Variational Pair-Density Functional Theory : Dealing with Strong Correlation at the Protein Scale
(2023) In Journal of Chemical Theory and Computation- Abstract
Multiconfigurational pair-density functional theory (MC-PDFT) offers a promising solution to the challenges faced by traditional density functional theory (DFT) in addressing molecular systems containing transition metals, open-shells, or strong correlations in general. By utilizing both the density and on-top pair-density, MC-PDFT can make use of a more flexible multiconfigurational wave function to capture the necessary static correlation, while the pair-density functional also includes the effect of dynamic correlation. So far, MC-PDFT has been used after a multiconfigurational self-consistent field (MCSCF) step, using the orbitals and configuration interaction coefficients from the converged MCSCF wave function to compute PDFT... (More)
Multiconfigurational pair-density functional theory (MC-PDFT) offers a promising solution to the challenges faced by traditional density functional theory (DFT) in addressing molecular systems containing transition metals, open-shells, or strong correlations in general. By utilizing both the density and on-top pair-density, MC-PDFT can make use of a more flexible multiconfigurational wave function to capture the necessary static correlation, while the pair-density functional also includes the effect of dynamic correlation. So far, MC-PDFT has been used after a multiconfigurational self-consistent field (MCSCF) step, using the orbitals and configuration interaction coefficients from the converged MCSCF wave function to compute PDFT energies and properties. Here, instead, we propose to perform a direct optimization of the wave function using the pair-density functionals, resulting in a variational formulation of MC-PDFT. We derive the expressions for the wave function gradient and illustrate their similarity to standard MCSCF equations. Furthermore, we illustrate the accuracy on a set of singlet-triplet gaps as well as dissociation curves. Our findings highlight one of MC-PDFT’s standout features: a reduced dependency on the active space size compared to conventional multiconfigurational wave function methodologies. Additionally, we show that the computational cost of MC-PDFT is potentially lower than MCSCF and often on-par with standard Kohn-Sham DFT, which is demonstrated by performing a MC-PDFT calculation of the entire ferredoxin protein with 1447 atoms and nearly 12 000 basis functions.
(Less)
- author
- Scott, Mikael LU ; Rodrigues, Gabriel L.S. LU ; Li, Xin and Delcey, Mickael G. LU
- organization
- publishing date
- 2023
- type
- Contribution to journal
- publication status
- epub
- subject
- in
- Journal of Chemical Theory and Computation
- publisher
- The American Chemical Society (ACS)
- external identifiers
-
- pmid:38217859
- scopus:85182570838
- ISSN
- 1549-9618
- DOI
- 10.1021/acs.jctc.3c01240
- language
- English
- LU publication?
- yes
- id
- 4b7b1b10-147b-4f87-ad6d-c0cefde59943
- date added to LUP
- 2024-02-15 14:55:06
- date last changed
- 2024-04-30 15:28:24
@article{4b7b1b10-147b-4f87-ad6d-c0cefde59943, abstract = {{<p>Multiconfigurational pair-density functional theory (MC-PDFT) offers a promising solution to the challenges faced by traditional density functional theory (DFT) in addressing molecular systems containing transition metals, open-shells, or strong correlations in general. By utilizing both the density and on-top pair-density, MC-PDFT can make use of a more flexible multiconfigurational wave function to capture the necessary static correlation, while the pair-density functional also includes the effect of dynamic correlation. So far, MC-PDFT has been used after a multiconfigurational self-consistent field (MCSCF) step, using the orbitals and configuration interaction coefficients from the converged MCSCF wave function to compute PDFT energies and properties. Here, instead, we propose to perform a direct optimization of the wave function using the pair-density functionals, resulting in a variational formulation of MC-PDFT. We derive the expressions for the wave function gradient and illustrate their similarity to standard MCSCF equations. Furthermore, we illustrate the accuracy on a set of singlet-triplet gaps as well as dissociation curves. Our findings highlight one of MC-PDFT’s standout features: a reduced dependency on the active space size compared to conventional multiconfigurational wave function methodologies. Additionally, we show that the computational cost of MC-PDFT is potentially lower than MCSCF and often on-par with standard Kohn-Sham DFT, which is demonstrated by performing a MC-PDFT calculation of the entire ferredoxin protein with 1447 atoms and nearly 12 000 basis functions.</p>}}, author = {{Scott, Mikael and Rodrigues, Gabriel L.S. and Li, Xin and Delcey, Mickael G.}}, issn = {{1549-9618}}, language = {{eng}}, publisher = {{The American Chemical Society (ACS)}}, series = {{Journal of Chemical Theory and Computation}}, title = {{Variational Pair-Density Functional Theory : Dealing with Strong Correlation at the Protein Scale}}, url = {{http://dx.doi.org/10.1021/acs.jctc.3c01240}}, doi = {{10.1021/acs.jctc.3c01240}}, year = {{2023}}, }