Fact-checked by Grok 2 weeks ago

Quantum Monte Carlo

Quantum Monte Carlo (QMC) encompasses a family of stochastic simulation techniques that apply Monte Carlo methods to quantum mechanical problems, transforming quantum degrees of freedom into classical configurations for probabilistic sampling to compute properties like ground-state energies and correlation functions with high accuracy.^[1] These methods address the many-body Schrödinger equation by using random walks to sample distributions derived from the system's Hamiltonian, enabling ab initio calculations for complex systems where deterministic approaches fail due to exponential scaling with particle number.^[2] Key variants include variational Monte Carlo (VMC), which evaluates expectation values using a trial wave function sampled from its squared modulus to provide an upper bound to the ground-state energy; diffusion Monte Carlo (DMC), which projects the ground state via imaginary-time evolution simulated as branching random walks, often with the fixed-node approximation to handle fermionic antisymmetry; and path integral Monte Carlo (PIMC), which maps finite-temperature quantum statistical mechanics to classical polymer chains for studying thermal properties.^[2]^[3] QMC excels in providing benchmark results for electronic structure in molecules, solids, and materials like graphene or high-pressure hydrogen, as well as nuclear systems and quantum liquids such as helium-4.^[3]^[4] Despite their precision—often achieving chemical accuracy (1 kcal/mol) for ground states—QMC methods face challenges like the fermion sign problem, which biases sampling in systems with negative wave function signs, and computational costs scaling as N^3 to N^4 for N particles, though parallelization and pseudopotentials mitigate these for practical use.^[2] Ongoing advances integrate QMC with machine learning for improved trial functions and extend applications to excited states, real-time dynamics, and quantum computing hybrids.^[3]

Introduction

Definition and Scope

Quantum Monte Carlo (QMC) encompasses a class of stochastic methods that utilize random sampling techniques to approximate solutions to the many-body Schrödinger equation for quantum systems. These approaches leverage Monte Carlo integration to evaluate high-dimensional integrals arising in quantum mechanics, enabling the computation of ground-state energies and other properties for systems with strong electron correlations. The scope of QMC extends to both bosonic and fermionic many-body systems, facilitating studies of ground states, excited states, finite-temperature thermal properties, and real-time dynamics. While QMC is exact in principle for bosonic systems without approximations, fermionic calculations typically incorporate practical constraints like the fixed-node approximation to address the fermion sign problem, yielding highly accurate results with controlled errors.^[5] At the core of QMC methods are the many-body wavefunction \Psi(\mathbf{r}_1, \mathbf{r}_2, \dots, \mathbf{r}_N) and the Hamiltonian operator H, which governs the system's quantum evolution. Observables are obtained through expectation values defined as

\langle O \rangle = \frac{\int \Psi^* O \Psi \, d\tau}{\int |\Psi|^2 \, d\tau},

where d\tau represents the integration over all particle coordinates. In contrast to deterministic methods such as density functional theory (DFT), which depends on approximate exchange-correlation functionals and mean-field assumptions, or configuration interaction (CI), which involves exact diagonalization but suffers from exponential scaling, QMC directly samples the multidimensional wavefunction configuration space to capture strong correlations without relying on single-particle approximations, achieving chemical accuracy for systems comprising up to thousands of particles.^[5]

Historical Development

The origins of Quantum Monte Carlo (QMC) methods trace back to the 1940s, when Nicholas Metropolis and Stanislaw Ulam developed the Monte Carlo technique for simulating neutron transport during the Manhattan Project, leveraging random sampling to model complex stochastic processes on early computers like the ENIAC.^[6] This foundational work, formalized in their 1949 paper, extended to quantum problems by proposing Monte Carlo solutions to the Schrödinger equation through diffusion-like random walks.^[7] In the 1950s, Mark Kac and others further connected random walks to the time-dependent Schrödinger equation, laying theoretical groundwork for quantum applications.^[7] The 1960s marked the transition to explicit quantum many-body simulations, with Malvin Kalos pioneering diffusion-based methods for small bosonic systems, such as three- and four-body nuclei, using Green's function Monte Carlo (GFMC) to project onto the ground state.^[7] William McMillan introduced Variational Monte Carlo (VMC) in 1965, applying it to the ground state of liquid helium-4 with a Jastrow trial wavefunction and the Metropolis sampling algorithm to compute expectation values stochastically. By the 1970s, David Ceperley and Berni Alder advanced quantum applications, with Ceperley developing fixed-node approximations in his 1977 thesis to handle fermions, addressing the emerging fermion sign problem that causes exponential decay in signal-to-noise ratios for antisymmetric wavefunctions.^[7] Key milestones in the 1980s solidified QMC as a benchmark for ground-state properties. Ceperley and Alder's 1980 diffusion Monte Carlo (DMC) calculation of the homogeneous electron gas provided unprecedented accuracy for fermionic systems, highlighting the sign problem's severity while demonstrating fixed-node DMC's efficacy. James B. Anderson contributed pioneering VMC studies of atomic systems, such as hydrogen molecules, emphasizing correlated trial functions for chemical accuracy in the late 1970s and 1980s. For finite-temperature extensions, E. L. Pollock and Ceperley introduced Path Integral Monte Carlo (PIMC) in 1984, enabling simulations of quantum Boltzmann statistics for bosons like helium. Kalos's ongoing refinements to DMC, including importance sampling, further improved efficiency for condensed matter. The 1990s saw intensified focus on mitigating the fermion sign problem, with constrained-path and released-node techniques explored but limited by computational cost, while fixed-node approximations became standard for solids and molecules. In the 2000s, auxiliary-field QMC (AFQMC) emerged as a powerful projector method for strongly correlated electrons, decoupling interactions via Hubbard-Stratonovich transformations and phaseless approximations to control phase oscillations, enabling applications to Hubbard models and transition metal oxides.^[8] Ceperley's DMC work on helium liquids continued to set benchmarks, achieving near-exact energies for quantum fluids.^[9]

Basic Principles

Monte Carlo Integration in Quantum Contexts

Monte Carlo integration provides a powerful numerical technique for evaluating multidimensional integrals that arise frequently in statistical mechanics and quantum mechanics, where deterministic methods become computationally infeasible due to the curse of dimensionality. In its classical form, the method approximates the integral \int f(\mathbf{x}) \, d\mathbf{x} over a domain by generating N random samples \mathbf{x}_i from a probability distribution p(\mathbf{x}) and computing the average \frac{1}{N} \sum_{i=1}^N \frac{f(\mathbf{x}_i)}{p(\mathbf{x}_i)}, which converges to the true value as N \to \infty by the law of large numbers. To reduce variance and improve efficiency, importance sampling is employed by choosing p(\mathbf{x}) to concentrate samples where |f(\mathbf{x})| is large; the optimal distribution is p(\mathbf{x}) \propto |f(\mathbf{x})|, minimizing the variance to zero in the ideal case. In quantum mechanics, Monte Carlo methods are adapted to compute expectation values of operators in the many-body Schrödinger equation, which involve integrals over $3N-dimensional configuration spaces for N particles, rendering exact solutions impractical for large systems. A key adaptation involves reinterpreting the time-independent Schrödinger equation in imaginary time \tau = it, transforming it into a diffusion-like equation \frac{\partial \Psi}{\partial \tau} = -([\hat{H}](/page/Hat) - E_T) \Psi, where \hat{H} is the Hamiltonian and E_T a reference energy; as \tau \to \infty, the solution projects onto the ground-state wavefunction \Psi_0. This imaginary-time propagation e^{-\tau \hat{H}} \Psi facilitates stochastic simulation akin to a random walk with branching, allowing ground-state properties to be extracted through sampling. Central to quantum Monte Carlo is the stochastic estimation of expectation values using a trial wavefunction \Psi_T, which guides the sampling to focus on relevant regions of configuration space. The expectation value of an operator \hat{O} is given by

\langle \hat{O} \rangle = \frac{\langle \Psi_T | \hat{O} | \Psi \rangle}{\langle \Psi_T | \Psi \rangle},

where \Psi is the exact wavefunction (or its imaginary-time propagated form); this ratio is approximated by averaging over N "walker" configurations \mathbf{R}_i sampled from the distribution p(\mathbf{R}) = |\Psi_T(\mathbf{R})|^2 / \int |\Psi_T(\mathbf{R})|^2 d\mathbf{R}, yielding \langle \hat{O} \rangle \approx \frac{1}{N} \sum_{i=1}^N \frac{\Psi_T^*(\mathbf{R}_i) \hat{O} \Psi(\mathbf{R}_i)}{\Psi_T(\mathbf{R}_i) \Psi(\mathbf{R}_i)} after normalization. The trial wavefunction \Psi_T thus serves as an importance-sampling guide, reducing statistical noise by aligning the sampling distribution with the sought-after quantum state (detailed further in the role of trial wavefunctions). Efficient sampling in these high-dimensional spaces relies on Markov chain Monte Carlo techniques, which generate sequences of configurations that explore the probability distribution ergodically, ensuring that the chain's long-time average equals the ensemble average regardless of initial conditions. Ergodicity guarantees that the Markov chain visits all accessible states with the correct frequency, provided the transition probabilities satisfy detailed balance. The Metropolis-Hastings algorithm implements this by proposing moves from current configuration \mathbf{R} to \mathbf{R}' according to a proposal distribution q(\mathbf{R}' | \mathbf{R}) and accepting with probability \min\left(1, \frac{p(\mathbf{R}') q(\mathbf{R} | \mathbf{R}')}{p(\mathbf{R}) q(\mathbf{R}' | \mathbf{R})}\right), producing uncorrelated samples after sufficient equilibration.

The Role of Trial Wavefunctions and Stochastic Sampling

In quantum Monte Carlo (QMC) methods, the trial wavefunction \Psi_T(\mathbf{R}) serves as a guiding ansatz to approximate the ground-state wavefunction while incorporating essential physical symmetries and correlations. For fermionic systems like electrons, a common form is the Slater-Jastrow ansatz, \Psi_T(\mathbf{R}) = D(\mathbf{r}_1, \dots, \mathbf{r}_N) \exp[J(\mathbf{R})], where D is a Slater determinant ensuring antisymmetry and J is a Jastrow correlation factor that accounts for electron-electron interactions through two-body terms like u(r_{ij}). This form leverages single-particle orbitals from mean-field theories (e.g., Hartree-Fock) to respect Pauli exclusion while adding explicit correlation effects, thereby reducing the statistical error in QMC estimates.^[10] The local energy, defined as E_L(\mathbf{R}) = \frac{\hat{H} \Psi_T(\mathbf{R})}{\Psi_T(\mathbf{R})}, where \hat{H} is the Hamiltonian, plays a central role in evaluating the quality of \Psi_T. For an exact eigenstate, E_L is constant and equal to the eigenvalue, but for approximate \Psi_T, it fluctuates, with its expectation value providing an upper bound to the ground-state energy via the variational principle. Configurations \mathbf{R} (electron positions) are stochastically sampled from the distribution |\Psi_T(\mathbf{R})|^2 using algorithms like Metropolis to compute averages such as \langle E_L \rangle. In diffusion-based QMC methods, importance sampling guided by \Psi_T introduces a drift term, while population control is achieved through branching (replication of walkers where E_L < E_T, with E_T a reference energy) and killing (elimination where E_L > E_T) to maintain numerical stability and converge to the ground state.^[10] To enhance efficiency, variance control is crucial, as statistical errors scale with the variance of E_L. The energy variance is given by

\sigma_E^2 = \langle E_L^2 \rangle - \langle E_L \rangle^2,

where averages are over the distribution |\Psi_T|^2; minimizing \sigma_E^2 reduces the number of samples needed for a given precision. The zero-variance principle identifies the optimal importance function as the exact ground state, where \sigma_E^2 = 0 because E_L becomes constant, minimizing statistical error without bias. This principle guides improvements in \Psi_T and estimator design, as derived from the condition for variance-free estimators in Monte Carlo integration.^[11]^[10] For fermions, the sign problem arises from oscillating wavefunctions, but the fixed-node approximation mitigates this by enforcing the nodal structure (zero surfaces) of \Psi_T to approximately preserve antisymmetry. Walkers are restricted to one side of the nodes, preventing sign changes and yielding a variational upper bound to the energy, though introducing a systematic fixed-node error typically recovering 95% of correlation energy in practice. This constraint transforms the problem into a bosonic-like diffusion process while bounding the exact fermionic energy from above.^[10]

Ground State Methods

Variational Monte Carlo

The variational Monte Carlo (VMC) method approximates the ground state energy of a quantum many-body system by optimizing a parameterized trial wave function \Psi_T(\mathbf{R}; \{\alpha\}), where \mathbf{R} denotes the particle coordinates and \{\alpha\} are variational parameters. According to the variational principle, the expectation value of the Hamiltonian E = \frac{\langle \Psi_T | \hat{H} | \Psi_T \rangle}{\langle \Psi_T | \Psi_T \rangle} provides an upper bound to the true ground state energy E_0, i.e., E \geq E_0, with equality achieved only if \Psi_T is the exact ground state wave function. Minimization of E with respect to the parameters \{\alpha\} yields the best possible approximation within the chosen functional form of \Psi_T. The VMC algorithm evaluates the variational energy E stochastically by sampling configurations from the probability distribution \rho(\mathbf{R}) = \frac{|\Psi_T(\mathbf{R})|^2}{\int d\mathbf{R}' |\Psi_T(\mathbf{R}')|^2}.^[12] Configurations \mathbf{R}_k (for k = 1, \dots, M) are generated using the Metropolis-Hastings algorithm, which proposes random moves and accepts them with probability \min\left(1, \frac{|\Psi_T(\mathbf{R}_{new})|^2}{|\Psi_T(\mathbf{R}_{old})|^2}\right). At each sampled configuration, the local energy E_L(\mathbf{R}_k) = \frac{\hat{H} \Psi_T(\mathbf{R}_k)}{\Psi_T(\mathbf{R}_k)} is computed, and the energy estimator is the average \bar{E}_L = \frac{1}{M} \sum_{k=1}^M E_L(\mathbf{R}_k), which converges to E as M \to \infty with statistical error scaling as $1/\sqrt{M}.^[12] Optimization of the parameters \{\alpha\} proceeds iteratively to minimize \bar{E}_L. Common techniques include steepest descent, where updates follow \delta \alpha_i = -\eta \left\langle \frac{\partial E}{\partial \alpha_i} \right\rangle with step size \eta, and stochastic reconfiguration (SR), a more stable method approximating the inverse Hessian for faster convergence. In SR, the parameter update is given by \delta \alpha_i \propto -\left\langle \frac{\partial E}{\partial \alpha_i} \right\rangle / \left\langle \frac{\partial^2 E}{\partial \alpha_i^2} \right\rangle, where the averages are Monte Carlo estimates over samples from \rho(\mathbf{R}), effectively preconditioning the gradient with a covariance matrix derived from wave function fluctuations. VMC is particularly effective for bosonic systems, as sampling from |\Psi_T|^2 avoids the fermion sign problem entirely. It scales favorably to large systems, with computational cost dominated by the evaluation of E_L at each sample, enabling studies of dozens to hundreds of particles. For example, VMC calculations on the helium atom achieve ground state energy accuracies of 1 mHartree using optimized Jastrow-Slater trial wave functions.

Diffusion Monte Carlo

Diffusion Monte Carlo (DMC) is a stochastic projector method that solves the time-independent Schrödinger equation for the ground state of many-body quantum systems by propagating an initial wave function in imaginary time, effectively filtering out excited-state contributions. Unlike variational methods, which provide an upper bound to the ground-state energy, DMC projects to energies below this bound, yielding highly accurate results when combined with a good trial function. The approach relies on interpreting the imaginary-time evolution operator as a diffusion process with branching, enabling efficient sampling of the ground-state density in high-dimensional configuration space. The core of DMC is the imaginary-time Schrödinger equation,

\frac{\partial \Psi(\mathbf{R}, \tau)}{\partial \tau} = -(\hat{H} - E_T) \Psi(\mathbf{R}, \tau),

where \tau = it is imaginary time, \hat{H} = -\frac{1}{2} \nabla^2 + V(\mathbf{R}) is the Hamiltonian (in atomic units), E_T is an energy offset chosen to normalize the population, and \mathbf{R} denotes the 3N-dimensional configuration of N particles. As \tau \to \infty, \Psi(\mathbf{R}, \tau) converges to the ground-state wave function \Psi_0(\mathbf{R}) (up to a constant), assuming the initial state has a nonzero overlap with \Psi_0. The evolution is discretized into short-time steps \Delta \tau, using the approximate Green's function

G(\mathbf{R}', \mathbf{R}; \Delta \tau) \approx \left(2\pi \Delta \tau\right)^{-3N/2} \exp\left[ -\frac{(\mathbf{R}' - \mathbf{R} - \mathbf{v}(\mathbf{R}) \Delta \tau)^2}{2 \Delta \tau} - (E_L(\mathbf{R}') + E_L(\mathbf{R})) \frac{\Delta \tau}{2} + E_T \Delta \tau \right],

where \mathbf{v}(\mathbf{R}) = \nabla \ln |\Psi_T(\mathbf{R})| is the drift velocity from the trial wave function \Psi_T, and E_L(\mathbf{R}) = [\hat{H} \Psi_T(\mathbf{R})]/\Psi_T(\mathbf{R}) is the local energy.^[12] This Green's function captures the short-time propagator \exp[-\Delta \tau (\hat{H} - E_T)], with errors of order (\Delta \tau)^2 that can be extrapolated to zero. The algorithm employs a population of random walkers to sample the mixed distribution \Psi_T(\mathbf{R}) \Psi_0(\mathbf{R}). Each walker at position \mathbf{R} undergoes diffusion via a Gaussian displacement with variance $2 \Delta \tau per dimension, guided by the drift term to regions of high |\Psi_T|, followed by branching: the walker's weight is multiplied by \exp[-(E_L(\mathbf{R}) - E_T) \Delta \tau], leading to replication for E_L < E_T or death for E_L > E_T. Population control stabilizes the number of walkers (typically thousands to millions) through techniques like weighted branching or pure diffusion, where weights are adjusted periodically to minimize bias from finite populations, which scales as O(1/\sqrt{M}) for M walkers. The ground-state energy is estimated using the mixed estimator \langle E \rangle = \langle \Psi_T | \hat{H} | \Psi_0 \rangle / \langle \Psi_T | \Psi_0 \rangle, but pure \Psi_0-only estimators for properties like densities are obtained via forward walking, averaging descendant walker positions over multiple future steps to eliminate trial-function bias. For fermionic systems, the antisymmetric nature of \Psi_0 introduces a nodal surface where \Psi_0 = 0, causing the sign problem in unrestricted DMC due to oscillating densities. The fixed-node approximation resolves this by constraining walkers to one side of the nodal hypersurface defined by \Psi_T, equivalent to adding infinite repulsive potentials at the nodes of \Psi_T, yielding a variational upper bound to the true fermionic energy. This introduces a small fixed-node error (typically 1-5% of the correlation energy), which can be reduced by improving \Psi_T (e.g., via multi-Slater determinants) or corrected through extrapolation schemes, such as linear fits to nodal volume variations. The method, originally proposed for random walks by Anderson, was adapted to DMC for molecular systems in early applications. Fixed-node DMC routinely achieves chemical accuracy (errors < 1 kcal/mol) for ground-state properties of small molecules, often matching or exceeding coupled-cluster benchmarks. For instance, benchmark studies report mean absolute deviations of about 0.6 kcal/mol in atomization energies for sets including first-row hydrides using optimized trial functions.^[13] These accuracies stem from DMC's ability to capture strong correlations beyond mean-field approximations, though locality errors from pseudopotentials and time-step biases require careful control.

Finite Temperature Methods

Path Integral Monte Carlo

Path Integral Monte Carlo (PIMC) is a stochastic method for computing finite-temperature properties of quantum many-body systems, especially bosons, by leveraging the path integral representation of the thermal density matrix. The partition function is expressed as Z = \mathrm{Tr}[e^{-\beta \hat{H}}], where \beta = 1/(k_B T) and \hat{H} is the many-body Hamiltonian; this trace is reformulated using Feynman's imaginary-time path integral as an integral over all closed paths \mathbf{R}(\tau) in configuration space, weighted by the Euclidean action S_E[\mathbf{R}] = \int_0^\beta d\tau \left[ \frac{1}{2} m \dot{\mathbf{R}}^2 + V(\mathbf{R}) \right].^[14] To make the path integral tractable, imaginary time is discretized into P slices (or "beads"), with time step \epsilon = \beta / P, transforming the problem into a classical statistical mechanics integral over P N-dimensional configurations of ring polymers, where N is the number of particles. The short-time density matrix is approximated as \rho(\mathbf{R}', \mathbf{R}; \epsilon) = \langle \mathbf{R}' | e^{-\epsilon (\hat{K} + \hat{V})} | \mathbf{R} \rangle, typically using the primitive estimator \rho(\mathbf{R}', \mathbf{R}; \epsilon) \approx \left( \frac{m}{2\pi \epsilon \hbar^2} \right)^{3N/2} \exp\left[ -\epsilon \left( V(\mathbf{R}') + V(\mathbf{R}) \right)/2 - \frac{m |\mathbf{R}' - \mathbf{R}|^2}{2 \epsilon \hbar^2} \right], where \hat{K} and \hat{V} are kinetic and potential operators; higher accuracy is achieved with pair-product approximations that incorporate two-body interactions directly into the propagator, reducing the required P to around 50 for systems like helium.^[14] As P \to \infty, the Trotter product converges to the exact density matrix, enabling unbiased estimators for observables like energy and pressure via virial theorems.^[14] The multidimensional integral is evaluated using Metropolis Monte Carlo sampling of polymer configurations, with moves designed to explore permutation sectors for identical bosons. Early implementations relied on staging algorithms, which generate correlated bead displacements along the polymer chain to efficiently sample the free-particle Gaussian while satisfying detailed balance.^[14] For improved efficiency in handling Bose exchanges and off-diagonal properties, worm algorithms introduce dynamic "worms" that break and reconnect polymer segments, allowing direct sampling of open paths and permutation cycles without fixed topology constraints; this approach scales well for systems up to hundreds of particles and has become standard for superfluid simulations. PIMC excels in applications to quantum fluids like superfluid ^4He, where it accurately predicts the λ-transition at 2.17 K from simulations on early supercomputers, computes the condensate fraction via the one-body density matrix, and evaluates superfluid density using the winding number estimator \rho_s / \rho = \frac{ m k_B T \langle W^2 \rangle}{d \hbar^2 N}, with W the net path windings and d spatial dimensions, matching experimental values within 1-2%.^[14] It has also enabled detailed simulations of Bose-Einstein condensation in harmonically trapped dilute gases, yielding precise radial density profiles and interaction-induced shifts in the critical temperature T_c for up to 10^4 atoms, confirming mean-field predictions with quantum corrections of order 10-20% for realistic scattering lengths.

Determinantal Quantum Monte Carlo

Determinantal Quantum Monte Carlo (DQMC) is a stochastic method for simulating finite-temperature properties of interacting fermionic systems, particularly lattice models, by decoupling interactions via the Hubbard-Stratonovich transformation and evaluating fermionic traces through determinants. Developed by Blankenbecler, Scalapino, and Sugar in 1981, the approach operates in the grand canonical ensemble and provides numerically exact results within statistical error bars, making it a benchmark tool for strongly correlated electron systems. It builds on path integral representations of the partition function but specializes to fermions by introducing auxiliary fields to handle antisymmetry. The method relies on an auxiliary-field decomposition to manage the two-body interactions. The short-time propagator for the interaction term is transformed using the Hubbard-Stratonovich decoupling: e^{-\Delta \tau \hat{V}} \approx \int d\sigma \, \exp\left(-\Delta \tau V(\sigma)\right), where \hat{V} is the interaction operator, V(\sigma) is a one-body potential linear in the auxiliary fields \sigma, and the integral is over fluctuating fields that couple to density or spin operators. The kinetic energy \hat{K} is incorporated via non-interacting fermionic propagators, leading to a Trotter-discretized imaginary-time evolution. The partition function is then cast as Z \approx \int D\sigma \, \exp\left(-S_{\rm cl}(\sigma)\right), where S_{\rm cl}(\sigma) is an effective classical action over the fields, and the fermionic trace yields a product of determinants for spin-up and spin-down sectors: \det\left(I + B(\sigma)\right), with B(\sigma) the matrix product of short-time propagators including the fields. In the algorithm, configurations of the auxiliary fields \sigma are sampled using Metropolis Monte Carlo updates, weighted by \exp\left(-S_{\rm cl}(\sigma)\right) \left| \det\left(I + B(\sigma)\right) \right|, with observables obtained from stochastic estimates of Green's functions derived from ratios of determinants. The fermionic sign problem arises from oscillations in the determinant phase, which is handled by including the phase in the measurement average, though this limits simulations to parameter regimes where phase fluctuations are manageable, such as at half-filling or with particle-hole symmetry. Continuous-time variants, developed in the 2010s, eliminate Trotter discretization errors by directly sampling interaction events in continuous imaginary time, enhancing accuracy for quantum impurity and lattice models. DQMC has been extensively applied to the Hubbard model to study phenomena in doped Mott insulators, such as the crossover from antiferromagnetic order to strange metallic behavior upon doping away from half-filling.^[15] For instance, simulations at intermediate doping levels (n \approx 0.8) and temperatures around the pseudogap scale reveal linear-in-temperature resistivity and Planckian scattering rates, aligning with cuprate high-temperature superconductors and validating the method's predictive power.^[15] These results demonstrate DQMC's ability to capture correlation effects with controlled finite-size scaling, providing benchmarks for approximate theories in condensed matter physics.^[16]

Time-Dependent Methods

Real-Time Evolution in Closed Systems

In closed quantum systems, real-time evolution is governed by the unitary propagator U(t) = e^{-i \hat{H} t / \hbar}, where \hat{H} is the Hamiltonian, enabling the simulation of non-equilibrium dynamics such as wave packet propagation or coherent oscillations. Unlike imaginary-time evolution, which benefits from exponential decay that suppresses high-energy contributions and facilitates convergence in Monte Carlo sampling, real-time propagation introduces highly oscillatory phases that lead to destructive interference and severe sign or phase problems, making direct stochastic sampling inefficient or infeasible for large systems. These challenges arise because the real-time path integral representation involves complex actions without the damping effect of imaginary time, often resulting in exponentially growing variance in estimators for observables. To address these issues, time-dependent Variational Monte Carlo (t-VMC) extends the variational principle to real-time dynamics by parameterizing a trial wave function |\Phi(\theta(t))\rangle and evolving its parameters \theta(t) according to the time-dependent Schrödinger equation (TDSE) in a linearized form. The method solves the Dirac-Frenkel variational principle, minimizing the distance between the exact time derivative i \hbar \partial_t |\psi\rangle = \hat{H} |\psi\rangle and its projection onto the variational manifold, yielding equations for the parameter velocities \dot{\theta}_k = \sum_{k'} (G^{-1})_{k k'} F_{k'}, where G_{k k'} is the quantum geometric tensor and F_k involves energy gradients, both estimated via Monte Carlo sampling of the local energy and logarithmic derivatives. This approach captures many-body correlations through ansatze like Jastrow factors and backflow transformations, enabling accurate simulations of unitary evolution without explicit time propagation of the full wave function. For improved accuracy, techniques such as projected t-VMC (p-t-VMC) mitigate statistical biases from approximate nodal structures by solving an implicit optimization problem at each time step, reducing sample complexity by orders of magnitude compared to explicit schemes. Complementary real-time path integral Monte Carlo methods employ short-time approximations, such as Trotter decomposition of the propagator into e^{-i \hat{H} \Delta t / \hbar} \approx \prod e^{-i \hat{H}_j \Delta t / \hbar}, to discretize paths and sample configurations stochastically while managing phase oscillations through importance sampling or stochastic unwrapping. Observables like \langle O(t) \rangle are then computed as averages over paths weighted by e^{i S[\mathbf{x}(\tau)] / \hbar}, where S = \int_0^t L(\mathbf{x}, \dot{\mathbf{x}}) d\tau is the action with Lagrangian L = \frac{1}{2} m \dot{\mathbf{x}}^2 - V(\mathbf{x}) for non-relativistic systems, though convergence requires careful control of discretization errors and variance. These techniques have been applied to study quench dynamics in spin chains, where sudden Hamiltonian changes reveal entanglement growth and thermalization, with t-VMC accurately reproducing fidelity metrics such as F(t) = |\langle \psi(0) | \psi(t) \rangle|^2 up to long times in one-dimensional models. For instance, in Ising spin chains, such simulations demonstrate revivals and decay patterns consistent with exact diagonalization for small systems, scaling to larger lattices via efficient parameter optimization.

Nonequilibrium Dynamics Simulations

Nonequilibrium dynamics simulations in quantum Monte Carlo (QMC) address processes in driven quantum systems and open quantum systems coupled to external reservoirs, capturing phenomena such as relaxation to steady states and transport under bias. These methods extend beyond unitary evolution by incorporating time-dependent Hamiltonians H(t), which model external drives like voltage biases or laser fields, often starting from equilibrium initial states and evolving via quantum quenches or periodic perturbations. For instance, in strongly correlated electron systems, diagrammatic QMC techniques map lattice models onto impurity problems coupled to fermionic baths, enabling the study of current dynamics after a sudden bias application across a range of temperatures. Hybrids with linear response theory or exact diagonalization are employed for validation in small clusters, providing benchmarks for larger-scale QMC results. A key framework for open systems is the Lindblad master equation, which governs the reduced density matrix \rho of the system:

\frac{d\rho}{dt} = -i [H(t), \rho] + \sum_k \left( L_k \rho L_k^\dagger - \frac{1}{2} \{ L_k^\dagger L_k, \rho \} \right),

where L_k are jump operators representing dissipative processes from reservoirs. This equation is unraveled into an ensemble of stochastic trajectories using stochastic Schrödinger equations (SSEs), where each trajectory evolves a normalized state vector |\phi(t)\rangle under a non-Hermitian effective Hamiltonian plus stochastic noise terms corresponding to jumps. QMC sampling, such as variational Monte Carlo, is then applied to average over these trajectories, computing expectation values like \langle \hat{O} \rangle = \lim_{N \to \infty} \frac{1}{N} \sum_i \langle \phi_i(t) | \hat{O} | \phi_i(t) \rangle, with thousands of realizations ensuring convergence. For steady states in nonequilibrium impurity models, direct steady-state QMC formulations avoid time evolution altogether, solving for the fixed point of the dynamics in the presence of multiple leads at different chemical potentials. Techniques like real-time diagrammatic QMC incorporate reservoirs through Keldysh contour integrals, treating thermal baths as non-interacting fermions and sampling bold diagrams for vertex corrections. In fermionic systems near quantum critical points, such as Dirac criticality, QMC reveals universal scaling in short-time relaxation dynamics from nonequilibrium initial states, with dynamic exponents like \theta = -0.84(4) indicating anomalous diffusion unlike classical cases. Representative applications include nonequilibrium transport in ultracold atomic gases, where path-integral QMC computes shear viscosity and spin conductivity via analytic continuation of imaginary-time correlators, showing minima near twice the Kovtun-Son-Starinets bound in the unitary Fermi gas. In light-matter coupled systems, wormhole QMC algorithms map out phase diagrams of Dicke-Ising models, uncovering superradiant and ferromagnetic phases induced by cavity-mediated interactions mimicking laser drives. Recent 2020s advances integrate hybrid classical-quantum QMC with machine learning to simulate dynamics in non-Hermitian open systems, extending imaginary-time evolution for efficient trajectory unraveling in PT-symmetric setups.

Applications

In Quantum Chemistry and Materials Science

In quantum chemistry, Quantum Monte Carlo (QMC) methods excel at computing accurate binding energies and ionization potentials for molecular systems, often serving as benchmarks against coupled-cluster methods like CCSD(T). For instance, phaseless auxiliary-field QMC (ph-AFQMC) calculations on small molecules, including first-row atoms and transition metal complexes, yield ionization potentials within 0.1 eV of experimental values, surpassing CCSD(T) in some cases due to better handling of strong correlations.^[17] A notable example is the dissociation of the ozone molecule, where full configuration interaction QMC (FCIQMC) accurately maps the potential energy surface along the O₂ + O → O₃ pathway, predicting barrier heights and dissociation energies to within chemical accuracy (1 kcal/mol) of experiment, highlighting QMC's ability to capture multireference character absent in single-reference methods.^[18] In materials science, QMC provides precise predictions of electronic properties in solids, such as band gaps in semiconductors and correlation effects in strongly interacting systems. Diffusion Monte Carlo (DMC) calculations determine fundamental band gaps in materials like silicon and gallium arsenide with errors below 0.2 eV compared to experiment, offering a systematic improvement over density functional theory (DFT) underestimations.^[19] For high-temperature superconductors like cuprates, DMC applied to the Hubbard model elucidates electron correlations, reproducing antiferromagnetic ordering and doping-dependent spectral functions that align with angle-resolved photoemission spectroscopy data.^[20] These applications leverage DMC's projective nature for ground-state properties in extended systems. QMC methods offer key advantages in these fields, including size consistency—ensuring additive energies for non-interacting subsystems—and flexibility with basis sets, allowing seamless use of plane waves or Gaussian orbitals without complete basis set extrapolation penalties common in post-Hartree-Fock approaches.^[21] Integration with DFT-generated trial wave functions and pseudopotentials further enhances efficiency, enabling all-electron accuracy for valence electrons in large systems while treating core electrons approximately.^[22] A landmark demonstration is the 2014 QMC study of diamond's cohesive energy, computed at 7.327(3) eV/atom (zero-point energy corrected) versus the experimental 7.346 eV/atom, achieving ~0.3% relative accuracy and validating QMC for covalent solids.^[23] Recent post-2020 applications extend QMC to two-dimensional materials, such as graphene, where self-healing diffusion Monte Carlo quantifies correlation contributions to total energies.^[24] These studies underscore QMC's role in predicting optoelectronic properties for emerging van der Waals heterostructures.

In Nuclear and Atomic Physics

In atomic physics, quantum Monte Carlo (QMC) methods have been instrumental in studying the excited states of helium, providing high-accuracy benchmarks for few-electron systems. For instance, correlation-function QMC calculations have yielded energies for the ground and low-lying excited states of helium atoms under extreme conditions, such as those in neutron-star magnetic fields, with errors below 0.1 hartree compared to exact results.^[25] Additionally, QMC has advanced the computation of van der Waals forces in helium dimers, where variational and diffusion QMC techniques compute the full potential energy curve, capturing dispersion interactions with chemical accuracy and demonstrating the method's reliability for weakly bound atomic clusters.^[26] In nuclear physics, Green's function Monte Carlo (GFMC) serves as a cornerstone for ab initio calculations of light nuclei, enabling exact solutions for systems up to mass number A=12 using realistic nucleon-nucleon potentials. GFMC propagates trial wave functions in imaginary time to project the ground state, achieving binding energies accurate to 1-2% for nuclei like the alpha particle and triton. For neutron matter, auxiliary-field diffusion Monte Carlo (AFDMC) computes the equation of state at zero temperature, revealing stiff behavior at high densities relevant to neutron stars, with results consistent with chiral effective field theory (EFT) interactions up to next-to-next-to-leading order.^[27]^[28]^[29] Key techniques in these applications include chiral EFT potentials, which provide a systematic hierarchy of nuclear interactions derived from quantum chromodynamics, integrated into QMC for soft, local Hamiltonians that avoid the fermion sign problem in light systems. Hyperspherical coordinates facilitate few-body nuclear calculations by separating the center-of-mass motion and exploiting rotational invariance, allowing QMC to efficiently sample correlated wave functions in hyperradial and hyperangular variables for systems like the triton. A notable example is a 2022 perturbative QMC study of the triton binding energy, yielding 8.48 MeV with less than 1% deviation from experiment using chiral EFT potentials.^[30] Recent advances extend ab initio QMC to medium-mass nuclei (A≈16-40), employing neural-network trial functions and no-core shell model hybrids to achieve ground-state energies with uncertainties below 0.5 MeV per nucleon.^[31]^[32]

Challenges and Advances

The Fermion Sign Problem

The fermion sign problem arises in quantum Monte Carlo (QMC) simulations of fermionic systems due to the antisymmetry of the many-body wave function under particle exchange, which introduces oscillatory signs in the Boltzmann weights. For identical fermions, the wave function Ψ(R) for configuration R = {r₁, r₂, ..., r_N} is antisymmetric, expressible via a Slater determinant where the sign originates from the sum over permutations P of the single-particle orbitals: Ψ(R) ∝ ∑_P (-1)^P ∏i φ{P(i)}(r_i). This leads to phase oscillations in Ψ(R), causing the effective weights in the path integral or projector formulations to alternate in sign, resulting in an exponential growth in statistical variance with increasing particle number N or inverse temperature β. The problem is particularly severe because the average sign ⟨s⟩ decays exponentially as ⟨s⟩ ≈ exp(-β N Δf), where Δf is the free-energy difference between the fermionic system and a bosonic reference, making simulations computationally infeasible for large systems or low temperatures. In diffusion Monte Carlo (DMC), the sign problem manifests through negative local energies E_L(R) = H Ψ / Ψ(R) when walks cross nodal surfaces where Ψ(R) = 0, leading to unphysical branching and population instability without approximations. In determinantal QMC (DQMC), it appears as phase fluctuations or negative values in the fermion determinant det(M(σ)), where M is the matrix from Hubbard-Stratonovich fields σ, causing the weights w(σ) ∝ det(M(σ)) to become negative and the average sign to vanish exponentially with system size. To handle this, simulations often sample from an effective positive distribution p(R) = |Ψ(R)|², with observables computed as weighted averages incorporating sign(Ψ(R)), but the noise in estimators scales as exp(β ΔE) / √N_b, where ΔE is the excitation gap to the first bosonic state and N_b is the number of samples, leading to prohibitive error bars. The impact of the sign problem severely restricts QMC applications for fermions, confining reliable simulations primarily to half-filled systems on bipartite lattices, such as the Hubbard model at particle-hole symmetry, where particle-hole transformation renders weights positive. Away from these conditions, such as in doped systems or frustrated lattices, the exponential scaling precludes accurate results for realistic parameters. Recent post-2020 developments have explored hybrid quantum-classical approaches, using quantum computers to prepare sign-free trial states or unbias constrained-path estimators, thereby resolving the sign problem for systems up to 120 orbitals in proof-of-principle demonstrations. Common mitigations include the fixed-node approximation, which enforces Ψ(R) ≥ 0 by constraining walks to regions defined by a trial nodal structure, providing an upper bound to the ground-state energy with errors typically below 1% for molecular systems. The fixed-phase variant extends this to complex wave functions by fixing arg(Ψ(R)). Constrained-path Monte Carlo (CPMC) projects the ground state while discarding paths that change the sign relative to a trial function, reducing bias through importance sampling. Semi-stochastic projection methods combine deterministic Lanczos diagonalization in a low-rank subspace with stochastic sampling elsewhere, suppressing sign oscillations and enabling studies of the Hubbard model beyond half-filling.

Recent Developments and Computational Improvements

In the 2020s, neural network-based trial wave functions have emerged as a powerful advancement in variational Monte Carlo (VMC) and diffusion Monte Carlo (DMC), enabling more expressive representations of complex many-body correlations beyond traditional Slater-Jastrow forms. These architectures, such as feed-forward and deep neural networks, optimize wave functions for continuous systems, achieving accuracies comparable to or exceeding coupled-cluster methods for small molecules while scaling to larger systems like atomic clusters. For instance, neural networks have facilitated fixed-node DMC calculations for ground states of atoms and molecules, reducing variational errors by up to an order of magnitude in challenging cases like beryllium. Similarly, they have been extended to excited states, providing parameter-free estimates with chemical accuracy for systems up to dozens of electrons.^[33]^[34]^[35] Transcorrelated Hamiltonians, which incorporate explicit correlation factors into the operator via similarity transformations, have seen renewed development to enhance QMC accuracy for strongly correlated electrons. Recent implementations optimize polynomial correlation factors to mitigate cusp errors and improve energy convergence, yielding orders-of-magnitude reductions in fixed-node errors for molecular systems compared to standard pseudopotentials. These methods transfer correlation effects from the trial wave function to the Hamiltonian, allowing diffusion QMC to achieve near-exact results for second-row atoms and small molecules without sign problem exacerbation.^[36]^[37] Computational efficiency has advanced through GPU acceleration, particularly for DMC, where parallelizable walks and branching operations benefit from massive concurrency. The CASINO code, for example, now leverages GPUs to speed up real materials simulations by factors of 10-100, enabling routine calculations for systems with hundreds of electrons. QMCPACK, an open-source framework, integrates batched GPU drivers for mixed-precision DMC and auxiliary-field QMC (AFQMC), supporting exascale platforms like Frontier for distributed sampling across thousands of nodes. Machine learning enhancements to importance sampling further optimize walker distributions, using neural networks to predict high-probability configurations in VMC for statistical physics models.^[38]^[39]^[40] Hybrid approaches integrating QMC with quantum computing address longstanding issues like the fermion sign problem by augmenting trial states. In 2022, variational quantum eigensolvers (VQEs) were combined with AFQMC to unbias fermionic projections, outperforming standalone VQE for Hubbard models by reducing phase biases through quantum-generated constraints. Quantum-assisted VMC variants, such as those using shallow-circuit ansatze enhanced by classical post-processing, have demonstrated ground-state accuracies for small fermionic systems beyond classical limits. These hybrids scale to noisy intermediate-scale quantum devices, with tensor network extensions enabling distributed quantum sampling.^[41]^[42]^[43] Recent improvements in continuous-time QMC solvers for dynamical mean-field theory (DMFT) focus on efficient hybridization expansions and diagrammatic sampling, with 2024-2025 advancements incorporating tensor networks to handle multiorbital impurities at finite temperatures. These enable accurate self-energy computations for correlated materials like cuprates, reducing computational cost for real-frequency spectra by optimizing Monte Carlo updates. Jackknife resampling has been refined for error bounds in large-scale QMC, providing unbiased variance estimates for binning strategies in AFQMC and DMC, crucial for systems approaching 1000 electrons. Looking ahead, exascale integrations in codes like QMCPACK target such scalabilities, with distributed algorithms achieving near-linear weak scaling for 1000+ electron solids on petascale clusters, paving the way for routine QMC in materials design.^[44]^[45]^[46]^[47]

References

[1]
Introduction (Chapter 1) - Quantum Monte Carlo Methods
The Monte Carlo method is not a specific technique but a general strategy for solving problems too complex to solve analytically or too intensive numerically ...
[2]
[PDF] An Overview of Quantum Monte Carlo Methods David M. Ceperley
In this brief article, I introduce the various types of quantum Monte Carlo (QMC) methods, in particular, those that are applicable to systems in extreme ...
[3]
A brief introduction to the diffusion Monte Carlo method and the fixed ...
Dec 30, 2024 · Quantum Monte Carlo (QMC) methods represent a powerful family of computational techniques for tackling complex quantum many-body problems ...INTRODUCTION · III. DIFFUSION MONTE CARLO · VI. DIFFUSION MONTE...
[4]
Quantum Monte Carlo Methods in Nuclear Physics: Recent Advances
Jan 15, 2019 · In this review, we discuss these recent developments and give an overview of the exciting results for nuclei, neutron matter and neutron stars, and electroweak ...
[5]
[PDF] Introduction to Quantum Monte Carlo Methods Ap
In these lectures, I will briefly review some of the Quantum Monte Carlo (QMC) methods that have been used to calculate properties of the “electron gas” and ...
[6]
[PDF] The Monte Carlo Method - Nicholas Metropolis; S. Ulam
Jan 25, 2006 · The probability of pro- ducing a given particle with a given energy in any given collision is dependent on the energy of the incoming particle.Missing: transport | Show results with:transport
[7]
[PDF] The Early Years of Quantum Monte Carlo (1): the Ground State - arXiv
At the end of the fifties, the two main methods of classical molecular simulation, Monte Carlo and Molecular Dynamics, had been invented and an impressively ...
[8]
Sign-Problem-Free Fermionic Quantum Monte Carlo - Annual Reviews
Mar 10, 2019 · Here, we concisely review recent progress made in developing new sign-problem-free QMC algorithms, including those employing Majorana ...
[9]
Path integrals in the theory of condensed helium | Rev. Mod. Phys.
Apr 1, 1995 · In this review, the author introduces this picture of a boson superfluid and shows how superfluidity and Bose condensation manifest themselves.
[10]
Quantum Monte Carlo simulations of solids | Rev. Mod. Phys.
Jan 5, 2001 · This article describes the variational and fixed-node diffusion quantum Monte Carlo methods and how they may be used to calculate the properties of many- ...
[11]
Zero-Variance Principle for Monte Carlo Algorithms | Phys. Rev. Lett.
Dec 6, 1999 · By writing down the zero-variance condition a fundamental equation determining the optimal choice for the renormalized observable is derived ( ...
[12]
Introduction to the variational and diffusion Monte Carlo methods
Aug 12, 2015 · We provide a pedagogical introduction to the two main variants of real-space quantum Monte Carlo methods for electronic-structure calculations.
[13]
Towards the ground state of molecules via diffusion Monte Carlo on ...
Apr 3, 2023 · In this work, we apply the neural-network based trial wavefunction in fixed-node DMC, which allows accurate calculations of a broad range of atomic and ...Missing: seminal | Show results with:seminal
[14]
Fast and accurate quantum Monte Carlo for molecular crystals - PNAS
Feb 5, 2018 · We demonstrate that diffusion quantum Monte Carlo (DMC) delivers subchemical accuracy for a diverse set of molecular crystals at a surprisingly moderate ...
[15]
https://www.science.org/doi/10.1126/science.aau7063
[16]
Strange metallicity in the doped Hubbard model - Science
Here we demonstrate strange metallic transport in the doped two-dimensional Hubbard model using determinantal quantum Monte Carlo calculations.
[17]
Mott and band-insulator transitions in the binary-alloy Hubbard model
Apr 19, 2007 · In this paper we will reexamine the physics of the binary-alloy Hubbard model using determinant quantum Monte Carlo (DQMC) simulations and exact ...
[18]
Benchmark Phaseless Auxiliary-Field Quantum Monte Carlo Method ...
Jul 20, 2023 · Using the recently developed phaseless auxiliary-field quantum Monte Carlo (AF QMC) method, we examine bond stretching in the well-studied mols.
[19]
Comparisons of full configuration interaction quantum Monte Carlo ...
Sep 6, 2017 · The association/dissociation reaction path for ozone (O2 + O ↔ O3) is notoriously difficult to describe accurately using ab initio ...
[20]
Electronic band gaps from quantum Monte Carlo methods
Feb 11, 2020 · We develop a method for calculating the fundamental electronic gap of semiconductors and insulators using grand canonical quantum Monte Carlo simulations.Abstract · Article Text · INTRODUCTION · COMPUTATIONAL METHODS
[21]
Solutions of the Two-Dimensional Hubbard Model: Benchmarks and ...
Dec 14, 2015 · Solutions of the Two-Dimensional Hubbard Model: Benchmarks and Results from a Wide Range of Numerical Algorithms ... diffusion Monte Carlo within ...
[22]
QMCPACK: Advances in the development, efficiency, and ...
Quantum Monte Carlo (QMC) methods are an attractive approach for accurately computing and analyzing solutions of the Schrödinger equation.1–3 The methods form a ...
[23]
Trail-Needs pseudopotentials in quantum Monte Carlo calculations ...
We report a systematic analysis of the performance of a widely used set of Dirac-Fock pseudopotentials for quantum Monte Carlo (QMC) calculations.Missing: flexible | Show results with:flexible
[24]
Cohesion Energetics of Carbon Allotropes : Quantum Monte Carlo ...
Dec 31, 2013 · ... diamond and graphite, respectively, when the zero-point energies ... The cohesive energy of \gamma-graphyne, the most energetically ...
[25]
Multireference diffusion Monte Carlo reaches 2D materials - Nature
Sep 26, 2025 · Here, we estimate the magnitude of impact of correlation on the total energy for the quintessential 2D material, graphene, by performing and ...
[26]
Fixed-phase correlation-function quantum Monte Carlo calculations ...
Mar 21, 2013 · We apply the correlation-function quantum Monte Carlo (CFQMC) method to the calculation of the energies of ground and excited states for helium in neutron-star ...
[27]
Quantum Monte Carlo calculations of the potential energy curve of ...
Mar 19, 2008 · Quantum Monte Carlo (QMC) methods offer the promise of a unified approach with the accuracy required to describe van der Waals interactions. In ...
[28]
Green's function Monte Carlo study of light nuclei | Phys. Rev. C
Nov 1, 1987 · Green's function Monte Carlo calculations of the alpha particle provide the first test of variational methods for systems with spin-dependent ...
[29]
Quantum Monte Carlo calculations of light nuclei - EPJ A
Our Green's function Monte Carlo calculations are accurate to ~1-2% for the ... Some of these results are presented to show that ab initio calculations of light ...
[30]
Quantum Monte Carlo calculation of the equation of state of neutron ...
May 18, 2009 · Quantum Monte Carlo calculation of the equation of state of neutron matter ... Equation of state of neutron matter. We employed the fixed ...
[31]
Quantum Monte Carlo Calculations with Chiral Effective Field ...
Jul 18, 2013 · We present the first quantum Monte Carlo (QMC) calculations with chiral effective field theory (EFT) interactions.
[32]
Universal few-body physics and cluster formation | Rev. Mod. Phys.
Aug 28, 2017 · The few-body physics in chemical sciences will be presented through classical trajectory calculations in hyperspherical coordinates involving ...
[33]
Perturbative Quantum Monte Carlo Method for Nuclear Physics
Jun 14, 2022 · Perturbative Quantum Monte Carlo Method for Nuclear Physics. Bing ... triton binding energy. We also implement a static Coulomb force V ...
[34]
Deep-neural-network approach to solving the ab initio nuclear ...
Mar 31, 2023 · ... ab initio studies of medium-mass nuclei with quantum Monte Carlo approaches. II. ARCHITECTURE. At the core of our approach is a deep-learning ...
[35]
Artificial Neural Networks as Trial Wave Functions for Quantum ...
Jan 25, 2021 · Feed-forward neural networks are proposed as a general purpose trial wave function for quantum Monte Carlo simulations of continuous many-body systems.
[36]
Towards the ground state of molecules via diffusion Monte Carlo on ...
Apr 3, 2023 · In this work, we apply the neural-network based trial wavefunction in fixed-node DMC, which allows accurate calculations of a broad range of atomic and ...Results · Single Atoms · Methods
[37]
Accurate computation of quantum excited states with neural networks
Aug 23, 2024 · We present an algorithm to estimate the excited states of a quantum system by variational Monte Carlo, which has no free parameters and requires ...
[38]
Studies on the Transcorrelated Method | Journal of Chemical Theory ...
Aug 28, 2023 · We investigate the possibility of using a transcorrelated (TC) Hamiltonian to describe electron correlation.
[39]
Orders of magnitude increased accuracy for quantum many-body ...
Transcorrelated methods provide an efficient way of partially transferring the description of electronic correlations from the ground-state wave function ...
[40]
Acceleration of the CASINO quantum Monte Carlo software using ...
Oct 26, 2025 · We describe how quantum Monte Carlo calculations using the CASINO software can be accelerated using graphics processing units (GPUs) and ...
[41]
Quantum Monte Carlo Methods — QMCPACK Manual documentation
Consequently, batched drivers allow mixing and matching CPU-only and GPU-accelerated features in a way that is not feasible with the legacy GPU implementation.
[42]
Performance of machine-learning-assisted Monte Carlo in sampling ...
Oct 7, 2025 · Recent years have seen a rise in the application of machine learning techniques to aid the simulation of hard-to-sample systems that cannot ...
[43]
Unbiasing Fermionic Quantum Monte Carlo with a Quantum Computer
Jun 30, 2021 · Our results demonstrate a new paradigm of hybrid quantum-classical algorithm, surpassing the popular variational quantum eigensolver in ...
[44]
Variational Quantum-Neural Hybrid Eigensolver | Phys. Rev. Lett.
Mar 24, 2022 · We introduce the variational quantum-neural hybrid eigensolver (VQNHE) in which the shallow-circuit quantum Ansatz can be further enhanced by classical post- ...
[45]
Quantum computing quantum Monte Carlo with hybrid tensor ...
Jun 6, 2024 · We propose an algorithm combining QC-QMC with a hybrid tensor network to extend the applicability of QC-QMC beyond a single quantum device size.Results · Methods · Quantum Computing Quantum...<|control11|><|separator|>
[46]
Algorithm for computing perturbation series of dynamical mean field ...
Apr 27, 2025 · We show how to use diagrammatic techniques to compute the weak-coupling perturbation series of the self-consistent solution to a Dynamical ...
[47]
Quantum Impurity Solvers
Aug 4, 2025 · CT-QMC and Hybridization Expansion Algorithms. Continuous-time quantum Monte Carlo methods represent a cornerstone of modern impurity solvers.2. Ct-Qmc And Hybridization... · 3. Advanced Methods: Tensor... · 5. Performance, Benchmarking...
[48]
Strategies for improving the efficiency of quantum Monte Carlo ...
Jun 27, 2011 · We describe a number of strategies for minimizing and calculating accurately the statistical uncertainty in quantum Monte Carlo calculations ...
[49]
QMCPACK - Exascale Computing Project
QMC allows for highly accurate calculations and enables feasible testing of any approximations, rendering computational models of this type truer to nature than ...