id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
610936ae8804439fdde55e7b | 11 | Data Mining. We extracted the synthesis conditions from MOF publications using different NLP techniques. To select synthesis paragraphs, we developed a decision tree algorithm based on a keyword list selected from 100 MOF synthesis papers. To analyse the synthesis paragraph and identify information about chemical entities, experimental steps, and corresponding conditions associated with those steps, we applied the ChemicalTagger software. When precursors, solvents and additives, as well as solvothermal synthesis conditions were extracted, we compared the metal element from the automatically formed synthesis protocol to the CoRE MOF database to eliminate mismatched conditions. The results of this fully automated data extraction are collected in the SynMOF-A database. |
610936ae8804439fdde55e7b | 12 | Machine Learning. We developed a code to extract the MOF linker from the CIF. The RDKit library was further used to evaluate the molecular fingerprint of the extracted linker. The MOF metal nodes were represented by their full electronic configuration. The molecular fingerprint of the linker and the full electronic configuration of the metal node, accounting for its oxidation state, were combined to form the input of the ML model. This input representation was compared to the MOF representation developed by Kulik and coworkers, relying on autocorrelation features of the metal cores and the linkers. The output of the ML model was the MOF synthesis conditions, namely temperature, synthesis time, solvent properties and additive type. Depending on the specific synthesis conditions, we evaluated several regression models, in particular random forest regression and neural networks. The scikit-learn library in Python was used for the implementation of the ML models. 70% of the full dataset was used to train the ML model, while the remaining data was used to test the model. In the case of solvent property prediction, we limited the data to MOFs with single-solvent synthesis. To quantify the accuracy of the trained ML model, we calculated the mean absolute error and the correlation coefficient r 2 of the training and test dataset for the regression tasks. The accuracy of the ML model for the classification tasks were quantified by calculating the normalized confusion matrix. |
6334a9fbea6a22e834017b28 | 0 | Manganese oxides are a uniquely diverse class of materials both structurally and chemically. Over 30 Mn (oxy)(hydr)oxide minerals are known to form via natural processes with a wide range of possible crystal structures, Mn oxidation states, and intercalated ions. An even greater variety can be synthesized with simple laboratory methods. In nature, this rich assortment of structure and ion insertion chemistry plays a vital role in regulating biological as well as geochemical cycles. It has also sparked significant interest in using Mn oxides in next generation energy storage and conversion, chemical separations, and pollution control technologies. Mn atoms can adopt oxidation states ranging from 0 to +7 depending on their environment. In solid (oxy)(hydr)oxide phases, Mn cations are generally found in the +2, +3 or +4 oxidation states, octahedrally coordinated by oxygen. However, Mn can also appear in tetrahedral coordination in phases like Mn 3 O 4 . A wide variety of Mn oxide crystal structures arises from different configurations of MnO 6 octahedra (shown in Figure ). These octahedra assemble into linear chains or 2D planes via edge sharing (i.e. adjacent Mn ions sharing two oxygen atoms). Extended crystal structures with 1D tunnels are formed when these chains attach through corner sharing oxygens, and 2D layered structures are formed when planes of octahedra stack. Common Mn oxide polymorphs include the 𝛼 (2×2 tunnels), β (1×1 tunnels), R (2×1 tunnels), 𝛿 (layered) and λ (spinel) structures (Fig. ). Amorphous Mn oxides, intergrowths (𝛾-MnO x with both 2×1 & 1×1 tunnels) and larger tunnel structures (𝜏-MnO x with 3×3 tunnels) also exist (structures not shown). |
6334a9fbea6a22e834017b28 | 1 | The remarkable chemical diversity of Mn oxides comes from the ability of these structures to adopt multiple Mn oxidation states and host a variety of intercalated ions. When H + is the charge compensating ion to Mn redox, the reduction potential of MnO 2 is near the oxidative voltage stability window of water (1.23 V vs. RHE). This makes MnO 2 a suitable positive electrode for aqueous batteries like the ubiquitous alkaline Zn-MnO 2 primary battery. In addition to protons, MnO x crystal structures can accommodate various monovalent (e.g. Na + , K + ) and multivalent (e.g. Mg 2+ , Zn 2+ ) cations inside their tunnels or between their layers. As a result, Mn oxides are also targeted for low-cost "beyond-Li" battery electrodes, both aqueous and non-aqueous, with large theoretical capacities (617 mAh/g MnO 2 ) if the full 2-electron redox (Mn 2+ ⇌ Mn 4+ + 2e -) can be cycled reversibly. Higher energy densities, lower costs and greater safety could potentially all be achieved relative to current-generation Li-ion batteries. The sheer number of possible combinations of structures and ions in the Mn oxide system frustrates understanding its electrochemical ion selectivity. Exhaustive sampling of this large phase space would be prohibitively time-consuming, so experimental works typically focus on a single ion-polymorph couple. In contrast, computational approaches can efficiently evaluate the thermodynamics of materials, enabling holistic comparison of a wide range of chemistries. Comparing different ions and Mn oxidation states with a single approach requires a special computational scheme. Previous approaches, however, have been tailored to computing material properties at constant Mn oxidation state, like structure selection during polymorph synthesis. To compare between crystal structures with identical oxidation state, these works used the meta-GGA SCAN functional, which was shown to reproduce the experimentally-observed energetic ordering of b and R-MnO 2 . The behavior of the polymorphs with changing voltage was not considered. Coarser methodologies have been used to screen MnO 2 polymorphs as potential Caion battery cathodes, and study the proton-coupled redox behavior of select polymorphs in alkaline Zn-MnO 2 batteries. Yet, the absence of a tailored benchmarking scheme with Hubbard-U corrections limits the predictive power of these approaches across the broad scope of ion insertion electrochemistry and manganese redox. Conversely, computational investigations with Hubbard-U corrections have typically been limited to a single polymorph/electrolyte pair. In this work, we compare the accuracy of the SCAN and PBE functionals including the Hubbard-U parameter and develop a rigorous correction scheme to compute accurate redox energetics. We find that the PBE+U functional is a less computationally expensive and slightly higher accuracy approach to calculate ion insertion voltages in the Mn oxide system. Using this efficient approach, we investigate a broad range of Mn oxide electrochemical behavior, directly comparing the voltages and ion capacities of 𝛼, β, R, 𝛿 and λ-MnO 2 in H + , Li + , Na + , K + , Mg 2+ , Ca 2+ , Zn 2+ and Al 3+ electrolytes. We place particular emphasis on the implications of our calculations for energy storage and conversion in aqueous electrolytes, whereby we compare the affinity of MnO 2 polymorphs to insert protons versus non-protonic cations. While the selectivity of non-protonic cation insertion hinges on oxygen coordination environments, protons are usually favored when present in the electrolyte. We also evaluate the stability of these polymorphs with respect to dissolution and structural reorganization to elucidate promising design strategies for nextgeneration batteries. Finally, we examine the effect of ions inserting with partial solvation shells in 𝛼 and 𝛿-MnO 2 , which are large enough to accommodate water molecules. |
6334a9fbea6a22e834017b28 | 2 | All calculations were performed with the Vienna Ab-initio Simulation Package (VASP) 36 using projector-augmented wave pseudopotentials (PAW) to approximate the core electrons. A plane wave basis set with a cutoff of 600 eV was chosen to represent the valence states. Calculations were performed on a gamma-centered Monkhorst-Pack grid with a minimal density of k-points in each orthogonal direction specified by the equation: 50 Å/(unit cell length). Energies were relaxed to a 10 -5 eV convergence threshold, while the forces on each atom were converged to 0.02 eV/Å. An effective Hubbard-U (U eff ) value was applied to the Mn 3d-electrons (described below) to better represent the semiconducting nature of the partially occupied Mn d-band. |
6334a9fbea6a22e834017b28 | 3 | To obtain accurate enthalpies, we added zero-point energies (ZPE) and phonon vibrational energies (U vib ) to the DFT-derived electronic energies. ZPE and U vib were calculated under the independent harmonic oscillator approximation as implemented in the ASE Thermochemistry class. We found that the combined vibrational energies are significant for protonated (H x MnO 2 • yH 2 O) structures (Fig. ). In fact, a linear relationship exists between ZPE+U vib and the number of hydroxyl groups in H x MnO 2 • yH 2 O compounds. For ion-inserted structures without hydroxyl groups (A x MnO 2 , A ≠ H), we approximated the combined vibrational energies as that of pure b-MnO 2 . |
6334a9fbea6a22e834017b28 | 4 | To calculate formation enthalpies, all compounds were referenced to their constituent elements at standard state. A correction scheme was applied to rectify well-known DFT errors: namely, overbinding of O 2 and incomplete cancelation of self-interaction errors. To address the former, our correction scheme adjusts the energy of O 2 to reproduce the experimental formation enthalpy of water, consistent with previously reported schemes. To minimize error arising from incomplete self-interaction cancelation, the energy of Mn metal was set to the average value calculated from five experimental formation energy reactions (eq. 1-5, Table ). Critically, the five reactions include Mn ions with +2, +3, and +4 oxidation states. Averaging the Mn metal reference energy across these reactions is likely to achieve a reasonable balance in the accuracy of DFT-computed formation energies across all oxidation states of AxMnO2 • yH2O computed in this work. |
6334a9fbea6a22e834017b28 | 5 | 3 Mn (") + 2O $ (&) ↔ Mn ' O ( ∆H = -1387.8 kJ/mol [5] Previously reported DFT energy correction schemes have minimized errors by using computationally-expensive hybrid functionals, fitting U eff , mixing GGA and GGA+U energies, and fitting elemental reference energies. Other strategies have included fitting the DFT-computed energy of O 2 to reproduce the formation energies of binary oxides or of water. We chose a synergistic strategy of fitting U eff (discussed below), adjusting the Mn reference energy, and referencing the energy of O 2 to the formation energy of water. Similar schemes have been shown to compare favorably for the formation of high fidelity Pourbaix diagrams. Our scheme also keeps bulk DFT calculations maximally compatible with surface calculations. |
6334a9fbea6a22e834017b28 | 6 | To calculate free energies, experimental entropies were taken from thermochemical tables (Table ). Three approximations were made to fill gaps in available experimental entropy data for A x MnO 2 • yH 2 O phases. First, the difference in entropy between polymorphs with identical stoichiometry was assumed to be negligible. For example, the entropy of all pure MnO 2 structures was approximated as that of b-MnO 2 . Second, additional entropy arising from non-protonic ion insertion was assumed to be small, such that the entropy of A x MnO 2 (A ≠ H) was approximated as that of MnO 2 . Finally, the configurational entropies of intercalated H + and solvation shell H 2 O were assumed to be small so the entropies of H x MnO 2 • yH 2 O (0 < x < 1 & 1 < x < 2) were approximated as linear combinations of the entropies of b-MnO 2 , 𝛾-MnOOH, and 𝛿-Mn(OH) 2 according to the number of O-H bonds present (Fig. ). These approximations likely slightly underestimate the stability of A x MnO 2 • yH 2 O phases. However, we find that entropy accounts for approximately 10% or less of the free energy of Mn oxides at room temperature, consistent with previous work. Approximations made for unknown entropies are therefore unlikely to significantly affect our conclusions. |
6334a9fbea6a22e834017b28 | 7 | While Mn oxides are known to display complex non-collinear spin structures, experiments show most phases exhibit predominately antiferromagnetic (AFM) interactions. Previous computational work has also shown that Mn oxides can be reasonably approximated as collinear antiferromagnets. Therefore, we adopt the AFM magnetic structures previously shown to have the lowest energy. 29 |
6334a9fbea6a22e834017b28 | 8 | Redox energetics via Hubbard-U. Before computing the energies of A x MnO 2 • yH 2 O phases for which experimental data is unavailable, we compared the accuracy of the SCAN and PBE functionals on a smaller set of compounds (Fig. ), hereafter referred to as the calibration set. The compounds chosen for the calibration set meet two criteria: i) they are predicted to be equilibrium (or very close to equilibrium) phases within water's voltage stability window, and ii) experimental thermochemical data exists for benchmarking (Table ). |
6334a9fbea6a22e834017b28 | 9 | Despite the good performance of the SCAN functional in predicting the relative energies of the pure MnO 2 polymorphs (Fig. ), in the absence of a correction scheme, both SCAN and PBE functionals display considerable errors in predicting Mn oxide formation energies (Fig. ). This finding is consistent with previous investigations of transition metal oxides. While DFT errors tend to cancel when calculating the relative energy of two compounds with similar electronic structures, errors persist in formation energy calculations (Eqs. 1-5), which compare compounds with different oxidation states. Implementing the correction scheme detailed in the methods section improves the accuracy of the computed formation energies by a factor of 5 (Fig. , averaged across the three functionals). However, mean absolute error of over 100 meV/atom remains (Fig. at U = 0), frustrating reliable investigation of redox energetics. To further counteract self-interaction errors, we fit an effective Hubbard-U value to the calibration set of Mn oxides (Fig. ). Unfortunately, no single U eff can exactly reproduce the energies of all Mn oxides simultaneously. The computed energies of small bandgap semiconductors with high covalency (small Mn 3d -O 2p charge transfer gaps) like b-MnO 2 increase with increasing U eff , while materials with greater ionic-character like Mn(OH) 2 exhibit decreasing energy with increasing U eff . For both PBE and SCAN, the Hubbard-U value that minimizes errors across the entire calibration set is between 2.5 eV and 2.75 eV (Fig. ). Since the error is equivalent at these two values, we chose to perform all subsequent calculations with U eff = 2.75eV, corresponding to mean absolute errors (MAE) of 12.2 meV/atom for PBE and 18.2 meV/atom for SCAN. |
6334a9fbea6a22e834017b28 | 10 | The optimal U-values we find are smaller than those used by the Materials Project and those reported in other works. These previously reported U-values are typically derived from smaller sets of binary Mn oxides, which exclude Mn (oxy)hydroxides. Therefore, the optimal Uvalue we report likely yields better overall accuracy in predicting bulk electrochemical redox behavior, particularly in aqueous environments. Notably, we find that PBE+U systematically displays lower errors in predicted formation energies than SCAN+U regardless of U-value. At their respective minima, PBE+U displays almost half the error of SCAN+U (Fig. ). However, the difference in error between the two functionals is small compared to the absolute error incurred by both functionals with no Hubbard-U corrections. The PBE(+U) functional is less computationally expensive than SCAN(+U), which is important for investigating large electrochemical phase spaces, and is known to perform well in surface calculations. For these reasons, we chose to perform all subsequent calculations using PBE+U with U eff = 2.75 eV. There are tradeoffs to using PBE+U over SCAN+U, however. For example, SCAN+U displays lower error in computed lattice parameters (Table ), although the difference is less than 1% of the experimental values. SCAN(+U) also correctly predicts β-MnO 2 to be the ground state polymorph, whereas PBE(+U) does not. Caution should therefore be exercised when analyzing small differences in energy between MnO x structures with identical stoichiometry. However, these errors have minimal effect on the topotactic electrochemical properties of a given MnO x polymorph. 2 . Applying the DFT correction scheme detailed above, we compute the energies of H + , Li + , Na + , K + , Mg 2+ , Ca 2+ , Zn 2+ and Al 3+ ions inserting into 𝛼, β, R, 𝛿 & λ-MnO 2 (Fig. ). These ions were chosen for their relevance to advanced energy storage and conversion technologies. While the ion insertion behavior of 𝛾-MnO 2 (i.e. electrolytic MnO 2 , "EMD", the common intergrowth polymorph used in alkaline batteries) was not computed directly, it is expected to be similar to that of its constituent polymorphs, β and R-MnO 2 . A full list of computed structures and their energies is given in Table and is available online. To distinguish solid-solution from two-phase intercalation behavior, we conducted a convex hull analysis for each structure-ion pair (Fig. ). From the resulting set of equilibrium structures, ion insertion voltages were calculated according to the following equation: To enable a direct comparison between ions, all voltages were referenced to the standard hydrogen electrode (SHE). Figure displays these ion insertion voltages and their corresponding Mn oxidation states calculated at 1 M concentration for all ions except H + . Proton concentration was set to 10 -14 M, corresponding to a pH of 14 in aqueous conditions. For easier comparison of the voltages and capacities that could be achieved in non-aqueous batteries with metal anodes, Fig. plots the ion-insertion voltages vs. A/A n+ . In both figures, higher voltages signify more favorable insertion events. On the left, MnO 2 crystal structures are shown with representative ions occupying their lowest-energy crystallographic sites. |
6334a9fbea6a22e834017b28 | 11 | The maximum ion capacity of a given polymorph was taken to be the highest degree of intercalation in which the original MnO 2 polymorph structure is maintained. Beyond this maximum capacity, continued ion insertion causes dramatic structural rearrangement, which can lead to irreversible capacity degradation. Examples of such rearrangements include broken Mn-O bonds (considered in this work to be at distances greater than 3 Å) and Mn atom migration to non-octahedral sites. Because we exclude such distorted phases, the maximum ion capacities we calculate for each polymorph differ from previous work. In Fig. , voltages are plotted against x in A x MnO 2 for direct comparison of maximum ion capacity. |
6334a9fbea6a22e834017b28 | 12 | In 𝛼-MnO 2 (Fig. ), large cations like Na + , K + and Ca 2+ are favored to insert at low capacity. These large cations occupy the 4-coordinated 2a (Na + & Ca 2+ ) and 8-coordinated 2b (K + ) Wyckoff sites at the center of the polymorph's [2×2] tunnels. For K + and Na + , the ionic bonds formed with the surrounding oxygens are strong enough that removing these cations from their sites entirely requires applied voltages beyond the oxidative stability limit of water in alkaline electrolytes. This finding agrees well with experiment, as 𝛼-MnO 2 is typically synthesized with small amounts of intercalated K + , Na + or Ba 2+ , and extraction of these ions is achieved with acid treatment. However, the large sizes of Na + , K + , and Ca 2+ severely limit the maximum capacities that can be achieved, with the sites filling completely at < 200 mAh/g. Higher capacities can be achieved with smaller ions (e.g. Mg 2+ , Zn 2+ , Al 3+ ), which preferentially fill the 8h sites. However, beyond Mn 3.875+ , proton insertion is thermodynamically favored relative to these other ions. Notably, because the large [2×2] tunnels limit the number of available host sites, we find that the full 2-electron capacity of MnO 2 can only be achieved with protons in the 𝛼 polymorph. |
6334a9fbea6a22e834017b28 | 13 | In the β and R polymorphs (Fig. ), H + insertion is favored, even at pH 14. The smaller [1×1] tunnels of the rutile β structure prevent ions larger than Zn 2+ (e.g. Na + , Ca 2+ , K + ) from inserting while the [2×1] tunnels of R-MnO 2 are just large enough to accommodate Na + at low concentrations. The voltages and maximum capacities of both Zn 2+ and Mg 2+ insertion into the 2b octahedral sites of β-MnO 2 are low, however. Insertion beyond Mn 3.75+ leads to broken Mn-O bonds and structural distortions (Fig ). The slightly larger tunnels of the R polymorph are a more suitable host for these ions. Like the 𝛼 polymorph, the full 2-electron capacity of β-MnO 2 is only accessible via proton insertion. While the R polymorph could, in theory, achieve 2-electron redox with Mg 2+ and Zn 2+ ions, H + will insert preferentially, if available. |
6334a9fbea6a22e834017b28 | 14 | In 𝛿-MnO 2 , ion insertion voltages are closely spaced at low capacities (i.e. high Mn oxidation states) (Fig. ), as the distance between Mn layers can expand or contract to accommodate inserted ions in their preferred coordination environment. This finding helps explain the pseudocapacitive behavior of layered MnO 2 in aqueous electrolytes. In λ-MnO 2 (Fig. ), Zn 2+ and Li + are favored to insert at low capacities into the 8a tetrahedral sites. At higher capacities, once the tetrahedral sites are full, continued insertion forces these ions to occupy octahedral sites (above Mn 3.5+ for Li + and above Mn 3+ for Zn 2+ ), leading to a corresponding drop in the insertion voltages. The relatively high stability of Zn 2+ in tetrahedral sites explains why hetaerolite (spinel ZnMn 2 O 4 ) is a common degradation product in primary alkaline batteries with Zn anodes and in secondary Zn 2+ shuttle batteries with MnO 2 cathodes. Spinel LiMn 2 O 4 is also a well-known Li-ion battery cathode and was the first demonstrated positive electrode in aqueous Li-ion batteries, although its lower voltage compared to Ni and Co-based cathodes limits its desirability for volumeconstrained applications. |
6334a9fbea6a22e834017b28 | 15 | Our calculations suggest that, in 𝛿-MnO 2 , the full 2-electron Mn redox can be accessed topotactically with Al 3+ at high voltage. Similarly, R and λ-MnO 2 can achieve close to the full 2electron redox. While batteries with Al 3+ shuttle mechanisms may suffer from poor kinetics owing to strong electrostatic interactions between Al 3+ cations and anions in the electrolyte or in the MnO 2 framework, strategies like vacancy engineering and electrolyte tuning may be able to overcome these limitations. The high theoretical energy densities predicted by our calculations merit further investigation. |
6334a9fbea6a22e834017b28 | 16 | In neutral electrolytes, proton insertion is thermodynamically favored regardless of polymorph (Fig ). In basic electrolytes, however, competition exists between the insertion of protons and most non-protonic cations. Even when [H + ] = 10 -14 M, only a few ions are strongly favored to insert over protons, namely Na + & K + in 𝛼-MnO 2 , and Li + & Zn 2+ in λ-MnO 2 . The insertion voltages of other ions in 𝛼 and λ-MnO 2 are close to or lower than that of H + , and in the polymorphs with small tunnels (β & R, Figs. ), H + is always favored. This last finding is consistent with the wellknown proton insertion mechanism of alkaline primary batteries with highly concentrated KOH electrolytes. 𝛾-MnO 2 , the electrode material in these batteries, has [1×1] and [2×1] tunnels identical to the β and R polymorphs. |
6334a9fbea6a22e834017b28 | 17 | No ion displays preferential insertion relative to H + over the entire 2-electron capacity of MnO 2 in any of the polymorphs. This finding has important implications for MnO 2 electrodes in batteries, pseudocapacitors, and other ion insertion applications. Notably, in the presence of protic solvents, either proton-only or mixed ion/H + insertion reactions are likely to occur depending on pH and polymorph. Great interest exists in developing secondary aqueous batteries with MnO 2 cathodes, particularly those paired with Zn/Zn 2+ anodes. While promising performance improvements have been made, the precise operating mechanism of these batteries remains a subject of debate. Experimental reports on different polymorphs have hypothesized proton insertion, 17,70 Zn 2+ insertion, and Zn 2+ /H + co-insertion mechanisms. Our results introduce polymorph-specific explanations for these findings and demonstrate that co-insertion with H + is a general feature of the MnO 2 system, not just in Zn 2+ electrolytes. In fact, in protic electrolytes like water, dual ion insertion is likely to be the only possible charge storage mechanism other than a proton-only mechanism, and an effective ion-shuttling battery should pair a proton-discharging anode for maximum energy density. |
6334a9fbea6a22e834017b28 | 18 | Dual ion insertion reflects the competition between a driving force to form O-H bonds and an electrostatic driving force to form ionic bonds between inserted non-protonic cations and oxygen anions in the Mn oxide framework. This latter driving force can be evaluated across cations and available Wyckoff sites using bond valence theory. |
6334a9fbea6a22e834017b28 | 19 | We use bond valence (BV) theory to rationalize why certain non-protonic cations preferentially occupy certain crystallographic sites. This relatively simple electrostatic model of crystal chemistry modifies Pauling's second rule by relating the strength of a cation-anion bond (s i ) to its length (R i ) via two bond-specific empirically fitted constants (R o & b). Summing the strengths of the bonds to a cation (details in Table ) as defined by Eq. 7, |
6334a9fbea6a22e834017b28 | 20 | In the Mn oxide system, we find that comparing values of the site preference index (SPI) across ion-site pairs can predict the preferred ion for a given site and the preferred site for a given ion. The site preference index is a measure of how close a cation's BV sum is to its formal charge, with lower values signifying a more favorable occupation. The results of this inexpensive calculation for 𝛼-MnO 2 (Fig. ) show why K + prefers the 8-coordinated 2b site, whereas Na + and Ca 2+ prefer the 4-coordinated 2a site and Li + , Zn 2+ , Mg 2+ and Al 3+ occupy the smaller 5-coordinated 8h site. They also reveal why, for example, in the 2a site, Na + is preferred relative to Ca 2+ (reflected in their respective insertion voltages) despite its slightly larger ionic radius. The ordering of site preference index values for different ions in the same site rationalize the ordering of the insertion voltages of these cations at low capacity (Fig. ). The trend from lower to higher index values going from Li + to Zn 2+ to Mg 2+ to Al 3+ in the 8h site explains the opposite trend (from higher to lower) in insertion voltages of these ions. At higher capacities, Jahn-Teller distortions and ioninduced structural relaxations, which BV theory does not capture, can complicate the predictive power of this approach. Nevertheless, its success, particularly at low capacities, highlights the electrostatic interactions between cation and site that determine the energy of non-protonic cation insertion into MnO 2 in the dilute limit. |
6334a9fbea6a22e834017b28 | 21 | A n+ ] = 1 M concentration, we expand our computational analysis to investigate the stability of MnO 2 polymorphs across a broad range of solvent pH and voltage. Pourbaix diagrams, which plot the most stable phases at each value of pH and applied voltage, indicate how MnO 2 electrodes are likely to evolve under various conditions. However, the original Pourbaix diagram for the Mn oxide system published in the Pourbaix Atlas considered only select phases with Mn-O-H composition. It therefore remains unknown whether A x MnO 2 compounds are thermodynamically stable under relevant electrochemical conditions. It is also unclear how these compounds evolve with changing pH and voltage. |
6334a9fbea6a22e834017b28 | 22 | Using experimental values where available (calibration set, Table ) and filling in the gaps with our DFT-computed energies, we construct Pourbaix diagrams for 1 molal aqueous solutions of Li + , Na + , K + , Mg 2+ , Ca 2+ , Zn 2+ and Al 3+ . The energies of ionic species are taken from reference data. We find that of the 100+ A x MnO 2 compounds we calculated, only three (𝛼-K 0.125 MnO 2 , 𝛼-Na 0.125 MnO 2 , and λ-Li 0.5 MnO 2 ) are thermodynamically stable at any value of pH and voltage (Fig ). All other ion-polymorph pairs are found to be metastable with respect to β-MnO 2 , Mn 2 O 3 , Mn 3 O 4 , and 𝛿-Mn(OH) 2 . The stability windows of these three compounds exist predominately at high pH, in agreement with the finding that proton redox dominates at lower pH (Fig ). The inserted ions in these compounds are also the same ones found previously to be strongly favored over proton insertion at pH 14 (Fig. ). The conditions chosen for this Pourbaix analysis are reflective of those experienced by Mn oxide electrodes in electrochemical devices. Previous work has shown, however, that a greater number of ion-inserted phases can be stabilized with supersaturated solutions that have even greater ion concentrations. While only three ion-inserted polymorphs are thermodynamically stable under electrochemical conditions, several others are highly metastable. Figure shows the free energies of five of these low-lying metastable phases (𝛿-Ca 0.25 MnO 2 , 𝛿-K 0.25 MnO 2 , λ-Mg 0.5 MnO 2 , 𝛿-Na 0.25 MnO 2 , and λ-Zn 0.5 MnO 2 ) as a function of applied voltage. For example, at pH 14 and 950 mV vs. RHE, 𝛿-Ca 0.25 MnO 2 and 𝛿-K 0.25 MnO 2 are within 35 meV of the ground state, while λ-Mg 0.5 MnO 2 and 𝛿-Na 0.25 MnO 2 are within 65 meV. Such low driving forces to form the ground state phases implies that these metastable ion-inserted polymorphs can be trapped kinetically over a wider range of conditions than our Pourbaix analysis would suggest. Therefore, to gain a comprehensive understanding of MnO x electrode behavior, possible reaction pathways involving metastable phases and the associated driving forces to form more stable phases must be evaluated. The energies tabulated in Table enable this type of analysis for many different ions in the Mn oxide system. In general, the small energy differences between many different Mn oxide phases suggests that electrochemical reaction pathways in this class of materials depends heavily on kinetic barriers. Estimation of these barriers with, for example, MD simulations or with the recently developed surface-stochastic walking method, will be critical to gain a complete understanding of device performance and degradation modes in A x MnO 2 electrodes. |
6334a9fbea6a22e834017b28 | 23 | Interestingly, the three thermodynamically stable compounds in Figure all contain singly valent cations. However, Figure indicated that multivalent ion insertion could be favorable relative to proton redox in certain polymorphs given a low H + concentration (e.g. Zn 2+ in λ-MnO 2 , Al 3+ in 𝛿-MnO 2 ). The Pourbaix diagrams in Figure reveal precipitated (hydr)oxides to be a more stable configuration for multivalent cations in water at high pH. Phases like ZnO, Ca(OH) 2 and Mg(OH) 2 are thermodynamically preferred relative to their corresponding ion-inserted manganese oxides. As a result, developing a multivalent ion shuttle battery that can access the full 2 electron capacity of Mn oxide may require significant engineering of the electrolyte to limit formation of these precipitated phases. |
6334a9fbea6a22e834017b28 | 24 | Outside of oxidative voltages at high pH, the thermodynamic behavior of Mn oxides is independent of the solution conditions. At low pH, solid phases dissolve into Mn 2+ at all but the highest applied voltages, whereas at high pH, solids are stable. Energy storage devices could therefore be designed with dissolution-precipitation mechanisms in acidic conditions, whereas at higher pH, intercalation or conversion mechanisms would dominate. Dissolution-precipitation benefits from high reversibility and long cycle-life whereas intercalation/conversion can achieve higher energy densities given the limited solubility of Mn salts. Stabilizing specific crystal structures at low pH would require suppression of Mn dissolution, for example, by pre-saturating electrolytes with Mn 2+ . The oxidative stability of MnO 2 increases at lower pH, suggesting that MnO 2 is a more stable oxygen evolution catalyst in acid than in base. At high pH, MnO 2 is predicted to dissolve to the highly oxidized MnO 4 -and MnO 4 2-ions at typical oxygen evolution overpotentials. However, because of Mn 2+ dissolution in acidic electrolytes, stable oxygen reduction likely requires more basic conditions. |
6334a9fbea6a22e834017b28 | 25 | While several highly metastable ion-inserted solids exist between Mn 3+ and Mn 4+ , metastability is less pronounced at lower oxidation states. Among Mn 2+ solids, layered 𝛿-Mn(OH) 2 is the most stable, while the next lowest energy phase, λ-Mn(OH) 2 , is close to 100 meV less stable than 𝛿-Mn(OH) 2 and the β and 𝛼 hydroxides are less stable by more than 150 meV. This larger difference in stability compared to the Mn 3+ and Mn 4+ solids rationalizes why layered phases often form at reducing voltages regardless of the polymorph present initially. Lastly, we note that debate exists over whether bixbyite 𝛼-Mn 2 O 3 is thermodynamically stable with respect to MnOOH. 𝛼-Mn 2 O 3 appears in the original Pourbaix diagram, but no MnOOH phases were considered. Recent DFT calculations have predicted MnOOH to be more stable, which agrees qualitatively with experimental observations that 𝛼-Mn 2 O 3 formation requires elevated temperatures and that MnOOH forms when 𝛼-Mn 2 O 3 and Mn 3 O 4 are aged in aerated water. However, solution-phase potential-pH measurements of MnOOH phases have reported free energies slightly less stable than that of 𝛼-Mn 2 O 3 . To the best of our knowledge, no hightemperature heat capacity measurements of the Mn oxyhydroxides have been reported to resolve this debate. Therefore, in our analysis, we use the experimental values that have been reported thus far, which our calculations reproduce (Table ). |
6334a9fbea6a22e834017b28 | 26 | Solvation effects. Ions in aqueous electrolytes are surrounded by large solvation shells of hydrogen-bonded water molecules. Because of their small radii and high polarity, water molecules that directly coordinate solvated ions can be tightly bound, particularly to smaller ions with higher charges. This strong coordination makes it possible for ions to insert into MnO 2 crystal structures with partial solvation shells. However, of the 6 most common MnO 2 polymorphs, only the 𝛼 and 𝛿 phases possess channels that are large enough to accommodate water molecules. These polymorphs are typically synthesized with water molecules inside their channels, but whether water intercalates dynamically during redox reactions remains a subject of debate. To investigate the effects that water can have on Mn oxide electrochemistry, we calculate the voltages and corresponding capacities of ions inserting into MnO 2 with partial solvation shells (Fig. ). In all cases, solvation shells were initialized with the oxygens of the water molecule residing approximately 2 Å away from the inserted ion and oriented according to the results of previous neutron diffraction experiments. Like in Fig. , higher voltages correspond to more favorable insertion events. |
6334a9fbea6a22e834017b28 | 27 | In the 𝛼 polymorph, solvation shells have a slight stabilizing effect for smaller ions that preferentially occupy the 8h sites when water is not present. This can be attributed to two effects: a more stable coordination environment formed in part by the oxygen atoms of water molecules, and hydrogen bonding between the water molecule and the MnO 2 framework oxygen anions. The insertion of water molecules into the [2×2] tunnel creates new quasi-tetrahedral sites that small ions like Li + , Mg 2+ , Zn 2+ and Al 3+ occupy instead of the 8h sites (Fig. on the left). Because small, highly charged atoms also tend to have strong solvation energies, 98 it's highly probable that these ions will insert into the [2×2] tunnels with water molecules, particularly at low capacities. Na + , K + , and Ca 2+ , however, which are too large for the newly formed tetrahedral sites, don't exhibit a preference for inserting with water. In fact, Na + and K + exhibit a large destabilization with coinserted water because the extra water molecules displace these ions from the center of the 2a and 2b sites respectively. |
6334a9fbea6a22e834017b28 | 28 | For all ions, insertion with partial solvation shells decreases the maximum ion capacity that can be achieved (Fig. ). Water molecules either occupy or crowd out crystallographic sites that could otherwise host an inserted cation. If water molecules cannot easily exit the tunnels to make way for new ions, then the maximum achievable energy density will be limited. Therefore, polymorphs with smaller tunnels that can't accommodate water or with layers that can swell to host more inserted ions may yield better performance. |
6334a9fbea6a22e834017b28 | 29 | In 𝛿-MnO 2 , the stacked layers of Mn octahedra expand and slide upon insertion of water molecules. This rearrangement, along with the presence of extra oxygen atoms in the interlayer gap, disrupts the regular bonding between inserted ion and Mn oxide framework. Therefore, partial solvation shells have a slight destabilizing effect, with ion insertion voltages being slightly lower with water than without. This finding suggests that co-insertion with water in the 𝛿 polymorph is primarily a kinetic effect, governed by high barriers to ion desolvation rather than by strong interactions with the Mn oxide structure. Previous experiments have also shown a relationship between temperature-driven water loss in 𝛿-MnO 2 and the hydration energies of inserted ions, demonstrating that interlayer water is tightly-bound. Large ions like Na + , K + , and Ca 2+ reside in the same plane as the co-inserted water molecules whereas smaller ions sit in between the water layer and the layer of Mn octahedra. In both cases, 𝛿-MnO 2 undergoes a c-axis expansion from ~4 Å to ~7 Å upon ion insertion, consistent with experimental data on the birnessite family of structures. While it is well known that highly charged ions like Zn 2+ and Mg 2+ can insert with multiple water layers, leading to c-axis expansions as large as 10 Å, the energetic effect of water co-insertion should be well-captured by calculations containing a single water layer. The ability of these layers to expand is what gives 𝛿-MnO 2 a higher maximum ion capacity than the 𝛼 polymorph. |
6334a9fbea6a22e834017b28 | 30 | In this work, we have investigated the electrochemical ion insertion redox behavior of A x MnO 2 compounds (A = H + , Li + , Na + , K + , Mg 2+ , Ca 2+ , Zn 2+ & Al 3+ ) in aqueous and non-aqueous environments using ab-initio density functional theory. We show that Hubbard-U corrections are necessary for computing accurate redox energetics, and that these corrections are more important than choice of exchange-correlation functional (PBE vs. SCAN). Using PBE+U, we find that nonprotonic cation insertion into MnO 2 strongly depends on how effectively ions are coordinated by oxygen atoms inside the tunnel and layered structures. A competing driving force to form O-H bonds exists whenever H + is present and is usually favored over non-protonic cation insertion. This finding explains experimental reports of H + and H + /Zn 2+ co-insertion mechanisms in aqueous Znion batteries. 𝛼-K 0.125 MnO 2 , 𝛼-Na 0.125 MnO 2 , and λ-Li 0.5 MnO 2 are the only phases with inserted nonprotonic cations that are thermodynamically stable within water's voltage stability window, while other aqueous ion insertion in MnO 2 relies on metastability. We identify Al 3+ insertion into the 𝛿, R, and λ polymorphs as promising candidates to achieve the full 2 electron redox of MnO 2 , however, multivalent ion battery electrolytes that can prevent the precipitation of insoluble (hydr)oxides and facilitate ion desolvation will likely have to be developed. We also show that coinsertion with water is favored for small ions in the 𝛼 polymorph due to confinement effects, while solvation energies and kinetic effects dictate water insertion in 𝛿-MnO 2 . In both cases, capacity decreases due to water molecules occupying sites that could otherwise host an inserted cation. The calculations performed in this work rationalize experimental observations of MnO x electrochemical behavior and serve as a guide for the rational design of Mn oxide ion-insertion devices. |
60c74a7c9abda2bbb1f8ce65 | 0 | The development of smart and controlled nanocarriers for targeted and triggered drug delivery provides therapeutic opportunities for the treatment of major diseases such as infections, inflammation, and cancer. Synthetic amphiphilic copolymers, able to undergo self-assembly into nanostructures, are ideal carriers for drug delivery due to the flexibility and diversity of polymer chemistry. Various carrier morphologies can be engineered with tuneable stabilities and surface chemistries, which include facile functionalisation with targeting ligands, in addition to triggerable drug release via stimuli-responsive copolymers. Typical morphologies of polymer-based structures used for drug delivery include micelles (spherical and worm-like), vesicles (polymersomes), capsules, polyion complex (PIC) micelles and vesicles, nanoparticles, and nanogels. The key advantages offered by nanogels are their ease of preparation that yields uniform and tuneable sizes, relatively high encapsulation efficiency, stability in the presence of serum, and the possibility of simple introduction of stimuli responsiveness. To yield high level of control of drug loading and triggered release, temperature and/or enzyme-responsive motifs can be incorporated in polymer-based drug delivery. With respect to temperature-triggered systems, poly(N-isopropylacrylamide) (PNIPAM) remains one of the most studied polymers incorporated in the design of temperature-responsive materials for biomedical applications due to its fast temperature-induced phase transition at a lower critical solution temperature (LCST) of ~32 °C. Copolymers of PNIPAM with a temperatureindependent hydrophilic block allow rapid assembly of nanostructures upon heating either after or during the polymerisation because of the thermally induced collapse of the PNIPAM block at temperatures above the LCST. However, PNIPAM-based nanoassemblies are limited to transition around 32 °C; the LCST cannot be changed easily. Further, temperature is not an ideal trigger for cargo release given our fixed body temperature. Hence, modulation of the self-assembly temperatures (LCST) using a range of different copolymers, combined with a suitable stimuli-responsive cross-linking strategy, represents an interesting alternative platform that could be employed both above and below the LCST. |
60c74a7c9abda2bbb1f8ce65 | 1 | Our strategy uses temperature changes as a simple and fast organic solvent-free nanogel formation method to encapsulate biomolecules. Poly(N-cyclopropylacrylamide) (PNCPAM), closely related to PNIPAM, consists of pendant cyclopropylamine groups instead of the isopropylamine groups in PNIPAM and possesses a LCST of ~49 °C. We propose herein that the copolymerisation of NCPAM and NIPAM to form block copolymers together with a hydrophilic block -consisting of a short block with functional reactive handles and polyethylene glycol (PEG) -provides an efficient method to precisely control the LCST, and hence the self-assembly temperature, from 32 up to 49 °C. Fine-tuning of the transition temperature using a library of copolymers, with precise control of the LCST around body temperature, enables application-dependent selection of a certain copolymer from the library to create customised vehicles for loading and release of sensitive therapeutic protein cargoes. |
60c74a7c9abda2bbb1f8ce65 | 2 | A cross-linking strategy was further included to (i) stabilise the nanocarrier, (ii) avoiding premature release of cargo, and (iii) incorporating stimuli-responsiveness. This additional cross-linking strategy serves as an additional trigger besides temperature for cargo release, through degradation of the responsive cross-links. These are important considerations when taking into account the immensely complex environment faced by nanocarriers upon in vivo administration. We employed a copper-catalysed azide-alkyne cycloaddition (CuAAC) crosslinking strategy, which remains one of the most popular "click" reactions for cross-linking and functionalisation in the biomaterial and biomedical fields due to high efficiency, specificity, and simplicity, to incorporate the necessary functional handles. Our covalent cross-linking strategy provided the fundamental advantage of keeping our nanogels intact above and below the LCST, which in turn allows fast and simple loading of cargo and subsequent purification. |
60c74a7c9abda2bbb1f8ce65 | 3 | With respect to cross-linker backbone chemistry, peptides are particularly attractive due to their simple solid phase synthesis, incorporation of non-canonical functional groups for the introduction of "clickable" handles, and the possibility of using enzyme-specific cleavable peptide sequences. Due to their intrinsic involvement in biological and metabolic processes, enzymes have emerged as suitable triggers for drug delivery. Different classes of enzymes (e.g. proteases, phospholipases, oxidoreductases) have been leveraged for the purpose of triggered drug release from nanocarriers. A specific class of proteolytic enzymes, matrix metalloproteinases (MMPs), are associated with several disease conditions, including infections, inflammation, cardiovascular diseases and cancer. Overexpression of MMPs at disease sites can be exploited to release drugs at the target site by selective rupture of engineered MMP-responsive delivery platforms. Herein, we successfully synthesised libraries of PNIPAM-, PNCPAM-and P(NIPAM-co-NCPAM)-based triblock copolymers with tuneable compositions via reversible additionfragmentation chain-transfer (RAFT) polymerisation. The structure of the final block copolymers was carefully designed to include these temperature-responsive blocks, a functional block for nanogel cross-linking and a hydrophilic outer block. The thermoresponsive block is hydrophilic below its LCST and turns hydrophobic above the LCST, thus switching the block copolymer from hydrophilic to amphiphilic, which promotes the formation of self-assembled nanostructures. The choice over amphiphilicity or hydrophilicity of degradation products at body temperature, through selecting a copolymer with a transition temperature below or above 37 °C, offers means of potentially modulating elimination pathways upon in vivo administration in the future. To increase nanoparticle stability and incorporate responsiveness to a model MMP enzyme (MMP-7), azido-bi-functionalised peptides were synthesised and used to cross-link the nanoparticles via the alkyne groups present in the functional middle block using CuAAC. |
60c74a7c9abda2bbb1f8ce65 | 4 | A considerable challenge when deciphering the exact morphology of copolymer assemblies is the structural ambiguity that is provided by certain routine characterisation techniques such as conventional dynamic light scattering (DLS) and negative stain transmission electron microscopy (TEM). However, accurate information on morphology is of great importance since it directly affects choice of suitable disease targets, drug cargo, delivery route, stability and release mechanism. Thus, the use of additional complementary characterisation techniques is needed to characterise the morphology of soft self-assembled nanostructures. |
60c74a7c9abda2bbb1f8ce65 | 5 | TEM at cryogenic temperatures (cryo-TEM) and small-angle neutron scattering (SANS) are two complementary methods to study these types of nanostructures in their native hydrated state. A third technique, fluorescence correlation spectroscopy (FCS) is a diffusion-based single-molecule analysis technique. FCS allows accurate characterisation of nanoparticle assembly/disassembly -including in the presence of enzymatic triggerssurface functionalisation, cargo loading, and release. Our thoroughly characterised nanogel system is shown to serve as a versatile protein delivery platform with a simple and gentle temperature induced formation and loading protocol, and triggered release through stimuli-responsiveness provided by enzymatically cleavable cross-links. This modular platform can be adjusted for disease-specific applications by choosing the appropriate copolymer, peptide cross-linker sequence, and therapeutic protein cargo. |
60c74a7c9abda2bbb1f8ce65 | 6 | We designed triblock copolymers to confer the final self-assembled nanogels with favourable properties for enzyme-triggered drug delivery applications (Scheme 1). Conceptually, an inert hydrophilic block (Scheme 1, dark blue), which we expected to form the corona of the assembled nanogels, was connected to a thermo-sensitive block (Scheme 1, light blue/green) which would drive self-assembly at temperatures above its LCST. We chose to use PEG as the hydrophilic block and either PNIPAM, PNCPAM or copolymers of the two P(NIPAM-co-NCPAM) as the thermo-responsive block. Between these two blocks we incorporated a functional block (Scheme 1, red) composed of a mixture of methacrylic acid (MAA) and trimethylsilylpropargylmethacrylate (TMSPMA) to provide a latent handle for cross-linking with azide-bi-functionalised peptide after self-assembly and desilylation. We used RAFT polymerisation for the synthesis of these polymers, due to the ease of preparing well-defined multi-block copolymers and PNIPAM via this route. The detailed synthetic procedure is shown in Scheme 2. |
60c74a7c9abda2bbb1f8ce65 | 7 | The first step consisted of the synthesis of the RAFT agent 1, following a procedure previously reported in the literature, which was then conjugated to a PEG block (113 units) via esterification to form a macroRAFT agent. Chain extension of this macroRAFT agent with a random copolymer of MAA and TMSPMA enabled the introduction of carboxylic groups and terminal alkyne functionalities (in their protected form) as functional handles. Good control over the polymerisation of PEG-b-P(MAA-co-TMSPMA) 3 was evidenced by low dispersity Đ = 1.07, and a clean shift in the gel permeation chromatography traces (GPC, Figure ). The same polymer (with 113 PEG units, 5 MAA units and 19 TMSPMA units) was used throughout the remainder of the study and chain extended to produce three different libraries of triblock copolymers: (i) a library of PNIPAM-based block copolymers (5a1-5a5) with increasing number of NIPAM units; (ii) a library of PNCPAM-based block copolymers (5b1-5b5) with increasing number of NCPAM units; (iii) a library of P(NIPAM-co-NCPAM)-based block copolymers (5c1-5c3) with different NIPAM/NCPAM (NI/NC) ratios. We hypothesised that the incorporation of cyclopropyl moieties (NCPAM) would lead to an increase in the LCST providing a Tcp (cloud point temperature as measured by DLS) for nanogel self-assembly/disassembly at temperatures just above body temperature. Successful synthesis of the three block copolymer libraries was confirmed by 1 H NMR and GPC yielding degree of polymerisation (DP), number average molecular weight (Mn), and dispersity (Đ) (Table ). Full composition and characterisation details of selected 5a4/5c1-3 block copolymers are reported in Table , whilst data for the remaining block copolymers can be found in Table of the supporting information (SI). For the P(NIPAM-co-NCPAM)-based copolymers (5c1-3), full chain extension was observed for all the copolymers with a minimal amount of macro-RAFT agent left behind and minimal shouldering in the GPC traces (Figure ). Due to the structural similarity and similar degree of reactivity between NIPAM and NCPAM, the formation of a random block stemming from their simultaneous polymerisation rather than two distinct blocks was assumed. In general, the dispersity was Đ < 1.5. The length of the thermo-responsive block was successfully modulated in the separate NIPAM and NCPAM libraries (Table ), whilst the copolymers shown in Table were of similar length, as shown by NMR and GPC. The evolution of Mn and Đ versus DP is shown in Figure . We also determined the block copolymer cloud point temperatures Tcp via temperature sweeps and monitoring by dynamic light scattering (DLS). |
60c74a7c9abda2bbb1f8ce65 | 8 | Temperature-responsive assembly/disassembly of nanostructures around body temperature is a valuable means of controlling encapsulation/release of therapeutic cargo. Modulation of the transition temperature allows for application-specific optimisation of cargo loading and delivery, which are critical considerations in the drug delivery field. The synthesised P(NIPAMco-NCPAM)-based triblock copolymers were designed to exhibit increasing transition temperatures. DLS was used to confirm successful modulation of the self-assembly temperature for the P(NIPAM-co-NCPAM)-based triblock copolymers containing increasing amounts of the cyclopropyl-based repeating unit (NCAPM from 0 to 60%) (5a4/5c1-3) (Figure ). By dissolving these copolymers in PBS at 25 °C and increasing the temperature to 60 °C (above the LCST), a clear increase in the turbidity of the solution could be observed by eye and by light scattering (derived count rate, kcps) recorded using DLS, which suggests a coilto-globule transition (Figure ). Since larger structures scatter more light than dissolved copolymers, the self-assembly of the copolymers could be followed by monitoring the scattering intensity. The inflection points of each curve represent the Tcp of the corresponding copolymers. DLS measurements at temperatures below/above the LCST indicate formation of stable and well-defined nanostructures when heated above the LCST (Figure ), whilst bigger structures were formed with increasing amounts of NCPAM (mean hydrodynamic diameters between 74 and 295 nm). Due to the formation of large structures/aggregates that led to some precipitation for 5c3 when heated/cooled slowly (1 °C temperature ramp with 5 min equilibration time), the intensity curves are not smooth (Figure ) and no accurate Tcp can be given for this copolymer. However, it appears to be between 45 to 50 °C when estimating based on both curves (up/down, heating and cooling). The dependence of particle size on the heating rate during the self-assembly process was also confirmed for two copolymers that form smaller nanoparticles in general (5a4 and 5c1, Figure ). The final assembled particles tended to be larger when subjected to slower heating. The cyclopropyl-dependent size increase might be caused by the higher rigidity and lower rotational freedom of the cyclopropyl group when compared to the isopropyl group, which might affect the packing parameters to yield larger structures. The self-assembly temperature of the resulting triblock copolymers was successfully tuned from 33 °C, corresponding to the Tcp of the 5a4 copolymer with 0% NCPAM content, to a Tcp of about 47 °C for the 5c3 copolymer with 60% NCPAM content. It can be concluded that by increasing the number of cyclopropyl units in the copolymers, the self-assembly temperature increases proportionally. However, when the NCPAM content exceeds 50%, the stability of the resulting self-assembled structures is compromised due to the formation of large particles/aggregates, which can lead to sedimentation (Figure ). The self-assembly process of PNIPAM-and P(NIPAM-co-NCPAM)-based copolymers can be reversed by cooling the solution below the transition temperature (down to 25 °C) causing complete nanogel dissolution (Figure ). The globuleto-coil transition is slower than the coil-to-globule transition, which is indicated by the slight hysteretic behaviour. According to this initial DLS analysis, the selected block copolymers (5a4/5c1-3) revealed an LCST-type, reversible thermo-responsive behaviour that yielded nanogels at temperatures above their LCST. We also tested all the other copolymers from the separate PNIPAM (5a1 to 5a5) and PNCPAM (5b1 to 5b5) series for the formation of nanogels at temperatures above the LCST (Table ). All copolymers except 5b1 assembled into nanogels, however, no clear trends regarding polymer lengths and corresponding nanostructure sizes could be found. |
60c74a7c9abda2bbb1f8ce65 | 9 | Accurate shape, size and morphology characterisation of drug delivery systems is crucial for understanding their behaviour, especially for dynamic self-assembled systems. This will help to define their potential as drug delivery vehicles. As an example, morphology and shape can affect cargo loading and cellular uptake. Conventional DLS and TEM can give an indication of nanostructure size, polydispersity and shape, but complementary techniques are needed to decipher the internal morphology of self-assembled particles. SANS allows more detailed structural information to be acquired in an artefact-free manner to obtain bulk properties such as morphology and polydispersity. Cryo-TEM is a useful tool for imaging self-assembled nanostructures to obtain a snapshot in their native hydrated state. Often, the drying step in the sample preparation procedure needed for negative stain TEM introduces drying artefacts, which may prevent accurate morphology determination. A combination of SANS and cryo-TEM is often used when characterising block copolymer assemblies in solution, e.g. when studying vesicle shape transformations at the nanoscale. We used these two techniques to further analyse the self-assembly process by first studying the copolymers before ('coil' state) and after self-assembly ('globule' state) (Figure ). The data obtained for 5a4 suggests that this block copolymer has fully swollen chains in good solvent conditions at 25 °C (below the LCST) in agreement with literature on similar copolymers. A power law exponent of 1.96 ± 0.01 block copolymer 5c1 is explained by the presence of polymer chains in theta conditions with equally favoured polymer-polymer and polymer-solvent interactions. |
60c74a7c9abda2bbb1f8ce65 | 10 | When heated above the LCST (45 °C), the copolymers are expected to transition from hydrated and flexible coils to collapsed globules. SANS data for copolymers 5a4 and 5c1 above the LCST were successfully fitted with a Sphere model. This indicates formation of dense spherical polymer nanogels rather than vesicles (polymersomes), which confirms the conclusions from the cryo-TEM images (Figure ). The insertion of 20% cyclopropyl repeating units (5c1) did not affect the final morphology and the scattering data could again be fitted with the simple Sphere model. The best-fit diameters were 47.6 ± 1.0 nm (5a4) and 78.0 ± 0.5 nm (5c1), respectively. Data for size obtained by SANS, cryo-TEM and DLS are similar and the trend of forming larger structures with 5c1 compared to 5a4 was consistent across all the measurements. SANS typically shows the smallest diameter of all the three methods, because it does not account for the hydration shell in the measurement. |
60c74a7c9abda2bbb1f8ce65 | 11 | None of our synthesised block copolymers self-assembled into vesicular morphologies, which would have been observed as membranous structures in the cryo-TEM micrographs and distinct SANS curves that could be fitted with a Hollow sphere model as shown previously. This is in disagreement with previous studies on similar PEG-b-PAA-b-PNIPAM block copolymer assemblies, which highlights the challenge of using theoretical rules to predict the final structure and encourages the use of various complementary techniques to precisely evaluate morphology. A combination of techniques such as DLS, cryo-TEM and scattering techniques like SANS, small-angle X-ray scattering (SAXS) and combining dynamic and static light scattering (DLS/SLS), is necessary to obtain a clear picture of the size and morphology of self-assembled nanostructures, which allows identification of the type of nanostructure formed. |
60c74a7c9abda2bbb1f8ce65 | 12 | Several cross-linking strategies have been adopted in the drug delivery field to increase the stability of self-assembled nanostructures and introducing stimuli-responsive properties to the delivery system. The dynamic nature of self-assembled nanostructures can compromise their stability in biological environments due to the presence of other entities (e.g. proteins) and dilution upon injection in the body. This can lead to premature payload leakage. Stabilisation of amphiphilic assemblies by covalent bonds is a versatile method to address the above-mentioned challenges, whilst stimuli-responsive motifs can be incorporated simultaneously. A stimuli-responsive cross-linker domain that is cleaved in the presence of the desired stimuli allows triggered-structure disassembly and controlled payload delivery. Herein, we leveraged natural enzymes as triggers for cargo release due to their inherent involvement in disease progression. The functional alkyne handles, introduced into the copolymers as described above, were used as cross-linking points using azido-bifunctionalised peptides bearing a protease sensitive sequence. A model matrix metalloproteinase was selected as our initial target due to their involvement in many diseases. Matrix metalloproteinase 7 (MMP-7) was chosen due to its small size with a Mw of only 19 kDa in its active form, which allows sufficient penetration into the polymer network of the nanogels to cleave the cross-links. |
60c74a7c9abda2bbb1f8ce65 | 13 | Three of the block copolymers (5a4/5c1/5c2) were used for the development of peptide crosslinked nanogels. The peptide cross-linker was prepared by solid phase peptide synthesis inserting an azido-lysine residue at both the C-and N-termini according to literature procedures. An MMP-7-cleavable sequence (Figure ) was installed into the peptide backbone enabling enzyme-triggered nanogel degradation. LC-MS spectra of the functionalised MMP-7-cleavable peptide after purification is shown in Figure . In order to prepare the cross-linked nanogels, we developed a two-step protocol. First, the formation of self-assembled nanogels was induced by heating the block copolymer solution to 45 °C (above the LCST). A solution of the cross-linker was then added, together with small amounts of CuSO4•5H2O and (+)-sodium-L-ascorbate, leading to covalent crosslinking; the mixture was subsequently purified by size exclusion chromatography. Upon cooling below the LCST, our nanogels therefore retained their structure, despite the solubilisation of the polymer chains. |
60c74a7c9abda2bbb1f8ce65 | 14 | Details of the cross-linking strategy are schematically shown in Figure . After the cross-linking reaction, the resulting polymer nanogels were analysed by DLS, both above and below the LCST, which is shown in Figure . A small difference in the hydrodynamic diameter was observed between the cross-linked (93 ± 2 nm) and selfassembled 5a4 nanogels (74 7 nm, before cross-linking) above the LCST. In contrast, DLS measurements at 25 °C (below the LCST) revealed successful cross-linking had taken place. |
60c74a7c9abda2bbb1f8ce65 | 15 | The non-cross-linked polymer nanogels were disassembled when the temperature was lowered below the LCST, as described above (Figure ), but nanogels remained stable when cross-linked. When comparing the DLS traces of the cross-linked 5a4 nanogels at 37 °C and 25 °C, a size increase can be observed (93 ± 2 nm compared to 133 ± 9 nm). The slight size increase with decreasing temperature can be attributed to the thermo-responsive phase transition of the PNIPAM block, which switches from hydrophobic nature above the LCST to hydrophilic below the LCST. This could cause relaxation and water-induced swelling of the polymer chains, as previously observed in the literature. These networks are expected to swell and relax up to a limit determined by constraints imposed by the crosslinking, which will be dependent on the length and chemical structure of the peptide, and the cross-linking density. |
60c74a7c9abda2bbb1f8ce65 | 16 | Figure shows DLS characterisation of 5c1 (NI0.8NC0.2)-based nanogels after cross-linking. 5c1based nanogels could be cross-linked successfully, yielding larger sizes compared to 5a4-based nanogels when measured below the LCST (181 ± 4 nm vs 133 ± 9 nm). Cross-linking of 5c2based nanogels could not be achieved due to severe aggregation and precipitation. Hence, detailed morphological analysis was restricted to the two copolymers that could also be crosslinked successfully (5a4 and 5c1). Long-term stability of cross-linked nanogels made from these two block copolymers when incubated in PBS at 37 °C was demonstrated by DLS (Figure ). |
60c74a7c9abda2bbb1f8ce65 | 17 | During SANS measurements, 5c1 block copolymers and the corresponding cross-linked nanogels were subjected to a temperature ramp from 25 to 45 °C and from 32 to 42 °C, respectively (Figure , Table and Table ). Gaussian chains to a more entangled polymer network (Rg of 52.0 ± 0.6 nm) indicating the start of the coil-to-globule transition. At higher temperatures (36 and 45 °C) data were fitted using a triaxial ellipsoid model, which suggests the formation of non-perfect spherical nanogels, with an observable trend of ellipse transitioning to more sphere shape with increasing temperature (Table ). From these experiments a Tcp of 35-36 °C was obtained, which was similar to the value found by DLS (37 °C). Multiplying the scattering intensity by Q 2 allowed us to obtain the Kratky plot with linear x-and y-axes (Figure ). This plot is useful for studying nanogel degree of globularity, flexibility and compactness. The peak appearing in the Kratky plot at higher temperatures suggests the formation of globular particles above 35 °C. As the temperature further increases, the peak becomes more defined suggesting an increase in globularity. |
60c74a7c9abda2bbb1f8ce65 | 18 | The effect of cross-linking can be seen in the temperature-dependent SANS data of peptide cross-linked 5c1 nanogels (Figure ,d, Table ) when compared to the purely self-assembled nanogels (Figure2f, Figure ,b, Table ). The power law exponent value below 2 for the non-cross-linked copolymer compared to a value of ~2.4 for the cross-linked nanogels, reveals the difference of the two populations at lower temperatures. As expected, by cross-linking the polymer chains a stable polymer network was formed, and the chains remained interconnected between each other also at temperatures below the LCST. When comparing the two Kratky plots (Figure ) it can be concluded that the self-assembled and cross-linked nanogels both transition from coil to globule at 36 °C, whilst the cross-linked particles were slightly smaller after the transition as seen by the shift of the peak to the higher Q region. This trend is also confirmed by cryo-TEM (Figure , Figure ). An explanation for this smaller size could be that the peptide cross-linkers restrict chain flexibility, keeping the polymer chains tightly connected. 51 |
60c74a7c9abda2bbb1f8ce65 | 19 | MMP-triggered nanogel disassembly was analysed using MMP-7 as a model MMP. We first confirmed that the free peptide in solution can be cleaved by MMP-7 as shown using LC-MS (Figure ). We next established the necessary concentrations of MMP-7 to successfully degrade the cross-links of the nanogels, which revealed that physiologically relevant nanomolar concentrations are sufficient for degradation (Figures S4). The kinetics of MMP-7responsive nanogel disassembly at a fixed enzyme concentration (90 nM) were subsequently studied by DLS (normalised derived count rate (in kcps) vs incubation time). The kinetic study (Figure ) reveals a drop in the scattering intensity over time only in the presence of the enzyme, indicating a decrease in nanogel concentration and/or size. The kcps value remained unvaried in the control nanogel sample without enzyme (red symbols). A further control, incubating the cross-linked nanogels at 37 °C in 5 % (v/v) fetal bovine serum (FBS) revealed very high stability of the nanogels in the presence of proteins over at least two days (Figure ). This confirms the ability of the nanogels to undergo enzyme-triggered degradation. From the evolution of scattering intensity in Figure it is apparent that degradation has already occurred to some extent after 1 h incubation with MMP-7. The number mean size of the nanogels has already decreased (measured below the LCST), which suggests the presence of differently sized fragments and partial degradation. Within the next few hours of incubation with enzyme, a mixture of differently sized particles/fragments/polymers (large errors) was detected. An equilibrium is reached after 24 h incubation, when a clear size difference between the control and enzyme incubated sample became apparent (Figure ). |
60c74a7c9abda2bbb1f8ce65 | 20 | Following DLS analysis, we also monitored nanogel degradation by a single-molecule detection method called fluorescence correlation spectroscopy (FCS). We previously used FCS for a detailed study of MMP-responsive degradation using another peptide-based nanostructure for diagnostic applications. Herein, FCS was employed to first observe nanogel degradation and subsequently to quantitatively analyse model biomolecule loading and release. In order to study MMP-7 triggered degradation of nanogels by FCS, cross-linked nanogels were labelled with Cy5-azide and incubated with excess MMP-7 (9 µM). Successful labelling of nanogels with Cy5-azide can be seen when calculating the hydrodynamic diameters from the FCS autocorrelation analysis (Figure ) that yielded nanogel size rather than free dye (Figure ). Further, non-crosslinked Cy5-azide labelled copolymers in nonassembled form revealed an intermediate diffusion time between free dye and nanogels (Figure ). FCS curves of control experiments to check for non-specific interactions of dyes with nanogels/copolymers yielded diffusion times similar to free dye (Figure ), which confirms successful specific labelling in the case of Cy5-azide (Figure ). |
60c74a7c9abda2bbb1f8ce65 | 21 | After MMP-7 incubation, the corresponding autocorrelation curves shifted to faster diffusion times indicating cleavage of the peptide bonds (Figure ). In the presence of MMP-7, a twocomponent fit of the autocorrelation curves revealed complete disappearance of any structures with original nanogel sizes. We only detected the presence of free Cy5-polymer chains (54 % of total diffusing species) and some bigger, possibly entangled degradation products (DH value of 35 ± 6 nm, 46%, shown in Figure ), demonstrating MMP-7-dependent peptide cleavage and complete nanogel disassembly below the LCST. By comparing the nanogel signal intensity (counts per particle, CPP in kHz) to the dye intensity, the number of dye molecules per single nanogel could be determined. Upon enzyme-mediated nanogel degradation, the calculated number of dyes per particle changed from about 54 dyes per particle before degradation, to about 2 dyes per particle for the degradation products. This is a further indication of nanogel disassembly. Together with the DLS analysis (Figure ,b), enzyme-responsiveness of our nanogel system was successfully demonstrated. |
60c74a7c9abda2bbb1f8ce65 | 22 | After confirming the possibility of enzyme-triggered degradation of the herein developed peptide-cross-linked nanogels using a disease-relevant enzyme of the MMP class, loading and release properties of a model protein cargo (fluorescent Oregon Green-labelled bovine serum albumin, OG-BSA) was studied using one selected copolymer (5a4). UV-Vis spectroscopy revealed a loading efficiency of 18 ± 8 % (average ± SD of three independent batches) after purification by size exclusion chromatography. This represents better loading efficiencies for nanogels compared to polymersomes made via standard film rehydration, that often have low loading efficiencies of few percent for hydrophilic cargo such as proteins. Our rapid procedure to yield protein-loaded, well-defined nanogels at high polymer concentrations without the use of any organic solvent represents a suitable strategy to load sensitive therapeutic proteins for biomedical applications in the future. Loading and release of our model protein OG-BSA was further studied in detail by FCS (Figure ). Successful loading of OG-BSA into peptide cross-linked 5a4 nanogels was observed by FCS. |
60c74a7c9abda2bbb1f8ce65 | 23 | Free dye, OG-BSA and OG-BSA incorporated in nanogels can already be distinguished clearly in the fitted autocorrelation curves (Figure ). When calculating the corresponding diffusion times (τD) and hydrodynamic diameters (DH) the size difference between the diffusing species was obtained (Figure ). In the case of the nanogel measurements, the diffusing species correspond to the nanogel hydrodynamic diameter obtained in other measurements (Figure and Figure ), which indicates successful incorporation of OG-BSA. Comparing the signal intensity per nanogel (counts per particle, CPP in kHz) with that of free OG-BSA, the number of loaded OG-BSA per single nanogel was determined and shows that multiple protein cargos were loaded per nanogel (Figure ). Release studies were performed in triplicate by incubating OG-BSA loaded nanogels (batches A-C) with various MMP-7 concentrations in a physiologically relevant nanomolar range. FCS curves recorded at different time points in the presence or absence of MMP-7 revealed the release kinetics (Figure , Figure ). This allowed us to compare diffusion-driven release (non-specific release from the nanogels over time) to MMP-7 triggered release. |
60c74a7c9abda2bbb1f8ce65 | 24 | FCS curves at the different time points were fitted using a two-component fit to account for both diffusing species (loaded nanogels and released OG-BSA). Further, a correction was included to account for the disproportional effect of bright slow diffusing species (loaded nanogels) compared to the less bright free OG-BSA. We propose herein (see Materials and Methods in SI) to simply combine the fraction of free OG-BSA (directly from the twocomponent fit) with the decrease in the number of cargo for the remaining nanogels (via CPP), which is a second indicator for OG-BSA release (number of proteins per nanogel is decreasing). This correction is necessary because nanogels with several loaded proteins overlap with the free protein signal, as demonstrated in the literature for mixtures of differently sized nanoparticles of varying brightness. This finally yielded the percentage of free OG-BSA and allowed us to further calculate the cumulative release profile over time (Figure ). A clear difference between 90 nM MMP-7-triggered release and diffusion-based profiles was detected, whilst the release profile in the presence of 9 and 0.9 nM MMP-7 gave intermediate results as expected for an enzyme-triggered release. This successful enzymetriggered release profile justifies the further study of these copolymer libraries using therapeutic cargo and studying release behaviour at various temperatures, accompanied by corresponding delivery data at the cell and tissue level but this is beyond the scope of this paper. It is also clear that MMP-7 triggered release profiles will vary significantly in vivo compared to the in vitro data, due to a much higher level of environmental complexity that will cause differences in cross-linker degradation and hence release. Reasons for these differences include the presence of other substrates and specific cofactors, as well as continuous turn-over of enzymes. Release studies in more complex environments, including in the presence of cells and tissue, are needed in the future to choose an optimal copolymer system for the delivery of a specific drug for a chosen disease that involves upregulation of MMP enzymes at the disease site. Simply changing the peptide cross-linker sequence will further allow application of our nanogel platform using other enzymes as triggers, as relevant in specific diseases. |
60c74a7c9abda2bbb1f8ce65 | 25 | In this work, we took advantage of the versatility of polymer chemistry for developing a tuneable, temperature-responsive, peptide-crosslinked polymeric nanogel platform for enzyme-triggered biomolecule delivery. Via copolymerisation of NIPAM with NCPAM we could tune the thermal behaviour of the resulting copolymers resulting in a range of Tcp values around body temperature (from 33 to 44 °C). A peptide-based cross-linking approach was developed to covalently constrain the nanogel-forming copolymer chains to increase stability and equip the nanocarrier platform with enzyme-responsiveness. We obtained high-quality morphological information on the self-assembled structures, successfully studied dynamic behaviours upon temperature changes and cross-linking, as well as model protein loading and enzyme-triggered release by employing a combination of cryo-TEM, SANS and FCS. This detailed analysis yielded a morphology of filled spheres (imaged with cryo-TEM) that relaxed to an interconnected polymer network upon cross-linking and lowering the temperature below the LCST as found by SANS measurements. Even further, the coil-to-globule transition of one selected copolymer was successfully monitored in real-time by SANS. Model enzymetriggered (MMP-7) particle disassembly and accompanying increase in cargo release kinetics was determined by employing FCS. Our modular platform features the following key characteristics: reproducible, simple and fast temperature-induced nanogel assembly, modulation of the self-assembly temperature via copolymerisation of NIPAM and NCPAM in the thermo-sensitive block, containing two distinct handles (carboxylic groups and alkynes) |
62d8007e581987674d48b516 | 0 | Organic lighting-emitting diodes (OLEDs) have emerged as an exciting display technology that has steadily gained market share in a number of consumer electronic markets, from mobile phones and smart watches to televisions and monitors. The quality of an OLED depends on three key factors: its stability, its efficiency and its colour purity. All of these parameters are linked in part to the intrinsic properties of the emitter. Two strategies to achieve high colour purity in the device are to (1) employ a narrowband emissive material and/or (2) employ color filters to select out the desired emission. The use of color filters leads inevitably to light being lost by absorption, and a lower efficiency device, thus there is a strong desire to develop narrowband emitters. A high-efficiency device also requires the use of emitter materials that can efficiently convert both singlet and triplet excitons into light. Two classes of emitters are capable of realizing 100% internal quantum efficiency (IQE) in the device, these are phosphorescent compounds and compounds that emit via thermally activated delayed fluorescence (TADF). Organic TADF emitters harvest both singlet and triplet excitons via an endothermic reverse intersystem crossing (RISC) process that converts non-emissive triplet excitons into singlets prior to light generation. RISC is enabled when the energy gap between the lowest-lying singlet and triplet excited states (ΔEST) is sufficiently small. A molecular design that showcases a small exchange integral between molecular orbitals related to the excited state is responsible for the small ΔEST in the emitter. This is typically accomplished using a twisted structure containing donor (D) and acceptor (A) fragments. However, the charge-transfer (CT) excited state and the large reorganization energy in the excited state lead to broad emission, characterized by large full width half-maximum (FWHM) > 80 nm. The corresponding TADF OLEDs do not show the desired high color purity. |
62d8007e581987674d48b516 | 1 | A solution is to use multi-resonant TADF (MR-TADF) emitters. MR-TADF emitters are typically based on p-and n-doped nanographenes and so possess a rigid molecular structure. The ingenious molecular design places each of the frontier molecular orbital (FMOs) on adjacent atoms thus leading to short-range charge transfer (SRCT) between adjacent atoms, inducing the small ΔEST to turn on TADF. The SRCT excited state that results from the transition from the HOMO to the LUMO coupled with the small reorganization energy result in narrowband emitters. Since the first examples of this class of emitter, documented in 2016 by Hatakeyama and co-workers, more than 200 distinct examples have been reported, most of which show blue and green emission. We previously demonstrated that decorating a MR-TADF core with peripheral donor groups can result in emission tuning. However, when too strong donors are employed then the lowest energy excited states are no longer SRCT and the narrowband emission is lost. Rather, the emissive excited state becomes long-range CT (LRCT) that is typically observed in donor-acceptor TADF compounds. The aforementioned reports evidence of emission tuning from 462 to 608 nm using a peripheral decoration strategy on the MR-TADF core; nevertheless, examples of red-emitting MR-TADF emitters remain rare. |
62d8007e581987674d48b516 | 2 | Fluorobenzene, tert-butylcarbazole, tert-butylbenzene, and cyanobenzene are the most frequently used substituents to decorate MR-TADF emitters. These examples are either weak electron-withdrawing and/or electron-donating groups and so their incorporation does not disrupt the MR-TADF character. Diphenylamine (DPA) possesses an intermediate electrondonating ability between carbazole, which allows retention of MR-TADF when it is added to a core, and DMAC, which results in a D-A emitter when it is decorated about a MR-TADF core. |
62d8007e581987674d48b516 | 3 | The incorporation of peripheral DPA groups has been shown to red-shift the emission of MR-TADF compounds by Yasuda and co-workers. For instance, DACz-B (Figure ) shows bright yellow narrowband emission (photoluminescence emission peak wavelength, lPL = 576 nm; full width half-maximum (FWHM) = 44 nm; photoluminescence quantum efficiencies (FPL) of 87%), which is red-shifted compared to Cz-B (lPL= 484 nm; FWHM = 30 nm; FPL of 97%, Figure ) in 1 wt%-doped mCBP film. The OLED with DACz-B produces yellow electroluminescence (peak electroluminescence wavelength, lEL= 571 nm) with a maximum external EL quantum efficiency (EQEmax) of 19.6%. Using a different MR-TADF skeleton, TOAT, Adachi and co-workers introduced different numbers of DPA groups para to the central nitrogen atom of the TOAT core (3, 4 and 5, see Figure ). This led to the realization of red emission with a FWHM 45 nm in toluene for emitter 5 (Figure ). Using the same TOAT core, Zhang and co-workers reported DPA analogues mBDPA-TOAT and pBDPA-TOAT that contained tert-butyl groups on the periphery of the DPA groups, which showed orange emission of 571 (FWHM = 34 nm) and 563 nm (FWHM = 48 nm) in toluene, respectively. However, none of the OLEDs using a TOAT derivative shows an EQEmax>18%. Here we report two emitters containing either three diphenylamine (DPA) or triphenylamine (TPA) donor groups decorating a central MR-TADF core, DiKTa: 3DPA-DiKTa and 3TPA-DiKTa (Figure ). Both emitters behave as MR-TADF compounds, characterized by small DEST, narrow FWHM, and high FPL in 1,3-bis(N-carbazolyl)benzene (mCP) doped films. |
62d8007e581987674d48b516 | 4 | 3TPA-DiKTa shows an emission maximum at λPL = 561 nm (FWHM = 65 nm) with FPL = 93% while the emission of 3DPA-DiKTa is red-shifted (λPL = 613 nm, FWHM = 54 nm, FPL = 60%) in the 2 wt% mCP doped films. Benefitting from both the high ΦPL and the enhanced light- outcoupling associated with a preferential horizontal orientation of its transition dipole moment, the vacuum-deposited device with 3TPA-DiKTa exhibits an EQEmax = 30.8% at an electroluminescence maximum, lEL, of 553 nm (FWHM of 62 nm). The device with 3DPA-DiKTa shows an EQEmax of 16.7% with a lEL of 613 nm (FWHM of 60 nm). This work demonstrates how molecular engineering can lead to the most efficient ketone-containing MR-TADF emitters discovered so far. |
62d8007e581987674d48b516 | 5 | The synthesis routes to 3DPA-DiKTa and 3TPA-DiKTa are shown in Schemes S1-S2. The previously reported intermediate 3Br-DiKTa was elaborated with either diphenylamine or triphenylamine donors via a three-fold palladium-catalyzed Buchwald-Hartwig or Suzuki-Miyaura cross-coupling reaction, respectively. The identity and purity of the two emitters were determined using a combination of 1 H and 13 C NMR spectroscopy, high-resolution mass spectrometry, element analysis, high-pressure liquid chromatography (HPLC), single-crystal-XRD and melting point determination (Figures ). |
62d8007e581987674d48b516 | 6 | The DiKTa core in both emitters showed a low degree of planarity because of the presence of the puckered acridone moieties, fused with a common nitrogen. The degrees of puckering between the two emitters are somewhat similar, with angles between the common phenyl ring plane and the peripheral ring planes in the fused acridone of 13.96° and 17.09° in 3TPA-DiKTa and 15.85° and 27.80° in 3DPA-DiKTa. This results in an angle between the peripheral phenyl ring planes of the fused acridones of 24.95° for 3TPA-DiKTa and 41.49° for 3DPA-DiKTa. |
62d8007e581987674d48b516 | 7 | The dihedral angles between the mean-plane of the DiKTa core and the phenyl rings of diphenyl amine in 3DPA-DiKTa range between 10.66° and 42.24°. In 3TPA-DiKTa the bridging phenylene rings are close to co-planar with the DiKTa core, the torsion angles between the three bridging phenylene rings and the mean-plane of the DiKTa core being 7.86, 14.20 and 16.47°. |
62d8007e581987674d48b516 | 8 | The dihedral angles between the bridging phenylene rings and the peripheral phenyl rings of the triphenyl amine in 3TPA-DiKTa range between 11.35 and 32.00°. Due to the presence of the phenylene bridge in 3TPA-DiKTa, the distance between DiKTa nitrogen atom and diphenylamine nitrogen atom is 9.9 Å, larger than for 3DPA-DiKTa (5.5 Å). The degree of planarity of the geometries of both molecules was then calculated by PBE0/6-31G(d,p) based on the crystal structures. Overall, both molecules show planarity ratios of 67% for 3TPA-DiKTa 68% for 3DPA-DiKTa (Figure ), indicating that the compounds in the crystal structure geometry are reasonably planar. |
62d8007e581987674d48b516 | 9 | The nature of the lowest-energy excited states can play a significant role in the color purity of the emitter. In particular, the narrowband of MR-TADF emitters is conserved when the S1 state possesses SRCT, whereas the emission can broaden in some cases when the MR-TADF core is decorated with donor groups due to the stabilization of the LRCT state that dominates in donoracceptor systems. To gain insight into the relative energies of these states and to rationalize the optoelectronic properties of the emitters, the FMOs were first modelled at the optimized ground state geometry in the gas-phase using the density functional theory (DFT) at the PBE0/6-31G(d,p) level of theory. In contrast to the planarities calculated from the X-ray structures, those calculated from the DFT-optimized structures show planarity differences between the two compounds (Figure ). 3TPA-DiKTa adopts a more planar conformation, with a planarity ratio of 74%, compared to 3DPA-DiKTa where the planarity ratio is 42%. |
62d8007e581987674d48b516 | 10 | According to Figure compounds. We thus proceeded to model the excited states using the second order algebraic diagrammatic construction (ADC(2)) with the cc-pVDZ basis set using the spin component scaling (SCS-) approximation. The different density plots of the low-lying singlet and triplet excited states are shown in Figure . The S1 difference densities of both emitters show similar patterns to that of DiKTa (Figures and). Based on the charge-transfer distance, DCT < 0.6 Å, we assign these excited states to be SRCT (Table ). For both emitters, there are small contributions to the difference density from the peripheral donors. Specifically, the decreased density of 3DPA-DiKTa is mostly localized on the nitrogen atoms of the DPA groups, which will increase the electron-donating strength of the substituent, contributing to a more stabilized S1 state, while the decreased density of 3TPA-DiKTa is mostly located on the phenyl rings of the TPA groups, thereby having less of an influence on the energy of the S1 state relative to the DPA groups in 3DPA-DiKTa. The presence of the phenylene bridge between the DPA groups and the DiKTa core in 3TPA-DiKTa reduces the HOMO and LUMO overlap but also increases the oscillator strength (f) in 3TPA-DiKTa in comparison to 3DPA-DiKTa (0.29 vs 0.18, respectively). The S2 state of 3TPA-DiKTa has a typical n-p * character, while for 3DPA-DiKTa S2 has similar difference density to S1. According to our previous study, the SRCT should dominate the S1 emission process under a low polarity environment. The predicted S1/T1 energies for 3TPA-DiKTa and 3DPA-DiKTa are 3. |
62d8007e581987674d48b516 | 11 | Cyclic voltammetry (CV) and differential pulse voltammetry (DPV) measurements were carried out in dichloromethane (DCM) to experimentally ascertain the HOMO and LUMO levels (Figure ). The electrochemical data is compiled in Table . The reduction waves of ). The corresponding electrochemical gap for 3DPA-DikTA is 2.12 eV and for 3TPA-DikTA is 2.29 eV, both significantly smaller than that of DiKTa at 3.12 eV. |
62d8007e581987674d48b516 | 12 | Spectroscopic measurements (absorption and photoluminescence) in dilute toluene (10 -5 M) at room temperature were undertaken to understand the photophysical properties of the monomolecular species. The spectra are shown in Figure . For both emitters the absorption band below 400 nm is attributed to a locally excited (LE) π-π* transition of the whole skeleton. The longer wavelength SRCT absorption bands at 494 nm for 3TPA-DiKTa and 554 nm for 3DPA-DiKTa are red-shifted compared to that of DiKTa at 433 nm. The PL maxima of 3DPA-DiKTa occurs at 597 nm and 3TPA-DiKTA occurs at 537 nm, both significantly red-shifted compared to that of DiKTa (λPL= 453 nm); indeed, 3DPA-DiKTa |
62d8007e581987674d48b516 | 13 | shows the reddest emission among the DiKTa-based MR emitters. The Stokes shift of 3TPA-DiKTa is 42 nm and of 3DPA-DiKTa is 41 nm. The emission of both emitters is narrow, with FWHM of 54 nm or energy width at half maxima (EWHM) of 252 meV for 3TPA-DiKTa and FWHM of 47 nm or EWHM of 168 meV for 3DPA-DiKTa. These values reveal only a modest degree of reorganization in the excited state compared to the ground state. The slightly broader emission of 3TPA-DiKTa than 3DPA-DiKTa is due to a greater admixture of CT character to the SRCT emissive excited state. The EWHM value of 3DPA-DiKTa is smaller than that of 3TPA-DiKTa yet similar to that of DiKTa (172 meV or 27 nm), indicating the more pronounced LRCT character of the SRCT excited state of 3TPA-DiKTa, which aligns with the coupled cluster calculations analysis (Figure ). Figure shows the effect of oxygen on the steady-state PL in toluene. The FPL values in degassed toluene are 44% for 3TPA-DiKTa and 59% for 3DPA-DiKTa, which decreased in the presence of oxygen to 33% |
62d8007e581987674d48b516 | 14 | We next determined the singlet/triplet (S1/T1) energies of both emitters in 2-MeTHF glass at 77K (Figure ). The S1/T1 energies are 2.48/2.27 eV for 3TPA-DiKTa and 2.17/1.95 eV for 3DPA-DiKTa, resulting in small DEST values of 0.21 eV for 3TPA-DiKTa and 0.22 eV for 3DPA-DiKTa. These values are almost identical to that of DiKTa (~ 0.20 eV) and its derivatives. There is a notable small degree of positive solvatochromism observed for 3DPA-DiKTa (Figure ), which is characteristic of MR-TADF compounds. There is a more pronounced positive solvatochromism for 3TPA-DiKTa, consistent with a more significant admixture of CT character to the SRCT emissive excited state. Full-width at half-maximum; c Photoluminescence quantum yield of thermally evaporated thin films, measured using an integrating sphere, under N2 at lexc = 340 nm. Measured at lexc = 379 nm and 300 K under vacuum; e Obtained from the onset of the prompt spectrum (1-100 ns) at 77 K. f Obtained from the onset of the delayed spectrum (1-8.5 ms) at 77 K (lexc = 343 nm). g DEST = E(S1) -E(T1). |
62d8007e581987674d48b516 | 15 | We next investigated the photophysical behavior of both emitters in the OLED relevant host, mCP, which has a suitably high triplet energy of 2.91 eV. First, we identified 2 wt% as the optimal concentration by measuring the FPL of spin-coated films of varying concentration from 1 wt% to 10 wt% in mCP. We then compared the FPL of 2 wt% films in a range of hosts such as bis[2-(diphenylphosphino)phenyl]ether oxide (DPEPO) and 4,4'-bis(9-carbazolyl)-1,1'biphenyl (CBP), and the FPL in mCP was found to be the highest at 86%. (Figure ). The u. |
62d8007e581987674d48b516 | 16 | FPL decreased along with a red-shifted emission (Figure and Table ) upon an increase of the doping concentration of each emitter in mCP host. The red-shifted emission can be ascribed in part to aggregate formation. As shown in Figure , 3TPA-DiKTa emits at lPL = 551 nm with FWHM of 58 nm (245 meV) in 2 wt% mCP evaporated film, whereas 3DPA-DiKTa shows a more red-shifted emission at lPL = 617 nm with FWHM of 56 nm (198 meV). |
62d8007e581987674d48b516 | 17 | Notably, their emission spectra in mCP are red-shifted compared to those in dilute toluene, which given the low polarity of this host, implies the presence of host/guest interactions and likely contribution from aggregates in the solid state even at the low doping concentration employed. Benefiting from the delicate balance between CT and SRCT, 3TPA-DiKTa shows a higher FPL of 93% than 3DPA-DiKTa (FPL = 60%) in 2 wt% mCP evaporated film. |
62d8007e581987674d48b516 | 18 | The time-resolved PL decays in mCP show a nanosecond prompt emission and a microsecond long delayed emission at room temperature (Table ). 3TPA-DiKTA possesses a shorter delayed lifetime, tDF, of 131 μs in comparison to 3DPA-DiKTa (tDF of 323 μs) in 2wt% mCP doped films (Figures ). The temperature-dependent time-resolved PL decays are shown in Figure . The prompt emission is insensitive to temperature while the delayed emission is thermally activated, the latter behaviour consistent with TADF. The S1/T1 energies were determined from the onsets of the prompt fluorescence and phosphorescence spectra of 2 wt% mCP doped films of 3TPA-DiKTa and 3DPA-DiKTa at 77 K (Table ). The DEST value is 0.13 eV for 3TPA-DiKTa and 0.20 eV for 3DPA-DiKTa (Figure ). The former is smaller than that of DiKTa (0.20 eV) and accompanied by a shorter tDF, which is desirable for a TADF material. Based on the promising optoelectronic properties, we next proceeded to fabricate vacuumdeposited OLEDs using 3TPA-DiKTa and 3DPA-DiKTa as emitters. As shown in Figure and Figure , the optimised OLED stack (device B) consisted of: indium-tin-oxide (ITO, 112 nm)/ 1,4,5,8,9,11-hexaazatriphenylenehexacarbonitrile (HATCN) (5 nm)/ 1,1-bis[(di-4tolylamino)phenyl]cyclohexane (TAPC) (40 nm)/tris(4-carbazoyl-9-ylphenyl)amine (TCTA) (10 nm)/ 1,3-bis(N-carbazolyl)benzene (mCP) (10 nm)/emissive layer (20 nm)/ 1,3,5-tri[(3pyridyl)-phen-3-yl]benzene (TmPyPB) (50 nm)/ LiF (0.6 nm)/ Al (100 nm), where HATCN is the hole injection layer (HIL), TAPC and TCTA play the role of hole transporting layers (HTL) and mCP acts as an electron/exciton blocking layer (EBL). TmPyPB acts both as an electron transport layer and a hole blocking layer due to its deep HOMO (-6.7 eV), and LiF acts as The EL spectra are similar to their corresponding PL spectra in the mCP doped thin film (Figure ) reflecting an emission from the same SRCT excited state. In comparison to the previously reported OLED with DiKTa (lEL = 465 nm, FWHM = 39 nm), both of the devices showed a red-shifted emission with a slightly broader electroluminescence. As shown in Table and DiKTa-based MR-TADF emitters (Figure ). We note that, besides the DiKTa based red MR-TADF OLEDs, there are only a small number of red MR-TADF OLEDs. These include emitters based on TOAT such as mBDPA-TOAT (lEL = 600 nm, FWHM = 45 nm, EQEmax =17.3%), pBDPA-TOAT (lEL = 624 nm, FWHM = 62 nm, EQEmax = 11.3%), and 5 (lEL = 595 nm, FWHM = 45 nm, EQEmax = 2%), based on BCz such as BBCZ-tert-butyl (5) (lEL = 615 nm, FWHM = 26 nm, EQEmax = 22%) and boron-nitrogen embedded polycyclic heteroaromatics such as R-BN (lEL = 664 nm, FWHM = 48 nm, EQEmax = 28.1%) and R-BNT (lEL = 686 nm, FWHM = 49 nm, EQEmax = 27.6%). The corresponding Commission Internationale de l'Éclairage (CIE) coordinates are (0.409, 0.577) and (0.633, 0.365) for the devices with 3TPA-DiKTa and 3DPA-DiKTa, respectively. The 3TPA-DiKTa based device showed a much higher EQEmax in comparison to the device with 3DPA-DiKTa as well as other reported DiKTa based devices (Table ). Considering the FPL of the evaporated 2 wt% mCP doped films (Table ) and assuming 25% out coupling efficiency, the EQEmax of the 3TPA-DiKTa device is expected to be 23.3%, which is much lower than the observed EQEmax of 30.8% whilst the calculated EQEmax of the 3DPA-DiKTa device is 15.0%, and very close to the observed EQEmax of 16.7%. One potential explanation of the much-improved 3TPA-DiKTa OLED efficiency would be that the transition dipole moment (TDM) of the emitter is preferentially horizontally oriented, parallel to the substrate surface. We therefore measured the orientation of the TDM of the emitters in 2 wt% evaporated doped films in mCP of 50 nm thickness, emulating the thickness in the device. The angular dependent PL measurement results are shown in Figure and supporting refractive index measurements and modelling are shown in Figures . The anisotropy factors, a, which is defined by the ratio of emitted power by vertical dipoles to total emitted power by all dipoles and extracted from the p-polarized emission, were found to be 0.189 for 3TPA-DiKTa, and 0.278 for 3DPA-DiKTa; notably, previously reported DiKTa derivatives showed anisotropy factors close to isotropic at 0.33. The relatively horizontal transition dipole moment orientation in 3TPA-DiKTa is due to its high molecular weight and higher degree of planarity in contrast to 3DPA-DiKTa (based on the DFT calculated optimized structure in the gas phase). Considering the device structure and the measured anisotropic factors, the corresponding simulated out-coupling efficiency calculated for 3TPA-DiKTa is 28.3% and for 3DPA-DiKTA is 20.5%. Combining the measured FPL of the films and the simulated outcoupling efficiencies, the calculated EQEmax values are 26.4% for 3TPA-DiKTa and 12.3% for 3DPA-DiKTa, which are slightly lower than the observed values. Therefore, the emitter orientation alone cannot explain this inconsistency. Similar higher than expected efficiencies of OLEDs have also been reported in many other MR-TADF device studies. Although the origin of the higher-than-expected EQEmax is not clear here, we can envision that a potential cause is the Purcell effect associated with the reflective electrodes in the OLED stack, which can result in a higher than expected IQE. A second potential explanation could be microcavity effects in the OLED stack leading to light emission that is directed forwards more than for a Lambertian emitter and hence increasing the apparent EQE when measured from the forward direction. An important contribution to the outstanding EQE of the 3TPA-DikTa-is horizontal alignment of the transition dipole moment of the emitter in the evaporated film. These results demonstrate that simple decoration of the DiKTa acceptor with a TPA and DPA substituent is an effective approach to attaining efficient green/red TADF OLEDs. |
60c753c8567dfe6b0dec5f81 | 0 | Introduction: Selective hydrogenation of olefin bond of conjugated carbonyl compounds in the presence of other function groups is an important task in organic synthesis. As hydrogen 1,2 from gas cylinder is potentially explosive and requires to be used with caution in a special apparatus, search for alternative methods to conventional hydrogenation procedures such as heterogeneous and homogeneous catalytic transfer hydrogenation is still going on. Surprisingly, there are only a few methods available in the literature for effecting the reduction of the olefin in ,-unsaturated carbonyl/nitro molecules. The most common used reagents for the reduction of conjugated carbonyl compounds are palladium assisted hydrogen transfer by ammonium formate, 3 formic acid, Diphenyl sulfide, 5 polystyrene, or potassium formate, Palladium acetate, 8 metal catalyzed reduction by sodium hydrogen telluride, copper nanoparticles, 10 samarium iodide, lithium aluminum hydride with SbCl3, sodium boroacetate, sodium borohydride reduced hydroxy ester, nickel chloride, iridium, ruthenium, silane hydride and Yeast in organic solvent system. Recently, O-mesityl(sulfonyl)hydroxylamine is used for aziridination of simple olefins. It is a versatile class of reagents and has been used for the various functional group transformation, migration of O-Phenylhydroxylamine, N-Unsubstituted Arylsulfilimines, Hydroxy-amine, Hydrazine, and Beckman Rearrangement. Surprisingly, no significant effort has yet been made to utilize this potential reagent for the regio and stereoselective hydrogenation of double bonds of entry 7). The structure was initially assigned by 1 H, C-NMR and HRMS analyses and further confirmed by single crystal X-ray analysis. To test the generality of the procedure, various substates were subjected to the reaction conditions (see table ) and it is concluded that the ease of reduction is sensitive to steric and electronic factor. For example, when cinnamoyl ester (entry 8, Method-A) high yield is obtained, due to two electron withdrawing group on either side. Whereas low chemical yield (<10%) was obtained with -methyl (entry 9, Method-A) or -bromo (entry 10, Method-A) substituent while no desired products were isolated when ,dimethyl (entry 11, Method-A) or ,dimethyl (entry 12, Method-A) substrates were used. Even more interestingly the chemical yields were significantly improved in the presence of catalytic amount of Lewis acid (entries 1 and 5-8, Method-B). In case of ,-dimethyl (entry 11, Method-B) or ,dimethyl (entry 12, Method-B), even increase the catalyst load to 0.5% and prolong reaction time also did not gave the products. The used of triethylamine or DBU is ineffective on the rate of reaction. |
60c753c8567dfe6b0dec5f81 | 1 | To further explore the utility of O-mesityl(sulfonyl)hydroxylamine as a new and convenient entry to hydrogenation methods, a variety of ,-unsaturated carbonyl derivatives were studied, and the results are summarized in Table . The carbon carbon double bond in several structurally varied amides underwent hydrogenation to give the corresponding saturated analogues moderate to excellent yields by this procedure. |
60c753c8567dfe6b0dec5f81 | 2 | The substituents on the olefins apparently have influence on the course of hydrogenation. Other reducible groups such as ethyl ester or amide, halogen, nitro groups were fully compatible with the reaction conditions and remained unaltered. The mechanistic explanation of this reaction remained unclear and it is believed that self coupling of reagent occurs to form the diimide compounds with liberation of hydrogen gas, which is consumed by olefins during reaction. In separate experiments, the compounds with nitrile group and azide were treated under same reaction conditions, we recovered starting materials only, it further enhances the selective reduction of unsaturated carbonyl compounds over nitrile and azides. |
60c753c8567dfe6b0dec5f81 | 3 | Four sulfonyl amines that is O-( In summary, the mild condition, high yields of products, operational simplicity, easy availability of reagent and no external use of explosive hydrogen gas makes this methodology a more useful and practical alternative to the existing methods for reduction of ,-unsaturated carbonyl compounds. This extends the synthetic application to O-mesityl(sulfonyl)hydroxylamine. We believe that this method will find useful applications in the field of organic synthesis. |
64de5b5101042bc1cc37933c | 0 | Linear Free Energy Relationships (LFERs) quantitatively relate changes in molecular structure to changes in chemical reactivity and represent a well-established and robust approach to deriving mechanistic insight from multiple observations. Historically, LFERs have been carried out using parameters or descriptors obtained from experimental observations for a set of reference processes, such as acid ionization constants (thermodynamic measurements) or relative rate constants (kinetic measurements). These quantitative parameters can, in turn, be related to the outcomes of entirely different reactions. The Hammett relationship is a paradigmatic example in which benzoic acid acidity (an experimentally determined thermodynamic parameter) correlates with the rate for various reactions involving substrates, reagents, or catalysts bearing substituted aromatic rings. A linear correlation between reactivity and the inductive and resonance contributions from metaor para-substituents on aromatic rings (σm and σp Hammett parameters, respectively) reveal the sensitivity of a reaction to these electronic influences. Deviations from linearity may be related to changes in the rate-determining step (e.g., a concave downward curve) or in the mechanism or transition state of the reaction (e.g., a concave upward curve). The ability to relate substituent electronic effects to chemical reactivity is a cornerstone of physical organic chemistry. Several variants of the Hammett parameter, σ, have been derived for experimental reference processes, including Brown's σp + parameters, based on the solvolysis of t-cumyl chlorides, Arnold's σ⍺ parameters, based on spin delocalization of substituted benzyl radicals, Creary's σC parameters, based on methylenecyclopropane rearrangements, and Jiang and Ji's σjj parameters, based on trifluorostyrene cyclodimerizations. Alternative experimental measures, for example electrode polarization effects, have also been related to Hammett parameters. However, as quantum chemical methods have improved, computationally derived parameters, for example obtained from density functional theory (DFT) calculations, have emerged as an alternative for the quantitative description of molecules and their substituents. Given the broad utility of Hammett relationships in correlating reactivity and selectivity with substituent electronic effects, there have been numerous efforts to obtain parameters/descriptors computationally that avoid the need for measurements of equilibrium constants or which can be employed beyond aromatic substituents. 12 Additional Hammett-like equations have also been derived to capture electronic substituent effects using computational concepts like molecular softness and Fukui functions. Computationally derived electronic parameters are increasingly attractive since they can be obtained for structures and substituents without available experimental data and, due to recent advances in computational workflow automation, are readily accessible for large numbers of functional groups. Another advantage of using computed electronic descriptors vs. tabulated parameters is that additivity does not need to be assumed -multiply substituted aromatics can instead be computed directly. Combined steric and electronic effects have been insightful in describing Hammett relationships, however, many studies rely solely on electronic descriptors. Nevertheless, the departure from experimental descriptor acquisition, obtained from macroscopic "top-down" experimental observations to using computational "bottom-up" descriptors presents several challenges for practitioners. For example, the choice of computational protocol(s) and treatment of the conformational ensemble are critical considerations, particularly where descriptor values are obtained as an ensemble average. More fundamentally, since the quantum mechanical wavefunction/density is inherently delocalized across an entire structure and atomic charges are not themselves experimentally observable, myriad theoretical approaches exist to partition the electronic density, which does not by definition belong to any atom, over the different atoms. Our work is focused on establishing the most appropriate computational protocol for obtaining a correct qualitative and quantitative description of substituent electronic effects at the macroscopic level. To the best of our knowledge, no benchmark comparisons of these protocols against a broad dataset of experimental electronic parameters has been previously reported. |
64de5b5101042bc1cc37933c | 1 | Atomic charges play a fundamental role in qualitative pictures of reactivity and selectivity, and are integral ingredients used to construct quantitative relationships between chemical structure and reaction outcomes. The theoretical foundations of different charge models have been described extensively, and significant effort has gone into comparing the various methods for computing atomic charge. Cramer and Truhlar charges; (III) those obtained by fitting to reproduce a physical observable, (such as dipole moment or a computed electrostatic potential) seen in the CHarges from ELectrostatic Potentials using a Grid-based method (CHELPG), Merz-Kollman charges using Universal force field radii (MKUFF), Hu, Lu, Yang charges with standard atomic densities (HLYGat) and Atomic Polar Tensor (APT) 27 charges; or (IV) those based on semi-empirical adjustments to Class II or III methods, such as in Charge Model 5 (CM5). With the above categories another important distinction arises between methods that project the molecular wavefunction onto atom-centered basis functions, like NPA, and those that are based directly on the electron density as a function of space, such as the Hirshfeld scheme. In this work we observe a distinct performance difference between these two groupings. |
64de5b5101042bc1cc37933c | 2 | Each charge method operates on its own scale (although atomic contributions do sum to total molecular charge) and all types have been used for various applications. In this work, we explore the ability of various charge models from Classes II-IV (what we term a bottom-up description) to describe empirically determined (i.e., top-down) Hammett parameters. While important general considerations -for example, regarding the more stable performance of Hirshfeld or NPA charges vs. Mulliken for different basis set sizes -have been reported previously, less specific information is available to guide the choice of computational protocol in describing macroscopic electronic parameters. Our quantitative comparisons against Hammett parameters addresses this need. |
64de5b5101042bc1cc37933c | 3 | In addition to atomic charges, the relationship of computed atomic properties such as NMR chemical shifts to Hammett parameter is appealing due to their ability to capture local electronic effects and having a corresponding experimental observable. Perhaps due to more complex nuclear coupling relationships captured in this experimental measurement, correlations to partial atomic charge values may not be as high. Still, successful studies have utilized NMR shifts in their relationship to Hammett parameters and partial atomic charges. A variety of methods also exist to compute NMR shifts, including Gauge-Independent Atomic Orbital (GIAO), Individual Gauges for Atoms in Molecules (IGAIM), and the Continuous Set of Gauge Transformations (CSGT) 36 methods. Machine learning (ML) models have been trained to compute partial atomic charges and NMR shifts. When utilized within their domain of applicability, these models give a highly accurate surrogate prediction of QM or experimental properties much faster than traditional QM calculations. These models typically use simple molecular representations, such as a SMILES string, as inputs. The qmdesc Python library utilizes a model trained to predict Hirshfeld charges and GIAO NMR shifts, among other QM properties. The CASCADE Python library also can be utilized to predict experimental 1 H and 13 C NMR shifts. In this work, we perform a quantitative comparison of a wide range of computed parameters obtained from different protocols against experimental Hammett parameters. We examine the performance of various computational methods, including density functional theory (DFT), semi-empirical QM, and ML models (Figure ). Due to their broad generality and widespread availability of empirical values, our study focuses on Hammett's original definition of σm and σp values. |
64de5b5101042bc1cc37933c | 4 | We assess the suitability of different charge models and computational strategies, such as analyzing the meta/para carbon atoms, the attached meta/para hydrogen atoms, and additional ring substituents in deriving computed parameters that are correlated with Hammett parameters. In doing so, we demonstrate a wide range of performance across the different computational protocols and observe stark and surprising differences in the ability of computational parameters to capture para vs. meta-electronic effects. Additionally, we apply these methods to parametrize multiply-substituted aromatics, compare against experimental rate data, and develop suggested best practices for "bottom-up" computational models of inductive and resonance effects as encapsulated by Hammett parameters. We propose linear regression formulae to derive Hammett parameters for substituents without experimental data, and for multiply substituted aromatic rings from computed descriptors. |
64de5b5101042bc1cc37933c | 5 | Our studies use an experimental dataset curated by Ertl 12b based on the 200 most common aryl substituents in the ChEMBL database, of which 89 have experimentally determined Hammett parameters available (Figure ). Molecules (in which each substituent is appended to a phenyl group) were first converted from SMILES format to 3D structures with added hydrogens using RDKit. Conformer ensembles for each structure were then generated with CREST, from which all structures were fully optimized with DFT. We used the B3LYP functional with a Becke-Johnson damped Grimme D3dispersion correction and def2-TZVP basis set in chloroform using SMD implicit solvation for geometry optimizations. All structures were verified as minima by analysis of vibrational frequencies. The DFT-computed charges and NMR shielding tensors were obtained at the same level of theory. Single-point analyses with other density functionals (M06-2X, wB97XD) and solvent models yield parameters that are highly correlated with those discussed in the manuscript; the statistical comparisons made in the main text are independent of the level of theory used (see Supporting Information, section S2 for detailed comparison). For the DFT optimized structures, Class II, III, and IV atomic charges along with NMR isotropic shielding tensors (proportional to computed NMR chemical shifts) were obtained at the same level of theory, all of which are widely available in modern electronic structure packages. Variations of these charge models were also computed, including Minimal-Basis |
64de5b5101042bc1cc37933c | 6 | Mulliken charges (MBS-Mulliken), iterative Hirshfeld, and iterative CM5 methods. ML predicted Hirshfeld charges were calculated by the qmdesc ML package and NMR chemical shifts were predicted using both qmdesc and CASCADE models. Additionally, semi-empirical xTB Mulliken and CM5 charges were also computed. For species with multiple conformers, the atomic charges and chemical shifts were Boltzmann-averaged using the computed Gibbs free energy values. |
64de5b5101042bc1cc37933c | 7 | Computed charges and chemical shift values were parsed from output files (see the repository in the Supporting Information for all Python workflows) for carbon and hydrogen atoms meta and para to the substituent of interest. Initially, we attributed the poor performance of several protocols in describing electronic effects at the meta-position to the difference in experimental scales between σp (-0.83 to 0.78) and σm (-0.24 to 0.71), making the latter intrinsically harder to predict computationally. However, upon further investigation using the computed charge and NMR shift values at the hydrogen (i.e., rather than carbon) atoms, we found this explanation to be incorrect. Indeed, for the meta-H atoms, we obtained an overall improvement in correlation for σm across nearly all methods tested, except for Class III charges and NMR values (Figure ). Correlations for σp values remained relatively consistent relative to those obtained from the analysis of C atoms. These results are perhaps surprising since the absolute differences between computed charges/chemical shifts are smaller for H than for C, being one bond further away from the substituent of interest, nor are these atoms directly influenced by resonance effects. Nevertheless, computed descriptors for H atoms statistically outperform C atoms, and we propose that it is the closer proximity of the meta-C atom to the substituent of interest that results in numerical differences between the different methods. |
64de5b5101042bc1cc37933c | 8 | While NPA charges remove the problem of orbital overlap through orthogonalization, nevertheless, we find these values are significantly peturbed by the spatial proximity of meta-substituents yielding unrealistic results that correlate poorly with experimental Hammet parameters. In contrast, integration of the electron density over atomic domains, as in the Hirshfeld scheme, avoids problems inherent to basis set based approaches yielding atomic charges that perform well across para-and meta-substitution patterns and which correlate best to observable Hammet parameters. Hirshfeld or other density-based schemes (CM5, Voronoi) are therefore strongly recommended as electronic parameters/descriptors for future physicalorganic or statistical studies, over the widely-used NPA charges. Additionally, where the practitioner has a choice of atom(s) |
64de5b5101042bc1cc37933c | 9 | for which to collect electronic parameters, we observe that locations closer to the site of change (which would be expected to show the largest variations) may, in fact, produce descriptors less correlated with macroscopic observations. To further test the hypothesis that computed descriptors for aromatic substituents perform better than the ring atoms themselves in describing electronic effects, we sought a more challenging set of experimental data. We selected the oxidative cleavage of olefins with photoexcited nitroarenes, developed by Leonori and co-workers. In this work, the electronic influence of nitroarene substituent(s) on the rate of cycloaddition with a model alkene (adamantylideneadamantane) was explored by an experimental Hammett study using tabulated sigma values. Importantly, several of these substrates have multiple meta-and para-substituents, for which electronic effects were assumed to be additive in the original study (Figure ). The nitroarenes are shown with the logarithm of the experimental reaction rates relative to (unsubstituted) nitrobenzene (log(kX/kH)). Another advantage of using computed electronic descriptors vs. tabulated parameters is that additivity does not need to be assumed -multiply substituted aromatics can instead be computed directly. We applied the same computational protocols discussed above to obtain charge and NMR shift values for the fourteen nitroarenes. We separately obtained parameters at the ipso-C attached to the nitro group and the N atom itself. When compared against experimental relative rates (Figure ), |
64de5b5101042bc1cc37933c | 10 | for the C atom parameters we observe similar trends in the performance as observed above for meta-substitution. The bestperforming methods are Hirshfeld (R 2 = 0.94), CM5 (R 2 = 0.86), and MBS-Mulliken (R Our computed parameters and results above were obtained for ground singlet state nitroarenes; however, we also computed parameters for the triplet excited state, which is the reactive intermediate involved in the cycloaddition (see Supporting |
64de5b5101042bc1cc37933c | 11 | Class III charges such as CHELPG, MKUFF, and HLYGAT correlate poorly, while Hirshfeld and CM5 charges yield the highest correlations with σm or σp values (R 2 = 0.8-0.9). Another factor influencing the correlation between computed descriptors and experiment is the choice of probe atom: meta-and para-carbon atoms in the aromatic ring produce inferior results to the attached H atoms or other ring substituents. This effect is most pronounced for the prediction of metasubstituent effects, where computed results are the most variable and most methods evaluated at carbon atoms correlate very poorly indeed with σm values. These observations suggest that density-based partition schemes such as Hirshfeld should be generally preferred over basis-set based approaches in obtaining electronic descriptors (such as Hammett parameters) for organic structures, since they are more stable to unphysical perturbations of charge that result from nearby functional groups. |
64de5b5101042bc1cc37933c | 12 | From the perspective of predicting unknown Hammett constants, or those for multiply-substituted systems, we have obtained linear scaling relationships between experimental parameters and all computed protocols studied here. Hirshfeld qmdesc Hirshfeld charges evaluated at carbon atoms perform well globally (R 2 = 0.88), while CASCADE chemical shifts perform well (R 2 = 0.84) in predicting σp values. Overall, these observations lead us to conclude that electronic parameterization strategies should carefully consider the impact of location of probe atoms and how partial atomic charges are derived, since we observe a larger impact than from the choice of density functional and basis set used. Our studies were restricted to the computational description of Hammett parameters, however, as archetypal electronic descriptors that have been applied to LFERs across thousands of studies, we suggest that our findings and recommendations should be considered as part of general computational approaches to electronic parameterization. |
6605e28ce9ebbb4db9d320a5 | 0 | phosphate groups. Bridges between phosphate and solvent-exposed nucleobase atoms have also been identified. The 2 ′ -hydroxyl group, typical of the ribose ring, has been under special focus, with water molecules suggested to form bridges between the 2 ′ -OH and phosphate or base atoms. However, while early studies suggested the 2 ′ -OH group to be involved in long-lived water bridges, molecular dynamics simulations showed that there is no long lived hydrogen-bond between water and the 2 ′ -OH group, and that water molecules with long residence times (up to several hundreds of ps) are found instead next to the phosphate oxygen atoms, with long-lived water bridges between successive O R phosphate oxygen atoms. |
6605e28ce9ebbb4db9d320a5 | 1 | However, a general understanding of the molecular features that govern DNA and RNA hydration dynamics and that explain their differences has remained elusive. The broad variety of RNA folds makes it difficult to draw general conclusions on RNA hydration and the extent of its differences with that of DNA from these studies on specific sequences. |
6605e28ce9ebbb4db9d320a5 | 2 | Hence, in this work, we compare RNA and DNA hydration reorientation dynamics using short model double-stranded sequences adopting the typical helical conformations (A-form for RNA, B-form for DNA), and obtain for the first time a picture of the hydration dynamics at a single-site resolution. Combining our simulations with an analytic model to describe the mechanism of water reorientation and the impact of a biomolecular interface, we then provide an unprecedented molecular level rationalization of the main differences between ds-DNA and ds-RNA hydration dynamics. |
6605e28ce9ebbb4db9d320a5 | 3 | We performed 100 ns-long molecular dynamics (MD) simulations of two 18-mers (GC-CGCGCGCGCGCGCGGC and GCGGGGGGGGGGGGGGGC, later named "GCGC" and "GGGG"), based on the systematic study conducted by the ABC consortium, 42 using the Amber force field for nucleic acids in its parmbsc0 version with χ OL3 modifications for RNA molecules, and SPC/E water molecules. Following the same strategy as in previous works by Auffinger and Westhof, we chose the 18-mers to contain only GC base pairs, so that the only chemical difference between the DNA and RNA systems is the sugar-ribose in RNA and deoxyribose in DNA. The preparation of the systems, simulation setup and structural analysis are detailed in Supporting Information. |
6605e28ce9ebbb4db9d320a5 | 4 | Despite very similar chemical compositions, the ds-DNA and ds-RNA 18-mers adopt strikingly different helical conformations (Figure and S1). Consistently with previous structural studies, the simulated DNA B-form helices have a very narrow minor groove (width around 7.5 Å), while the major groove is much wider (see SI Figures and Table ). In contrast, for the RNA A-helix, the major groove is narrow (width around 6 Å) and very deep (typical depth of about 10 Å), whereas the minor groove is wide (width around 10 Å) and shallow. The surface of the nucleic acid is colored according to the τ reor value computed at each site. Snapshots were prepared using the VMD software. Following a strategy previously successfully applied to proteins and DNA, we performed a site-resolved analysis of water dynamics in the hydration shell of these RNA and DNA 18-mers. Water dynamics in biomolecular hydration shells can be monitored in different ways. Commonly used residence times focus on the time spent by water molecules in the vicinity of a given solute site. However, such times can only be probed indirectly in the experiments and very sensitively depend both on the exact site definition and on the time allowed for transient excursions outside the site. Here, we rather focus on water reorientation dynamics, which is measurable by NMR. This rotational dynamics is very closely linked to the reorganization dynamics of the H-bond network, and thus reports on the lability of the hydration shell. |
Subsets and Splits