HIGHLIGHTS

new, improved method for optimizing molecular geometries

Scientific Achievement

Optimization of molecular geometries is best done in internal coordinates: bonds lengths, bond angles, and dihedral angles. However, these coordinates are coupled, creating complicated constraints and consequently loss of efficiency for widely used optimization algorithms. Our new optimization strategy interprets the space of all possible molecular geometries as a manifold, and molecular geometry displacements are taken along the geodesics on these manifold, inherently satisfying the necessary constraints. Compared to the traditional approach, this substantially reduces the number of steps required to reach convergence on a molecular geometry optimization benchmark

Significance and Impact


Geometry optimization is a key aspect in any computational study of molecules. This step may be ratelimiting in automatic potential energy surface exploration frameworks suitable for use on exascale computational resources. Our geodesic optimization method dramatically speeds up optimization of molecules, thereby increasing the throughput in such applications.


Research Details


  • Our method has been implemented in our open-source geometry optimization code, Sella.

  • The method is a drop-in replacement for existing optimization codes using redundant internal coordinates.

  • It can also be used in saddle point optimization for reaction network exploration.


Eric D. Hermes, Khachik Sargsyan, Habib N. Najm, and Judit Zádor, The Journal of Chemical Physics, 2021, 155, 094105. https://doi.org/10.1063/5.0060146


https://github.com/zadorlab/sella

more accurate approach for adsorbate thermophysical properties

Scientific Achievement

We found that contributions of translational anharmonicity are significant for H adsorbed on a Cu(111 ) surface, by computing thermodynamic properties via a phase space integration (PSI) approach. The approach is in excellent agreement with a quantum state counting benchmark over the temperature range of interest, especial ly compared to the most commonly used models, the harmonic oscillator (HO) and the free translator (FT). Obtaining more accurate adsorbate thermophysical properties favorably impacts microk inetic mechanisms for heterogeneous catalysis, as surface coverages depend heavily on these estimates.

Significance and Impact


The method is to be a part of the Sandia National Laboratories’ ECC computational framework for automated chemistry, an effort to provide accurate microkinetic mechanisms. As we transition into the exascale, accurate methods that require computational power become relevant and useful tools to considerably improve chemical process predictions.



Research Details


  • We produced training data for the potential energy surface by performing DFT calculations on the ANL Theta system.

  • A minima-preserving neural network (MP-NN) is constructed and used within the PSI routine.

  • Our method has been implemented in our open-source phase space integration code, AdTherm.


Katrín Blöndal, Khachik Sargsyan, David H. Bross, Branko Ruscic and C. Franklin Goldsmith, The Journal of Physical Chemistry C, doi: 10.1021/acs.jpcc.1c04009

New, Efficient Method for Locating Saddle Points Demonstrated

Scientific Achievement

A new method has been developed for locating saddle points on potential energy surfaces in atomistic simulations using iterative Hessian diagonalization. Our method scales better with respect to system size and converges more reliably than traditional approaches. The method is implemented in a new open-source software package, Sella, which provides a convenient way to use our method in combination with more than 40 electronic structure packages for molecules, solids and other atomic systems.

Significance and Impact


Automated workflows to explore reactive potential energy landscapes combined with current peta- and upcoming exascale computing resources promise an unprecedented acceleration of the rate at which we can understand complex chemical phenomena at the atomic scale. Robust and efficient optimizers are the linchpins of these frameworks. Especially difficult in this context is to have reliable search methods for first order saddle points, these lowest energy passage points between the intermediates of a reaction networks, which ultimately determine the importance and rate of the various chemical processes.




Research Details


We use iterative diagonalization to update an approximate Hessian. Efficiency is gained by using it both to progress towards the saddle point and as a preconditioner for subsequent Hessians. Balancing the cost of diagonalization and the desired accuracy to make the right geometry steps results in a substantial reduction in the total cost of optimization compared to other low-scaling algorithms. Sella is being incorporated into automated potential energy landscape exploration codes, including KinBot (for gas-phase molecules) and the newly-developed pynta (for heterogeneous catalysis).


Eric D. Hermes, Khachik Sargsyan, Habib N. Najm, and Judit Zádor, Journal of Chemical Theory and Computation, 2019, 15, 6536-6549.

https://github.com/zadorlab/sella

Parametric Uncertainty In Automatic Mechanism Generation

Scientific Achievement

A sophisticated microkinetic model of the CO2 methanation for a Ni(111) facet was automatically compiled using the Reaction Mechanism Generator (RMG). 5000 mechanisms were generated in a probabilistic approach by perturbing all energetic parameters in a correlated manner over their entire uncertainty range to explore all the possible chemistry. Propagation of this uncertainty through microkinetic modeling reveals a broad spectrum of possible activity of the Ni(111) facet for the CO2 methanation, with some microkinetic models in quantitative agreement with experimental results.

Significance and Impact

The study highlights the effect of DFT-based uncertainties on the construction of the mechanism. Further, we demonstrate that with the current error bars of ab-initio heats of formation or activation barriers, it is challenging to make conclusive statements about the activity of a metal facet or the rate-determining step. However, the study reveals that only a few adsorbates and elementary steps govern the activity in the uncertainty range. Computationally expensive higher-level theory methods can be employed to determine their energetic parameters more accurately. The procedure provides a universal framework for investigating other heterogeneously catalyzed reactions.

Research Details

  • Mechanism Generation for CO2 methanation on Ni(111) with RMG using linear-scaling and Brønsted-Evans-Polanyi relations

  • Mean-field microkinetic modeling and comparison to temperature scanning methanation experiments from a Ni/SiO2 catalyst

  • Degree of rate control and global uncertainty analysis reveal important adsorbates and elementary steps


Kreitz, B.; Sargsyan, K.; Blöndal, K.; Mazeau, E. J.; West, R. H.; Wehinger, G. D.; Turek, T.; Goldsmith, C. F., Quantifying the Impact of Parametric Uncertainty on Automatic Mechanism Generation for CO2 Hydrogenation on Ni(111). JACS Au 2021, 1, 1656-1673., https://doi.org/10.1021/jacsau.1c00276

Using computational singular perturbation for dynamical analysis in heterogeneous catalysis

Scientific Achievement

The computational singular perturbation (CSP) method has been applied in gas phase chemical kinetic modeling, as well as biochemical modeling, but not hitherto in heterogeneous catalysis. It has been used for diagnosis of dynamical system response, time integration of stiff ordinary differential equations (ODEs), and for providing means for chemical model reduction. In this work, we extend its formulation to handle differential-algebraic equation (DAE) systems that are typically used in heterogeneous catalysis models. Using this new formulation, we demonstrated dynamical analysis of a catalytic system, highlighting important fast/slow processes, and identifying cause-and-effect relationships. We also used the DAE analysis results to examine the quality of different quasi steady state approximations, identifying choices resulting in minimal errors relative to the ODE formulation.

Significance and Impact

CSP analysis is useful for dynamical systems that exhibit stiffness resulting from a wide range of time scales, as is the case e.g. in heterogeneous catalysis. Its main strategy is to decouple fast and slow dynamics through an alternative representation of the original system of equations, thereby providing a framework for dynamical analysis and reduction. Its application in catalytic DAE systems enables improved precision in the analysis of system dynamics, which facilitates diagnostic studies of chemical kinetic model behavior, assessment of the quality of DAE approximations, extraction of understanding of system response, and chemical model reduction in catalytic systems.

Research Details

  • CSP analysis for DAE/ODE of catalysis systems has been implemented in open-source code, CSPlib: https://github.com/sandialabs/CSPlib

  • CSP computations can be executed on modern computing platforms, i.e, GPUs and many-core CPUs


Oscar Díaz-Ibarra, Kyungjoo Kim, Cosmin Safta, Judit Zádor & Habib N. Najm (2021) Using computational singular perturbation as a diagnostic tool in ODE and DAE systems: A case study in heterogeneous catalysis, Combustion Theory and Modelling, https://doi.org/10.1080/13647830.2021.2002417