This work presents a high-throughput in silico screening of ruthenium(II) pincer complexes as potential catalysts for pyridine hydrogenation To explore how backbone architecture, substituent sterics, and hemilability govern substrate binding energetics, this study screened 24 Ru(II) pincer complexes comprising four distinct lutidine-based pincer backbones (PNpyP, PONpyOP, NONpyON, NNpyN) and six substituents (Me, iPr, tBu, Bu, Ph, Cy). Additionally, 18 complexes representing three specific coordination modes from the PNpyP ligand family were screened using pyridine as a model substrate.

MACE was employed to generate stereoisomer and conformer libraries without conformational bias. Structures were screened with the Universal Force Field, refined via DFT (PBE0-D3(BJ)/def2-SVP), and analysed using ensemble-averaged steric descriptors (percent buried volume and bite angle) and pyridine binding free energies across three coordination pathways: (i) carbonyl auxiliary substitution, (ii) central nitrogen donor dissociation, and (iii) phosphorus side-arm dissociation.

The results reveal that pyridine binding to form RuH₂(py)(𝜅³-PNP), RuH₂(CO)(𝜅²-P,N-PNP)(py), or RuH₂(CO)(𝜅²-P,P-PNP)(py) complexes is generally unfavorable. However, binding becomes more accessible via the hemilabile dissociation of either the nitrogen or phosphorus pincer donor. Dissociation of the nitrogen donor emerged as the most energetically accessible pathway (Δ𝐺 = 30–105 kJ/mol), particularly with phenyl substituents where 𝜋–𝜋 stacking stabilises binding (Δ𝐺 = 29.9 kJ/mol). Conversely, dissociation of the phosphorus pincer donor (Δ𝐺 = 60–105 kJ/mol) presents the most favorable binding energetics in the presence of bulky substituents, as the enhanced flexibility of the phosphine side-arm accommodates steric bulk more effectively. Substitution of the carbonyl remains largely inaccessible (Δ𝐺 = 100–150 kJ/mol), as it requires the coordination of pyridine to a fully occupied complex.

Analysis of the ensemble-averaged steric descriptors revealed that an optimal balance between ligand hemilability, steric constraints, and geometric flexibility governs substrate accessibility. A percent buried volume in the range of 40–50% and P–Ru–P bite angles near 95–115^◦ showed significantly improved pyridine binding energetics. This was exemplified by the RuH₂(CO)(𝜅²-P,P-PNP)(py) complex, where structural reorganisation upon nitrogen dissociation allows the complex to exploit conformational flexibility and stabilise substrate binding through non-covalent interactions. This creates a confined,welldefined pocket adjacent to the metal center that effectively accommodates the substrate. These findings warrant further investigation to bridge computational insights with improved catalytic performance.
... Read more

Addressing the resources needed to produce sustainable and environmentally friendly products is key within the field of catalysis. One of the key applications of catalysis is the storage of renewable hydrogen. This could be achieved through liquid organic hydrogen carriers (LOHCs), which participate in homogeneous catalysis. This field emphasizes high selectivity through the use of transition-metal complexes. A suitable LOHC candidate could be pyridine, a type of N-heterocycle, which serves as a benchmark for successful binding to the metal complex. The binding of pyridine can be enhanced by substituting noble metals with transition metals, such as manganese, which has shown promising catalytic activity. We perform high-throughput screening with DFT calculation for the following systems: PNP, PONOP, NNN, and NONON, which vary in their backbones. Every metal complex consists of three various configurations: the alignment of the auxiliary carbonyl ligand with the hydride atom (config 1), pyridine (config 2) and the lutidine part of the nitrogen pincer atom (config 3). The catalytic performance is studied by determining the stable and reactive complexes, as well as their specific configurations. Only positive binding energies, in terms of the ΔGreaction, are observed. This indicates that no complexes show strong binding of the metal to pyridine. However, the phenyl-substituted system exhibits the lowest binding energies, of which the third configuration is the most favored for all ligand types. This holds for the trans-positioning of the auxiliary carbonyl ligand with the lutidine part of the nitrogen pincer atom,mainly for the NNN-based complex. Still, the preferred configurations do not correlate with the strengthened metal-pyridine binding and weakening of the metal-auxiliary carbonyl ligand bond length. In terms of reactivity, the nitrogen-based complexes show the highest hydride charges, which could be assumed to provide high reactivity. However, the hydridic behavior of the complexes does not correspond with the stability of the phenyl substituents and all remaining complexes. After the reactivity of the complexes is considered, we can enable fine-tuning of specific backbones to forecast trends observed in reactivity. The phenyl-substituent can be used as a starting point due to its delocalized system and electron-donating property for fine-tuning. Nitrogen-based complexes enable fine-tuning of the reactivity because they depend on the ligand scaffold rather than the substituents. In addition, variation in binding energies for every substituent is most commonly observed for nitrogen-based complexes. The nitrogen-based complex can therefore be used to performimproved ligand design with the fine-tuning ability of the phenyl substituent. ... Read more

This study is an in silico screening of zinc PNP and NNN pincer complexes with variation of the R groups in the ligands for the homogeneous catalysis of the hydrogenation of pyridine. This was done by investigating geometry, hemilability, and binding energies. The scope lies in identifying which complexes are thermodynamically capable of binding pyridine, which is researched using the Gibbs free energy of the binding reaction. Energies and optimized geometries were obtained using Density Functional Theory with ORCA and the supercomputer Snellius. The screening revealed that PNP-R complexes favor tetrahedral geometries after optimization, but three and five coordinated structures are also possible. Hemilability of the phosphorus and/or nitrogen arm were observed. All PNP-complexes showed thermodynamically unfavored binding of pyridine. Additionally, NNN-R complexes did not bind to pyridine. Only when phenyl was used as the R group in the NNN backbone did pyridine bind. This is believed to be due to a hydride migrating to one of the phenyl rings, but even then, the binding energy was not thermodynamically favorable. Moreover, research on ligand substitution by pyridine showed that in most complexes this reaction is unfavorable. Only in three coordinated structures of NNN was this reaction thermodynamically feasible. These findings contribute to understanding the pincer ligand dynamics of PNP and NNN complexes and suggest directions for future research. ... Read more

As the transition to cleaner energy intensifies, N-heterocycles as liquid organic hydrogen carriers (LOHCs) offer a promising approach. However, their reliance on noble metals such as ruthenium, iridium, and platinum poses sustainability challenges.
In this study, 60 Mo(I) pincer complexes were screened using a combination of MACE and DFT. MACE was used to generate initial 3D molecular structures via force field optimization. These were further analysed using DFT calculations with the PBE0-D3BJ functional and def2-SVP basis set under standard conditions. Gibbs free energies were computed and used to evaluate pyridine binding energies across the ligand set.
The results indicate that pyridine binding to Mo(I) complexes is unfavourable, despite several complexes exhibiting sufficiently hydridic Mo-H bonds to suggest potential catalytic activity. The hydride charge appears to be conformation dependent, with certain configurations decreasing or increasing the hydridic charge. Oxidation of Mo(I) was observed in some systems, leading to pincer ligand decomposition. Additionally, nitrogen donor arms often dissociate from the metal centre, resulting in 5-coordinated geometries.
The unfavourable binding energy suggest that inner sphere mechanisms are unlikely. Instead, the findings support the plausibility of outer sphere pathways, especially given the lack of correlation between pyridine binding energy and the hydridic charge. Furthermore, the observed dissociation of nitrogen donor arms hint at potential ligand hemilability, which may influence catalytic dynamics and warrants further investigation.
... Read more

This thesis investigates why EEG-based neurophysiological research in architecture remains primarily confined to the design guideline phase, despite its potential for direct integration into design development. Through historical, philosophical, and empirical analysis, the study argues that neuroarchitecture possesses a sufficiently mature theoretical foundation for application across all design phases. A systematic review analysis demonstrates that research complexity increases substantially in design development, requiring advanced computational methods such as machine learning and real-time brain-computer interfaces. The findings suggest that the limited adoption of neuroscience during the act of designing is not due to theoretical shortcomings, but rather to the absence of accessible computational tools, pipelines, and shared datasets. The thesis concludes that broader integration of neuroscience in architecture depends on the development of practical and accessible infrastructures for designers. ... Read more

Bidentate ligand-coordinated transition metal complexes are often used as homogeneous catalysts, as they have the ability to produce enantioselective compounds. These compounds are of high interest in the pharmaceutical and food industries. However, identifying high performing catalysts relies on trial-and-error approaches, which is time-consuming and costly. The use of data-driven predictive models could improve this process significantly by shifting most of the work from experimental work to computational work. Previous work from the group has attempted to develop such a predictive model using Machine Learning (ML), a representation of a manually generated static structure, and a database generated through High-Throughput Experimentation (HTE). However, these models faced challenges in terms of model performance and consistency between different substrates. This research aims to enhance these models by improving the representations used in ML to achieve more accurate predictions. To bring the representations closer to reality, both dynamic and new static approaches are tested, using conformer ensembles (CEs) generated by CREST. These structures were then used in DFT calculations to obtain accurate properties of these complexes. Additionally, new HTE data, which is closer to the complexes used in the simulation, was incorporated to improve training data for the ML models. The investigated reaction is the hydrogenation of norbornadiene (NBD) using Rh-NBD complexes. The performance of both classification and regression was compared across different representations: a cheap topological connectivity fingerprint (ECFP), semi-empirical DFT representations, and expensive fully DFT-optimized representations. The results conclude that none of the DFT-based representations outperforms the cheap topological fingerprint for this specific reaction. The study also highlights the importance of high-quality data in training the models. Ultimately, while the representation was improved, the much simpler topological method was the most effective for prediction of catalyst performance. ... Read more

Transition metal complexes are important in homogeneous catalysis reactions, especially in asymmetric hydrogenation reactions. The coordinated ligands in the complexes provide stability and selectivity. With the right combination of ligands and metal centre a high selectivity can be reached. To find optimal metal-ligand combinations, the chemical space is explored with high throughput computational methods. In these methods descriptors are obtained, usually from a single structure with one specific ligand configuration, which might not represent reality very well. In earlier research the chemical space of iridium(III), ruthenium(II) and manganese(I) complexes has been explored. And in this research additional analysis was done to look at possible relations between ligand configurations and descriptors. This was done by using three unsupervised dimensionality reduction methods, i.e. PCA, t-SNE and UMAP. PCA showed that the ligand configuration could have an influence on mainly electronic descriptors, but failed to show clusters in terms of relative stability. t-SNE and UMAP showed some clusters for the stability, as well as overlapping between certain ligand configurations. However, no definitive relations have been found, thus optimising the analysis methods and performing other statistical analysis on the descriptors might give different outcomes. ... Read more

Asymmetric hydrogenation is a field of major interest for the pharmaceutical industry. Using these catalyzed reactions instead of traditional stoichiometric reactions can reduce waste and energy, and can open up possibilities to new intermediates, products, and synthesis pathways. Finding an optimal catalyst to produce a selected enantiomer remains a struggle, however, requiring large time and resource investments. Determining ligand performance can be done experimentally using HTE campaigns, supplemented with predictive methods, either mechanism-based or mechanism-agnostic. Recent advancements in mechanism-agnostic predictive methods include a large range of studies using Machine Learning approaches, relying mostly on molecular descriptors to represent the catalyst structure to the models. More recently, with the rise of NLP models, string-based structural identifiers are used to train a Language Model to predict catalyst performance. In a recent study, an LSTM model was trained and used to predict the enantiomeric excess of a range of ligands for an asymmetric hydrogenation reaction. This work is based on the workflow used by them and validates the performance of this model as shown in their paper. Furthermore, this model was applied and tuned to predict the enantiomeric excess of a range of ligands, based on a dataset from the ISE research group. ... Read more

Transition metal complexes as homogeneous catalysts enable high enantioselectivity in hydrogenation reactions, making them especially beneficial for the pharmaceutical industry. The development of data-driven prediction models enhances high-throughput catalyst design. However, these models often focus solely on static molecular representation, neglecting the dynamic behavior of the system, such as the formation of conformer ensembles. Currently, no method is available to systematically account for these conformational effects at reasonable costs. In light of this, the study aimed to develop a practical tool that allows predictive models to incorporate the dynamic characteristics of catalysts via conformer ensembles. A dataset of Rh-based precatalysts with mainly bidentate ligands was utilized. Three cheminformatic tools—RDKit, OpenBabel, and CREST—were explored for reliable, automated conformer ensemble generation. Among them, only CREST proved feasible, although it exhibited several limitations and required manual modification. A mapping between the conformer geometries obtained from GFN2-xTB and DFT calculations was achieved based on the relative energies and root mean square deviations. This revealed that many conformers generated by CREST converge into the same DFT local minimum. A classification method was developed to bridge the gap between conformers obtained from the two quantum chemical calculations by selecting a subset of conformers from the CREST ensemble that appear as distinct conformers in the DFT ensemble. This approach allows DFT calculations to be performed only on conformers that would result in different DFT minima on the potential energy surface, thereby eliminating redundant calculations and saving significant costs. This unsupervised DBSCAN clustering algorithm was applied to the GFN2-xTB energy and RMSD of the conformers, reducing the number of redundant conformers by 46% in the original dataset of Rh-based precatalyst structures. ... Read more

Various industries rely upon transition metal complexes to efficiently catalyze chemical reactions. These transition metal complexes often consist of precious metals, which are scarce and expensive. Therefore, a shift towards catalysts containing earth-abundant metals is necessary. Computational catalyst screening has the potential to accelerate this shift by reducing catalyst discovery time. The first step of computational catalyst screening consists of obtaining a digital representation of the catalyst. Usually, a fixed a priori ligand configuration of TM-complexes is assumed to represent the catalyst. However, this approach of catalyst representation might not capture the influence of alternative ligand configurations on the observed catalytic behavior. In the context of high-throughput in-silico catalyst screening, this study aims to evaluate the influence of ligand configurations on the stability and physicochemical properties of transition metal complexes. An automated workflow for the generation of complexes, complex sorting based on ligand configuration, DFT geometry optimization, and descriptor extraction is employed. Ensembles of ligand configurations are generated for iridium(III), ruthenium(II), and manganese(I) complexes featuring 88 bisphosphine bidentate ligands. Based on DFT-optimized geometries, analysis reveals a preference for a specific ligand configuration for iridium(III) complexes. However, this preference is not observed for ruthenium(II) and manganese(I) complexes. Furthermore, for the majority of ruthenium(II) and manganese(I) complexes, multiple ligand configurations are found within a 10 kJ/mol range from the most favorable one. The analysis of thermodynamic, electronic, geometric, and steric descriptors reveals that none of the descriptors consistently correlates to the preferred ligand configuration. These findings indicate that a priori selection of ligand configuration may result in insufficient coverage and representation of key catalyst features for predictive in-silico chemical space exploration. ... Read more

Computational chemistry is about making models that simulate the behaviour of real chemical entities. These models can then be used as a predictive tool. For example, in high throughput screening of a library of computationally optimized molecules to find catalysts for drug design. In the screening there need to be values on which it is screened, one type of possible value can be descriptors, a numerical representation of the molecule. It is thus important that these molecules are as accurate as possible to get representative descriptors, but also relatively fast. As to make a library a lot of optimizations are performed, thus driving the costs up if the optimization takes a long time. To optimize a molecule its energy is needed. Calculating the energy of a system is done using quantum mechanics and is an integral part of computational chemistry. Many methods can be used to approximate the energy of a system. One such way of calculating the energy is DFT. DFT depends on a functional and a basis set, which the user must choose. This study is the descriptor-based benchmarking of three basis sets, def2-SV(P), def2-TZVPP, def2-QZVPP and five functionals, PBE, TPSS, PBE0, B3LYP and MN15. The optimisations’ results are compared using different methods using three molecular descriptors (bite angle, buried volume, and HOMO-LUMO gap). The third part of this study uses a combination of optimisation methods to try and improve the previous results. The different methods were compared by comparing the optimisation times and the descriptors to the standard of PBE0/def2-SV(P). The structures were optimised using Gaussian, and the descriptors were calculated using the in-house workflow OBeLiX. When comparing the basis sets, it was noted that def2-QZVPP took too much time to be of use and was thus not used in further comparisons. def2-TZVPP took substantially longer to complete than def2-SV(P). Looking at the descriptors, there was no difference between them. This led to the conclusion that def2-SV(P) was the optimum basis set for these 192 complexes as it was faster but had the same accuracy. When comparing the functionals, there was the surprising result that the choice of functional did not impact the chosen geometric and steric descriptors of bite angle and buried volume. The electronic descriptor, the HOMO-LUMO gap, differed greatly per method. The lower-level theory PBE and TPSS had a very low value compared to the hybrid functionals but were close to each other. MN15 had a HOMO-LUMO gap that was substantially higher than B3LYP and PBE0. The assumption was thus made that B3LYP and PBE0 were the most accurate functionals in this case. Looking at the time needed for the bulk of the optimisations to complete, PBE was by far the faster functional and MN15 the slowest, PBE0 was located in the middle of the pack. As PBE0 has a shorter optimisation time than B3LYP, the conclusion was that PBE0 was the optimal functional to use in this case. The third part of the study looked at combining optimisationmethods to see if a faster optimisation could be achievedwith the same accuracy. Here the base was a fastway, such as GFN2-xTB and PBE, to calculate the geometric and steric properties and then use a PBE0 calculation to make the electronic descriptor as accurate as when doing a general PBE0 optimisation. The best option was doing a single-point PBE0 calculation after a PBE optimisation. It was much faster than a PBE0/def2-SV(P) optimisation and as accurate. Making it an excellent way to optimise the TM complexes. ... Read more

Homogeneous transition metal-based (TM) catalysts are crucial to producing chemically pure drugs, stemming from their ability to obtain high product selectivity. However, experimental screening of TM-based complexes is expensive, so computational methods are leveraged instead. Especially machine learning (ML) approaches show promise due to being efficient as well as unbiased. ML of homogeneous TM-based catalysts is based on physiochemical properties named descriptors. Descriptors are dependent on the method of simulation and the simulated complex itself. Methods with a higher level of theory are more accurate, but also more resource intensive. Similarly, larger complexes simply demand more computational resources. Two general methods to minimize the number of resources needed are: 1) using the lowest level of theory containing reasonable accuracy and 2) using the simplest representative complex. In this thesis, possible simplifications were investigated for a homogeneous TM-based catalyst screening workflow. Objective 1 was investigating the effect of levels of theory for geometry optimization on descriptors. Structures were optimized for four levels of theory relevant to this workflow, namely: MACE, GFN-FF, GFN2-xTB, and DFT. Subsequently, xTB level descriptors were calculated for the first three levels of theory and were then correlated against xTB level descriptors of the benchmark, DFT. In addition, it was investigated how descriptors obtained from xTB and DFT single-point calculations differ. Objective 2 was investigating the effect of the chemical structure on descriptors. To do so, a set of octahedral complexes and a set of simplified structures were generated, and descriptors of both sets were correlated against each other. Regarding objective 1, it was observed that solely descriptors from the GFN2-xTB level of theory correlated well with DFT, at least for the majority of descriptors. Next to that, it was found that GFN2-xTB geometries more or less coincide with DFT geometries. Regarding objective 2, it was found that the bidentate ligands in the model set deform towards the metal centre, which leads to decreased correlations among the majority of the descriptors. Additionally, it was found that clustering occurred due to the presence of two different ligand classes in the dataset. The primary conclusion of this research was that geometries originating from GFN2-xTB geometry optimization are structurally comparable to geometries originating from DFT geometry optimization. However, descriptors obtained from GFN2-xTB single-point calculations are not comparable to descriptors obtained from DFT single-point calculations. As such, to accurately extract descriptors, DFT single-point calculations are necessitated. ... Read more

Catalysts play an essential role in industry and for the general progress of mankind. With the parallel energy and technological transformations, it is important to create tools that aid in the development of better catalysts. To achieve this feat, it is firstly required to have a fully automated approach for in silico structure generation. Thus, in this study the OBeLiX workflow has been developed. The designed package includes a scaffold generation tool, a substituent placement tool, GFNn-xTB optimization and conformer search tools completed by a fully automated descriptor calculator. Even though descriptor databases can be found in literature, their reproducibility is limited. Consequently, the ability to reconstruct proposed approaches for new chemical reactions is hindered. OBeLiX has been used to investigate a series of hydrogenation reactions catalyzed by rhodium phosphine complexes. The approach begins with the creation of a structure database for 192 such complexes. To simplify this process, it was opted to use a mechanistically relevant model catalyst structure. In the first step of the catalytic cycle, π-complexation occurs between the substrate and the metal center. Thus, a symmetric chelating norbornadiene molecule has been chosen to model the asymmetric substrates. The generated database of model catalysts has been featurized through OBeLiX. The use of model structures underlined that the substrates have to be quantified as well. While for the complex model catalysts a series of chemically descriptive features have been created, the substrates were converted to two-dimensional fingerprints, and Sterimol parameters that describe the 3D size of the substrate around the double bond that is to be hydrogenated. Therefore, featurization of the chemical reaction has been achieved. Training machine learning algorithms on these features, yielded high correlations including out-of-sample binary reactivity classification for substrates outside the training set. ... Read more

Many drugs cannot be made without homogeneous catalysis. To increase the yield of drug synthesis, the search for new catalysts continues. Computational catalysis is becoming a more prominent tool since it en- ables to screen many catalysts without performing (m)any laboratory experiments. In this research a com- putational workflow has been created to calculate both electronic (e.g. dipole, ionisation potential, nucle- ophilicity, etc.) and steric (bite angle, buried volume, cone angle, etc.) molecular descriptors. The structures were automatically created starting from the metal centre, some bidentate phosphorus ligands, auxiliary lig- ands, and the substrate. An in-house computational workflow, MACE, is used for for high-throughput gen- eration of structures from the starting bidentate phosphorus ligands, by generating stereo-isomers around to metal centre. Afterwards small substituents (H, CH3, Ph, etc.) are changed by ChemSpaX increasing the number of structures combinatorically. To reduce the computational cost of this workflow, it has been re- searched whether properties of the octahedral geometry could be predicted using properties from a simple model structure containing only the metal centre and the bidentate phosphorus ligand. This model structure did not show a correlation except for the electronic energy and the solvent accessible surface area, which are both primarily influenced by the number of electrons. Other correlations may be found if some other descriptors were calculated which where excluded now, like the HOMO-LUMO gap and the substrate bind- ing energy. The workflow could be extended to machine learning and improved by including symmetry and optical isomerism. ... Read more

Manual data processing is a tedious task that should be automated. Besides saving time, automated data processing also fights other problems in chemistry. Automated data processing in a normalized way makes data analysis between different experiments possible and can remove biases. In this study, kinetic data workflow is studied and a python script for automated data processing is made. Different experiments on the hydrogenation reaction of ethyl hexanoate using a ruthenium-PNN catalyst have been performed to obtain kinetic hydrogenation data. Reaction temperature was set to either 50, 70 or 110 °C and catalyst loading was either 50, 100 or 200 ppm. In total, nine experiments were performed. Gas chromatography and pressure readings during the reaction are used for analysis. Analysis is done on two different machine that report data in a different template and format. Therefore, a python script was written to automatically process the raw data obtained by these analysis methods. The written script imports obtained data, calculates new normalized parameters, concentration of different species for example, creates different plots of processed data and exports an excel file containing normalized data. This exported excel file was used to further examine catalyst kinetics. It was found for example, that at 50 and 70 °C, the reaction order of the catalyst is below one and at 90 and 110 °C catalyst reaction order is above one. The made processing script has improved data workflow, while data processing time has been reduced. ... Read more

The search for better catalysts for imine hydrogenation is an ongoing challenge. Understanding of the underlying mechanism can play an important role in the evolution of catalyst design. The goal of this research is to get more insight in the asymmetric hydrogenation (AH) of 2-Methyl-1-pyrroline with an Ir-based catalyst with a phosphorous bidentate ligand (JosiPhos) using density functional theory (DFT) calculations. In this research, two mechanisms are investigated. The inner sphere C-migration mechanism is explored for all possible options for the coordination of the substrate to the catalyst complex. A conformer search with Conformer–Rotamer Ensemble Sampling Tool (CREST) is done on three transition states (TS) to investigate if ensemble representations are important for the determination of the enantioselectivity of a certain path. The outer sphere mechanism is only partially explored. Using DFT, the energy landscape of each reaction path could be mapped out.

Of the inner sphere C-migration mechanism, ten possible reaction paths were obtained. The two most thermodynamically favourable paths are the paths where the substrate binds on one specific equatorial coordination site. There is no substantial preference for the formation of one enantiomer over the other. However, CREST calculations generated possible conformers of certain transition states, some with a lower energy. This resulted in a preference for the formation of the R-enantiomer for the inner sphere C-migration mechanism.

Little can be said about the outer sphere mechanism since there is not enough data yet to form a clear conclusion about the favourability of this mechanism. One of the findings was that another type of outer sphere mechanism was found other than the outer sphere mechanism proposed based on literature. To be able to state what mechanism is the most likely to occur, more quantitative research is needed on the outer sphere mechanism and other possible mechanisms. ... Read more