⚛️ quantum physics

Parameter estimation for quantum jump unraveling

This paper presents a comprehensive framework for estimating parameters in continuously monitored quantum systems under jump unraveling by deriving computable Fisher Information expressions for multi-channel renewal processes, introducing a hybrid algorithm combining monitoring operators and the Gillespie method for non-renewal processes, and providing tools to account for information loss in data compression or post-selection.

Original authors: Marco Radaelli, Joseph A. Smiga, Gabriel T. Landi, Felix C. Binder

Published 2026-02-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Marco Radaelli, Joseph A. Smiga, Gabriel T. Landi, Felix C. Binder

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

Imagine you are trying to guess the secret settings of a mysterious machine. You can't open the machine to look at the dials, and you can't ask the machine directly. The only way to learn about it is to watch a stream of "blips" or "jumps" it makes on a screen. Sometimes the blips are red, sometimes blue, and they happen at random times.

This paper is about the best way to figure out the machine's secret settings (like temperature or energy levels) just by watching these blips. The authors provide a new "toolkit" for scientists to calculate exactly how much information they can get from these blips, even when the blips are tricky and remember things from the past.

Here is a breakdown of their ideas using simple analogies:

1. The Problem: The "Memory" of the Machine

In many simple experiments, every blip is independent. It's like rolling a die: if you roll a 6, it doesn't change the odds of the next roll. But in quantum systems, the blips often have memory.

The Analogy: Imagine a game of "Hot and Cold." If you take a step and get "hotter," that tells you something about where the treasure is. But if you take another step and get "colder," that new information depends on where you were before. The path you took matters.
The Challenge: Because each blip depends on the previous ones, it's very hard to calculate exactly how much information you are gathering. The math gets messy because the "blips" are tangled together in time.

2. The Two Types of Machines

The authors realized there are two main types of machines (quantum processes) and they needed different tools for each:

Type A: The "Reset" Machine (Renewal Processes)

How it works: Every time a blip happens, the machine completely forgets its past and starts fresh. It's like a vending machine that, after you get a snack, resets to the exact same state, ready for the next person.
The Solution: For these machines, the authors found a simple formula. They showed that you can calculate the total information just by looking at the average time between blips and which color blip appeared. It's like counting how many red and blue marbles you pull from a bag that refills itself perfectly every time.

Type B: The "Forgetful" Machine (Non-Renewal Processes)

How it works: This is the tricky one. When a blip happens, the machine changes, but it doesn't fully reset. It keeps a "ghost" of what happened before. The next blip depends on the whole history of the machine.
The Solution: Since there is no simple formula for this, the authors invented a new computer algorithm called the "Fisher-Gillespie Algorithm."
- How it works: Instead of trying to solve a giant, impossible math equation, the computer simulates the machine thousands of times. It watches thousands of fake "movies" of the blips, calculates the information for each movie, and then averages them all together.
- The Benefit: This is like trying to guess the weather pattern by simulating a million days of weather on a computer rather than trying to predict the future with a single equation. It's fast, efficient, and handles the "memory" problem perfectly.

3. The "Stochastic" Scorecard

The authors also introduced a cool concept called Stochastic Fisher Information.

The Analogy: Imagine you are a detective solving a case. Usually, you wait until the end of the case to see how good your clues were. But this new method gives you a "score" after every single clue you find.
Why it matters: Sometimes, a specific sequence of blips might be a "bad" path that gives you very little information, while another path is a "goldmine." This scorecard tells you, in real-time, how much you are learning from the specific path the machine is taking right now.

4. What Happens When You Lose Data?

The paper also asks: "What if we don't record everything?"

The Analogy: Imagine you are watching a game, but you only remember when the goals were scored, not which team scored them. Or maybe you only remember the team, but forgot the time.
The Result: The authors show that throwing away information (like forgetting the time or the color of the blip) always lowers your ability to guess the secret settings. They provide tools to calculate exactly how much "precision" you lose when you compress your data.

5. Real-World Examples

To prove their tools work, they tested them on real physics scenarios:

Thermometry: Measuring the temperature of a tiny quantum object by watching how it emits energy.
Coupled Qubits: Two tiny quantum bits interacting, where one "leaks" energy. This is a "non-renewal" case where the memory effect is strong, and their new algorithm was essential to solve it.
The Maser: A device where atoms fly through a cavity. They showed how the initial state of the atoms changes how much information you can get about the machine's settings.

Summary

In short, this paper gives scientists a complete set of instructions for measuring the unknown in quantum systems that emit random signals.

If the system resets after every signal, use their simple formula.
If the system remembers its past, use their new computer simulation algorithm.
If you lose some data, use their compression tools to know how much accuracy you sacrificed.

This allows researchers to know exactly how precise their measurements can be before they even build the experiment.

Technical Summary: Parameter Estimation for Quantum Jump Unraveling

Problem Statement
The paper addresses the challenge of estimating a physical parameter $\theta$ encoded in the measurement record of a continuously monitored quantum system, specifically within the framework of quantum jump unraveling. In this scenario, an experimenter observes a single-shot trajectory consisting of discrete jumps occurring at random times ( $\tau_i$ ) and in random channels ( $k_i$ ). Unlike standard quantum metrology which often assumes independent and identically distributed (i.i.d.) copies or allows for optimization over measurement strategies, this work focuses on "correlated classical metrology." The experimenter has no control over the measurement strategy and must estimate $\theta$ based solely on the observed stochastic record $\omega_{1:N} = \{(\tau_1, k_1), \dots, (\tau_N, k_N)\}$ .

The primary difficulty arises from intricate temporal correlations and memory effects inherent in the detection record. For general processes, the joint probability distribution of the trajectory does not factorize, making the direct calculation of the Fisher Information (FI)—the standard metric for estimation precision—computationally intractable due to high-dimensional sums and integrals. Furthermore, while bounds on the unconditional state exist, they are generally not saturable in the jump unraveling scenario.

Methodology
The authors develop a comprehensive framework to compute the Fisher Information for two distinct classes of quantum jump processes:

Multi-channel Renewal Processes:
For processes where the system state is reset to a specific state $\sigma_k$ after a jump in channel $k$ (satisfying the renewal condition $J_k\rho = \text{Tr}[J_k\rho]\sigma_k$ ), the authors demonstrate that the measurement record forms a Markov chain. They derive an analytical expression for the FI that relates it directly to the Waiting Time Distribution (WTD), $W(\tau, k|q)$ , which describes the probability of a jump occurring at time $\tau$ in channel $k$ given the previous jump was in channel $q$ . The total FI is shown to be linear in the number of jumps $N$ and can be decomposed into contributions from the sequence of channels and the conditional jump times.
Non-Renewal Processes:
For general processes where the post-jump state depends on the pre-jump state (e.g., optical cavities or coupled qubits), the memory structure is long-lived, and the process is not a simple Markov chain. To address this, the authors introduce a new numerical algorithm combining:
- The Monitoring Operator Formalism: A method to compute the FI by sampling over quantum trajectories using a monitoring operator $\xi_N = \partial_\theta \tilde{\rho}_N / \text{Tr}(\tilde{\rho}_N)$ , where $\tilde{\rho}_N$ is the unnormalized state after $N$ jumps. The FI is given by the ensemble average $F_{1:N}(\theta) = E[\text{Tr}(\xi_N)^2]$ .
- The Quantum Gillespie Algorithm: A stochastic simulation method adapted to sample jump trajectories efficiently. The authors integrate the monitoring operator evolution into the Gillespie steps, allowing for the efficient sampling of a "stochastic Fisher Information" along individual trajectories. This approach avoids the numerical instabilities associated with calculating likelihoods for long trajectories by working with the trace of the monitoring operator.

Additionally, the paper provides tools for analyzing data compression and post-selection. Using the data-processing inequality, the authors show how to compute the FI when information is lost, such as when jump channels are indistinguishable, time tags are discarded, or only summary statistics (like average waiting times) are used.

Key Results

Analytical Formula for Renewal Processes: The authors derive Eq. (17), a computable expression for the FI of multi-channel renewal processes based on the WTD. They further demonstrate that the FI can be split into channel-only and time-conditional components (Eq. 21), allowing for the assessment of which aspect of the data carries more information.
Stochastic Fisher Information: For non-renewal processes, the paper establishes that the stochastic FI, sampled via the proposed algorithm, satisfies useful properties for single-shot estimation. The algorithm allows for the calculation of the FI rate and the variance of estimators without requiring the explicit construction of the full joint probability distribution.
Maximum Likelihood Estimation (MLE): The monitoring operator formalism is utilized to perform MLE on quantum jump trajectories. The authors show that finding the parameter $\theta$ that maximizes the likelihood is equivalent to finding the $\theta$ that makes the trace of the monitoring operator zero ( $\text{Tr}[\xi_{\omega_{1:N}}^\theta] = 0$ ). This provides a numerically stable method for parameter estimation, circumventing the exponential decay of likelihoods in long trajectories.
Information Loss: The paper quantifies the reduction in Fisher Information when data is compressed. For instance, it shows that for renewal processes, marginalizing over time tags retains the renewal property, whereas marginalizing over channels generally destroys it, requiring the numerical approach for FI calculation.

Illustrative Examples
The framework is validated through four examples:

Qubit Thermometry: A renewal process where the FI for estimating temperature is decomposed into channel and time contributions, showing that channel information dominates at low temperatures.
Resonant Fluorescence: A single-channel renewal process where the FI of the full record is compared to that of the sample mean, demonstrating that summary statistics often yield significantly lower precision.
Coupled Qubits: A non-renewal process where one qubit resets but the other retains memory. The algorithm successfully computes the FI and shows that while the average FI increases monotonically, individual stochastic trajectories can exhibit decreases in the squared trace of the monitoring operator, serving as an indicator of trajectory atypicality.
Single-Atom Maser (Micromaser): A complex non-renewal collision model. The study reveals that the FI rate for estimating the coupling constant depends critically on the initial state of the incoming atoms, vanishing when the atom is in the ground state due to the lack of interaction with the near-vacuum cavity field.

Significance and Claims
The paper claims to provide a "full set of solutions" for parameter estimation in quantum jump unravelings, covering both renewal and non-renewal regimes. Its primary contribution is the development of a practical, computationally efficient toolkit (the Gillespie-Fisher algorithm and monitoring operator formalism) that makes the calculation of Fisher Information feasible for complex, correlated quantum trajectories where analytical solutions are unavailable.

The authors position their work as a bridge between theoretical bounds and experimental reality, where data is often limited to single-shot records with memory effects. They state that these tools are essential for:

Computing thermodynamic and kinetic uncertainty relations in quantum jump processes.
Designing and optimizing quantum sensors by understanding how information is encoded in jump times and channels.
Providing a rigorous framework for classical metrology on quantum systems where the measurement strategy is fixed by the experimental setup.

The paper concludes modestly, noting that while the methods are general, they are specifically tailored to the jump unraveling scenario and rely on the specific structure of the quantum master equation. The results are illustrated with examples from quantum optics and condensed matter, demonstrating the applicability of the framework across different physical platforms.