MC tools

Detailed Description

Tools to implement and perform Monte Carlo simulations.

The MC tools module provides a generic framework for writing Markov-chain Monte Carlo simulations. It does not implement any particular physical algorithm; instead it supplies the scaffolding common to all such simulations — driving the Markov chain, dispatching user-defined updates, accumulating user-defined observables, and aggregating results across MPI ranks.

A simulation is described by three ingredients supplied by the user:

• a sign type (real or complex, see triqs::mc_tools::DoubleOrComplex) that encodes the per-configuration weight,
• a set of moves that propose updates of the Markov chain and report their Metropolis acceptance ratio (see MC moves and triqs::mc_tools::MCMove),
• a set of measurements that accumulate observables along the chain (see MC measurements and triqs::mc_tools::MCMeasure).

These ingredients are plugged into a triqs::mc_tools::mc_generic instance, which manages warmup (equilibration, optional move calibration) and accumulation (measurement) phases, organised into cycles of MC steps. The driver is MPI-aware: random seeds, exception propagation, cycle counters and measurement reduction are all coordinated through an mpi::communicator, so the same simulation code runs serially or in parallel.

Topics
	MC concepts
	Concepts that specify the requirements for MC moves and measurements.
	MC measurements
	Abstraction for Monte Carlo measurements.
	MC moves
	Abstraction for Monte Carlo moves.
	MC simulation
	Run and manage MC simulations.
	MC utilities
	Various utilities used with MC tools.