Mail:
Dept. of Chemistry
Ohio State University
100 W. 18th Ave.
Columbus, OH 43210

Office:
412 CBEC

Email:
herbert@
chemistry.ohio-state.edu

John Herbert research group

Development of TD-DFT methods

Natural transition orbitals for the aqueous electron
Natural transition orbitals for several
electronic transitions of the
aqueous electron

Overview

Time-dependent density functional theory (TD-DFT) is currently the only tractable ab initio method for calculating electronic excitation energies in systems with more than 20–30 atoms. For a wide variety of organic molecules, the (statistical) accuracy of these excitation energies is ≈ 0.3 eV (at the ground-state geometry), good enough to make the method useful in many chemistry applications. However, the high level of accuracy exhibited by TD-DFT in small-molecule calculations does not always carry over to larger systems. Moreover, the calculation of excited states in very large molecules (those containing hundreds of atoms described at the DFT level), while technically feasible, is neither trivial nor routine. QM/MM methods, with the chromophore as the QM region, are an obvious way to reduce the cost of such calculations, but sometimes very large QM regions are necessary, in order to include explicit QM solvent molecules, or to incorporate electronic coupling between multiple chromophores. Thus, we are working to extend both the computational feasibility (via improved algorithms) and the accuracy (by means of improved functionals) of large-molecule TD-DFT calculations. Much of this work is tightly integrated with other ongoing projects in our group, including the development of QM/MM methods and continuum solvation models.

Noteworthy Accomplishments

Spurious CT states
Intensity borrowing by a spurious,
low-energy CT state of uracil in water,
computed using TD-PBE0.
LRC-DFT Coulomb operator
Decomposition of the Coulomb
potential for LRC-DFT

In some of our group's early work applying TD-DFT to relatively large molecules and (especially) clusters or other systems with explicit solvent molecules, we pointed out that the appearance of spurious, low-energy charge-transfer (CT) excited states is practically inevitable in large systems, when standard functionals such as BLYP, B3LYP, or PBE0 are employed. The density of spurious states grows with the size of the molecule or cluster, and these spurious states can (anomalously) borrow intensity from the real bright states. This problem is substantially ameliorated by the use of "long-range corrected" (LRC) density functionals, also known as "range-separated hybrid" functionals. The LRC approach partitions the electron-electron Coulomb operator into long- and short-range components, treating the latter with density-functional exchange, while using 100% Hartree-Fock exchange for the long-range component. These functionals have the correct asymptotic distance dependence for a charge-separated state, and thereby eliminate spurious CT states, at least in the low-energy UV region of the spectrum. As an example, we have used TD-LRC-DFT to compute the electronic absorption spectrum of the "aqueous electron", using a QM/MM model with a large QM region, and we find that the calculated spectrum is in near-quantitative agreement with experiment. Such a calculation would not have been possible using standard functionals (e.g., TD-B3LYP), because a plethora of spurious CT states would have made it prohibitively expensive to calculate enough real excited states to obtain the high-energy tail in this spectrum.

TD-LRC-DFT absorption spectrum for e-(aq)
TD-LRC-DFT absorption spectrum
for the aqueous electron

A variety of LRC functionals have been developed by several different research groups. We have incorporated functionals based on a short-range version of Becke '88 exchange ("μB88") and two different short-range versions of Perdew-Burke-Ernzerhof exchange ("μPBE" and "ωPBE") into Q-Chem, and have characterized their accuracy for both ground-state properties and vertical excitation energies. An important observation in these studies is that the optimal parameters for ground-state thermochemistry are often quite different from those that afford the best excitation energies, and to obtain a robust functional one must carefully balance the parameterization. The statistically-best LRC-ωPBE and LRC-ωPBEh functionals that emerged from our work are the first functionals to achieve comparable accuracy (≈ 0.3 eV) for both localized valence excitations (e.g., nπ*, ππ*, or πσ* states) as well as CT excitations. At the same time, the accuracy of ground-state predictions is not significantly degraded.

RMSE
PBE0 LRC-ωPBEh
Atomization energies (kcal/mol) 7.3 6.6
Reaction barrier heights (kcal/mol) 6.0 4.7
Electron affinities (kcal/mol) 2.9 2.5
Ionization energies (kcal/mol) 3.9 4.7
Localized excitation energies (eV) 0.3 0.3
Charge-transfer excitation energies (eV) 3.0 0.3
Benchmark comparison of PBE0 and LRC-ωPBEh.

Conical intersections in H3
Conical intersections around the D3h degenerate ground state geometry of H3,
computed using restricted open-shell CIS versus spin-flip CIS.
Recently, we have derived and implemented analytic derivative couplings for the "spin-flip" variants of configuration interaction singles (CIS) and TD-DFT. These are the couplings between excited states that are needed for nonadiabatic molecular dynamics simulations, and our algorithm allows one to perform such calculations, and to locate minimum-energy crossing points along conical seams and minimum-energy photochemical pathways, at DFT cost. Unlike derivative couplings computing using ordinary (spin-conserving) CIS or TD-DFT, spin-flip methods correctly describe the topology of the two-dimensional branching space around a conical intersection. (The example shown here is the conical intersection corresponding to the Jahn-Teller effect in the H3 radical; in this particular case, there is a symmetry-required ground-state degeneracy, and a conical intersection that lifts the degeneracy, when the molecule has D3h symmetry.) While H3 is only a toy model, these methods are actively being used to understand excited-state dynamics in complex systems such as DNA.

Representative Publications


This material is based upon work supported by the National Science Foundation under grant nos. CHE-0748448 and CHE-1300603, and by the Dept. of Energy (Award No. DE-SC0008550). Some calculations were performed at the Ohio Supercomputer Center, under Project Nos. PAS0291 and PAA0003. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding agencies.

Last modified February 20, 2016. Proudly powered by Words. By which we mean, hand-written HTML.

This page best viewed with a browser