Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session B17: Density Functional Theory in Chemical Physics IIFocus
|
Hide Abstracts |
Sponsoring Units: DCP Chair: Carsten Ullrich, University of Missouri Room: Room 209 |
Monday, March 6, 2023 11:30AM - 12:06PM |
B17.00001: New results from semiclassics: chemical accuracy in orbital-free DFT Invited Speaker: Kieron Burke The successes and failures of approximate density functionals are due to their connection with semiclassical expansions. In the semiclassical limit, relative errors in local density approximations vanish. Carefully derived corrections to that limit have been shown to be far more accurate than our usual DFT approximations. I will discuss important new results in our 20-year-long quest to derive density functional approximations as expansions in hbar. These include both a new correction to the expansion of the exchange energy of atoms and an orbital-free calculation with sub-milli-Hartree accuracy. |
Monday, March 6, 2023 12:06PM - 12:18PM |
B17.00002: Effect of Electron-electron Cusp Condition on the Exchange-Correlation (XC) Hole Studied with Hooke's atoms Lin Hou, Tom J Irons, James W Furness, Andrew M Teale, Jianwei Sun The exchange correlation (XC) hole is critical for understanding density functional theory and the associated XC density functional approximations. The XC hole of a system can be calculated with accurate wavefunction methods. CCSD is a wavefunction method widely used for ordinary atoms and molecules and exact for 2 electron systems at the basis set limit, while it misses the electron-electron cusp condition with finite conventional basis sets. Here, using the Hooke's atom for which the analytic wavefunction solutions are available, we study the effect of cusp condition on the XC hole calculated from CCSD. We show that the cusp effect-driven error in CCSD is negligible with the t-aug-cc-pV5Z basis set, and becomes smaller for more diffusive electron densities. We also studied the coupling constant averaged and system and angle averaged XC hole against the PBE and LDA XC holes. |
Monday, March 6, 2023 12:18PM - 12:30PM |
B17.00003: Testing orbital-free models of the kinetic energy density in semiconductors Akinfolarin V Akinola, Antonio C Cancio, Brielle Shope Meta-generalized gradient approximations (mGGAs) for the exchange-correlation (XC) energy in density functional theory (DFT) conventionally depend upon the Kohn-Sham kinetic energy density (KED). Use of the KED makes mGGAs more accurate than generalized gradient approximations (GGAs) but also more computationally expensive for applications such as ab initio molecular dynamics. Deorbitalizated mGGAs replace the KED with a pure density functional. Through visualization we explore how well the exact KED can be represented by a single KE mGGA functional dependent upon the scaled density, scaled density gradient and density Laplacian. We calculate the KE and electron density of semiconductor solids with varying ionicity and atomic number using the ABINIT DFT plane-wave pseudopotential code. We find a near-universal linear correlation with the density Laplacian and density gradient for regions outside the atomic bond. That is consistent with a modification of the second-order gradient expansion. Non-self-consistent calculations of structural properties using several KED functionals and the Kohn-Sham density were performed. We find best performance of mGGA KED models for those that approximate this linear correlation. |
Monday, March 6, 2023 12:30PM - 12:42PM |
B17.00004: Efficient Implementation of Machine Learning-Based Nonlocal Functionals for Molecules and Solids Kyle Bystrom, Boris Kozinsky Machine Learning (ML) has recently gained attention as a means to develop more accurate exchange-correlation (XC) functionals, and in particular could be used to improve XC functionals for solids without using the computationally expensive exact exchange energy. Feature design is one of the main challenges of this approach because the features must include enough nonlocality to capture the complex nonlocal nature of the exact XC functional while also allowing for computationally efficient and scalable implementations. To address this challenge, the CIDER model for designing nonlocal ML functionals is introduced and used to create an exchange functional that obeys the uniform scaling rule for exchange. In addition, two efficient methods for evaluating CIDER features are implemented: A quadratic-scaling algorithm for all-electron, molecular density functional theory (DFT) and a quasi-linear-scaling algorithm for plane-wave DFT with the PAW method. Efficiency and accuracy benchmarks for molecular periodic systems are presented, and the importance of the kinetic energy as a feature for learning the exchange functional is discussed. |
Monday, March 6, 2023 12:42PM - 12:54PM |
B17.00005: Optimization of kinetic energy functionals for orbital-free exchange-correlation meta-GGAs. Antonio C Cancio, Bishal Thapa Orbital-free (OF) meta-GGA functionals for the exchange-correlation energy [1,2] have shown promise as a cost-effective alternative to conventional meta-GGAs. These replace the noninteracting kinetic energy (KE) density used in a meta-GGA with a pure functional of the density. One family of KE functionals, based on the Perdew-Constantin functional (Phys. Rev. B 75, 155109 (2007)), uses the Laplacian of the density to switch from the slowly varying electron gas to the von Weizsacker or localized electron-pair limits, resulting in an accurate OF model for the KE density. Unfortunately use of the density Laplacian creates unphysically spiky Pauli potentials that are numerically slow to converge and lead to noisy results. We discuss the construction of a measure of smoothness of functional potentials and the implementation of an optimization procedure for minimizing fluctuations in the potential based on the Poisson equation. This is used to generate modifications of the PC and related functionals. Tests on small atoms demonstrate the issue of noise in Laplacian-based potentials and the value of incorporating potential smoothness as a design goal for developing such functionals. |
Monday, March 6, 2023 12:54PM - 1:30PM |
B17.00006: Density Functional Theory of molecular fragments: Strongly-correlated electrons in weakly-interacting fragments Invited Speaker: Adam Wasserman To gain flexibility in the search for the minimum energy in ground-state DFT, and to improve on the accuracy of density functional approximations for stretched bonds, we explore a generalized DFT in which the ground-state energy of a molecule is decomposed into a sum of fragment energies and a partition energy. The latter is defined as a functional of the set of fragment spin densities and is amenable to simple, physically-motivated approximations designed to satisfy exact constraints that functionals of the total molecular spin densities cannot easily meet. Employing standard XC approximations for the fragments, this approach is successful at dissociating challenging molecules correctly without artificially breaking their spin symmetries. We discuss implications for strongly-correlated electronic systems, and challenges ahead. |
Monday, March 6, 2023 1:30PM - 1:42PM |
B17.00007: Analysis of the Large Z Exchange Expansion of Atoms and Ions Beyond the Leading Order JEREMY J REDD, Antonio C Cancio, Kieron Burke, Nathan Argaman Lieb-Simon zeta-scaling, which describes the scaling of neutral closed-shell atoms as Z approaches infinity, has been used with marked success to provide theoretical constraints on density functionals and to help generate perturbative expansions with Z. The exact large Z expansion for the exchange energy has never been explicitly calculated beyond the leading order Z5/3 term. We have recently provided consistent numerical and analytical evidence of a ZlnZ term beyond the LDA. Further the numerical evidence suggests the LDA has a Z4/3 term. We propose an analytical model for ions, and compare with numerical results for N=1 and N/Z=1/2 atoms, and find that the model agrees with the results within 10%. These results improve as Z increases. We also provide an analysis of the Z4/3 behavior in the LDA using the Thomas-Fermi potential and Kohn-Sham orbitals. |
Monday, March 6, 2023 1:42PM - 1:54PM |
B17.00008: Deorbitalization of two-indicator-function meta-GGA exchange-correlation functionals Hector I Francisco Rodriguez, Samuel B Trickey, Antonio C Cancio De-orbitalization of orbital-dependent meta-GGA exchange-correlation (XC) approximations is, by now, a demonstrably successful strategy to reduce the cost of ab initio MD simulations with such functionals and reconnect them to pure Kohn-Sham (local potential) analysis. De-orbitalization replaces the KS kinetic energy density in a meta-GGA with an explicit function of the density n(r), its gradient, and its Laplacian. The success is incomplete however: neither TPSS, original Tao-Mo, revised Tao-Mo (revTM), or revised-regularized TM (rregTM) functionals deorbitalize accurately with the prescription of Phys. Rev. A 96, 052512 (2017) for molecules. Strangely, solid deorbitalization [Phys. Rev. B 102, 121109 (2020)] fails completely for original TM, but is not too bad for rregTM. Both TPSS and all the TM variants are distinct in having two iso-orbital indicators, which seems to be the barrier. We discuss the implications, as well as the effects of pseudo-potential choice and refinement of deorbitalizers. |
Monday, March 6, 2023 1:54PM - 2:06PM |
B17.00009: Smoothing techniques to tame Laplacian-dependent exchange-correlation functionals Samuel B Trickey, Tun Sheng Tan, Antonio C Cancio Inclusion of dimensionless, density-Laplacian variable dependence has |
Monday, March 6, 2023 2:06PM - 2:18PM |
B17.00010: Reassessing the role of exact conditions in density functional theory Ryan D Pederson, Kieron Burke Exact conditions have long been used to guide the construction of density functional approximations. Nowadays hundreds of approximations are in common use, many of which neglected these conditions in their design. Such approximations yield extremely accurate results on comprehensive molecular benchmarks and are often more accurate than their more constrained counterparts. To resolve this paradox, we carefully parse the logic of exact conditions, finding that many enforced conditions are too strong for real matter. A computational scanning procedure finds violations, coupled with the construction of corresponding reasonable (but not realistic) densities. Half a dozen exact conditions and hundreds of approximate functionals are analyzed. Several forgotten conditions are revived and analyzed, while even well-known conditions yield surprising new twists. The role of exact conditions in density functional development is revisited. |
Monday, March 6, 2023 2:18PM - 2:30PM |
B17.00011: Physics Informed Neural Nets for Prediction of KS Potentials Vincent Martinetto, Attila Cangi, Aurora Pribram-Jones Kohn-Sham density functional theory is one of the most successful electronic structure methods for molecules and materials, and density-to-potential inversions can provide insights into the exact formalism underlying this approach. This work looks to circumvent normal inversion schemes by employing Physics Informed Neural Nets (PINNs) in their place. PINNs help to improve predictive transferability and reduce the requisite amount of data to properly train a neural network. Implementations of a convolutional PINN and its application to exactly solvable models, such as soft-Coulomb systems, will be presented. Extensions of the network into ensemble density functional theory and realistic systems will be discussed. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700