Application of Semiempirical Methods to Transition Metal Complexes: Fast Results but Hard-to-Predict Accuracy

Minenkov, Yury; Sharapa, Dmitry I.; Cavallo, Luigi

doi:10.1021/acs.jctc.8b00018

Cited by 46 publications

(56 citation statements)

References 138 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Narrower interval (1‐2 kcal/mol) would result in less reliable statistics (see below); be free of transition metals and/or any other poorly parameterized elements as Se, Te, etc. According to our recent study blind application of semiempirical methods might result in non‐reliable relative conformational energies and/or unrealistic geometries of transition metal complexes; contain not more than sixty atoms. We introduced this restriction to be able to perform single‐point energy reference DLPNO‐CCSD(T) calculations with triple‐ζ correlation consistent basis set and accurate TightPNO settings in reasonably short time (see below); to be of practical scientific/industrial interest.…”

Section: Methodology and Computational Detailsmentioning

confidence: 99%

A Robust and Cost‐Efficient Scheme for Accurate Conformational Energies of Organic Molecules

et al. 2018

Self Cite

View full text Add to dashboard Cite

Several standard semiempirical methods as well as the MMFF94 force field approximation have been tested in reproducing 8 DLPNO‐CCSD(T)/cc‐pVTZ level conformational energies and spatial structures for 37 organic molecules representing pharmaceuticals, drugs, catalysts, synthetic precursors, industry‐related chemicals (37conf8 database). All contemporary semiempirical methods surpass their standard counterparts resulting in more reliable conformational energies and spatial structures, even though at significantly higher computational costs. However, even these methods show unexpected failures in reproducing energy differences between several conformers of the crown ether 1,4,7,10,13,16‐hexaoxacyclooctadecane (18‐crown‐6). Inexpensive force field MMFF94 approximation groups with contemporary semiempirical methods in reproducing the correct order of conformational energies and spatial structures, although the performance in predicting absolute conformational energies compares to standard semiempirical methods. Based on these findings, we suggest a two‐step strategy for reliable yet feasible conformational search and sampling in realistic‐size flexible organic molecules: i) geometry optimization/preselection of relevant conformers using the MMFF94 force field; ii) single‐point energy evaluations using a contemporary semiempirical method. We expect that developed database 37conf8 is going to be useful for development of semiempirical methods.

show abstract

Section: Methodology and Computational Detailsmentioning

confidence: 99%

A Robust and Cost‐Efficient Scheme for Accurate Conformational Energies of Organic Molecules

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…While this task is mostly straightforward in ground state organic chemistry, it is far more challenging for open shell transition metal chemistry central to the development of new functional catalysts 4,[29][30][31][32][33][34][35][36] and materials [37][38][39][40][41][42][43][44][45] . Characterization of open shell transition metal complexes requires time-consuming first-principles simulation, typically 4,29,[46][47] with density functional theory (DFT), because few semi-empirical theories 48 or force fields [49][50] can describe the multiple accessible oxidation and spin states of transition metal complexes in a balanced manner. Such simulations typically require manual validation that the simulation was successful, e.g., if the structure optimized to the intended geometry 15,48 and if the electronic structure reflects the expected spin state.…”

Section: Introductionmentioning

confidence: 99%

“…Characterization of open shell transition metal complexes requires time-consuming first-principles simulation, typically 4,29,[46][47] with density functional theory (DFT), because few semi-empirical theories 48 or force fields [49][50] can describe the multiple accessible oxidation and spin states of transition metal complexes in a balanced manner. Such simulations typically require manual validation that the simulation was successful, e.g., if the structure optimized to the intended geometry 15,48 and if the electronic structure reflects the expected spin state. Automated approaches have only begun to be applied, e.g., for functional selection 51 , active space selection in wavefunction theory calculations 52 , or for interpreting reactivity [53][54][55] .…”

Section: Introductionmentioning

confidence: 99%

Learning from Failure: Predicting Electronic Structure Calculation Outcomes with Machine Learning Models

Duan

Janet

Liu

et al. 2019

J. Chem. Theory Comput.

153

View full text Add to dashboard Cite

High-throughput computational screening for chemical discovery mandates the automated and unsupervised simulation of thousands of new molecules and materials. In challenging materials spaces, such as open shell transition metal chemistry, characterization requires time-consuming first-principles simulation that often necessitates human intervention. These calculations can frequently lead to a null result, e.g., the calculation does not converge or the molecule does not stay intact during a geometry optimization. To overcome this challenge toward realizing fully automated chemical discovery in transition metal chemistry, we have developed the first machine learning models that predict the likelihood of successful simulation outcomes. We train support vector machine and artificial neural network classifiers to predict simulation outcomes (i.e., geometry optimization result and degree of deviation) for a chosen electronic structure method based on chemical composition. For these static models, we achieve an area under the curve of at least 0.95, minimizing computational time spent on non-productive simulations and therefore enabling efficient chemical space exploration. We introduce a metric of model uncertainty based on the distribution of points in the latent space to systematically improve model prediction confidence. In a complementary approach, we train a convolutional neural network classification model on simulation output electronic and geometric structure time series data. This dynamic model generalizes more readily than the static classifier by becoming more predictive as input simulation length increases. Finally, we describe approaches for using these models to enable autonomous job control in transition metal complex discovery. File list (3) download file view on ChemRxiv DuanClassifier.pdf (2.94 MiB) download file view on ChemRxiv SIClassifier_v5.pdf (5.28 MiB) download file view on ChemRxiv data_set.zip (35.83 MiB)

show abstract

“…The inclusion of systems with strong electron correlation in

scriptD

may lead to a bias in p which would at least partially explain the generally poor accuracy for transition‐metal complexes. Despite significant efforts, it was not yet possible to create an NDDO‐SEMO model which achieves a similar accuracy with respect to the reference data for transition‐metal complexes as for organic compounds …”

Section: Implicit Description Of Electron Correlation Effects Throughmentioning

confidence: 99%

Semiempirical molecular orbital models based on the neglect of diatomic differential overlap approximation

Husch

Vaucher

Reiher

2018

Int J of Quantum Chemistry

View full text Add to dashboard Cite

Semiempirical molecular orbital (SEMO) models based on the neglect of diatomic differential overlap (NDDO) approximation efficiently solve the self-consistent field equations by rather drastic approximations. The computational efficiency comes at the cost of an error in the electron-electron repulsion integrals. The error may be compensated by the introduction of parametric expressions to evaluate the electron-electron repulsion integrals, the one-electron integrals, and the core-core repulsion. We review the resulting formalisms of popular NDDO-SEMO models (such as the MNDO(/d), AM1, PMx, and OMx models) in a concise and self-contained manner. We discuss the approaches to implicitly and explicitly describe electron correlation effects within NDDO-SEMO models and we dissect strengths and weaknesses of the different approaches in a detailed analysis. For this purpose, we consider the results of recent benchmark studies. Furthermore, we apply bootstrapping to perform a sensitivity analysis for a selection of parameters in the MNDO model. We also identify systematic limitations of NDDO-SEMO models by drawing on an analogy to Kohn-Sham density functional theory. K E Y W O R D SNDDO approximation, semi-empirical methods | INTRODUCTIONThe driving force for the development of semiempirical molecular orbital (SEMO) models has always been the desire to accelerate quantum chemical calculations. At the outset of the development of SEMO models in the middle of the last century, [1][2][3][4][5][6][7][8][9][10] the goal was to carry out electronic structure calculations for small molecules, which was not routinely possible with ab initio electronic structure methods at that time. Since then, theoretical chemistry has seen a remarkable development not only in terms of computational resources but also in terms of ab initio methodology. [11] One must not forget that most electronic structure methods which we apply routinely today, such as Kohn-Sham density functional theory (KS-DFT) [12] and coupled cluster theory, [13] were developed concurrently with today's SEMO models. As a consequence of algorithmic and methodological developments, [11] accurate ab initio electronic structure methods have long replaced SEMO models in their original areas of application (electronic structure calculations for small molecules). Nevertheless, SEMO models did not become extinct. Instead, they opened up different areas of application which can broadly be divided into three categories (see also Ref. [14] for a recent review): (1) simulations of very large systems such as proteins [15][16][17][18][19][20][21][22] and those with thousands of small molecules, [23,24] (2) calculations for a large number of isolated and unrelated medium-sized molecules, for example, in virtual high-throughput screening schemes for materials discovery [25,26] and docking-and-scoring of potential drug candidates, [27][28][29][30][31][32] and (3) entirely new applications such as real-time quantum chemistry where ultra-fast SEMO models allow the perception of visual and h...

show abstract

Application of Semiempirical Methods to Transition Metal Complexes: Fast Results but Hard-to-Predict Accuracy

Cited by 46 publications

References 138 publications

A Robust and Cost‐Efficient Scheme for Accurate Conformational Energies of Organic Molecules

A Robust and Cost‐Efficient Scheme for Accurate Conformational Energies of Organic Molecules

Learning from Failure: Predicting Electronic Structure Calculation Outcomes with Machine Learning Models

Semiempirical molecular orbital models based on the neglect of diatomic differential overlap approximation

Contact Info

Product

Resources

About