Design of new family of minimal parameter spin-component-scaled double-hybrid density functional theory


Presently, Kohn-Sham density functional theory (DFT) is one of the most established theoretical methods due to its highly favorable system size to computational cost and inherent treatment of electron correlation. Conversely, this success critically depends on the quality of the exchange-correlation function. Recent scientific advances have seen the development of several techniques aimed at overcoming shortcomings in the treatment of weak interactions with the GGA functionals, which have a direct impact on the properties of a wide range of applications, including proteins folding, crystal packing motifs, and encapsulation energies in nanostructures. The double-hybrid scheme has shown to be a promising strategy for accurate representation of correlation effects. More specifically, dispersion-corrected spin-component-scaled double-hybrid, spin-opposite-scaled double-hybrid, and quadratic integrand double-hybrids, are of particular interest as they reach chemical accuracy. Unfortunately, there are still a number of features that result in an unphysical representation of the correlation effects, most notably the overweighting of the correlation contribution to the total energy of the functional in these methods.

Dr. Loïc Roch and Professor Kim Baldridge at University of Zurich in Switzerland addressed the major drawbacks of the general dispersion-corrected spin-component-scaled double-hybrid DFT methodology through the design of a new family of functionals, referred to as minimal parameter spin-component-scaled double-hybrid, or mSD-DFT. They anticipated that this methodology could retain a 0.5 kilocalorie per mole target accuracy, eradicate the unphysical correlation contributions, and provide a class of approaches with negligible empiricism. Their work is now published in the research journal, Phys.Chem.Chem.Phys.

The authors observed that for the specific case of mSD-PBEPBE, the mean absolute error was 0.4 kilocalorie per mole, which was within the chemical accuracy target. Moreover, with only two parameters, which are extrapolated to the complete basis set (CBS) limit, this novel family of double-hybrid DF ensures limited empiricism.

An important outcome of the work of Loïc Roch and Kim Baldridge is a generalized procedure towards the design of a new family of double-hybrid, referred to as mSD-DFTs. Their procedure is shown to be robust and lead to functionals that are relatively insensitive to  small variations in the parameters around the extrapolated minimums. Additionally, the mSD-DFT scheme was shown to be extendable to include the resolution of identity approximation for solving the computationally intensive integrals. This cost-effective counterpart, referred to as RI-mSD-DFT, drastically reduces the computational cost and resource usage to enable extension to large chemical and biological systems, for investigations of real-life chemical challenges. The scheme can therefore be recommended for the design of additional (RI-)mSD-DFT methods based on others of the plethora of density functional theory types. In fact, having a broader range of such functionals will enable a more thorough comparison of performance of this new family of double-hybrid density functional theories.

design of a new family of minimal parameter spin-component-scaled double-hybrid density functional theory. Advances in Engineering

About the author

Kim Baldridge (PhD, 1988) pursues research in the development and application of quantum mechanical methods for the prediction of molecular structure and physical properties, with special emphasis on aromatic materials, optoelectronic properties, methods for treating solvation, and the prediction of properties such as NMR spectra and acid dissociation constants (~210 publications). She is a leader in high-performance grid computing. She was Director of Computational Science at the San Diego, Supercomputer Center, then director for Grid Computing at UZH and now Co-Dean for Health Science and Director of HPC; from these posts she has made significant contributions to HPC and Grid Computing infrastructure and algorithm development that support QM Science globally.

She founded two major recurring international scientific symposia that highlight QM scientists (running for a combined quarter century), serves on the editorial advisory boards of Journal of Chemical Theory and Computation and Theoretical Chemistry Accounts, and has consulted as computational science expert to the U. S. Army Aberdeen Proving Grounds.

In 2000/2001, Prof. Baldridge was inducted as Fellow of the American Physical Society and the American Association for the Advancement of Science; and, she was awarded the Iota Sigma Pi – Agnes Fay Morgan Research Excellence Award.  Prof. Baldridge received the Chinese Government 1000Talent Award in 2014 and the Haihe Award in 2015.

About the author

Loïc M. Roch obtained a Master of Science MSc in Molecular and Biological Chemistry in 2012 from the École Polytechnique Fédérale de Lausanne (EPFL), Switzerland. After a 10-month stay at University College London (UCL) where his research interest was on the interaction of clay minerals with their surroundings, he moved to Zürich, Switzerland, to complete his PhD at the University of Zürich with Prof. Kim K. Baldridge. His thesis focused on the development of cost-effective methodologies for an accurate description of correlation energy, and on the design of highly redundant and reliable high performance computing (HPC) infrastructures. Since May 2017, Loïc M. Roch moved to Harvard University in the group of Prof. Alán Aspuru-Guzik as a Postdoctoral Fellow.

His work focuses on the design and development of robotic platforms to enable autonomous material discovery. This involves the formulation of complex orchestration software supplying the structured layers indispensable for operating autonomous laboratories, and the implementation of innovative machine learning algorithms for an efficient exploration of scalar, possibly non-convex, objective functions.


Loïc M. Roch, Kim K. Baldridge. General optimization procedure towards the design of a new family of minimal parameter spin-component-scaled double-hybrid density functional theory. Phys.Chem.Chem.Phys., 2017, volume 19, 26191


Go To Physical Chemistry Chemical Physics 

Check Also

Enhancing Photovoltaic Performance of Quaternary Cu-Ag-Bi-I Compounds: The Role of Annealing and Impurities in Cu2AgBiI6 Thin Films - Advances in Engineering

Enhancing Photovoltaic Performance of Quaternary Cu-Ag-Bi-I Compounds: The Role of Annealing and Impurities in Cu2AgBiI6 Thin Films