Automation of Active Space Selection for Multireference Methods via Machine Learning on Chemical Bond Dissociation.

Predicting and understanding the chemical bond is one of the major challenges of computational quantum chemistry. Kohn-Sham density functional theory (KS-DFT) is the most common method, but approximate density functionals may not be able to describe systems where multiple electronic configurations are equally important. Multiconfigurational wave functions, on the other hand, can provide a detailed understanding of the electronic structures and chemical bonds of such systems. In the complete active space self-consistent field (CASSCF) method, one performs a full configuration interaction calculation in an active space consisting of active electrons and active orbitals. However, CASSCF and its variants require the selection of these active spaces. This choice is not black box; it requires significant experience and testing by the user, and thus active space methods are not considered particularly user-friendly and are employed only by a minority of quantum chemists. Our goal is to popularize these methods by making it easier to make good active space choices. We present a machine learning protocol that performs an automated selection of active spaces for chemical bond dissociation calculations of main group diatomic molecules. The protocol shows high prediction performance for a given target system as long as a properly correlated system is chosen for training. Good active spaces are correctly predicted with a considerably better success rate than random guess (larger than 80% precision for most systems studied). Our automated machine learning protocol shows that a "black-box" mode is possible for facilitating and accelerating the large-scale calculations on multireference systems where single-reference methods such as KS-DFT cannot be applied.

Transition Metal Spin-State Energetics by MC-PDFT with High Local Exchange.

The energetics of the spin states of transition metal complexes have been explored with a variety of electronic structure methods, but the calculations require a compromise between accuracy and affordability. In this work, the spin splittings of several iron complexes are studied with multiconfiguration pair-density functional theory (MC-PDFT). The results are compared to previously published results obtained by complete active space second-order perturbation theory (CASPT2) and CASPT2 with coupled-cluster semicore correlation (CASPT2/CC). In contrast to CASPT2's systematic overstabilization of high-spin states with respect to the CASPT2/CC reference, MC-PDFT with the tPBE on-top functional understabilizes high-spin states. This systematic understabilization is largely corrected by revising the exchange and correlation contributions to the on-top functional using the high local-exchange approximation (tPBE-HLE). Moreover, tPBE-HLE correctly predicts the spin of the ground state in most cases, while CASPT2 incorrectly predicts high-spin ground states in all cases. This is encouraging for practical work because tPBE and tPBE-HLE are faster than CASPT2 by a factor of 50 even in a moderately sized example.

MC-PDFT can calculate singlet-triplet splittings of organic diradicals.

The singlet-triplet splittings of a set of diradical organic molecules are calculated using multiconfiguration pair-density functional theory (MC-PDFT), and the results are compared with those obtained by Kohn-Sham density functional theory (KS-DFT) and complete active space second-order perturbation theory (CASPT2) calculations. We found that MC-PDFT, even with small and systematically defined active spaces, is competitive in accuracy with CASPT2, and it yields results with greater accuracy and precision than Kohn-Sham DFT with the parent functional. MC-PDFT also avoids the challenges associated with spin contamination in KS-DFT. It is also shown that MC-PDFT is much less computationally expensive than CASPT2 when applied to larger active spaces, and this illustrates the promise of this method for larger diradical organic systems.

Systematic design of active spaces for multi-reference calculations of singlet-triplet gaps of organic diradicals, with benchmarks against doubly electron-attached coupled-cluster data.

Singlet-triplet gaps in diradical organic π-systems are of interest in many applications. In this study, we calculate them in a series of molecules, including cyclobutadiene and its derivatives and cyclopentadienyl cation, by using correlated participating orbitals within the complete active space (CAS) and restricted active space (RAS) self-consistent field frameworks, followed by second-order perturbation theory (CASPT2 and RASPT2). These calculations are evaluated by comparison with the results of doubly electron-attached (DEA) equation-of-motion (EOM) coupled-cluster (CC) calculations with up to 4-particle-2-hole (4p-2h) excitations. We find active spaces that can accurately reproduce the DEA-EOMCC(4p-2h) data while being small enough to be applicable to larger organic diradicals.

Origin of the Strong Interaction between Polar Molecules and Copper(II) Paddle-Wheels in Metal Organic Frameworks.

The copper paddle-wheel is the building unit of many metal organic frameworks. Because of the ability of the copper cations to attract polar molecules, copper paddle-wheels are promising for carbon dioxide adsorption and separation. They have therefore been studied extensively, both experimentally and computationally. In this work we investigate the copper-CO2 interaction in HKUST-1 and in two different cluster models of HKUST-1: monocopper Cu(formate)2 and dicopper Cu2(formate)4. We show that density functional theory methods severely underestimate the interaction energy between copper paddle-wheels and CO2, even including corrections for the dispersion forces. In contrast, a multireference wave function followed by perturbation theory to second order using the CASPT2 method correctly describes this interaction. The restricted open-shell Møller-Plesset 2 method (ROS-MP2, equivalent to (2,2) CASPT2) was also found to be adequate in describing the system and used to develop a novel force field. Our parametrization is able to predict the experimental CO2 adsorption isotherms in HKUST-1, and it is shown to be transferable to other copper paddle-wheel systems.

Systematic Expansion of Active Spaces beyond the CASSCF Limit: A GASSCF/SplitGAS Benchmark Study.

The applicability and accuracy of the generalized active space self-consistent field, (GASSCF), and (SplitGAS) methods are presented. The GASSCF method enables the exploration of larger active spaces than with the conventional complete active space SCF, (CASSCF), by fragmentation of a large space into subspaces and by controlling the interspace excitations. In the SplitGAS method, the GAS configuration interaction, CI, expansion is further partitioned in two parts: the principal, which includes the most important configuration state functions, and an extended, containing less relevant but not negligible ones. An effective Hamiltonian is then generated, with the extended part acting as a perturbation to the principal space. Excitation energies of ozone, furan, pyrrole, nickel dioxide, and copper tetrachloride dianion are reported. Various partitioning schemes of the GASSCF and SplitGAS CI expansions are considered and compared with the complete active space followed by second-order perturbation theory, (CASPT2), and multireference CI method, (MRCI), or available experimental data. General guidelines for the optimum applicability of these methods are discussed together with their current limitations.

Design of a metal-organic framework with enhanced back bonding for separation of N2 and CH4.

Gas separations with porous materials are economically important and provide a unique challenge to fundamental materials design, as adsorbent properties can be altered to achieve selective gas adsorption. Metal-organic frameworks represent a rapidly expanding new class of porous adsorbents with a large range of possibilities for designing materials with desired functionalities. Given the large number of possible framework structures, quantum mechanical computations can provide useful guidance in prioritizing the synthesis of the most useful materials for a given application. Here, we show that such calculations can predict a new metal-organic framework of potential utility for separation of dinitrogen from methane, a particularly challenging separation of critical value for utilizing natural gas. An open V(II) site incorporated into a metal-organic framework can provide a material with a considerably higher enthalpy of adsorption for dinitrogen than for methane, based on strong selective back bonding with the former but not the latter.