Trial Wave Function Research Articles

Artificial Neural Networks as Trial Wave Functions for Quantum Monte Carlo

AbstractInspired by the universal approximation theorem and widespread adoption of artificial neural network techniques in a diversity of fields, feed‐forward neural networks are proposed as a general purpose trial wave function for quantum Monte Carlo simulations of continuous many‐body systems. Whereas for simple model systems the whole many‐body wave function can be represented by a neural network, the antisymmetry condition of non‐trivial fermionic systems is incorporated by means of a Slater determinant. To demonstrate the accuracy of the trial wave functions, an exactly solvable model system of two trapped interacting particles, as well as the hydrogen dimer, is studied.

Pseudo-BCS wave function from density matrix decomposition: Application in auxiliary-field quantum Monte Carlo

We present a method to construct pseudo-BCS wave functions from the one-body density matrix. The resulting many-body wave function, which can be produced for any fermion systems, including those with purely repulsive interactions, has the form of a number-projected BCS form, or antisymmetrized geminal power (AGP). Such wave functions provide a better ansatz for correlated fermion systems than a single Slater determinant, and often better than a linear combination of Slater determinants (for example from a truncated active space calculation). We describe a procedure to build such a wave function conveniently from a given reduced density matrix of the system, rather than from a mean-field solution (which gives a Slater determinant for repulsive interactions). The pseudo-BCS wave function thus obtained reproduces the density matrix or minimizes the difference between the input and resulting density matrices. One application of the pseudo-BCS wave function is in auxiliary-field quantum Monte Carlo (AFQMC) calculations as the trial wave function to control the sign/phase problem. AFQMC is often among the most accurate general methods for correlated fermion systems. We show that the pseudo-BCS form further reduces the constraint bias and leads to improved accuracy compared to the usual Slater determinant trial wave functions, using the two-dimensional Hubbard model as an example. Furthermore, the pseudo-BCS trial wave function allows a new systematically improvable self-consistent approach, with pseudo-BCS trial wave function iteratively generated by AFQMC via the one-body density matrix.Received 19 October 2020Revised 4 January 2021Accepted 6 January 2021DOI:https://doi.org/10.1103/PhysRevResearch.3.013065Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.Published by the American Physical SocietyPhysics Subject Headings (PhySH)TechniquesApproximation methods for many-body systemsBCS theoryHubbard modelQuantum Monte CarloCondensed Matter, Materials & Applied Physics

Diffusion Monte Carlo method on small boron clusters using single- and multi- determinant-Jastrow trial wavefunctions.

Multireference character in some small boron clusters could be significant, and a previous all-electron fixed-node diffusion quantum Monte Carlo (FN-DMC) calculation with the single-determinant-Jastrow (SDJ) trial wavefunction shows that the atomization energy (AE) of B4 + is overestimated by about 1.4 eV compared with the coupled cluster method with single, doubles, and perturbative triples [CCSD(T)] results. All-electron FN-DMC calculations and those with the pseudopotential (PP) using SDJ and multi-determinant-Jastrow (MDJ) trial wavefunctions with B3LYP orbitals as well as CC calculations at different levels are carried out on Bn Q (n = 1-5, Q = -1, 0, 1) clusters. The obtained FN-DMC energies indicate that the node error of the employed SDJ trial wavefunction in all-electron calculations is different from that with the PP for some clusters. The error of AEs and dissociation energies (DEs) from all-electron FN-DMC calculations is larger than that with the PP when the SDJ trial wavefunction is employed, while errors of CC methods do not depend on whether the PP is used. AEs and DEs of the boron clusters are improved significantly when MDJ trial wavefunctions are used in both all-electron calculations and those with the PP, and their error is similar to that of CCSD(T) compared with CCSDT(Q) results. On the other hand, reasonable adiabatic electron detachment energies (ADEs) and ionization potentials (AIPs) are achieved with FN-DMC using SDJ trial wavefunctions and MDJ is less effective on ADEs and AIPs. Furthermore, the relative energy between two structures of B9 - is predicted reliably with FN-DMC using the SDJ trial wavefunction and the effect of MDJ is negligible, while density functional theory results using different exchange-correlation functionals differ significantly.

Some recent developments in auxiliary-field quantum Monte Carlo for real materials.

The auxiliary-field quantum Monte Carlo (AFQMC) method is a general numerical method for correlated many-electron systems, which is being increasingly applied in lattice models, atoms, molecules, and solids. Here, we introduce the theory and algorithm of the method specialized for real materials and present several recent developments. We give a systematic exposition of the key steps of AFQMC, closely tracking the framework of a modern software library we are developing. The building of a Monte Carlo Hamiltonian, projecting to the ground state, sampling two-body operators, phaseless approximation, and measuring ground state properties are discussed in detail. An advanced implementation for multi-determinant trial wave functions is described, which dramatically speeds up the algorithm and reduces the memory cost. We propose a self-consistent constraint for real materials, and discuss two flavors for its realization, either by coupling the AFQMC calculation to an effective independent-electron calculation or via the natural orbitals of the computed one-body density matrix.

Ground State, Isoelectronic Ions and Low-Lying Excited States of Lithium Atom in Strong Magnetic Field

In the framework of the variational Monte Carlo method, the ground states of the lithium atom and lithium like ions up to Z = 10 in an external strong magnetic field are evaluated. Furthermore, the two low-lying excited states , and of the lithium atom in strong magnetic field are also investigated. Simple trial wave functions for lithium are used.

Interaction between iron ion and dipole carbon monoxide inside spherical cavities

The interactions that occur within hemeproteins are important for biological systems and they are of interest for understanding living systems. In this way, it is important to know the vibrational and electrostatic interactions in this system. In this article, a study is made using a new approach to describe the interaction between iron ions and carbon monoxide inside spherical cavities that mimic volumes of protein cavities in three different media (vacuum, water and ice). We use an alternative trial wavefunction as an ansatz in the Variational Method for the calculation of the energy for a confined ion–dipole system. This trial function is inspired by Supersymmetric Quantum Mechanics. One of the results obtained is the value of the ground state energy of this interaction in a vacuum inside a spherical cavity of radius approximately equal to 12 Bohr radius obtained by the Variational Method. This result is compared with the energy value obtained by the second order Moller–Plesset perturbative method and there is a difference of approximately 1.9 10 $$^{-3}$$ hartree (3.87 $$ \%$$ ).

The role of high-order electron correlation effects in a model system for non-valence correlation-bound anions.

The diffusion Monte Carlo (DMC), auxiliary field quantum Monte Carlo (AFQMC), and equation-of-motion coupled cluster (EOM-CC) methods are used to calculate the electron binding energy (EBE) of the non-valence anion state of a model (H2O)4 cluster. Two geometries are considered, one at which the anion is unbound and the other at which it is bound in the Hartree-Fock (HF) approximation. It is demonstrated that DMC calculations can recover from the use of a HF trial wave function that has collapsed onto a discretized continuum solution, although larger EBEs are obtained when using a trial wave function for the anion that provides a more realistic description of the charge distribution and, hence, of the nodal surface. For the geometry at which the cluster has a non-valence correlation-bound anion, both the inclusion of triples in the EOM-CC method and the inclusion of supplemental diffuse d functions in the basis set are important. DMC calculations with suitable trial wave functions give EBE values in good agreement with our best estimate EOM-CC result. AFQMC using a trial wave function for the anion with a realistic electron density gives a value of the EBE nearly identical to the EOM-CC result when using the same basis set. For the geometry at which the anion is bound in the HF approximation, the inclusion of triple excitations in the EOM-CC calculations is much less important. The best estimate EOM-CC EBE value is in good agreement with the results of DMC calculations with appropriate trial wave functions.

Comparative study of the density matrix embedding theory for Hubbard models

We examine the performance of the density matrix embedding theory (DMET) recently proposed in [G. Knizia and G. K.-L. Chan, Phys. Rev. Lett. 109, 186404 (2012)]. The core of this method is to find a proper one-body potential that generates a good trial wave function for projecting a large scale original Hamiltonian to a local subsystem with a small number of basis. The resultant ground state of the projected Hamiltonian can locally approximate the true ground state. However, the lack of the variational principle makes it difficult to judge the quality of the choice of the potential. Here we focus on the entanglement spectrum (ES) as a judging criterion; accurate evaluation of the ES guarantees that the corresponding reduced density matrix well reproduces all physical quantities on the local subsystem. We apply the DMET to the Hubbard model on the one-dimensional chain, zigzag chain, and triangular lattice and test several variants of potentials and cost functions. It turns out that ES serves as a more sensitive quantity than the energy and double occupancy to probe the quality of the DMET outcomes. A symmetric potential reproduces the ES of the phase that continues from a noninteracting limit. The Mott transition as well as symmetry-breaking transitions can be detected by the singularities in the ES. However, the details of the ES in the strongly interacting parameter region depends much on these variants, meaning that the present DMET algorithm allowing for numerous variant is insufficient to fully characterize the particular phases that require characterization by the ES.

Large-scale simulations of particle-hole-symmetric Pfaffian trial wave functions

We introduce a family of paired-composite-fermion trial wave functions for any odd Cooper-pair angular momentum. These wave functions are parameter-free and can be efficiently projected into the lowest Landau level. We use large-scale Monte Carlo simulations to study three cases: Firstly, the Moore-Read phase, which serves us as a benchmark. Secondly, we explore the pairing associated with the anti-Pfaffian and the particle-hole-symmetric Pfaffian. Specifically, we assess whether their trial states feature exponentially decaying correlations and thus represent gapped phases of matter. For Moore-Read and anti-Pfaffian we find decay lengths of $\xi_\text{Moore-Read}=1.30(5)$ and $\xi_\text{anti-Pfaffian}=1.38(14)$, in units of the magnetic length. By contrast, for the case of PH-Pfaffian, we find no evidence of a finite length scale for up to $56$ particles.

The ground state of the lithium atom in dense plasmas using variational Monte Carlo method

In this paper, the variational quantum Monte Carlo method is applied to investigate the ground state of the lithium atom. Moreover, the energy eigenvalues of the lithium atom in dense plasma are also investigated by using the Debye–Huckel model and the exponential cosine screened Coulomb potential model. The calculations are carried out by using trial wave functions in the form of the Slater determinant wave function multiplied by a correlation function due to the interaction between the electrons. Three types of correlation functions are used—with two, three and four variational parameters—one of which satisfies the well-known cusp conditions. Interesting results are obtained in comparison with results obtained by using other trial wave functions.

Efficient local energy evaluation for multi-Slater wave functions in orbital space quantum Monte Carlo.

We present an algorithm for calculating the local energy of a multi-Slater wave function in orbital space quantum Monte Carlo (QMC). Recent developments in selected configuration interaction methods have led to increased interest in using multi-Slater trial wave functions in various QMC methods. For an ab initio Hamiltonian, our algorithm has a cost scaling of O(n5 + nc), as opposed to the O(n4nc) scaling of existing orbital space algorithms, where n is the system size and nc is the number of configurations in the wave function. We present our method using variational Monte Carlo calculations with the Jastrow multi-Slater wave function, although the formalism should be applicable for auxiliary field QMC. We apply it to polyacetylene and demonstrate the possibility of using a much larger number of configurations than possible using existing methods.

Toward a systematic improvement of the fixed-node approximation in diffusion Monte Carlo for solids-A case study in diamond.

While Diffusion Monte Carlo (DMC) is in principle an exact stochastic method for ab initio electronic structure calculations, in practice, the fermionic sign problem necessitates the use of the fixed-node approximation and trial wavefunctions with approximate nodes (or zeros). This approximation introduces a variational error in the energy that potentially can be tested and systematically improved. Here, we present a computational method that produces trial wavefunctions with systematically improvable nodes for DMC calculations of periodic solids. These trial wavefunctions are efficiently generated with the configuration interaction using a perturbative selection made iteratively (CIPSI) method. A simple protocol in which both exact and approximate results for finite supercells are used to extrapolate to the thermodynamic limit is introduced. This approach is illustrated in the case of the carbon diamond using Slater-Jastrow trial wavefunctions including up to one million Slater determinants. Fixed-node DMC energies obtained with such large expansions are much improved, and the fixed-node error is found to decrease monotonically and smoothly as a function of the number of determinants in the trial wavefunction, a property opening the way to a better control of this error. The cohesive energy extrapolated to the thermodynamic limit is in close agreement with the estimated experimental value. Interestingly, this is also the case at the single-determinant level, thus, indicating a very good error cancellation in carbon diamond between the bulk and atomic total fixed-node energies when using single-determinant nodes.

Fundamental gap of fluorographene by many-body GW and fixed-node diffusion Monte Carlo methods.

Fluorographene (FG) is a promising graphene-derived material with a large bandgap. Currently existing predictions of its fundamental gap (Δf) and optical gap (Δopt) significantly vary when compared with experiment. We provide here an ultimate benchmark of Δf for FG by many-body GW and fixed-node diffusion Monte Carlo (FNDMC) methods. Both approaches independently arrive at Δf ≈ 7.1 ± 0.1 eV. In addition, the Bethe-Salpeter equation enabled us to determine the first exciton binding energy, Eb = 1.92 eV. We also point to the possible misinterpretation problem of the results obtained for gaps of solids by FNDMC with single-reference trial wave functions of Bloch orbitals. We argue why instead of Δopt, in the thermodynamic limit, such an approach results in energy differences that rather correspond to Δf, and we also outline conditions when this case actually applies.

On Diffusion Monte Carlo in spaces with multi-valued maps, boundaries and gradient torsion

Importance sampling in Diffusion Monte Carlo has a long history. However, only recently, simulations of ground state properties have been extended to spaces mapped with non-Cartesian coordinates. We demonstrate that in spaces with nonzero advection the Smoluchowski operator for any nontrivial trial wavefunction does not converge to the exact result. Rather, every drift term is equivalent to some advection in a manifold that contains the physical space of the system. Since these manifolds may be formulated with gradient torsion we demonstrate with several numerical experiments that Diffusion Monte Carlo is possible in these as well.

Taming the fixed-node error in diffusion Monte Carlo via range separation.

By combining density-functional theory (DFT) and wave function theory via the range separation (RS) of the interelectronic Coulomb operator, we obtain accurate fixed-node diffusion Monte Carlo (FN-DMC) energies with compact multi-determinant trial wave functions. In particular, we combine here short-range exchange-correlation functionals with a flavor of selected configuration interaction known as configuration interaction using a perturbative selection made iteratively (CIPSI), a scheme that we label RS-DFT-CIPSI. One of the take-home messages of the present study is that RS-DFT-CIPSI trial wave functions yield lower fixed-node energies with more compact multi-determinant expansions than CIPSI, especially for small basis sets. Indeed, as the CIPSI component of RS-DFT-CIPSI is relieved from describing the short-range part of the correlation hole around the electron-electron coalescence points, the number of determinants in the trial wave function required to reach a given accuracy is significantly reduced as compared to a conventional CIPSI calculation. Importantly, by performing various numerical experiments, we evidence that the RS-DFT scheme essentially plays the role of a simple Jastrow factor by mimicking short-range correlation effects, hence avoiding the burden of performing a stochastic optimization. Considering the 55 atomization energies of the Gaussian-1 benchmark set of molecules, we show that using a fixed value of μ = 0.5 bohr-1 provides effective error cancellations as well as compact trial wave functions, making the present method a good candidate for the accurate description of large chemical systems.

Starting-point-independent quantum Monte Carlo calculations of iron oxide

Quantum Monte Carlo (QMC) methods are useful for studies of strongly correlated materials because they are many body in nature and use the physical Hamiltonian. Typical calculations assume as a starting point a wave function constructed from single-particle orbitals obtained from one-body methods, e.g., density functional theory. However, mean-field-derived wave functions can sometimes lead to systematic QMC biases if the mean-field result poorly describes the true ground state. Here, we study the accuracy and flexibility of QMC trial wave functions using variational and fixed-node diffusion QMC estimates of the total spin density and lattice distortion of antiferromagnetic iron oxide (FeO) in the ground state $B1$ crystal structure. We found that for relatively simple wave functions the predicted lattice distortion was controlled by the choice of single-particle orbitals used to construct the wave function, rather than by subsequent wave function optimization techniques within QMC. By optimizing the orbitals with QMC, we then demonstrate starting-point independence of the trial wave function with respect to the method by which the orbitals were constructed by demonstrating convergence of the energy, spin density, and predicted lattice distortion for two qualitatively different sets of orbitals. The results suggest that orbital optimization is a promising method for accurate many-body calculations of strongly correlated condensed phases.

A first-principles Quantum Monte Carlo study of two-dimensional (2D) GaSe

Two-dimensional (2D) post-transition metal chalcogenides (PTMCs) have attracted attention due to their suitable bandgaps and lower exciton binding energies, making them more appropriate for electronic, optical, and water-splitting devices than graphene and monolayer transition metal dichalcogenides. Of the predicted 2D PTMCs, GaSe has been reliably synthesized and experimentally characterized. Despite this fact, quantities such as lattice parameters and band character vary significantly depending on which density functional theory (DFT) functional is used. Although many-body perturbation theory (GW approximation) has been used to correct the electronic structure and obtain the excited state properties of 2D GaSe, and solving the Bethe-Salpeter equation (BSE) has been used to find the optical gap, we find that the results depend strongly on the starting wavefunction. In an attempt to correct these discrepancies, we employed the many-body Diffusion Monte Carlo (DMC) method to calculate the ground and excited state properties of GaSe because DMC has a weaker dependence on the trial wavefunction. We benchmark these results with available experimental data, DFT [local-density approximation, Perdew-Burke-Ernzerhof (PBE), strongly constrained and appropriately normed (SCAN) meta-GGA, and hybrid (HSE06) functionals] and GW-BSE (using PBE and SCAN wavefunctions) results. Our findings confirm that monolayer GaSe is an indirect gap semiconductor (Γ-M) with a quasiparticle electronic gap in close agreement with experiment and low exciton binding energy. We also benchmark the optimal lattice parameter, cohesive energy, and ground state charge density with DMC and various DFT methods. We aim to present a terminal theoretical benchmark for pristine monolayer GaSe, which will aid in the further study of 2D PTMCs using DMC methods.

Binding and excitations in SixHy molecular systems using quantum Monte Carlo.

We present high-accuracy correlated calculations of small SixHy molecular systems in both the ground and excited states. We employ quantum Monte Carlo (QMC) together with a variety of many-body wave function approaches based on basis set expansions. The calculations are carried out in a valence-only framework using recently derived correlation consistent effective core potentials. Our primary goal is to understand the fixed-node diffusion QMC errors in both the ground and excited states with single-reference trial wave functions. Using a combination of methods, we demonstrate the very high accuracy of the QMC atomization energies being within ≈0.07 eV or better when compared with essentially exact results. By employing proper choices for trial wave functions, we have found that the fixed-node QMC biases for total energies are remarkably uniform ranging between 1% and 3.5% with absolute values at most ≈0.2 eV across the systems and several types of excitations such as singlets and triplets as well as low-lying and Rydberg-like states. Our results further corroborate that Si systems, and presumably also related main group IV and V elements of the periodic table (Ge, Sn, etc), exhibit some of the lowest fixed-node biases found in valence-only electronic structure QMC calculations.

Systematic comparison and cross-validation of fixed-node diffusion Monte Carlo and phaseless auxiliary-field quantum Monte Carlo in solids

Quantum Monte Carlo (QMC) methods are some of the most accurate methods for simulating correlated electronic systems. We investigate the compatibility, strengths and weaknesses of two such methods, namely, diffusion Monte Carlo (DMC) and auxiliary-field quantum Monte Carlo (AFQMC). The multi-determinant trial wave functions employed in both approaches are generated using the configuration interaction using a perturbative selection made iteratively (CIPSI) technique. Complete basis set full configuration interaction (CBS-FCI) energies estimated with CIPSI are used as a reference in this comparative study between DMC and AFQMC. By focusing on a set of canonical finite size solid state systems, we show that both QMC methods can be made to systematically converge towards the same energy once basis set effects and systematic biases have been removed. AFQMC shows a much smaller dependence on the trial wavefunction than DMC while simultaneously exhibiting a much larger basis set dependence. We outline some of the remaining challenges and opportunities for improving these approaches.

Stochastic nodal surfaces in quantum Monte Carlo calculations.

Treating the fermionic ground state-problem as a constrained stochastic optimization problem, a formalism for fermionic quantum Monte Carlo is developed that makes no reference to a trial wave function. Exchange symmetry is enforced by nonlocal terms appearing in the Green's function corresponding to an additional walker propagation channel. Complemented by a treatment of diffusion that encourages the formation of a stochastic nodal surface, we find that an approximate long-range extension of walker cancellations can be employed without introducing significant bias, reducing the number of walkers required for a stable calculation. A proof-of-concept implementation is shown to give a stable fermionic ground state for simple harmonic and atomic systems.