In quantum field theory and many-body physics, we are often interested in how a system evolves or how it responds to a probe. These "real-time" properties—like the way a particle decays, how heat flows through a liquid, or the shape of a gravitational wave—are encoded in spectral functions and correlation functions.
However, calculating these directly is notoriously difficult. For decades, the standard workaround has been to perform calculations in "imaginary time" (Euclidean space), where the math is more stable, and then use a mathematical trick called analytic continuation to translate the results back to the real world. But this trick comes with a heavy price: it is an "ill-posed" problem. Small amounts of noise in the imaginary-time data can lead to massive errors in the real-time result, effectively blurring out sharp physical features like the "kinks" in electron dispersion or the narrow peaks of new particles.
A growing body of research is moving away from this "Euclidean detour." By developing methods that operate directly on the real-frequency axis or the complex momentum plane, researchers are finding ways to simulate quantum dynamics with unprecedented fidelity.
The Closed Time Path and Causality
The fundamental challenge of real-time simulation is ensuring causality—that the state of a system depends only on its past, not its future. Standard scattering theory often struggles with this when calculating expectation values of observables.
To solve this, researchers like Fred Cooper and Emil Mottola championed the Schwinger-Keldysh "Closed Time Path" (CTP) formalism. As detailed in Nonequilibrium quantum fields in the large-N expansion, this approach uses a temporal loop that flows forward from an initial state and then backward again. This "time-folded contour" ensures that equations of motion remain causal and real-valued.
This framework is not just for subatomic particles. Simon Caron-Huot and Holmfridur S. Hannesdottir have recently extended these ideas to "asymptotic observables," providing a unified way to calculate everything from LHC cross-sections to the gravitational waveforms measured by LIGO, as seen in What can be measured asymptotically?.
Flowing Through the Complex Plane
One of the most successful strategies for avoiding the analytic continuation headache is to perform the continuation before the calculation is finished. In the Functional Renormalization Group (FRG) framework, physicists track how the physics of a system changes as they zoom in from high energies to low energies using "flow equations."
Rather than solving these equations in imaginary time and trying to invert the result, researchers like Ralf-Arno Tripolt, Lorenz von Smekal, and Jan M. Pawlowski apply analytic continuation directly to the differential equations themselves. In Spectral functions from the functional renormalization group, they demonstrate that by substituting discrete imaginary frequencies with continuous real energy variables during the integration, they can extract spectral functions directly.
This "direct" flow allows for the study of transport coefficients in the "perfect liquid" of the quark-gluon plasma (Transport Coefficients in Yang-Mills Theory and QCD) and the decay of mesons (Real-time correlation functions in the O(N) model from the functional renormalization group) without the noise sensitivity of traditional reconstruction methods.
Real-Time Solvers for Strong Correlations
In condensed matter physics, the challenge often lies in "impurity models"—small quantum systems embedded in a larger environment, used to understand materials like Mott insulators. Traditional solvers for Dynamical Mean Field Theory (DMFT) often rely on Quantum Monte Carlo (QMC) in imaginary time, which blurs the very quasiparticle peaks that define the material's properties.
New "impurity solvers" are bypassing this by working entirely in the real-time domain:
- Stochastic Exact Diagonalization: Mats Granath and Hugo U. R. Strand developed Distributional Exact Diagonalization (Dist-ED), which averages over many small, exactly solvable systems on the real-frequency axis. This revealed hidden coherent quasiparticles at the edges of Mott insulator bands that were previously invisible (Signatures of coherent electronic quasiparticles in the paramagnetic Mott insulator).
- Matrix Product States: Lei Wang and colleagues adapted the time-dependent variational matrix product state (tVMPS) approach to act as a DMFT solver, bridging the gap between Hamiltonian-based and stochastic methods (An Impurity Solver Using the Time-Dependent Variational Matrix Product State Approach).
- Inchworm QMC: To overcome the "sign problem" that plagues real-time simulations, Emanuel Gull and Guy Cohen implemented the "inchworm" algorithm, which builds up the real-time propagator in small steps, allowing for the direct observation of phenomena like the Mott transition in real time (Quantum Monte Carlo solution of the dynamical mean field equations in real time).
Why Intuition Favors the Real Axis
The shift toward direct real-time calculation is driven by a simple realization: physical information is lost during the Euclidean-to-Minkowski transformation. Sharp resonances, decay widths, and high-energy excitations are high-frequency features that become exponentially suppressed in imaginary time. By staying on the real axis—whether through time-folded path integrals, complex-plane renormalization flows, or real-time impurity solvers—researchers can preserve the full dynamical fingerprint of quantum systems.