The accurate description of non-ideal quantum many-body systems is of prime importance for a host of applications within physics, quantum chemistry, material science, and related disciplines. At finite temperatures, the gold standard is given by \textit{ab initio} path integral Monte Carlo (PIMC) simulations, which do not require any empirical input, but exhibit an exponential increase in the required compute time for fermionic systems with increasing the system size $N$. Very recently, it has been suggested to compute fermionic properties without this bottleneck based on PIMC simulations of fictitious identical particles. In the present work, we use this technique to carry out very large ($N\leq1000$) PIMC simulations of the warm dense electron gas and demonstrate that it is capable of providing a highly accurate description of investigated properties, i.e., the static structure factor, the static density response function, and local field correction, over the entire range of length scales.
Ab initio Monte Carlo simulations have been performed to determine the equilibrium properties of liquid lithium and lithium clusters at different temperatures. First-principles density-functional methods were employed to calculate the potential-energy change for each proposed change of configuration, which was then accepted or rejected according to the Metropolis Monte Carlo scheme. The resulting structural properties are compared to data from experimental measurements and ab initio molecular dynamics simulations. It is shown that accurate structural information can be obtained with ab initio …
Real-space quantum Monte Carlo is used to calculate the total atomization energy of benzene. In contrast to orbital-space methods common in quantum chemistry, real-space methods allow results at near the complete-basis-set limit to be immediately obtained, all at a computational cost that scales with the fourth power of system size. We demonstrate this utility using the moderately sized benzene molecule, obtaining energies that agree with experiment and previous results from highly accurate composite methods including HEAT and W2. Due to the low scaling of these algorithms, it opens up the pos…
Although liquid water is ubiquitous in chemical reactions at roots of life and climate on the earth, the prediction of its properties by high-level ab initio molecular dynamics simulations still represents a formidable task for quantum chemistry. In this article we present a room temperature simulation of liquid water based on the potential energy surface obtained by a many-body wave function through quantum Monte Carlo (QMC) methods. The simulated properties are in good agreement with recent neutron scattering and X-ray experiments, particularly concerning the position of the oxygen-oxygen pe…
We describe an algorithm to reduce the cost of auxiliary-field quantum Monte Carlo (AFQMC) calculations for the electronic structure problem. The technique uses a nested low-rank factorization of the electron repulsion integral (ERI). While the cost of conventional AFQMC calculations in Gaussian bases scales as $\mathcal{O}(N^4)$ where $N$ is the size of the basis, we show that ground-state energies can be computed through tensor decomposition with reduced memory requirements and sub-quartic scaling. The algorithm is applied to hydrogen chains and square grids, water clusters, and hexagonal BN…
We present an \textit{ab initio} auxiliary field quantum Monte Carlo method for studying the electronic structure of molecules, solids, and model Hamiltonians at finite temperature. The algorithm marries the \textit{ab initio} phaseless auxiliary field quantum Monte Carlo algorithm known to produce high accuracy ground state energies of molecules and solids with its finite temperature variant, long used by condensed matter physicists for studying model Hamiltonian phase diagrams, to yield a phaseless, \textit{ab initio} finite temperature method. We demonstrate that the method produces interna…
Ab initio path-integral Monte Carlo results for the one-particle spectral function of the warm dense electron gas
Helmholtz-Zentrum Dresden-Rossendorf (HZDR)（德累斯顿-罗斯多夫亥姆霍兹中心）；Universität Rostock（罗斯托克大学）；Christian-Albrechts-Universität zu Kiel（基尔克里斯蒂安-阿尔布雷希特大学）；Center for Advanced Systems Understanding (CASUS)（先进系统理解中心）
In this article, we propose an efficient method for sampling the relevant state space in condensed phase reactions. In the present method, the reaction is described by solving the electronic Schrödinger equation for the solute atoms in the presence of explicit solvent molecules. The sampling algorithm uses a molecular mechanics guiding potential in combination with simulated tempering ideas and allows thorough exploration of the solvent state space in the context of an ab initio calculation even when the dielectric relaxation time of the solvent is long. The method is applied to the study of t…
We present a Nested Markov chain Monte Carlo (NMC) scheme for building equilibrium averages based on accurate potentials such as density functional theory. Metropolis sampling of a reference system, defined by an inexpensive but approximate potential, was used to substantially decorrelate configurations at which the potential of interest was evaluated, thereby dramatically reducing the number needed to build ensemble averages at a given level of precision. The efficiency of this procedure was maximized on-the-fly through variation of the reference system thermodynamic state (characterized here…
We study the efficiency, precision and accuracy of all-electron variational and diffusion quantum Monte Carlo calculations using Slater basis sets. Starting from wave functions generated by Hartree-Fock and density functional theory, we describe an algorithm to enforce the electron-nucleus cusp condition by linear projection. For the 55 molecules in the G2 set, the diffusion quantum Monte Carlo calculations recovers an average of 95% of the correlation energy and reproduces bond energies to a mean absolute deviation of 3.2 kcal/mol. Comparing the individual total energies with essentially exac…
QMCPACK is an open source quantum Monte Carlo package for ab-initio electronic structure calculations. It supports calculations of metallic and insulating solids, molecules, atoms, and some model Hamiltonians. Implemented real space quantum Monte Carlo algorithms include variational, diffusion, and reptation Monte Carlo. QMCPACK uses Slater-Jastrow type trial wave functions in conjunction with a sophisticated optimizer capable of optimizing tens of thousands of parameters. The orbital space auxiliary field quantum Monte Carlo method is also implemented, enabling cross validation between differ…
We here apply the recently developed initiator density matrix quantum Monte Carlo (i-DMQMC) to a wide range of chemical environments using atoms and molecules in vacuum. i-DMQMC samples the exact density matrix of a Hamiltonian at finite temperature and combines the accuracy of full configuration interaction quantum Monte Carlo (FCIQMC) - full configuration interaction (FCI) or exact energies in a finite basis set - with finite temperature. By way of exploring the applicability of i-DMQMC for molecular systems, we choose to study a recently developed test set by Rubenstein and coworkers: Be, H…
TurboRVB is a computational package for {\it ab initio} Quantum Monte Carlo (QMC) simulations of both molecular and bulk electronic systems. The code implements two types of well established QMC algorithms: Variational Monte Carlo (VMC), and Diffusion Monte Carlo in its robust and efficient lattice regularized variant. A key feature of the code is the possibility of using strongly correlated many-body wave functions. The electronic wave function (WF) is obtained by applying a Jastrow factor, which takes into account dynamical correlations, to the most general mean-field ground state, written e…
The auxiliary-field quantum Monte Carlo (AFQMC) method provides a computational framework for solving the time-independent Schroedinger equation in atoms, molecules, solids, and a variety of model systems. AFQMC has recently witnessed remarkable growth, especially as a tool for electronic structure computations in real materials. The method has demonstrated excellent accuracy across a variety of correlated electron systems. Taking the form of stochastic evolution in a manifold of non-orthogonal Slater determinants, the method resembles an ensemble of density-functional theory (DFT) calculation…
Nanomaterials made of the cerium oxides CeO$_2$ and Ce$_2$O$_3$ have a broad range of applications, from catalysts in automotive, industrial or energy operations to promising materials to enhance hadrontherapy effectiveness in oncological treatments. To elucidate the physico-chemical mechanisms involved in these processes, it is of paramount importance to know the electronic excitation spectra of these oxides, which are obtained here through high-accuracy linear-response time-dependent density functional theory calculations. In particular, the macroscopic dielectric response functions $\barε$ …
In this letter we report the singlet ground state structure of the full carotenoid peridinin by means of variational Monte Carlo (VMC) calculations. The VMC relaxed geometry has an average bond length alternation of 0.1165(10) Å, larger than the values obtained by DFT (PBE, B3LYP and CAM-B3LYP) and shorter than that calculated at the Hartree-Fock (HF) level. TDDFT and EOM-CCSD calculations on a reduced peridinin model confirm the HOMO-LUMO major contribution of the Bu+-like (S2) bright excited state. Many Body Green's Function Theory (MBGFT) calculations of the vertical excitation energy of th…
A novel method for simulating the statistical mechanics of molecular systems in which both nuclear and electronic degrees of freedom are treated quantum mechanically is presented. The scheme combines a path integral description of the nuclear variables with a first-principles adiabatic description of the electronic structure. The electronic problem is solved for the ground state within a density functional approach, with the electronic orbitals expanded in a localized (Gaussian) basis set. The discretized path integral is computed by a Metropolis Monte Carlo sampling technique on the normal mo…
Enabling ab initio geometry optimization of strongly correlated systems with transferable deep quantum Monte Carlo
FU Berlin, Department of Mathematics and Computer Science（柏林自由大学数学与计算机科学系）；Microsoft Research AI4Science（微软研究院AI4Science）；FU Berlin, Department of Physics（柏林自由大学物理系）；Rice University, Department of Chemistry（莱斯大学化学系）
Luminescent solar concentrators (LSCs) are devices that trap a portion of the solar spectrum and funnel it towards photon harvesting devices. The modelling of LSCs at a quantum chemical level however, remains a challenge due to the complexity of exciton and photon dynamic modelling. This study examines singlet exciton dynamics occurring within a typical LSC device. To do this, we use a rejection-free kinetic Monte Carlo method to predict diffusion lengths, diffusion coefficients, substrate anisotropy, and average exciton lifetimes of perylene diimide (PDI) and tetracene based substrates in the…
We propose a streamlined combination scheme of the transcorrelation (TC) and coupled cluster (CC) theory, which not only increases the convergence rate with respect to the basis set, but also extends the applicability of the lowest order CC approximations to strongly correlated regimes in the three dimensional uniform electron gas (3D UEG). With the correct physical insights built into the correlator used in TC, highly accurate ground state energies with errors $\leq 0.001 $ a.u./electron relative to the state-of-the-art quantum Monte Carlo results can be obtained across a wide range of densit…
Recently, it has been shown that neural networks not only approximate the ground-state wave functions of a single molecular system well but can also generalize to multiple geometries. While such generalization significantly speeds up training, each energy evaluation still requires Monte Carlo integration which limits the evaluation to a few geometries. In this work, we address the inference shortcomings by proposing the Potential learning from ab-initio Networks (PlaNet) framework, in which we simultaneously train a surrogate model in addition to the neural wave function. At inference time, th…
We investigate the static and dynamic behaviors of a Br adlayer electrochemically deposited onto single-crystal Ag(100) using an off-lattice model of the adlayer. Unlike previous studies using a lattice-gas model, the off-lattice model allows adparticles to be located at any position within a two-dimensional approximation to the substrate. Interactions with the substrate are approximated by a corrugation potential. Using Density Functional Theory (DFT) to calculate surface binding energies, a sinusoidal approximation to the corrugation potential is constructed. A variety of techniques, includi…
In order to find the equilibrium geometries of molecules and solids and to perform ab initio molecular dynamics, it is necessary to calculate the forces on the nuclei. We present a correlated sampling method to efficiently calculate numerical forces and potential energy surfaces in diffusion Monte Carlo. It employs a novel coordinate transformation, earlier used in variational Monte Carlo, to greatly reduce the statistical error. Results are presented for first-row diatomic molecules.
Accurate ab initio calculations are of fundamental importance in physics, chemistry, biology, and materials science, which have witnessed rapid development in the last couple of years with the help of machine learning computational techniques such as neural networks. Most of the recent efforts applying neural networks to ab initio calculation have been focusing on the energy of the system. In this study, we take a step forward and look at the interatomic force obtained with neural network wavefunction methods by implementing and testing several commonly used force estimators in variational qua…
Given access to accurate solutions of the many-electron Schrödinger equation, nearly all chemistry could be derived from first principles. Exact wavefunctions of interesting chemical systems are out of reach because they are NP-hard to compute in general, but approximations can be found using polynomially-scaling algorithms. The key challenge for many of these algorithms is the choice of wavefunction approximation, or Ansatz, which must trade off between efficiency and accuracy. Neural networks have shown impressive power as accurate practical function approximators and promise as a compact wa…
Modelling the inelastic scattering of electrons in water is fundamental, given their crucial role in biological damage. In Monte Carlo track-structure codes used to assess biological damage, the energy loss function, from which cross sections are extracted, is derived from different semi-empirical optical models. Only recently, first ab-initio results for the energy loss function and cross-sections in water became available. For benchmarking purpose, in this work, we present ab-initio linear-response time-dependent density functional theory calculations of the energy loss function of liquid wa…
Diffusion Monte Carlo (DMC) is one of the most accurate techniques available for calculating the electronic properties of molecules and materials, yet it often remains a challenge to economically compute forces using this technique. As a result, ab initio molecular dynamics simulations and geometry optimizations that employ Diffusion Monte Carlo forces are often out of reach. One potential approach for accelerating the computation of "DMC forces" is to machine learn these forces from DMC energy calculations. In this work, we employ Behler-Parrinello Neural Networks to learn DMC forces from DMC…
In the free energy of hydration of a solute, the chemical contribution is given by the free energy required to expel water molecules from the coordination sphere and the packing contribution is given by the free energy required to create the solute-free coordination sphere (the observation volume) in bulk water. With the SPC/E water model as a reference, we examine the chemical and packing contributions in the free energy of water simulated using different electron density functionals. The density is fixed at a value corresponding to that for SPC/E water at a pressure of 1 bar. The chemical co…
Computer simulation methods, such as Monte Carlo or Molecular Dynamics, are very powerful computational techniques that provide detailed and essentially exact information on classical many-body problems. With the advent of ab-initio molecular dynamics, where the forces are computed on-the-fly by accurate electronic structure calculations, the scope of either method has been greatly extended. This new approach, which unifies Newton's and Schrödinger's equations, allows for complex simulations without relying on any adjustable parameter. This review is intended to outline the basic principles as…
