and the same at the solid surface, 2)
there is no interaction between adsorbed gas molecules, and 3) the dynamic equilibrium state is
reached between adsorbed gas molecules and free gas molecules. Langmuir’s model can be shown
as the following form,
n KP
n max (1)
1KP
where n is the adsorbed amount under equilibrium temperature and pressure, n is the maximum
max
adsorbed capacity, P is the adsorption pressure, K is the Langmuir constant which is a function of
temperature. In the limit of low pressure, Langmuir’s model is equivalent to Henry’s model,
n KP
nlim max KP
(2)
P0 1KP
As supported by numerous experimental data for methane in coal and shale, the Langmuir model
is routinely used to model methane adsorption in coal and shale for estimating adsorbed methane
content at reservoir conditions for the CBM and shale gas industry.
2 |
Virginia Tech | 1.1.2 Gibbs excess adsorption concept
In the laboratory, measurements of adsorption using either manometric or gravimetric approaches
cannot measure the true adsorbed amount because both methods, in principle, ignore the occupied
volume of the adsorbed phase. Under low pressure conditions (<10MPa), this assumption works
well, however, this assumption becomes invalid under high pressure conditions (>15MPa) because
it is observed that the measured adsorption uptake increases up to a maximum and then decreases
with increasing pressure. This observation contradicts the fact that the true adsorbed amount
monotonically increases with pressure. In order to solve this issue, Josiah Willard Gibbs introduced
the concept of “excess sorption” (also called “Gibbs excess sorption,”) where he gives a simple
geometric explanation of the measured adsorbed quantity by considering the finite volume of
adsorbed phase [18],
n n V n (1 g ) (3)
e a ad g a
ad
where n is the Gibbs excess adsorbed amount, n is the true adsorbed amount (absolute adsorbed
e a
amount), V is the volume of adsorbed phase, is the density of adsorbed phase, and is the
ad ad g
bulk gas density. The Gibbs excess sorption concept is illustrated in Figure 1. Figure 1 shows a
simplified equilibrium sorption system with a single component gas adsorbed on the porous solid
at pressure (P) and temperature (T). The density of “gas” (also called “adsorbed phase”) near the
solid surface is higher than the bulk gas density and decreases with the distance away from the
solid surface. At a certain distance, the surface can no longer influence the bulk gas, and the density
is equal to bulk gas density.
3 |
Virginia Tech | Figure 1.1.1 Concept of Gibbs surface excess sorption for gas adsorption on solid. V is the
tot
sum of V *and V * which can be measured by non-adsorbed gas (Helium) intrusion test.
a gas
The density file shows the hypothetical density profile near the solid surface.
According to Gibbs excess concept, if the volume of the adsorbed phase is extremely small or the
density of adsorbed phase is much higher than the bulk gas under low pressure, the Gibbs excess
adsorbed amount is almost equivalent to the true adsorbed amount,
n n (4)
e a
This also explains why both the manometric and gravimetric method approximate the true
adsorbed amount, therefore, the Langmuir equation (equation (1)) works well for modeling gas
adsorption behavior at low pressure conditions.
Under high pressure conditions there will be distinguishable differences between the measured
data and the true adsorbed amount, especially when the measured adsorption uptake increases up
to a maximum and then decreases with increasing pressure. In this situation, the Langmuir model
loses its power, and the Gibbs excess sorption concept needs to be applied. Since the measurement
of true physical properties of the adsorbed phase such as density and volume is not possible using
current technology, assuming either the density or the volume of the adsorbed phase as a constant
may provide a solution.
If the Langmuir model (equation (1)) can describe the relationship between the true adsorbed
amount and pressure (most cases for a homogenous surface), the relationship between Gibbs
excess adsorbed amount and the true adsorbed amount can be obtained,
n KP
n max V (5)
e 1KP ad g
For real adsorbents, the heterogeneous surface may offer two (or more) types of adsorption sites
with different characteristic energies [19-21]. Under this situation, the single site Langmuir model
can be extended to a dual-site Langmuir model corresponding to different adsorption sites,
K P K P
nn [(1) 1 2 ]
(6)
max 1K P 1K P
1 2
4 |
Virginia Tech | where K and K corresponds to each type of adsorption sites weighted by a coefficient (0<
1 2
<1). In this way, another relationship between Gibbs excess adsorbed amount and true adsorbed
amount can be obtained,
K P K P
n n [(1) 1 2 ]V (7)
e max 1K P 1K P ad
1 2
Since equations (5) and (7) consider the volume effect of the adsorbed phase, either of them could
provide a practical way to obtain the true adsorbed uptake based on measured data, especially
when the measured adsorption uptake increases up to a maximum and then decreases with
increasing pressure.
1.1.3 Thermodynamics of adsorption
When a gas molecule is adsorbed on a surface, it changes from free gas to the adsorbed film and
therefore results in an energy release. At equilibrium, the change in enthalpy of the system due to
adsorption at a specific state of surface occupancy is referred to as the isosteric heat of adsorption
(H ). Generally, the isosteric heat of adsorption can vary as a function of the amount of
ads
adsorbate and the system conditions. It therefore serves as an important descriptor of the
physisorption system, and is directly related to the strength of the interaction between the gas
adsorbate and the solid adsorbent.
The isosteric heat of adsorption can be determined via the Clapeyron relationship which is relevant
to the equilibrium between two phases in a closed system,
dP dP
H ( ) Tv( ) T(v v ) (9)
ads dT n a dT n a a g
Where v is the volume of adsorbed phase, v is the volume of bulk gas phase, T is temperature.
a g
Since the pressure in a closed system is a function of temperature and quantity adsorbed, a general
dP
expansion of ( ) can be made such that [22],
dT n a
dP P dn (lnP)
( ) ( ) a P( ) (10)
dT n a n T dT T n a
a
If the bulk fluid is approximated as an ideal gas,Pv RT , it follows that,
g
5 |
Virginia Tech | (lnP) RT2 P dn P dn (lnP)
H RT2[( ) ] ( ) a [( ) a P( ) ]T v (11)
ads(n a) T n a P n T dT n T dT T n a a
a a
RT2 P dn
In right hand side (RHS) of equation (8), the second term, ( ) a , includes the behavior
P n T dT
a
P dn (lnP)
of the adsorbed phase mass, and the third term, [( ) a P( ) ]Tv , considers the
n T dT T n a a
a
volume effect of the adsorbed phase.
If the volume of the adsorbed layer is taken to be negligible and the influence of the adsorbed mass
is therefore ignored, the routinely used Clausius-Clapeyron (C-C) relationship is obtained,
(lnP)
H H RT2[( ) ] (12)
ads ads,cc T n a
Equation (12) is only valid when the gas behaves like ideal gas and the influence of the adsorbed
phase can be ignored. In low pressure conditions like Henry’s range, equation (12) is applicable.
However, when the gas behavior deviates from ideal gas or the influence of the adsorbed phase
cannot be neglected, equation (12) is not reliable. Figure 1-1 shows how the real gas like methane
and carbon dioxide deviates from ideal gas. Under this situation, equation (12) cannot be applied
to explore the true behavior of the isosteric heat of adsorption.
Figure 1.1.2 Compressibility of methane and carbon dioxide under different pressures and
temperatures. (Data is obtained from the NIST Standard Reference Database 23 (REFPROP:
Version 8.0.))
6 |
Virginia Tech | 1.1.4 Kinetics of adsorption
Numerical modeling of both the recovery process for CBM and shale gas and the carbon dioxide
injection process requires the kinetics information of the sorption process. The kinetics behavior
of gas usually determines the rate of methane desorption in shale and coal for primary recovery
and the rate of carbon dioxide adsorption in coal/shale for carbon dioxide storage.
Several sorption kinetics models have been applied for gas and solid interactions: the unipore
model [23], the bidisperse model [25-26], the dynamic diffusion model [27-30] and other semi-
empirical models [31-32]. Among these models, the unipore model is the most widely used. The
unipore model is established based on the following four assumptions: 1) coal particles are
spherically symmetric, homogeneous and isotropic, 2) all the pores are of the same size, 3) at the
surface of the spheres gas concentrations are constant throughout the sorption process, and 4) gas
diffusion process follows mass conservation law and the continuity principle. Based on Fick’s
second law and the above four assumptions, the unipore model for spherically symmetric flow is,
C 2C 2C
D( ) (13)
t r2 r r
where r is the radius, C is the adsorbate concentration, D is the diffusion coefficient, and t is time.
The solution of equation (6) for a constant surface concentration of the diffusing gas can be
expressed as follows [33],
Q 6 1 Dn22t
t 1- exp( ) (14)
Q 2 n2 r2
n1
where Q is the total volume of gas desorbed in time t and Q is the total gas adsorbed or desorbed
t ∞
in infinite time. Note, there is no analytical solution for equation (9) but the approximate numerical
solution has been applied by different researchers to obtain the constant diffusion coefficient to
evaluate the gas diffusion process [23, 24, 34-37].
1.2 Problem statement
A unique characteristics of shale gas is its high temperature and high pressure reservoir condition
(up to 27MPa and 360 K), which differentiates it from coal seam gas. This feature has resulted in
the Gibbs excess adsorption behavior of methane in shale, where the observed adsorption uptake
of methane first increases and then decreases with increasing pressure [38]. Under this situation,
7 |
Virginia Tech | the single-site Langmuir model losses its power to model the methane adsorption behavior.
Therefore, a new adsorption model is needed for describing methane adsorption behavior under
high pressure conditions in order to obtain the true adsorbed gas amount. The same problem also
existes in carbon dioxide sequestration field, where an adsorption model in needed to model
supercritical carbon dioxide adsorption behavior in coal and shale.
Considering the heat change is always associated with the physical adsorption process, the
thermodynamics feature of gas adsorption in shale and coal is necessary in order to understand the
adsorption process. Unfortunately, the classic Clausius–Clapeyron approximation cannot be used
to obtain isosteric heat of adsorption under high pressure conditions because it does not account
for the real gas behavior and the volume effect of the adsorbed phase [20-21, 39-40]. Furthermore,
it is also inappropriate to calculate the isosteric heat of adsorption by using experimental data
(Gibbs excess sorption data) especially under high pressure conditions because experimental data
usually underestimate the true adsorbed amount. Therefore, isosteric heat of adsorption for high
pressure gas adsorption in shale and coal needs to be further studied by considering the real gas
behavior and the volume effect of the adsorbed phase, and a uniform approach for obtaining the
absolute quantity of adsorption from measured adsorption isotherms is also needed.
There also existed many key research questions surrounding the geological sequestration process
related to the dynamic interaction between carbon dioxide and coal. For example, how quickly the
injected CO plume will migrate through a coal seam during injection, how the sorption process
2
will affect the transportation of carbon dioxide in the coal seam, and whether continuous injection
or intermittent injection is more effective for maximizing storage. Therefore, the interaction
between gases (carbon dioxide, nitrogen and methane) and coal are analyzed to study the pore
characterization of coal, gas adsorption kinetics behavior in coal, and adsorption thermodynamics.
1.3 Objectives of this dissertation
In order to accurately estimate the CBM/shale GIP and carbon dioxide storage capacity under in
situ reservoir conditions, the following studies were carried out:
Model and analyze high pressure methane adsorption in shale
Develop a methodology for accurate estimation of shale GIP
Model and analyze supercritical carbon dioxide adsorption in coal
Develop a methodology for accurate estimation of carbon dioxide storage capacity
8 |
Virginia Tech | Chapter 2 High pressure methane adsorption in shale for deep shale gas resource
estimation
2.1 Comparison of adsorption models for high pressure methane adsorption in shale
Xu Tang*, Nino Ripepi*,†, Kray Luxbacher*,†
(*Department of Mining and Minerals Engineering & †Virginia Center for Coal and Energy
Research, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, 24060, U.S)
Abstract: Describing true supercritical methane adsorption behavior in shale under high pressures
(>15 MPa) is challenging because the density or volume of adsorbed methane cannot be measured
directly. There are several models available to describe the observed adsorption isotherms, but a
consensus model has not been reached by researchers. Based on the assumption that the density of
the adsorbed methane is an unknown constant, the authors successfully describe observed
adsorption isotherms of methane in shale for pressure up to 27MPa and temperature up to 355.15K
using a dual-site Langmuir equation, and the density of the adsorbed methane in shale is found to
be 17.7 mmol/mL. This work then compares the nine currently available adsorption models for
describing high pressure methane adsorption behavior in shale in order to assess the efficacy of
each model. Three aspects of the adsorption model are compared: (1) the goodness-of-fit of each
adsorption model, (2) interpretation of the observed test phenomena, and (3) predicted isotherms
beyond test data. Comparison results show that even though the goodness-of-fit for each model is
comparable, the dual-site Langmuir model is still superior to other available models because it can
not only reasonably address all observed test phenomenon but can also extrapolate adsorption
isotherms without using an empirical relationship. The dual-site Langmuir model is recommended
for describing high pressure methane adsorption in shale, especially when the Gibbs excess
adsorption phenomenon is observable.
Key words: Methane, adsorption, shale, Langmuir model, high pressure
13 |
Virginia Tech | 2.1.1 Introduction
Shale gas has been considered as one of the most important energy resources in the world and
countries have launched different programs to estimate shale gas resources (Wang et al., 2014;
Andrew et al., 2013; Kuuskraa et al., 2013). Shale gas, the most significant component of which
is methane, exists in three different states in the subsurface: free gas, adsorbed gas and dissolved
gas. Current studies have shown the adsorbed gas accounts for 20-85% of the total shale gas-in-
place (GIP) content (Curtis, 2002). Therefore, it is important to understand the adsorption behavior
of methane in shale in order to accurately estimate shale gas resources in shale formations.
Knowing the exact ratio between adsorbed and free shale gas is also fundamental to understand
shale gas transport behavior and predict shale gas well production behavior (Tang, 2016). Since
most of shale formations are at depths from 1000m to 3000 m, the reservoir pressure of deep shale
formations can go up to 27MPa (Curtis, 2002). This in-situ feature of shale formations requires
high pressure methane adsorption studies for shales. Unfortunately, because of instrument
limitation, there are limited data for high pressure methane adsorption in shale (Rexer et al., 2013;
Luo et al., 2015; Weniger et al., 2010; Tian et al., 2016) which makes investigation and
characterization of methane adsorption in shale challenging.
In order to understand methane adsorption in shale under reservoir conditions it is essential to have
an accurate model for high pressure supercritical gas adsorption in shale. In order to build a
methane adsorption model in shale, the challenge is to describe observed adsorption isotherms
showing Gibbs excess phenomena (Zhou et al., 2000 & 2009). Some researchers use the molecular
simulation approach to simulate methane adsorption behavior in shale and synthetic materials
(Ambrose et al., 2012; Luo et al., 2011; Mosher et al., 2013; Zhang et al., 2014; Chareonsuppanimit
et al., 2012; Fitzgerald et al., 2003; Sudibandriyo et al., 2010; Bourrelly et al., 2005; Aukett et al.,
1992; Snurr et al., 1991; Wang, 2007; Chen et al., 1997; Akkutlu et al., 2013). These studies are
important to understanding the methane adsorption mechanism at a molecular scale. However,
since the simplified, homogeneous pore structure of the computational approach does not represent
the heterogeneous properties of shale, the molecular simulation method has not been widely used
in engineering applications. In addition, molecular simulation has not been used to interpret the
isothermal adsorption phenomenon such as the crossover of the isotherms under different
temperatures observed in experimental data. Other researchers have attempted to build a physical
model from observed adsorption isotherms based on either known constant density (density of
14 |
Virginia Tech | liquid methane) or constant volume assumptions of the adsorbed methane phase (Rexer et al., 2013;
Bae et al., 2006; Sakurovs et al., 2007; Luo et al., 2015; Ottiger et al., 2006; Herbst et al., 2002;
Weniger et al., 2010; Zhou et al., 2000 & 2001; Do et al., 1997; Tian et al., 2016; Bruns et al.,
2016). Unfortunately, most of the proposed adsorption models in the literature do not provide
satisfactory interpretation of the experimental data, and the assumptions used are uncertain. For
example, the crossover of the observed adsorption isotherms at high pressures has not been
reasonably explained, where the observed adsorption content increases with increasing
temperature beyond the Gibbs excess maximum. In addition, none of the models can be used to
extrapolate adsorption isotherms beyond test data without using empirical relationships. Therefore,
an optimized model is needed for accurately describing the adsorption behavior of methane in
shale.
In order to simulate the true methane adsorption behavior in shale under high pressure conditions,
the authors introduced a dual-site Langmuir model to describe high pressure methane adsorption
behavior in shale for temperatures up to 355.15K and pressures up to 27 MPa (Tang et al, 2016).
This work compared this model with other available models in literature to present the specific
characteristics of each model using the test data, which provides a clearer picture of the strengths
and weaknesses of each model. This study compares adsorption models used for engineering
applications, especially for the shale gas industry; therefore, molecular simulation for methane
adsorption in shale is not part of this work.
2.1.2 Adsorption model review
2.1.2.1 Dual-site Langmuir model
In any pure gas-solid adsorption system, the observed adsorption quantity, also called the Gibbs
excess adsorption uptake, is given by the Gibbs equation (1),
n n V n (1 g) (1)
e a g a a
a
where the excess adsorption quantity (n ) refers to the difference between the absolute adsorption
e
quantity (n ) and the quantity of adsorbate that would be present in the same volume (V ) of the
a a
adsorbed phase at the density of the bulk gas phase ( ). When V is very low or the density of
g a
15 |
Virginia Tech | the adsorbed phase ( ) is much higher than the bulk gas phase density ( ), the excess
a g
adsorption quantity is approximately equal to the actual adsorbed amount. However, this relation
is invalid at high pressure where the density of the adsorbed phase is similar to the density of the
bulk fluid, the point at which the observed adsorption quantity reaches a maximum and then
decreases. This Gibbs excess maximum phenomenon has also been observed in many other gas-
solid adsorption systems (Rexer et al., 2013; Bae et al., 2006; Sakurovs et al., 2007; Luo et al.,
2015; Ottiger et al., 2006; Herbst et al., 2002; Weniger et al., 2010; Zhou et al., 2000 & 2001; Do
et al., 1997; Tian et al., 2016; Bruns et al., 2016). Under such conditions, the conventional
adsorption models that neglect the real volume of the adsorbed phase cannot reasonably explain
such adsorption behavior. Therefore, it is imperative to use a more sophisticated approach to obtain
the absolute isotherms from observed Gibbs excess isotherms at high pressures.
For heterogenous adsorbent sites, the dual-site Langmuir model is more suitable than the single-
site Langmuir model for describing the gas adsorption behavior (Graham et al., 1953; Mertens,
2009; Stadie et al., 2013 & 2015). The dual-site Langmuir model assumes two different adsorption
sites in the heterogenous adsorbent. The adsorption energy of the adsorption sites will vary, where
the strongest adsorption energy sites will be filled first, followed by the weak adsorption energy
sites. When both sites reached equilibrium with the same adsorbed phases, each site can be
E
modelled by two separate equilibrium constants K(T) 1 and K(T) 2 ( K(T) A exp( 1 ) and
1 1 RT
E
K(T) A exp( 2 ) , A 1, and A 2 are prefactors, E 1 and E 2 are the binding energy of the two different
2 2 RT
adsorption sites, R is universal gas content, T is temperature) with a weighting coefficient for two
different adsorption sites in the Langmuir type relationship (Graham et al., 1953). Thus, the single
site Langmuir equation can be superposed as the following form (equation 2), where α is the
fraction of two different adsorption sites (0<α<1),
K(T) P K(T) P
n (P,T)n (1)( 1 )( 2 ) (2)
a max 1 K(T) P 1 K(T) P
1 2
Based on the assumption that the density of adsorbed methane is an unknown constant under test
conditions, the volume of the adsorbed layer can be obtained in equation (3),
16 |
Virginia Tech | n
V a (3)
a
a
Similarly, we can obtain the maximum volume of the adsorbed phase, Vmax,
n
V max (4)
a
a
By combining equation (2) and (4), the volume changes of the adsorbed layer in different
adsorption sites can obtained in equation (5),
n K(T) P K(T) P
V max (1)( 1 )( 2 ) (5)
a 1 K(T) P 1 K(T) P
a 1 2
Combining equation (1), (2) and (5), the excess adsorption equation for dual sites adsorbates can
be obtained as shown in equation (6) and the surface coverage () is shown in equation (7),
K(T) P K(T) P
n (P,T)(n V ) (1)( 1 )( 2 ) (6)
e max g max 1K(T) P 1K(T) P
1 2
n (P,T) K(T) P K(T) P
a (1)( 1 )( 2 ) (7)
n 1K(T) P 1K(T) P
max 1 2
It is clear that if the experimental adsorption (Gibbs excess adsorption) isotherms are obtained
through isothermal adsorption tests, the unknown parameters in equation (6) can be easily obtained
via curve fitting. The absolute adsorption uptake can then be calculated by equation (2). In addition,
the density of adsorbed methane can be obtained using equation (4).
2.1.2.2 Review of adsorption models
In order to describe the observed methane adsorption behavior in shale and coal under high
pressures, several researchers have proposed different models based on experimental data
summarized in Table 2.1.1. These models can be classified into three different groups: (1)
unknown constant density of adsorbed methane layers with changing volume of adsorbed layer
with increasing adsorption uptake: ④; (2) known density assumption of adsorbed methane layers:
①②③⑦⑧⑨, and; (3) constant volume assumption of adsorbed methane layers: ⑤⑥. These
models can also be classified as Langmuir-style equations, Toth-style equations, and Dubinin–
17 |
Virginia Tech | Radushkevich (D-R) (or Dubinin–Astakhov (D-A)) equations based on adsorption potential theory
as shown in Table 2.1.1.
Table 2.1.1 Comparison of methane adsorption models in shale and coal
In practice, it is impossible to measure all isotherms under in-situ conditions in order to predict
shale GIP content. Therefore, the use of limited test data to extrapolate isotherms under different
temperatures has been researched. Researchers have attempted to use D-R or its revised form to
predict isotherms under different temperatures, because the characteristic curve is unique under
different temperatures for gas adsorption in microporous media like activated carbon (Dubinin et
al., 1960; Dubinin et al., 1971; Amankwah et al., 1995). However, when D-R methods are applied
for describing methane adsorption in coal or shale, the characteristic curve is not unique (Huan, et
al., 2015; Xiong et al., 2015). This can be attributed to (1) the heterogenous properties of natural
geo-materials; (2) the fact that methane is a supercritical gas under reservoir conditions, and the
empirical saturation pressure assumption is invalid, and; (3) the fitting parameters of D-R equation
and its revised form are non-unique which contradicts its assumptions. Therefore, other researchers
use an empirical approach to predict isotherms under different temperatures (Tian et al., 2016;
Hildenbrand et al., 2006; Kronimus et al., 2008; Busch et al., 2016). First, each isotherm is fitted
independently using the proposed model. Then, the relationship between fitting parameters and
temperature is obtained empirically. Based on this empirical relationship, the isotherms beyond
18 |
Virginia Tech | test pressures and temperatures are predicted. Since this approach highly depends on the obtained
empirical relationship, only limited information can be obtained from the predicted data and these
results should also be treated with caution.
2.1.3 Model evaluation criteria
A physical model is typically developed by scientists and engineers to simplify the complexity of
the real world to better understand the real phenomena. Generally, the best models represent a
simplification, but are still complex enough to help one understand the phenomena and to solve
the problem. The best model should simplify complexity of real world phenomena while retaining
the most important parameters. Figure 2.1.1 shows the way a model can be developed in order to
better understand the real world phenomena. From this flowchart, one can gain several intuitive
perspectives about the development of the model. First, the model should describe the observed
phenomena based on real world observations. Second, the model should give one a reasonable
interpretation of the real phenomena. Third, the model should provide predictable capacity, which
can be validated by more real phenomena. If the model is developed following these three
approaches, it will become a reliable model.
Figure 2.1.1 Depiction of the physical modelling approach from real world to conceptual
world (revised from Dym et al., 2004)
In order to compare the current available adsorption model, the first and crucial step is to set
comparison criteria for each model. Three general criteria are used here. First, the goodness-of-fit
of the model to test data will be evaluated. This is a straightforward approach to show whether the
proposed model can describe the experimental measurements. An accurate model should closely
match the data using the minimal but the most significant assumptions. It should be noted that
goodness-of-fit should reflect such a physical fact that the experimental results are not only
19 |
Virginia Tech | determined by the experimental parameters but also influenced by experimental errors. This means
too precise fitting may not be the best result for a good model. If the proposed model has too many
fitting parameters, higher fitting precision can be achieved, but the whole model can lose its
physical meaning. Secondly, the proposed model should interpret part or all observed phenomenon
in the test and should improve one’s understanding of the mechanisms of methane adsorption
mechanism in shale. Lastly, the prediction ability of the model will be compared, where isotherm
adsorption curves are predicted beyond test temperature and pressure. The predicted isotherms
should show similar properties with the observed test phenomena. This means a good model should
extrapolate to situations or data beyond those originally described in the model. If the proposed
model meets the all of the above three standards, the model should be treated as valid.
2.1.4 Test results and data processing method
Shale samples from the Lower Silurian Longmaxi Formation (2400.8 meters deep) were obtained
from the Fuling #1 well in the Fuling region, Sichuan Province, China. The vitrinite equivalent
reflectance (Ro) of the sample is 2.2% - 2.5% (Tang et al, 2016). Methane adsorption
measurements were conducted using a Rubotherm Gravimetric Sorption Analyzer IsoSORP. The
methane density is obtained via the NIST package using Setzmann & Wagner equation (Setzmann
et al., 1991).The instrument is rated up to pressures of 35 MPa and temperatures up to 150°C±0.2℃,
and pure methane gas (99.99%) is used as the adsorbate. Equilibrium was determined as when the
adsorption time was longer than 2 hours or when the weight change of the sample was within 30
μg over a span of 10 min. The detailed characteristics of the instrument have been extensively
described elsewhere (Keller & Staudt, 2005). The test results are shown in Figure 2.2.2, where test
data are retrieved from Tang et al, 2016. All raw data can be reached at the Supplemental Material
file.
20 |
Virginia Tech | Figure 2.1.2 High pressure methane adsorption test in shale (a: observed adsorption uptake
as a function of pressure; b: observed adsorption uptake as a function of bulk gas density)
The least squares residual method is used to fit nine different models in Table 1,
N 2
Residual[(nfitted ntested) ] with i=1, 2, 3, …..,j (7)
e,i e,i
i1
For method ④, the global fitting method is used, which means the four Gibbs excess adsorption
isotherms under different temperatures are fitted simultaneously to the dual-site Langmuir model
(equation 6) by a least-squares residual minimization algorithm. This means the j in equation (6)
is equal to 63, corresponding to the total measured points from all four isotherms. The seven
independent fitting parameters were varied to achieve the global minimum of the residual squares
value within the following limits: 0<n <100 mmol/g, 0< V <10 cm3/g, 0<α<1, 0< E <100
max max 1
kJ/mol, 0< E <100 kJ/mol, A > 0, A > 0). Minimization was performed in excess of 100 times
2 1 2
by changing the random seed in order to assure that a global minimum was achieved.
For methods ①-③ and ⑤-⑨, the conventional independent fitting method is used, which means
each isotherm under different temperatures is fitted independently using the corresponding
equation by a least-squares residual minimization algorithm. The best fitting parameters for each
isotherm can be obtained by achieving the local minimum of the residual squares value without
using a boundary constraint. This means j is equal to either 15, 16, or 17, corresponding to the
measured points from each isotherm.
Since the least squares residual method cannot reflect the fitting error for individual measured
points from each isotherm, the fitted relative error is used here in order to evaluate the difference
between the predicted data and test data,
nfitted ntested
e e
Relative Error % (8)
ntested
e
The relative error reflects how the predicted value deviates from the measured data in a
straightforward way, and it can be used to evaluate the fitting goodness of the model.
2.1.5 Results and discussion
2.1.5.1 Goodness-of-fit evaluation
21 |
Virginia Tech | As illustrated in Figure 2.1.3, more fitting parameters allow for better fitting results. Method ①
and ⑦ are the poorest fit, with only two fitting parameters in their models. The other methods
have three or more fitting parameters and all have similarly improved results. All fitting models
show that there are crossovers of the isotherm beyond the maximum Gibbs excess adsorption
content. However, only method ④ shows a clear trend that after the crossover point the increasing
temperature results in higher observed adsorption uptake. This trend was reported for methane
adsorption in activated carbon up to 50 MPa (Herbst et al, 2002).
Figure 2.1.3 Comparison between fitting curve and test data for each model: symbols
represent test data, solid lines represent fitting curves.
Using equation (8), the relative error for each fitting model is shown in Figure 2.1.4. The relative
error for method ④ is comparable to the error of other methods. Furthermore, it is difficult to
distinguish which method is better only by the relative fitting error (Figure 2.1.4). The fitting error
can only show the goodness of a fitting model but cannot reflect the physical meaning of each
model. However, whether the proposed model can be used to interpret the observed phenomena is
the critical criteria.
22 |
Virginia Tech | Figure 2.1.4 Relative error between fitting data and test data for each method for all raw
data
2.1.5.2 Interpretation of test phenomena
The adsorption model is built in order to explain the test phenomena. For adsorption isotherms
under different temperatures, the most distinguishable phenomena is the crossover of the isotherms
under different temperatures as shown in Figure 2.1.2(a). At pressures below the Gibbs excess
maximum, the excess adsorption is always lower at higher temperatures. However, at a point
somewhere beyond the Gibbs excess maximum, the isotherms crossover and higher temperatures
result in higher excess uptake at equivalent pressures.
Method ④ gives a reasonable interpretation for this crossover phenomenon. A reasonable
interpretation of the crossover phenomenon can be made by examining the change of the
coefficient of equation (6). As pressure increases, the density of gaseous methane increases, but
the density of the adsorbed phase stays constant based on the assumption in equation (3); further,
the density of gaseous methane approaches the density of the adsorbed phase (shown in Figure
2.1.5). This results in a decrease of the coefficient, (n V ) , as pressure goes up.
max g max
Temperature also has a positive effect on the coefficient: the higher the temperature the higher the
value of the coefficient. Figure 2.1.6 shows the temperature has a negative effect on the surface
coverage (equation (7)): the higher the temperature the lower the surface coverage. As we multiply
the coefficient ((n V )) and the surface coverage using equation (6), we obtain the
max g max
observed (excess) adsorption content with the crossover of the isotherms under high pressure
conditions. Therefore, the observation of the crossover phenomenon in the measured data supports
the assumption that the density of the adsorbed phase is constant and the volume of adsorbed phase
changes with temperature and pressure following a dual-site Langmuir-like equation.
23 |
Virginia Tech | Figure 2.1.5 Modelled values of the density of gaseous (solid color lines, left axial), adsorbed
and liquid methane (solid black lines, left axial) and the coefficient of equation (7)
((n V ), dotted lines, right axial) on Longmaxi shale as a function of pressure
max g max
Figure 2.1.6 Surface coverage of the methane in shale
Generally, adsorption isotherms show the relationship between Gibbs excess adsorption content
and pressure, where pressure is the independent variable for most of the adsorption isotherms under
intermediate pressures (10-15MPa). Under high pressure conditions (>15MPa), density is
suggested as the independent variable (Ottiger et al., 2006; Pini, 2014). The observed adsorption
isotherms as a function of gas density clearly show the temperature dependent properties of
adsorption isotherms. The crossover of the excess uptake isotherms will not be observed when the
isotherms are plotted as a function of bulk gas density instead of pressure. The measured isotherms
show the same temperature dependence at all pressures, i.e. increasing excess uptake with
decreasing temperature. As shown in Figure 2.1. 7, only Method ④ can reproduce this phenomena
even though the test data fluctuates slightly. Figure 2.1.2(b) shows a slight fluctuation of the test
data under 318.15K, which is caused by some measurement errors. All fitting curves in the other
methods still show crossover of the isotherms, which cannot overcome the fluctuation from the
raw data. This on the other hand confirms the robustness of Method ④, which is relatively immune
to fluctuations in the raw data.
24 |
Virginia Tech | Figure 2.1.7 Adsorption model fitting results: Gibbs excess adsorption content as a function
of bulk methane density
2.1.5.3 Evaluation of predicted isotherms beyond test data
In the shale gas industry, due to laboratory instrument limitations, adsorption isotherms are
typically measured under intermediate pressures (10-15MPa) and temperatures higher than room
temperature. The high pressure (>15 MPa) adsorption test typically requires higher reliability and
accuracy of the instrument (Tang et al., 2015). The widely used approach is to use methane
adsorption measurements at intermediate pressure conditions (10-15 MPa) to predict the methane
adsorption behavior in the higher pressure region (>15 MPa). In addition, the commonly used
technique for constant temperature is to use a water bath which can maintain room temperature to
about 100℃, but it is difficult to reach temperatures lower than room temperature. For shallow
coalbeds and shale formations, the temperature is typically lower than room temperature. It is also
impractical to measure all adsorption isotherms at all in-situ geological conditions. Engineers
usually use isotherms under intermediate temperatures to predict both low temperature (lower than
room temperature) and high temperature adsorption isotherms based on an empirical relationship
between fitting parameters and temperatures. Since a good physical model can not only help one
25 |
Virginia Tech | interpret the observed phenomena but also has predictive capability, we present the predicted
isotherms beyond test data in this section.
As mentioned previously, predicting adsorption isotherms at different temperatures is of
fundamental interest for reservoir characterization of coalbed and shale formations. Therefore, the
temperature within and beyond the test ranges is extrapolated for each model at temperatures of
353.15K, 375.15K, 395.15K and 415.15K. As shown in Figure 2.1.8, all isotherms are plotted as
a function of bulk gas density. It is clear that the predicted Gibbs excess adsorption isotherms using
Method ④ are the only isotherms exhibiting similar properties for both the observed adsorption
isotherms and predicted isotherms. Method ① and ⑦ also show a clear trend but they are not
immune to errors in the raw data, where the isotherms still crossover. This conflicts with the fact
that temperature always has a negative effect on the true (absolute) adsorption uptake.
Figure 2.1.8 Extrapolated Gibbs excess adsorption isotherms of methane on Longmaxi shale
(dashed lines) and as a function of bulk methane density (Note: Method 6 cannot be used to
predict isotherms because there is no consistent empirical relationship between fitting
parameters and temperature)
From the previous discussions, it is noted that the dual-site Langmuir model is the only model that
passes the three criteria. This supports the hypothesis that the dual-site model (Method ④) is
26 |
Virginia Tech | superior when compared with the other models (①-③ and ⑤-⑨). The successful application of
the dual-site Langmuir model also sheds light on the true behavior of the adsorbed phase for
methane in shale: the volume of the adsorption layer depends on temperature and pressure and the
density of the adsorbed layer can be treated as a constant value.
2.1.6 Conclusions
This work compares nine adsorption models for high pressure methane adsorption in shale using
isotherm data at four temperatures (303.15 K, 318.15 K, 333.15 K and 355.15 K) and high
pressures (up to 27 MPa) based on three evaluation criteria: (1) fitting goodness of the adsorption
model for describing experimental raw data, (2) interpretation of the observed test phenomena,
and (3) prediction capability of the adsorption models beyond the test data. The dual-site Langmuir
model is the only one that passes these three criteria, which supports the robustness of the dual-
site Langmuir model. Therefore, the dual-site Langmuir model is recommended to use for methane
adsorption in shale under high pressure conditions, especially when the Gibbs excess adsorption
phenomenon is observable.
Acknowledgements
This research was supported in part by the U.S. Department of Energy through the National Energy
Technology Laboratory’s Program under Contract No. DE-FE0006827. The authors would like to
thank Dr. Nicholas P. Stadie for the help in curve fitting and Mr. Lingjie Yu for conducting
isothermal adsorption experiments.
References
Ambrose, R. J., Hartman, R. C., Diaz Campos, M., Akkutlu, I. Y., & Sondergeld, C. (2010,
January). New pore-scale considerations for shale gas in place calculations. In SPE
Unconventional Gas Conference. Society of Petroleum Engineers.
Amankwah, K. A. G., & Schwarz, J. A. (1995). A modified approach for estimating pseudo-vapor
pressures in the application of the Dubinin-Astakhov equation. Carbon, 33(9), 1313-1319.
Andrews, I. J. (2013). The Carboniferous Bowland Shale gas study: geology and resource
estimation.
27 |
Virginia Tech | 2.2 A dual-site Langmuir equation for accurate estimation of high pressure deep shale gas
resources
Xu Tang*, Nino Ripepi*,†, Nicholas P. Stadie‡, Lingjie, Yu§,¶, Matthew R Hall#,||
(*Department of Mining and Minerals Engineering & †Virginia Center for Coal and Energy
Research, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, 24060, U.S;
‡ETH Zürich, Laboratory of Inorganic Chemistry, Vladimir-Prelog-Weg 1, 8093 Zürich,
Switzerland; §Wuxi Research Institute of Petroleum Geology of Sinopec Exploration & Production
Research Institute & ¶Sinopec Key Laboratory of Petroleum Accumulation Mechanisms, Wuxi,
Jiangsu, 214151, China; #Nottingham Centre for Geomechanics, Faculty of Engineering,
University of Nottingham, Nottingham, NG7 2RD UK, ||British Geological Survey, Environmental
Science Centre, Keyworth, Nottingham, NG12 5GG UK)
Abstract: Adsorbed methane makes up a large portion of the total shale gas-in-place (GIP)
resource in deep shale formations. In order to accurately estimate the shale GIP resource, it is
crucial to understand the relationship between the adsorbed methane quantity and the free methane
quantity of shale gas in shale formations (under high pressure conditions). This work describes
and accurately predicts high pressure methane adsorption behavior in Longmaxi shale (China)
using a dual-site Langmuir model. Laboratory measurements of high pressure methane adsorption
(303 - 355 K and up to 27 MPa) are presented. Our findings show that for depths greater than 1000
m (> 15 MPa) in the subsurface, the shale gas resources have historically been significantly
overestimated. For Longmaxi shale (2500 – 3000 m in depth), classical approaches overestimate
the GIP by up to 35%. The ratio of the adsorbed phase compared to the free gas has been
significantly underestimated. The methods used herein allow accurate estimations of the true shale
GIP resource and the relative quantity of adsorbed methane at in situ temperatures and pressures
representative of deep shale formations.
Key words: Shale gas, methane, absolute adsorption, Langmuir
Published in Fuel: Volume 185, 1 December 2016, Pages 10–17.
33 |
Virginia Tech | 2.2.1 Introduction
Shale gas resources are globally abundant and shale gas production has continuously increased
over the past ten years as a result of horizontal drilling and hydraulic fracture techniques (1-7). It
is now recognized as a promising unconventional natural gas resource, and many countries have
attempted to accurately estimate their shale gas resources in an effort to meet their future energy
demands (4, 7-8). For example, shale gas production has grown very rapidly in the United States,
reaching nearly 40% of total natural gas production in 2013 (6). Despite its widespread importance,
substantial uncertainties exist in assessing the quantity of recoverable shale gas, and current
resource estimates should be treated with considerable caution (9, 10). This large and continuing
uncertainty significantly impacts the total gas-in-place (GIP) estimation at a majority of sites,
especially in terms of the often-neglected effects of high pressure and temperature in deeper shale
formations, e.g. Barnett shale. The future of the shale gas industry and worldwide energy policy
therefore depends on the development of a more accurate shale gas resource estimation
methodology. In addition, with the development of non-aqueous fracturing fluids such as carbon
dioxide in the hydraulic fracturing technique, deep shale formations may become a viable option
for carbon dioxide sequestration (11, 12). A reasonable assessment of the carbon dioxide
adsorption capacity of shale at high pressure and temperature geological conditions is of parallel
interest (13, 14).
Shale gas trapped within shale formations is different from conventional natural gas since the shale
formation is often both the source and the reservoir of the natural gas itself. Shale gas exists in
three different phases within the shale formation: (i) as free compressed gas, (ii) as adsorbed fluid
on the surface, and (iii) as a dissolved component in the liquid hydrocarbon and brine. The most
widely used approach for estimating shale GIP is to sum these three components. The adsorbed
phase accounts for 20% to 85% of the total amount based on current studies in five major shale
formations in the United States (1). Thus, the estimation of the adsorbed amount of natural gas,
the largest component of which is methane, significantly influences the final determination of the
geological GIP quantity and the working life of the shale gas producing well (9).
Unlike coalbed methane which usually occurs in shallow coal seams (at depths of <1000 m), shale
formations are typically much deeper and under significantly different geological conditions. For
example, the Barnett shale completions are up to 2500 m deep, where reservoir pressures can reach
34 |
Virginia Tech | 27 MPa and the reservoir temperature can be up to 360 K (1). Unfortunately, the effects of these
high pressure and temperature conditions on the quantity of adsorbed methane available in shale
gas reservoirs have rarely been appropriately considered in both academia and industry. The
standard practice for estimating shale GIP is to use methane adsorption measurements at
intermediate pressure conditions (10-15 MPa) to predict the methane adsorption behavior in the
higher pressure region (>15 MPa) (1, 2, 4, 5, 8). However, the methodology used in the standard
practice does not account for the difference between observed and absolute adsorption quantities.
This misinterpretation can significantly affect the shale GIP estimation, especially the contribution
of the adsorbed methane at high pressure geological conditions (high pressure refers to reservoir
pressures above 15 MPa in this work) (9) where the Gibbs excess adsorption phenomenon is very
pronounced. Even though this phenomenon has been observed and acknowledged in numerous
cases (15-24), several fundamental problems still remain to be addressed. These include the
development of physically reasonable methods to (i) accurately describe the observed (excess)
adsorption isotherms, (ii) predict the corresponding absolute adsorption isotherms, and (iii) predict
adsorption isotherms at pressures and temperatures beyond the measured data. Several adsorption
models have been proposed (15, 17-19, 21-23), but these models do not give a satisfactory
interpretation of the experimental data and excess adsorption phenomena, and the assumptions
used are unphysical in nature. Most notably, a common assumption is to treat the adsorbed layer
as having a constant volume independent of the adsorbed amount and/or pressure of the bulk phase
(15-19, 21-23). Although in some cases this volume is allowed to vary with temperature (15, 16),
it is generally not valid to assume that the volume will not change as the adsorbed phase increases
in occupancy. The simplified, homogeneous pore structures used in the computational approach
can also not be used to reasonably portray the heterogeneous properties of shale or coal (24- 26).
In addition, all of these proposed methods cannot predict adsorption isotherms at arbitrary
conditions in a robust and rational way, which inhibits their application for shale gas resource
estimation as a function of specific location (e.g., subsurface depth). All of these shortcomings are
compounded by a lack of measured data under high pressure conditions (well beyond the Gibbs
excess maximum). Therefore, both high-pressure adsorption measurements and an optimized
adsorption model are needed to accurately describe the adsorption behavior of methane in shale
under relevant subsurface conditions. This will in turn allow an accurate shale GIP estimation for
a plethora of worldwide shale resources under actual in situ conditions.
35 |
Virginia Tech | In this work, methane adsorption in a sample of Longmaxi shale from China was measured using
a gravimetric method at four temperatures (303.15 K, 318.15 K, 333.15 K and 355.15 K) and high
pressures (up to 27 MPa). A dual-site Langmuir adsorption model is introduced to describe both
the observed and absolute isotherms at high pressure, utilizing the assumption that the volume of
the adsorbed phase changes constantly with the number of adsorbed molecules following a dual-
site Langmuir-type equation. These results shed light on the true quantity of shale GIP that can be
applied over a large range of temperature and pressure, relevant to the geological conditions of
actual shale gas resources.
2.2.2 Dual-site Langmuir adsorption model
In any pure gas-solid adsorption system, the observed adsorption quantity, also called the Gibbs
excess adsorption uptake, is given by the Gibbs equation,
n n V n (1 g) (1)
e a a g a
a
where the excess adsorption quantity (n ) refers to the difference between the absolute adsorption
e
quantity (n ) and the quantity of adsorbate that would be present in the same volume (V ) of the
a a
adsorbed phase at the density of the bulk gas phase ( ). When V is very low or the density of
g a
the adsorbed phase ( ) is much higher than the bulk gas phase density ( ), the excess
a g
adsorption quantity is approximately equal to the actual adsorbed amount. However, this relation
is invalid at high pressure where the density of the adsorbed phase is similar to the density of the
bulk fluid, the point at which the observed adsorption quantity reaches a maximum and then
decreases. Under such conditions, the conventional adsorption models that neglect the real volume
of the adsorbed phase cannot reasonably explain such adsorption behavior. Therefore, it is
imperative to use a more sophisticated approach to obtain the absolute isotherms from observed
Gibbs excess isotherms at high pressures. The absolute adsorbed amount (n ) should always be a
a
monotonically increasing quantity with increasing pressure for a physical adsorption system. A
simple description of such a system is the widely used Langmuir equation (equation 2),
K(T)P
n n
(2)
a max 1K(T)P
36 |
Virginia Tech | where n is the absolute adsorption quantity under equilibrium temperature (T) and pressure (P),
a
n is the maximum adsorption capacity, K(T) is the temperature-dependent equilibrium constant,
max
which can be expressed as E , E is the energy of adsorption, A is the pre-
K(T) A exp( 0 ) 0 0
0 RT
exponential coefficient and R is the ideal gas content (where both E and A are independent of
0 0
temperature).
In order to obtain the absolute adsorption amount from the observed Gibbs excess adsorption
isotherms, V or must be known. However, it is not possible to measure either of these
a a
quantities directly. Therefore, the most widely used approach is to estimate the density of the
adsorbed layer based on one of numerous empirical relationships (15, 17-23). It is common to
assume that the volume of the adsorbed phase is always constant as a function of adsorption uptake,
or in some cases only dependent on temperature. This assumption does not have a basis in the
physical understanding of adsorption where the volume of the adsorbed phase must increase as
uptake increases.
An alternative approach is to assume that the adsorbed phase has a constant density and that its
volume is therefore a linear function of adsorbed amount. In this case, the fact that different
researchers use different values for the density of the adsorbed phase (e.g., that of the liquid
adsorbate) to obtain absolute isotherms from observed Gibbs excess isotherms is a significant issue,
and these values cannot be directly validated through laboratory approaches (15-23). The most
general approach is to allow the adsorbed density to be an independent parameter of the adsorption
model. This is adopted herein as shown in equation (3), by treating the adsorbed layer as constantly
increasing as a function of uptake up to a fitted maximum adsorbed phase volume (14, 27-30).
This can be expressed as,
K(T)P
V V (3)
a max 1K(T)P
where V is the volume of the adsorbed phase at maximum adsorption capacity. This unknown
max
volume (V ) can be left as an independent fitting parameter and varies from system to system
max
but often yields densities of the adsorbed phase that are close to that of the liquid adsorbate.
37 |
Virginia Tech | Combining equations (1), (2) and (3), the excess adsorption uptake can be obtained as shown in
equations (4)
K(T)P
n (P,T)(n V ) (4)
e max max g 1K(T)P
However, this single-site Langmuir equation cannot sufficiently describe a large number of real
gas-solid adsorption systems (31-32). For heterogeneous surfaces (as in almost all real-world
materials), the adsorption energy at each site will vary, depending on the local chemistry and
structure. The single site Langmuir model is limited in this application (31-32). The most favorable
sites will be filled first, followed by the less favorable sites. In order to address heterogeneous
adsorbents, the most simplified case is where only two different adsorption sites are available.
Each site can be modelled by a separate equilibrium constant, K (T) and K (T)
1 2
E
( K (T)A exp( E 1 ) andK (T) A exp( 2 ) ), weighted by a coefficient (). Thus, the dual-site
1 1 RT 2 2 RT
Langmuir equation can be written in the following form (equation 5), where α is the fraction of
the second type of site (0<α<1),
K (T)P K (T)P
n (P,T) n (1)( 1 )( 2 ) (5)
a max 1K (T)P 1K (T)P
1 2
In the same way as for the single-site equation, the excess uptake in the dual-site equation can be
obtained, shown in (6),
K (T)P K (T)P
n (P,T)(n V )(1)( 1 )( 2 ) (6)
e max max g 1K (T)P 1K (T)P
1 2
Both the single-site (equation 2, 4) and dual-site equations (equation 5, 6) shown herein are based
on the assumption that the volume of the adsorbed layer increases linearly with the adsorbed
amount, up to a monolayer completion (V ). Then, the absolute adsorption amount can be
max
obtained from the measured adsorption data via a least-squares fitting analysis. It should be noted
that the real-world material may have an abundance of different adsorption sites in actuality, but
that a two-site model has often been found to be sufficient for describing such a system owing to
the large number of independent fitting parameters (28-30), and when using a global fitting method
38 |
Virginia Tech | (see: Section 3 Materials and methods) it is desirable to decrease the number of unnecessary such
parameters (30).
2.2.3 Materials and methods
Shale samples from the Lower Silurian Longmaxi Formation (collected at a depth of 2400.8 m)
were obtained from the Fuling #1 well in the Fuling region, Sichuan Province, China. The shale
specimen was ground and sieved using 0.38-0.83 mm metal sifters and placed in a drying oven at
105 °C for 24 h to dehydrate. After dehydration, the prepared sample was stored in a desiccator
prior to adsorption measurements. Methane adsorption measurements were conducted using a
Rubotherm Gravimetric Sorption Analyzer (Rubotherm GmbH, Bochum, Germany) with research
grade methane gas (99.99%). Detailed experimental procedures and physical parameters of the
shale sample are given in the Supplemental Materials.
In this work, four methane adsorption isotherms were obtained at 303.15 K, 318.15 K, 333.15 K
and 355.15 K. All isotherms were measured up to 27 MPa and fluctuations in temperature during
a given isotherm were < 0.2℃. The data were processed using a previously developed Mathematica
script (28-30); the four Gibbs excess adsorption isotherms were fitted simultaneously to the dual-
site Langmuir model (equation 6) by a least-squares residual minimization algorithm based on the
Differential Evolution method. Each data point was given the same weight and none were
discarded. The density of the bulk fluid as a function of temperature and pressure was obtained
from the NIST REFPROP database. The seven independent fitting parameters were varied to
achieve the global minimum of the residual-squares value within the following limits: 0<n
max
<100 mmol/g, 0< V <10 cm3/g, 0<α<1, 0< E <100 kJ/mol, 0< E <100 kJ/mol, A > 0, A > 0).
max 1 2 1 2
Minimization was performed in excess of 100 unique times by changing the random seed in order
to assure that a global minimum was achieved. Once the seven fitting parameters were determined,
absolute and excess adsorption uptake could be easily calculated at any temperature and pressure
by use of equations 5 and 6.
2.2.4 Results and discussions
2.2.4.1 Modeling of observed Gibbs excess adsorption at high pressures
Equilibrium excess adsorption uptake of methane measured on Longmaxi shale between 303-355
K and 0.1-27 MPa is shown in Figure 2.2.1. In all isotherms, the observed Gibbs excess adsorption
39 |
Virginia Tech | uptake increases with increasing pressure up to a maximum value and then decreases. At pressures
below the Gibbs excess maximum, the excess adsorption is always lower at higher temperatures.
However, at a point somewhere beyond the Gibbs excess maximum, the isotherms crossover and
higher temperatures now result in higher excess uptake at equivalent pressures. As seen in Figure
2.2.1, the observed maximum Gibbs excess adsorption quantities are 0.0893 mmol/g, 0.0813
mmol/g, 0.0786 mmol/g and 0.0719 mmol/g at 303.15 K (8 MPa), 318.15 K (10 MPa), 333.15 K
(12 MPa) and 355.15 K (12 MPa), respectively. As the isotherm temperature increases, higher
pressure is needed to reach the Gibbs excess maximum. This is a well-known phenomenon of
supercritical gas adsorption (33). The dual-site Langmuir adsorption model (equation 6) gives a
good global fit to the observed data, and the corresponding best-fit parameters are: n =0.1715
max
mmol/g, V =0.0097 mL/g, α=0.2640, E =16.706 kJ/mol, A =0.0002 1/MPa, E =15.592 kJ/mol,
max 1 1 2
A =0.0032 1/MPa. It should be emphasized that these seven parameters apply to all the isotherms
2
measured, and that by performing a single global fit to all the data at once, a most general
understanding of the properties of the adsorbent-adsorbate system can be achieved.
Figure 2.2.1 Gibbs excess adsorption isotherms of methane on Longmaxi shale (symbols) and
dual-site Langmuir model fits (lines)
An explanation of the Gibbs excess maximum phenomenon can be made by examining the change
in the volume of the adsorbed phase as compared to the volume-density product, as shown in
Figure 2.2.2. The volume of the adsorbed methane phase changes with pressure and temperature
following a dual-site equivalent of equation 3. Higher temperature decreases the adsorbed quantity
of methane, which results in a decreased volume of the adsorbed phase. As pressure increases, the
40 |
Virginia Tech | volume-density term (V *ρ ) of equation (6) always increases but the difference of (V *ρ ) at
a g a g
different temperatures becomes more pronounced. The (V *ρ ) term at low temperature is always
a g
higher than that at high temperature and the maximum absolute adsorption quantity (n ) is
max
constant, which results in the crossover of the Gibbs excess adsorption isotherms. Therefore, the
observation of the crossover phenomenon in the measured data (Fig. 2.2.1) supports the
assumption that the volume of adsorbed methane changes with temperature and pressure following
a dual-site Langmuir-type equation. This is in distinct contradiction to the approximation that the
adsorbed phase is constant, an often used approximation in other work.
Figure 2.2.2 Modelled values of the volume of adsorbed methane (V ) (solid lines, filled
a
symbols, left major axis) and the volume-density term (V *ρ ) (dotted line, hollow symbols,
a g
right minor axis) on Longmaxi shale as a function of pressure
The crossover of the excess uptake isotherms is not observed when the isotherms are plotted as a
function of bulk gas density instead of pressure (Figure 2.2.3). The measured isotherms show the
same temperature dependence at all pressures, i.e. increasing excess uptake with decreasing
temperature. This behavior is also inherently predicted by the dual-site Langmuir equation (see the
fits in Figure 2.2.3). The small deviations from this trend in the measured data at 318.15 K can be
attributed to experimental error, and the overall trend remains clear. The same phenomenon (seen
when plotting excess uptake as a function of bulk fluid density) was also reported for carbon
dioxide, methane and nitrogen adsorption in different materials (15, 17, 21).
41 |
Virginia Tech | Figure 2.2.3 Gibbs excess adsorption isotherms of methane on Longmaxi shale (symbols)
and dual-site Langmuir equation fits (lines) as a function of bulk methane density
2.2.4.2 Prediction of absolute adsorption and extrapolation to higher temperatures
Absolute adsorption isotherms of methane on Longmaxi shale based on equation 5 are shown in
Figure 2.2.4. As is characteristic of the Langmuir equation, the adsorption quantity increases
monotonically up to 27 MPa, which is consistent with the physical nature of adsorption. The
absolute adsorption quantity is significantly higher than the observed Gibbs excess quantity,
especially at 27 MPa. This implies the significant contribution of the adsorbed phase volume of
methane in shale toward the absolute adsorption content, which is neglected in the observed Gibbs
excess adsorption isotherms. Figure 2.2.4 also shows that at higher temperatures, this contribution
becomes less pronounced.
Predicting adsorption isotherms at different temperatures is of fundamental interest for shale GIP
estimations in the deep subsurface, typically reservoirs at a depth over 1000 m. It is impractical to
measure a large number of isotherms at different temperatures for shale gas resource estimation.
Thus, another feature of the dual-site Langmuir model used herein is that it can be used to predict
isotherms at arbitrary temperatures near the measured isotherms. This is very notably not possible
when each isotherm is fitted individually, as is often the case in other studies, and a global fit
across numerous isotherms is therefore an extremely desirable feature of a particular model.
Interpolation of the measured data (i.e., predictions at temperatures between the measured
42 |
Virginia Tech | isotherms) is expected to be highly accurate, though extrapolation to higher or lower temperatures
than measured, while also possible, should be performed with caution. Nevertheless, extrapolation
can often shed valuable light on conditions outside of the region where measurement is possible.
Estimated absolute adsorption isotherms of methane on Longmaxi shale are also shown at different
temperatures up to 415.15 K in Figure 2.2.4. The predicted Gibbs excess adsorption isotherms
exhibit similar properties as the observed isotherms and therefore the extrapolation is determined
to be reasonably dependable. As the temperature increases, the contribution of the adsorbed phase
volume for the absolute adsorption gradually becomes less pronounced. Notably, the negative
effect of temperature on methane adsorption on shale remains clear at all temperatures.
Figure 2.2.4 Gibbs excess adsorption (solid lines, filled symbols) and absolute adsorption
(dashed lines) isotherms of methane on Longmaxi shale as fitted by a dual-site Langmuir
equation (measured up to 355.15 K), extrapolated up to 415.15 K (gradual grey lines)
2.2.4.3 Accurate shale gas-in-place estimations from adsorption measurements
Equilibrium methane adsorption measurements in shale can be used to estimate the geological gas-
in-place (GIP) content of subsurface shale formations. It is important to note that this method does
not take into account any moisture present in the shale which can reduce the methane adsorption
capacity. In addition, this GIP content does not include any contribution from dissolved methane
in the liquid hydrocarbon or brine, and also does not consider the presence of other gaseous
components of natural gas (e.g., higher alkanes and hydrogen sulfide).
43 |
Virginia Tech | The geological GIP is estimated herein as the total amount of methane present in the gaseous and
adsorbed phases in a homogeneous formation of shale. Conceptually, this amount is accessible via
the sum of the free gas phase content,n , and the absolute adsorbed phase content, n .
free a
GIPn n (7)
free a
The amount of gaseous methane is equal to the bulk methane density multiplied by the volume of
the gas phase alone (excluding the volume of the adsorbed phase) as shown in Figure 2.2.5.
However, the volume accessible to the free gas is not the same as the entire empty volume of the
shale since the adsorbed phase occupies a finite volume itself, which is significant at high pressure.
Figure 2.2.5 Schematic depiction of the quantities relevant to gas-solid adsorption in two
distinct regimes: in the dilute limit (left) and at high pressures (right) of the bulk gas
The total GIP amount can also be derived in a much simpler way as the sum of the excess adsorbed
amount and a product of the entire free volume of the empty shale with the bulk gas phase density,
because of the Gibbs definition (from equation 1):
GIPn n V V n (8)
free e a g tot g e
All three of the quantities in the final expression of equation 8 are directly measureable: the total
empty volume accessible to gas in the shale formation (V ), the density of pure gaseous methane
tot
at the equilibrium conditions of the formation (ρ ), and the excess adsorbed amount under these
g
conditions (n ). In practice, the excluded space within the shale (V =V -V ) and/or its
e tot bulk skeletal
44 |
Virginia Tech | skeletal density (ρ = m /V ) are generally measured using pycnometry with a probe gas
shale shale skeletal
such as helium, which is assumed to be non-adsorbing, or by other indirect approaches such as
well logging. This measurement is required in order to make adsorption measurements, for which
the experimental outcome is the excess adsorbed amount (n ). It must therefore be emphasized that
e
the simplest and most accurate approach to estimate the total shale GIP is via equation 8. This is
demonstrated in Figure 2.1.6 where the adsorption isotherm of methane on Longmaxi shale
measured in this work is directly converted to GIP content as a function of pressure at 355.15 K.
No adsorption model is necessary to arrive at the total GIP content in this way.
Figure 2.2.6 Directly calculated shale GIP content as a function of pressure using the
measured data at 355.15 K
Historically, the precise definition of the measured adsorbed amount has been a matter of
confusion. In some reports, the volume of the adsorbed layer is accounted for twice owing to the
incorrect method of summing the “free gas content” in the entirety of the empty pore and the
absolute adsorption content (9, 34,35), corresponding to:
GIP V n (9)
incorrect tot g a
In this approach, where the absolute adsorption isotherms are used in place of the excess quantity
for estimating GIP, the total shale gas content will be significantly overestimated.
This may suggest that the effort to extract the absolute adsorption isotherm from the measured data
is unnecessary for understanding and estimating total GIP since only the excess adsorption data
45 |
Virginia Tech | are required (9, 34, 35). However, in order to determine the relative amount of adsorbed methane
versus gaseous methane in this total figure, the absolute adsorption isotherm is required.
2.2.4.4 Geological gas-in-place resource estimation of a shale gas reservoir in Fuling, China
Generally, coal seams are shallower than shale formations (usually within depths up to 1000 m
below the surface), and therefore the pressure (below 10-15 MPa) is low enough that the
contribution of the volume of the adsorbed methane phase toward absolute adsorption content has
little influence. In this case, employing either the absolute adsorbed amount (equation 9) or the
measured Gibbs excess quantity (equation 8) is reasonable to estimate the total GIP content, though
it is still simpler to use the directly measured quantity. Methane in deep shales, on the other hand,
are in a different geological situation. For example, the Barnett shale completions are up to 2500
m deep, where the reservoir pressure reaches up to 27 MPa and the reservoir temperature can be
up to 360 K (1). Therefore, both pressure and temperature effects on the adsorbed methane content
cannot be neglected. In addition, the large difference between the observed adsorption uptake and
the absolute adsorption uptake at these pressures demonstrates the importance of using an accurate
model of methane content in subterranean shale formations. In other reports, the absolute adsorbed
amount is estimated by simply fitting the excess adsorption quantities along a single isotherm to a
single site (classical) Langmuir isotherm (1, 2, 4, 5, 8), which cannot accurately describe the
changing volume of the adsorbed phase that is taking place. In these cases, regardless of whether
equation 7 or 9 is used, the estimated GIP will be significantly incorrect. This result is
demonstrated in Figure 2.2.7. Logically, there is undeniably a large contribution to the adsorbed
amount at high pressures that is undetected by experiment since the bulk gas density approaches
that of the adsorbed phase and the excess adsorption quantity is no longer accurate.
46 |
Virginia Tech | Figure 2.2.7 Comparison of the Gibbs excess adsorbed methane content (solid line) to two
estimates of absolute adsorbed methane (dashed lines) on Langmaxi shale, at geological
conditions of one completion well (353.15 K and up to 37.69 MPa (34)).
Herein, the geological GIP content of shale gas resources in the Fuling region in China is estimated
as an example to determine the magnitude of the difference between conventional methods and
those employed in this work. The shale gas wells in the Fuling region are the first commercialized
shale gas resource in China (36, 37). The Longmaxi shale formation of the Fuling region is between
2000 to 3000 m deep; the pressure and temperature conditions as a function of depth can be
estimated by the pressure coefficient (15 MPa/km) and the geothermal gradient (27.3℃/km). The
average porosity and density of the shale rock are 4.5% and 2.4 g/mL, respectively (36, 37).
Figure 2.2.8 Comparison of methane adsorption capacity in Fuling region shale formations
under geological temperature and pressure conditions as they vary with depth. Predictions
are based on the following adsorption quantities: observed Gibbs excess adsorption, modeled
absolute adsorption uptake (this work) and the “Conventional Absolute Prediction” (refer
to Supplemental Materials).
47 |
Virginia Tech | Figure 2.2.9 Shale GIP content in Fuling region shale formations under geological conditions,
where temperature and pressure are varied as a function of depth. The Correct Method uses
Eq. 8 where n is calculated using Eq. 6; Incorrect Method 1 uses Eq. 9 where n is calculated
e a
using Eq. 5; Incorrect Method 2 uses Eq. 9 where n is calculated using the Conventional
a
Absolute Prediction (refer to Supplemental Materials).
In this case, both the temperature and the pressure of the actual shale reservoir at maximum depth
are out of the range of data measured in this work. Nevertheless, the dual-site Langmuir model can
be used to predict both the Gibbs excess adsorption isotherms and the absolute adsorption
isotherms under different temperature and pressures, as previously discussed. Figure 2.2.8 shows
that there are significant differences between the observed Gibbs excess adsorption quantity, true
absolute adsorption quantity (as determined by the dual-site Langmuir model), and a common
oversimplified approach to predict the absolute quantity, especially for formations over 1000 m
deep. The oversimplified prediction of absolute adsorption is two times larger or more than the
Gibbs excess adsorption amount, and the best estimate of absolute adsorption is three times larger
or more. This is because the Gibbs excess adsorption amount and the oversimplified prediction
(the “Conventional Absolute Prediction”, see Supplemental Materials) are always less than the
true (absolute) adsorption amount. When equation 9 is used to incorrectly predict GIP, this leads
to a significant overestimation of geological GIP content under real geological conditions as shown
in Figure 2.2.9. The correct method to estimate GIP content as a function of depth is via equation
8. Using the incorrect method 1 and method 2 (shown in Figure 2.2.9), shale gas resources at a
depth of 3000 m are overestimated by 35% and 16%, respectively.
48 |
Virginia Tech | To accurately determine the ratio of adsorbed methane to gaseous methane in the total GIP
resource, one must employ an accurate absolute adsorption quantity. If either the measured excess
adsorption quantity or an oversimplified absolute adsorption prediction (as in Figure 2.2.7) is used,
the result will be a significant underestimation of the contribution to the total GIP from adsorbed
methane. The correct method is to consider the absolute adsorption quantity as the total adsorbed
amount, modeled by a physically robust method such as the dual-site Langmuir equation used
herein. This is shown as a function of formation depth in Figure 2.2.10. The actual adsorbed
methane amount still accounts for 46% of the total GIP content at a depth of 4000 m. If only the
excess adsorption quantity is taken, the result is a very large underestimation of the contribution
of adsorbed methane to the total GIP content: less than 12% at a depth of 3000 m.
Figure 2.2.10 Comparison of the estimated contribution to total GIP content by adsorbed
methane in Longmaxi shale by three methods: where the actual adsorbed amount is
estimated as the excess uptake (solid red), absolute uptake (by a dual-Langmuir fit, dashed
red), and by a conventional prediction of absolute uptake (dashed black). For demonstration
purposes, the correct total GIP content is used in all cases (via Eq. 8).
2.2.5 Conclusions
In this work, laboratory measurements of high pressure methane adsorption (303 - 355 K and up
to 27 MPa) are presented. Then, the dual-site Langmuir model is applied to describe and accurately
predict high pressure methane adsorption behavior in Longmaxi shale (China). Finally, the shale
49 |
Virginia Tech | GIP resources in deep high pressure shale formation are accurately predicted. Several preliminary
conclusions can be made,
(1) The crossover of the adsorption isotherms under high pressures and high temperatures are
observed and reasonably interpreted.
(2) Dual-site Langmuir model can not only accurately describes observed adsorption isotherms
but also can extrapolate adsorption isotherms beyond test data without using any empirical
relationship.
(3) For depths greater than 1000 m (> 15 MPa) in the subsurface, the shale GIP resources have
historically been significantly overestimated, and the ratio of the adsorbed phase compared
to the free gas has been significantly underestimated.
(4) On the basis of the dual-site Langmuir model, the proposed method allows accurate
estimations of the true shale GIP resource and the relative quantity of adsorbed methane at
in situ temperatures and pressures representative of deep shale formations.
Acknowledgments
Financial assistance for this work was provided by the U.S. Department of Energy through the
National Energy Technology Laboratory’s Program under Contract No. DE-FE0006827, the State
Key Development Program for Basic Research of China (Grant No. 2014CB239102) and
Department of Science and Technology at China Petroleum & Chemical Corporation (Grant
No.P12002, P14156).
References
41. Curtis, J. B. (2002). Fractured shale-gas systems. AAPG bulletin, 86(11), 1921-1938.
42. Montgomery, S. L., Jarvie, D. M., Bowker, K. A., & Pollastro, R. M. (2005). Mississippian
Barnett Shale, Fort Worth basin, north-central Texas: Gas-shale play with multi–trillion
cubic foot potential. AAPG bulletin, 89(2), 155-175.
43. King, G. E. (2010). Thirty years of gas shale fracturing: what have we learned? In SPE
Annual Technical Conference and Exhibition. Society of Petroleum Engineers.
http://dx.doi.org/10.2118/133456-MS.
50 |
Virginia Tech | 2.3 Deep means different: concept of the deep shale gas reservoir and its influence on shale
gas development
Xu Tang*, Nino Ripepi*,†, Cheng Chen*, Lingjie Yu ‡, §
(*Department of Mining and Minerals Engineering & †Virginia Center for Coal and Energy
Research, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, 24060, U.S; ‡
Wuxi Research Institute of Petroleum Geology of Sinopec Exploration & Production Research
Institute & §Sinopec Key Laboratory of Petroleum Accumulation Mechanisms, Wuxi, Jiangsu,
214151, China)
Abstract: Misunderstanding of methane adsorption behavior in shales under high-pressure
conditions has resulted in inappropriate application of shale gas transport models and
overestimation of shale gas resources in shale gas reservoirs. This work first reviews current
fundamental issues in shale gas development. Then, the concept of the deep shale gas reservoir is
proposed to provide a new perspective on shale gas development based on high pressure (up to
27MPa) methane adsorption studies in shales under different temperatures. This concept is on the
basis that the dual-site Langmuir model can not only describe the methane adsorption behavior
under high pressure conditions but also differentiate the true adsorbed methane content and
gaseous methane content in deep shale gas reservoirs. The successful application of the dual-site
Langmuir model in describing methane adsorption behavior in shale lays the foundation for
understanding methane adsorption behavior in shale, assessing shale GIP resource in deep
formations, and optimizing shale gas transport models for deep shale gas reservoirs. Finally, the
implications of the deep shale gas reservoir concept on shale GIP resource estimation,
thermodynamic analysis of high pressure methane in shale, and shale gas transport model are
discussed.
Key words: shale gas, deep, transport, gas-in-place, Langmuir
55 |
Virginia Tech | 2.3.1 Introduction
Shale gas has played a major role for the United States natural gas production over the past ten
years and there remain significant reserves throughout the world in deep formations up to 2500 m
(NETL, 2009; Kuuskraa et al, 2013; Wang et al, 2014; Curtis, 2002; Montgomery et al, 2005).
Shale gas typically exists in three different phases within shale formations: (i) as free compressed
gas, (ii) as adsorbed fluid on the surface, and (iii) as a dissolved component in kerogen, liquid
hydrocarbon and brine. The adsorbed phase accounts for 20% to 80% of the total amount based
on current studies from five major shale formations in the United States (Curtis et al, 2002). Thus,
the estimation of the adsorbed amount of natural gas, the largest component of which is methane,
significantly influences the final determination of the geological GIP resource and the working life
of a shale gas producing well (Ambrose et al, 2012; Singh et al, 2016).
Since shale formations are typically deep, in-situ reservoir pressure and temperature can be as high
as 27MPa and 76℃, respectively (Curtis et al, 2002). It is still unclear whether the deep in-situ
condition (high pressure [>15 MPa] and high temperature [up to 76℃]) can change methane
adsorption behavior in shale. Because of the limited data for methane adsorption in shale under
high pressures, the shale gas industry still follows the methodology used in shallow coal seams
and shale formations to estimate the shale GIP resource in the subsurface without seriously
considering the in-situ high-pressure conditions (Curtis, 2002; Montgomery et al, 2005; Kuuskraa
et al, 2013; Andrews, 2013). The standard practice for estimating shale GIP is to use methane
adsorption measurements under intermediate-pressure conditions (10-15 MPa) modeled by the
two-parameter Langmuir equation to predict the methane adsorption behavior in the higher-
pressure region (>15 MPa) (Curtis et al, 2002; Montgomery et al, 2005; Kuuskraa et al, 2013;
NETL, 2009; Andrews et al, 2013). Whether the commonplace methodology is reasonable or not
needs more research. Even though it is known that the neglected volume of adsorbed layers under
in-situ conditions results in overestimation of shale GIP (Ambrose et al, 2012), methane adsorption
behavior under high-pressure conditions has not drawn researcher’s attention from either academia
or industry especially in modeling shale gas transport in the subsurface As evidenced by the fact
that the two-parameter Langmuir model is still the foundation for developing shale gas transport
model (Yu et al, 2014; Akkutlu et al, 2012; Civan et al, 2011; Singh et al, 2016; Wu et al, 2015;
Naraghi et al, 2015; Pan et al, 2015; Yang et al, 2015; Wu et al, 2016). The observed adsorption
isotherms are typically fitted using two-parameter Langmuir equation to differentiate the adsorbed
56 |
Virginia Tech | gas content and study the contribution of adsorbed gas content on shale gas production (Yu et al,
2014; Akkutlu et al, 2012; Civan et al, 2011; Wu et al, 2015; Singh et al, 2016; Naraghi et al, 2015;
Pan et al, 2015; Yang et al, 2015; Wu et al, 2016). However, extending these gas transport models
to high-pressure shale formations needs more evidence.
This work reviews studies in shale gas development with emphasis on the fundamentals of shale
GIP estimation and gas transport in shale and then points out current issues in shale gas studies.
Several misunderstood concepts are also clarified. This paper proposes a new concept, the deep
shale gas reservoir, in response to historical studies that describe high pressure methane adsorption
by the dual site Langmuir equation. Then, the implication of the deep shale gas reservoir concept
in shale gas development are discussed in detail.
2.3.2 Current fundamentals for shale gas development
2.3.2.1 Shale GIP estimation in shale formations
Generally, the geological GIP resource is estimated as the total amount of methane present in the
gaseous and adsorbed phases in a shale formation (assuming a negligible contribution from
dissolved methane in kerogen, liquid hydrocarbons and brine). Equilibrium methane adsorption
measurement in shale is needed in order to estimate the geological GIP content of deep shale
formations. It is important to note that this method does not take into account any moisture which
can reduce the methane adsorption capacity. In addition, this GIP content does not include any
contribution from dissolved methane in kerogens, liquid hydrocarbons and brine, and also does
not consider the presence of other gaseous components of natural gas (e.g., higher alkanes and
hydrogen sulfide) (Ji et al, 2014 & 2015; Rexer et al, 2013).
Shale GIP resource is calculated via the sum of the free gas phase content,n , and the absolute
free
adsorbed phase content, n (illustrated in Figure 2.3.2).
a
GIPn n V V (1)
free a gas free a a
where, and V are the free gas density and volume, respectively. and V are the density
gas free a a
and volume of adsorbed gas, respectively, which cannot be measured using current technologies.
57 |
Virginia Tech | Figure 2.3.1 Conceptual model for shale gas phases in formations: both V (skeletal
shale
volume of shale) and V (total volume of pore space) can be measured using Helium
tot
intrusion tests; V (volume of adsorbed layers) and V (free gas volume existing in the shale
a free
formation) are unmeasurable using current technologies.
Eliminating V in equation (1) using volume conservation (V V V ), one obtains equation
free free tot a
(8):
GIP V n V (2)
gas tot a gas a
Under low-pressure conditions (<15 MPa),V is very small and thus V can be ignored.
a gas a
Equation (8) is then rewritten as:
GIP V n (3)
gas tot a
Equation (3) is the standard equation for estimating the shale GIP resource (NETL, 2009; Kuuskraa
et al, 2013; Wang et al, 2014; Curtis, 2002; Montgomery et al, 2005). For term (n ), the standard
a
method uses the two-parameter Langmuir equation to fit the isotherm adsorption test data:
K(T)P
n n (4)
a max1K(T)P
where n is the absolute adsorption quantity under reservoir temperature and pressure, n is the
a max
maximum single-layer Langmuir adsorption capacity, and K(T) is the temperature-dependent
E
Langmuir equilibrium constant, written as K(T) A exp( 0 ) , where E is the energy of
0 RT 0
adsorption and A is the pre-exponential coefficient, both of which are independent of temperature.
0
58 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.