| Diagrammatic Monte Carlo | |
|---|---|
| Name | Diagrammatic Monte Carlo |
| Type | Computational method |
| Field | Quantum many-body theory |
| Introduced | 1990s |
| Developers | N. Prokof'ev, B. Svistunov, M. Kozlov |
| Related | Feynman diagram, Quantum Monte Carlo |
Diagrammatic Monte Carlo
Diagrammatic Monte Carlo is a stochastic numerical method that evaluates perturbative expansions in many-body physics by sampling Feynman diagrams directly. Combining concepts from Feynman diagrammatic perturbation theory and Monte Carlo methods, it enables nonperturbative access to properties of strongly correlated fermionic and bosonic systems where conventional deterministic summation is intractable. The method matters in Quantum Physics because it can compute spectral functions, self-energies and thermodynamic observables for models such as the Hubbard model, the Fröhlich polaron problem, and quantum impurity problems without uncontrolled truncations.
Diagrammatic Monte Carlo builds on the formal machinery of perturbation theory where interacting propagators and vertices are represented by Feynman diagrams. In the context of quantum many-body problems one typically expands the partition function or Green's function in powers of an interaction parameter using bare or dressed propagators (e.g., Matsubara Green's function formalism for finite temperature). Key theoretical ingredients include the Dyson equation, the self-energy, and vertex functions ordinarily treated in diagrammatic resummation schemes such as Dyson series, Ladder approximation, and Bold diagrammatic (skeleton) expansions. Historical developments trace through quantum field theory methods applied to condensed matter physics and the impurity solvers used in Dynamical mean field theory (DMFT).
The central idea is to interpret the sum over diagrams as a high-dimensional integral and to estimate it by stochastic sampling. A Markov chain Monte Carlo procedure explores diagram space via updates that add, remove, or change diagram topology, order, internal times, and momenta, accepting moves with Metropolis-like criteria to satisfy detailed balance. Common choices include expansions in interaction strength or in hybridization for impurity problems; examples of problems treated this way include the Anderson impurity model and the Holstein model. Sampling directly in diagram space allows automatic inclusion of arbitrarily high orders, constrained only by computational resources and statistical variance.
Multiple algorithmic frameworks exist. "Bare" Diagrammatic Monte Carlo samples expansions built from free propagators, while "bold" approaches sample skeleton diagrams expressed in terms of dressed propagators and self-consistent quantities. Variants include Worm algorithms adapted to diagram sampling, continuous-time diagrammatic Monte Carlo for quantum impurity solvers (e.g., CT-HYB and CT-INT styles), and determinant diagrammatic Monte Carlo that employs determinants to handle fermionic exchange. Implementations often exploit fast Fourier transforms for momentum-frequency transforms, importance sampling strategies, and cluster techniques drawn from auxiliary-field Monte Carlo methods. Software frameworks have been developed in research groups at institutions such as Moscow State University, Princeton University, and the Max Planck Institute for the Physics of Complex Systems.
Diagrammatic Monte Carlo has been applied to a range of models and experimental contexts: the polaron problem (e.g., Fröhlich polaron studies), the Fermi polaron in ultracold gases, equation-of-state and spectral properties of the Hubbard model, quantum criticality in bosonic and fermionic systems, and transport in nanoscale quantum dot and impurity setups. It has also informed studies of superconducting instabilities by computing pairing susceptibilities and has provided benchmarks for other numerical methods such as density matrix renormalization group (DMRG) and conventional Quantum Monte Carlo approaches. Results have been compared with experiments in ultracold atoms and solid-state spectroscopy.
Convergence of diagrammatic series is a central issue: many physical expansions are asymptotic or divergent at moderate coupling. Resummation techniques—Padé approximants, Borel resummation, conformal mapping, and analytic continuation from Matsubara to real frequencies—are commonly combined with diagrammatic Monte Carlo data to extract physical observables. The fermionic sign problem appears as oscillatory contributions from diagrams with alternating signs, causing large variance; variants such as determinant and bold schemes can mitigate but not eliminate sign-related scaling. Strategies include reweighting, optimized sampling of dominant diagram topologies, and exploiting cancellations analytically where possible. Understanding radius of convergence and singularity structure of series is an active research area connecting to analytic properties of Green's functions.
Practical use requires careful statistical analysis: autocorrelation estimates, error bars from blocking analysis, and variance reduction. Benchmarks against exact results (e.g., Bethe ansatz solvable models), high-precision diagrammatic determinant Monte Carlo studies, and experimental data validate implementations. Computational cost scales with reachable diagram order, complexity of internal degrees of freedom, and severity of sign cancellations; high-performance computing resources and parallel algorithms are often used. Reproducibility is aided by published codes and systematic protocol papers; prominent benchmark studies appeared from groups led by Nikolay Prokof'ev and Boris Svistunov and in collaborations involving Evgeny Kozik and others. Ongoing developments target hybrid methods combining diagrammatic sampling with machine learning for proposal distribution optimization and automatic detection of dominant diagram families.
Category:Computational physics Category:Quantum field theory