⚛️ quantum physics

Stochastic unravelings for Heisenberg picture and trace-nonpreserving dynamics

This paper introduces a general framework for stochastic unravelings that extends efficient simulation techniques to arbitrary trace-nonpreserving master equations, enabling the study of diverse open system dynamics including Heisenberg picture evolutions, non-Hermitian generators, and full counting statistics through stochastic disappearance and replication of realizations.

Original authors: Federico Settimo, Kimmo Luoma, Dariusz Chruściński, Bassano Vacchini, Andrea Smirne, Jyrki Piilo

Published 2026-04-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Federico Settimo, Kimmo Luoma, Dariusz Chruściński, Bassano Vacchini, Andrea Smirne, Jyrki Piilo

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Simulating a Leaky Quantum World

Imagine you are trying to predict the weather. In the old days, scientists used a simple rule: "The total amount of water in the atmosphere stays the same; it just moves from clouds to rain to rivers." This is like the Schrödinger picture in quantum physics. It assumes the system is "closed" and nothing is lost or gained. To simulate this, scientists use a method called Stochastic Unraveling.

Think of Stochastic Unraveling like a massive simulation game. Instead of calculating the exact weather for the whole world (which is too hard), you run thousands of tiny, separate "mini-weather scenarios" (called trajectories). Some scenarios have a storm, some have sunshine. If you average out all these mini-scenarios, you get the accurate, real-world forecast.

The Problem:
Real quantum systems aren't perfect closed boxes. They leak energy, particles disappear, or new ones appear.

The "Leaky" Problem: Sometimes the system loses particles (trace-decreasing), and sometimes it gains them (trace-increasing). The old simulation rules break because you can't just average them if the total number of "mini-scenarios" changes.
The "Backwards" Problem: Usually, we simulate how a state (the weather) changes over time. But sometimes, physicists want to know how a measurement tool (the thermometer) changes as it interacts with the system. This is the Heisenberg picture. The old simulation tools didn't work well here.

The Solution:
This paper introduces a new, super-flexible simulation toolkit. It allows scientists to simulate these "leaky" and "backwards" quantum systems by letting the number of mini-scenarios in their computer program grow or shrink dynamically.

The Core Metaphor: The "Copy-Paste" and "Delete" Buttons

To understand how the authors fixed the simulation, imagine you are managing a team of reporters (the stochastic realizations) covering a story.

1. The Old Way (Standard Schrödinger Picture)

In the old method, you have exactly 1,000 reporters. Every minute, they move around, change their stories, or jump to a new location. But the total number of reporters stays exactly 1,000. You average their reports to get the truth.

Limitation: This only works if the story doesn't get bigger or smaller.

2. The New Way (Trace-Non-Preserving)

In the new method, the story is dynamic.

If the story gets smaller (Trace-Decreasing): Imagine a reporter gets fired or quits. They disappear from the team.
- The Fix: The simulation has a "Delete" button. If a reporter's path becomes unlikely (probability < 1), they vanish. The remaining reporters' reports are weighted to account for the missing ones.
If the story gets bigger (Trace-Increasing): Imagine the story is so exciting that the news station clones a reporter.
- The Fix: The simulation has a "Copy-Paste" button. If a path becomes more likely (probability > 1), the reporter duplicates themselves. Now you have two identical reporters covering the same angle.

The Result: By letting the team size grow (cloning) or shrink (deleting), the simulation can perfectly track systems that gain or lose energy/particles, which was impossible with the old "fixed team size" rule.

The "Heisenberg" Twist: Watching the Camera, Not the Actor

In physics, there are two ways to watch a movie:

Schrödinger: You watch the actor (the quantum state) move around the stage.
Heisenberg: You watch the camera (the measurement operator) move around the stage. The actor stays still, but the camera zooms, pans, and focuses.

Usually, simulating the "camera" (Heisenberg) is much harder because the camera can point at things that aren't "real" (mathematically, they aren't positive numbers).

The Paper's Trick:
The authors realized that even if the "camera" is pointing at weird, non-physical things, you can break the camera down into four simple, positive parts (like breaking a complex color into Red, Green, Blue, and Yellow). You simulate each of these four parts separately using the "Copy-Paste/Delete" method described above, and then stitch them back together at the end.

Why is this cool?
Sometimes, a system looks "messy" and impossible to simulate when watching the actor (Schrödinger), but looks "clean" and easy when watching the camera (Heisenberg).

Analogy: Trying to track a chaotic swarm of bees (Schrödinger) is hard. But if you just track the wind patterns that move the bees (Heisenberg), the wind might be smooth and predictable.
The Benefit: This new method allows scientists to choose the "easier" perspective (Heisenberg) to run their simulations, saving massive amounts of computer power.

Real-World Applications Mentioned

The paper shows this isn't just theory; it works for real physics problems:

Non-Hermitian Hamiltonians: These are systems where energy is lost or gained in a weird way (like a laser that gets brighter or dimmer spontaneously). The new method simulates this by cloning reporters when the laser gets brighter and firing them when it gets dimmer.
Photon Counting (Counting Light Particles): Scientists want to know the statistics of how many photons (light particles) hit a detector. This often involves "tilted" equations that don't preserve the total number of particles.
- The Application: The paper shows how to use this "Copy-Paste" method to calculate the average number of photons and how much they fluctuate, which is crucial for quantum communication and sensing.

Summary

In a nutshell:
This paper gives physicists a new set of "simulation rules" for the quantum world.

Old Rule: Keep the number of simulation paths constant. (Only works for closed systems).
New Rule: Let the simulation paths clone themselves when the system gets "bigger" and delete themselves when the system gets "smaller."
Bonus: It works for the "Heisenberg picture" (watching the measurement tool), which is often computationally cheaper and more efficient than watching the state itself.

This allows scientists to simulate complex, open quantum systems—like those found in quantum computers or biological energy transfer—with much higher efficiency and accuracy than before.

1. Problem Statement

Stochastic unravelings (e.g., Monte-Carlo Wave Function or Rate Operator methods) are powerful tools for simulating open quantum system dynamics by averaging over stochastic pure-state trajectories. However, traditional applications are restricted to the Schrödinger picture and Trace-Preserving (TP) master equations (MEs), which describe standard Lindblad dynamics.

The paper addresses two significant limitations in current simulation methodologies:

Heisenberg Picture Dynamics: Many open system problems are naturally formulated in the Heisenberg picture (evolving operators rather than states). Standard unraveling techniques do not directly apply here because the dynamical map is unital but not necessarily trace-preserving, and operators are not necessarily positive or Hermitian.
Trace-Non-Preserving (TNP) Dynamics: Many physically relevant scenarios involve TNP MEs, such as:
- Dynamics generated by non-Hermitian Hamiltonians (interpolating between non-Hermitian and Lindblad regimes).
- Full counting statistics (tilted Lindbladians) used to derive moments of probability distributions for observables like photon counts.
- Systems with particle loss or gain.

Existing methods for TNP dynamics often rely on non-linear equations or fail to preserve the norm of state vectors, making them unsuitable for arbitrary TNP MEs in a general framework.

2. Methodology

The authors propose a general framework that extends piecewise-deterministic processes (PDPs) to the Heisenberg picture and arbitrary TNP MEs, requiring only that the dynamics preserve positivity and Hermiticity.

Core Concept: Variable Number of Realizations

In standard Schrödinger unravelings, the trace is preserved because the total number of stochastic realizations ( $N$ ) is constant. In the proposed framework, the trace evolution is mapped to the time-varying number of stochastic realizations.

Trace Decrease ( $p_T < 1$ ): If the total probability of jump and deterministic evolution is less than 1, the "missing" probability corresponds to the disappearance (deletion) of a stochastic realization.
Trace Increase ( $p_T > 1$ ): If the total probability exceeds 1, the excess corresponds to the creation (cloning) of a new copy of the current realization.

Handling Arbitrary Operators (Heisenberg Picture)

Since Heisenberg operators $X$ are not necessarily positive or Hermitian, they cannot be directly represented as density matrices. The authors decompose any operator $X$ into Hermitian and anti-Hermitian parts ( $X = X_h + iX_a$ ), and further into differences of positive operators ( $X_{h,a} = X^+ - X^-$ ). Each positive component is unraveled separately, and the final solution is reconstructed as a weighted sum.

Specific Unraveling Schemes

The framework is demonstrated using two primary schemes:

Monte-Carlo Wave Function (MCWF):
- Jumps are defined by operators $\tilde{H}, R_j, \Gamma_H$ from the Heisenberg ME.
- The algorithm includes three possible events per time step $dt$:
  1. Jump: $|\psi\rangle \to R_j|\psi\rangle / \|R_j|\psi\rangle\|$ .
  2. Deterministic Evolution: $|\psi\rangle \to (1-iK dt)|\psi\rangle / \|\dots\|$ .
  3. Trace Adjustment: Either disappearance (if trace decreases) or cloning (if trace increases).
- This method handles non-Markovian dynamics (negative rates) via Reverse Jumps (NMQJ), where a trajectory jumps back to a previous state.
Rate Operator (RO) Unraveling:
- Uses a spectral decomposition of a rate operator $R_\psi$ to define jumps.
- The sum of jump and deterministic probabilities ( $p_T$ ) determines the cloning/disappearance rate, similar to MCWF.
- Advantage: The RO method can handle non-CP-divisible dynamics (negative rates) without requiring the mutual dependence of trajectories (unlike NMQJ in MCWF), significantly improving numerical efficiency.

3. Key Contributions

Generalization to Heisenberg Picture: The first general method to unravel Heisenberg picture master equations, allowing for the simulation of operator dynamics where the map is unital but not TP.
Unified TNP Framework: A single framework that handles both trace-decreasing and trace-increasing processes through stochastic disappearance and replication, applicable to arbitrary TNP MEs (not just specific cases like tilted Lindbladians).
Efficiency in Non-Markovian Regimes: Demonstrated that the Heisenberg picture can be CP-divisible (allowing standard MCWF) even when the corresponding Schrödinger picture is P-indivisible (requiring complex reverse jumps). This allows for more efficient simulations in specific regimes.
Application to Full Counting Statistics: Showed how the method can be applied to tilted Lindbladians to compute moments of probability distributions (e.g., photon counting statistics) by solving inhomogeneous MEs for derivatives of the density matrix.

4. Results and Examples

The authors validated the framework through several numerical examples:

Heisenberg CP-Divisible Dynamics:
- Simulated a driven two-level system where the Heisenberg picture rates were positive (CP-divisible), but the Schrödinger picture rates were not.
- Result: The MCWF unraveling in the Heisenberg picture matched the exact solution without needing reverse jumps, whereas the Schrödinger picture would have required them. The trace of the observable oscillated correctly, matching the ratio of the number of surviving trajectories to the initial number.
Heisenberg Non-CP-Divisible Dynamics:
- Simulated phase-covariant dynamics where rates were negative (violating CP-divisibility) but satisfied P-divisibility.
- Result: The Rate Operator (RO) method successfully simulated the dynamics without reverse jumps, maintaining high efficiency. The trace of the operator increased monotonically, correctly captured by the cloning mechanism.
Photon Counting Statistics (Tilted Lindbladians):
- Applied the method to a bosonic system to compute factorial moments of the photon number distribution.
- Result: By solving a hierarchy of TNP MEs for the derivatives of the density matrix ( $\tau_k$ ), the method accurately reproduced the first four moments of the distribution, demonstrating its utility for full counting statistics beyond standard Heisenberg MEs.

5. Significance

This work significantly broadens the scope of stochastic simulation techniques in quantum optics and open quantum systems:

Computational Efficiency: By exploiting the different divisibility properties of the Schrödinger and Heisenberg pictures, simulations can be made more efficient (avoiding the storage overhead of reverse jumps).
Versatility: It enables the simulation of a wide class of non-standard dynamics, including non-Hermitian Hamiltonians and systems with particle exchange, which were previously difficult to simulate stochastically.
Statistical Analysis: It provides a robust tool for calculating full counting statistics and higher-order moments, which are crucial for understanding fluctuations in quantum transport and thermodynamics.
Theoretical Unification: It unifies the treatment of trace-preserving and trace-non-preserving dynamics under a single stochastic paradigm, bridging the gap between standard quantum trajectory methods and advanced statistical mechanics approaches (like cloning algorithms).