Uniform process tensor approach for the calculation of multi-time correlation functions of non-Markovian open systems

This paper introduces a uniform process tensor approach using time-translation invariant matrix product operators (uniTEMPO) to efficiently compute multi-time correlation functions and multi-dimensional spectra of non-Markovian open quantum systems directly in Fourier space, thereby avoiding the computational cost of explicit real-time evolution.

The Gist

Here is an explanation of the paper, translated from complex physics jargon into everyday language with some creative analogies.

The Big Picture: Listening to a Noisy Room

Imagine you are trying to understand a conversation happening in a very noisy, crowded room.

• The System: The two people talking (the "quantum system").
• The Environment: The crowd, the music, and the echo in the room (the "bath" or "reservoir").

In the world of quantum physics, we often want to know how a system behaves over time. But here's the catch: the environment isn't just background noise; it has a memory. If you shout in a cave, the echo comes back later. If the environment is "non-Markovian" (the fancy physics term), it remembers what the system did a moment ago and reacts to it. This makes predicting the system's future incredibly hard.

Scientists need to calculate multi-time correlation functions. Think of this as asking: "If I poke the system at 1:00, then poke it again at 1:05, and look at the result at 1:10, how are those three moments connected?"

Doing this math is usually like trying to solve a maze while blindfolded, because the "memory" of the environment makes the equations explode in complexity.

The Old Way vs. The New Way

The Old Way (PT-TEMPO):
Imagine you are trying to record a long movie of this conversation. To do it accurately, you have to record every single frame from the start. As the movie gets longer, the file size gets huge, and your computer runs out of memory. You have to re-calculate the whole history every time you want to know what happens at the end. This is slow and inefficient.

The New Way (uniTEMPO):
The authors of this paper (Garbellini, Mickiewicz, et al.) found a clever shortcut. They realized that because the "crowd" (the environment) is steady and doesn't change its nature over time, the way it reacts is time-translation invariant.

Think of it like a stamping machine.

• In the old method, you had to build a new stamp for every single second of the movie.
• In the new method (uniTEMPO), you build one perfect master stamp. Because the environment is consistent, this one stamp works for any amount of time. You just press it down as many times as you need.

The Magic Trick: Skipping the Movie, Going Straight to the Soundtrack

The most powerful part of this new method is how it handles spectra (which are like the "soundtrack" or frequency analysis of the system).

Usually, to find the frequency of a sound, you have to record the whole movie of the sound wave first, and then run a complex computer program (Fourier transform) to figure out the notes.

The uniTEMPO trick:
Because they built that "Master Stamp" (which they call a Matrix Product Operator), they can mathematically skip the movie entirely. They can look at the stamp and immediately write down the soundtrack.

• No real-time evolution: They don't have to simulate the system second-by-second.
• Direct access: They get the answer in the "frequency domain" (the notes) instantly.

This is like being able to look at a musical instrument and instantly know what note it will play, without actually plucking the string and waiting for the sound to travel.

What Did They Prove?

The team tested this method on a simple model (a three-level system interacting with a "bath" of vibrations). They showed that:

1. It's Fast: They could calculate complex 2D spectra (which are like 3D maps of how the system reacts to light) in seconds or minutes, even for systems with strong "memory" effects.
2. It Scales Well: If you want to see what happens after a long waiting time (say, 10 seconds instead of 1 second), the old methods get exponentially slower. The new method takes almost the same amount of time regardless of how long you wait.
3. It's Accurate: They proved that by adjusting a few knobs (like the size of the "stamp"), they could get results as precise as they wanted.

The Takeaway

This paper introduces a new "universal remote control" for quantum systems. Instead of simulating every tiny step of a quantum system's chaotic dance with its environment, this method compresses the environment's memory into a single, efficient mathematical object.

This allows scientists to instantly see the "colors" (frequencies) of quantum systems, making it much easier to design better solar cells, quantum computers, or understand how energy moves through biological systems like photosynthesis. It turns a marathon run into a teleportation jump.

Technical Summary

Here is a detailed technical summary of the paper "Uniform process tensor approach for the calculation of multi-time correlation functions of non-Markovian open systems."

1. Problem Statement

Multi-time correlation functions are fundamental for understanding quantum system properties, particularly in spectroscopy (e.g., 2D electronic spectroscopy) and response theory. However, calculating these functions for non-Markovian open quantum systems (where the environment has memory effects) is computationally challenging.

• Limitations of Existing Methods: Traditional methods like Hierarchical Equations of Motion (HEOM) or Quantum Regression Theory often struggle with strong system-bath coupling or require explicit real-time evolution over large time grids.
• Process Tensor Bottleneck: While the Process Tensor framework (using Matrix Product Operators, MPO) offers a general description, standard implementations (like PT-TEMPO) typically use a finite time window. This obscures time-translation invariance and limits the maximum evolution time, making the calculation of spectra (which require Fourier transforms) inefficient as it necessitates long real-time propagations.

2. Methodology

The authors propose a method based on the uniform time-evolving matrix product operator (uniTEMPO) to compute multi-time correlation functions directly in the frequency domain.

• Model Setup: The system is described as a small number of degrees of freedom coupled to a bosonic bath (environment). The total Hamiltonian is partitioned into system ($H_S$), bath ($H_B$), and interaction ($H_{SB}$) terms.
• uniTEMPO Framework:
  • Instead of a finite time grid, uniTEMPO exploits the time-translation invariance of the bath correlation function.
  • It constructs a time-independent short-time propagator, denoted as $Q$, acting on an extended state space (system Hilbert space $\otimes$ auxiliary memory space).
  • The propagator $Q$ is represented as a Matrix Product Operator (MPO) with a specific bond dimension $\chi$.
• Spectral Representation:
  • The core innovation is the eigendecomposition of the propagator $Q$: $Q = \sum_k q_k |q_k\rangle\langle\langle q_k|$.
  • The eigenvalues $q_k$ are mapped to complex frequencies $\lambda_k = -i\omega_k - \gamma_k$.
  • This allows the time evolution operator $Q^N$ to be expressed analytically as $e^{\lambda_k t}$.
• Frequency Domain Calculation:
  • By utilizing the spectral decomposition, the authors derive analytical expressions for the Fourier transforms of the correlation functions.
  • This eliminates the need for explicit real-time evolution steps. The multi-time correlation function $R(\tau_N, \dots, \tau_1)$ is calculated as a sum over eigenmodes, where time intervals are replaced by Lorentzian functions in the frequency domain.

3. Key Contributions

1. Direct Spectral Access: The paper demonstrates that uniTEMPO provides a direct spectral representation of non-Markovian dynamics. This allows for the calculation of linear and multi-dimensional spectra without performing explicit time evolution, significantly improving numerical scaling.
2. Efficient 2D Spectra Calculation: The method enables the calculation of 2D spectra for arbitrary waiting times ($T$) with minimal additional computational cost. In contrast, time-domain methods require re-propagating the system for every new $T$.
3. Scalability Analysis: The authors provide a rigorous analysis of numerical scaling:
   • uniTEMPO: Scaling is roughly $O((\chi d^2)^2 M^2)$ for 2D spectra (where $d$ is system dimension, $\chi$ is bond dimension, $M$ is frequency grid size).
   • PT-TEMPO (Standard): Scaling is $O((\chi d^2)^3 N^2)$, where $N$ is the number of time steps. This highlights a superior scaling for uniTEMPO, especially for long waiting times.
4. Convergence Properties: The study establishes that the method converges rapidly with respect to the bond dimension ($\chi$) with an error scaling of $O(\chi^{-3})$ and follows the expected $O(\Delta^2)$ scaling for the Trotter time step $\Delta$.

4. Results

The authors applied the method to a three-level system coupled to a bosonic bath with an Ohmic-like spectral density, simulating linear and 2D electronic spectra under three distinct coupling regimes:

• Weak Coupling: $\lambda = 0.03$ ps$^{-1}$, $\Omega = 2.0$ ps$^{-1}$.
• Intermediate Coupling: $\lambda = 0.6$ ps$^{-1}$, $\Omega = 2.0$ ps$^{-1}$.
• Strong Coupling: $\lambda = 2.4$ ps$^{-1}$, $\Omega = 0.2$ ps$^{-1}$.

Key Findings:

• Accuracy: The method successfully reproduced absorption spectra and 2D spectra (at $T=0$ and finite $T$) consistent with previous literature.
• Performance: Calculations were performed on a single Intel i9 core.
  • Constructing the propagator $Q$ took seconds (e.g., 0.2s to 6.3s depending on $\chi$).
  • Computing the response function over a $100 \times 100$ frequency grid took from 0.06s to 87s.
  • Memory usage remained constant regardless of the waiting time $T$, as only the single tensor $Q$ needs to be stored.
• Waiting Time Evolution: The 2D spectra for large waiting times (e.g., $T=10$ ps) were generated instantly from the same eigendecomposition used for $T=0$, demonstrating the efficiency of the spectral approach.

5. Significance

This work represents a significant advancement in the simulation of non-Markovian quantum dynamics:

• Computational Efficiency: By moving from time-domain propagation to frequency-domain spectral representation, the method bypasses the "curse of dimensionality" associated with long-time real-time evolution.
• Broad Applicability: While demonstrated on 2D spectroscopy, the framework is general and applicable to any multi-time correlation function, including higher-order responses relevant to femtosecond spectroscopy and quantum transport.
• Future Directions: The authors note that the method can be extended to systems with multiple operators and correlated environments, making it a powerful tool for studying complex quantum materials and biological light-harvesting systems where memory effects are dominant.

In summary, the paper establishes uniTEMPO as a superior, numerically exact tool for calculating multi-time correlation functions in non-Markovian regimes, offering a unique combination of high accuracy, low memory footprint, and direct access to frequency-domain observables.