Eric Neuscamman

Multireference quantum chemistry through a joint density matrix renormalization group and canonical transformation theory

We describe the joint application of the density matrix renormalization group and canonical transformation theory to multireference quantum chemistry. The density matrix renormalization group provides the ability to describe static correlation in large active spaces, while the canonical transformation theory provides a high-order description of the dynamic correlation effects. We demonstrate the joint theory in two benchmark systems designed to test the dynamic and static correlation capabilities of the methods, namely, (i) total correlation energies in long polyenes and (ii) the isomerization curve of the [Cu(2)O(2)](2+) core. The largest complete active spaces and atomic orbital basis sets treated by the joint DMRG-CT theory in these systems correspond to a (24e,24o) active space and 268 atomic orbitals in the polyenes and a (28e,32o) active space and 278 atomic orbitals in [Cu(2)O(2)](2+).

Quadratic canonical transformation theory and higher order density matrices.

Canonical transformation (CT) theory provides a rigorously size-extensive description of dynamic correlation in multireference systems, with an accuracy superior to and cost scaling lower than complete active space second order perturbation theory. Here we expand our previous theory by investigating (i) a commutator approximation that is applied at quadratic, as opposed to linear, order in the effective Hamiltonian, and (ii) incorporation of the three-body reduced density matrix in the operator and density matrix decompositions. The quadratic commutator approximation improves CTs accuracy when used with a single-determinant reference, repairing the previous formal disadvantage of the single-reference linear CT theory relative to singles and doubles coupled cluster theory. Calculations on the BH and HF binding curves confirm this improvement. In multireference systems, the three-body reduced density matrix increases the overall accuracy of the CT theory. Tests on the H(2)O and N(2) binding curves yield results highly competitive with expensive state-of-the-art multireference methods, such as the multireference Davidson-corrected configuration interaction (MRCI+Q), averaged coupled pair functional, and averaged quadratic coupled cluster theories.

Strongly contracted canonical transformation theory.

Canonical transformation (CT) theory describes dynamic correlation in multireference systems with large active spaces. Here we discuss CT theorys intruder state problem and why our previous approach of overlap matrix truncation becomes infeasible for sufficiently large active spaces. We propose the use of strongly and weakly contracted excitation operators as alternatives for dealing with intruder states in CT theory. The performance of these operators is evaluated for the H(2)O, N(2), and NiO molecules, with comparisons made to complete active space second order perturbation theory and Davidson-corrected multireference configuration interaction theory. Finally, using a combination of strongly contracted CT theory and orbital-optimized density matrix renormalization group theory, we evaluate the singlet-triplet gap of free base porphin using an active space containing all 24 out-of-plane 2p orbitals. Modeling dynamic correlation with an active space of this size is currently only possible using CT theory.

An Introduction to the Density Matrix Renormalization Group Ansatz in Quantum Chemistry

The Density Matrix Renormalisation Group (DMRG) is an electronic structure method that has recently been applied to ab-initio quantum chemistry. Even at this early stage, it has enabled the solution of many problems that would previously have been intractable with any other method, in particular, multireference problems with very large active spaces. Historically, the DMRG was not originally formulated from a wavefunction perspective, but rather in a Renormalisation Group (RG) language. However, it is now realised that a wavefunction view of the DMRG provides a more convenient, and in some cases more powerful, paradigm. Here we provide an expository introduction to the DMRG ansatz in the context of quantum chemistry.

A review of canonical transformation theory

Canonical transformation (CT) theory targets the description of dynamic correlation in multireference quantum chemistry problems. When combined with a static correlation quantum chemistry method, it enables the quantitative description of chemical processes involving electronic structure not described by a single electronic configuration. We argue that many multireference dynamic correlation methods display unsatisfactory characteristics, including lack of size-consistency, a low-order treatment of correlation, and a poor computational scaling. By contrast, CT theory is based on an exponential ansatz that is rigorously size-consistent, reduces in a single-reference limit to a coupled cluster theory, and has an n 6 computational scaling with system and active space size. The efficient formulation of CT theory has allowed it to be applied to difficult systems in conjunction with active spaces with more than 30 orbitals, beyond the reach of traditional methods, with an accuracy that far exceeds multireference perturbation theories. Here we review the basic motivation, formulation, and implementation of CT theory, as well as survey some of our recent applications and possible future directions.

Size consistency error in the antisymmetric geminal power wave function can be completely removed.

The accurate but expensive product of geminals ansatz may be approximated by a geminal power, but this approach sacrifices size consistency. Here we show both analytically and numerically that a size consistent form very similar to the product of geminals can be recovered using a network of location specific Jastrow factors. Upon variational energy minimization, the network creates particle number projections that remove the charge fluctuations responsible for size inconsistency. This polynomial cost approach captures strong many-electron correlations, giving a maximum error of just 1.8 kcal/mol during the double-bond dissociation of H2O in an STO-3G atomic orbital basis.

A study of cumulant approximations to n -electron valence multireference perturbation theory

We investigate the possibility of reducing the complexity of multireference perturbation theory through cumulant based approximations to the high-order density matrices that appear in such theories. Our test cases show that while the cumulant approximated forms are degraded in accuracy relative to the parent theory and exhibit intruder state problems that must be carefully handled, they may provide a route to a simple estimation of dynamic correlation when the parent perturbation theory is infeasible. Nonetheless, further work is clearly needed on better approximations to the denominators in the perturbation theory.

Optimizing large parameter sets in variational quantum Monte Carlo

We present a technique for optimizing hundreds of thousands of variational parameters in variational quantum Monte Carlo. By introducing iterative Krylov subspace solvers and by multiplying by the Hamiltonian and overlap matrices as they are sampled, we remove the need to construct and store these matrices and thus bypass the most expensive steps of the stochastic reconfiguration and linear method optimization techniques. We demonstrate the effectiveness of this approach by using stochastic reconfiguration to optimize a correlator product state wave function with a Pfaffian reference for four example systems. In two examples on the two dimensional Fermionic Hubbard model, we study 16 and 64 site lattices, recovering energies accurate to 1% in the smaller lattice and predicting particle-hole phase separation in the larger. In two examples involving an ab initio Hamiltonian, we investigate the potential energy curve of a symmetrically dissociated 4 × 4 hydrogen lattice as well as the singlet-triplet gap in free base porphin. In the hydrogen system we recover 98% or more of the correlation energy at all geometries, while for porphin we compute the gap in a 24 orbital active space to within 0.02 eV of the exact result. The number of variational parameters in these examples ranges from 4 × 10^3 to 5 × 10^5.

Communication: A Jastrow factor coupled cluster theory for weak and strong electron correlation

We present a Jastrow-factor-inspired variant of coupled cluster theory that accurately describes both weak and strong electron correlation. Compatibility with quantum Monte Carlo allows for variational energy evaluations and an antisymmetric geminal power reference, two features not present in traditional coupled cluster that facilitate a nearly exact description of the strong electron correlations in minimal-basis N2 bond breaking. In double-ζ treatments of the HF and H2O bond dissociations, where both weak and strong correlations are important, this polynomial cost method proves more accurate than either traditional coupled cluster or complete active space perturbation theory. These preliminary successes suggest a deep connection between the ways in which cluster operators and Jastrow factors encode correlation.

The Jastrow antisymmetric geminal power in Hilbert space: theory, benchmarking, and application to a novel transition state.

The Jastrow-modified antisymmetric geminal power (JAGP) ansatz in Hilbert space successfully overcomes two key failings of other pairing theories, namely, a lack of inter-pair correlations and a lack of multiple resonance structures, while maintaining a polynomially scaling cost, variational energies, and size consistency. Here, we present efficient quantum Monte Carlo algorithms that evaluate and optimize the JAGP energy for a cost that scales as the fifth power of the system size. We demonstrate the JAGPs ability to describe both static and dynamic correlation by applying it to bond stretching in H2O, C2, and N2 as well as to a novel, multi-reference transition state of ethene. JAGPs accuracy in these systems outperforms even the most sophisticated single-reference methods and approaches that of exponentially scaling active space methods.