⚛️ quantum physics

A fidelity metric for quantum annealing benchmarked by extreme scaling quantum Monte-Carlo simulations

This paper proposes a fidelity metric to evaluate the intrinsic quality of quantum annealing processes rather than just optimization outcomes, demonstrating through extreme-scale quantum Monte-Carlo simulations that current Rydberg atom annealers are indistinguishable from thermal classical counterparts and fall short of the precision required for genuine quantum advantage.

Original authors: Gabriel Gouraud, Miha Srdinsek, Xavier Waintal

Published 2026-06-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Gabriel Gouraud, Miha Srdinsek, Xavier Waintal

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 find the lowest point in a vast, foggy mountain range. This is what a Quantum Annealer tries to do: it searches for the perfect solution to a complex puzzle (like scheduling a train network or managing a financial portfolio) by slowly "cooling down" a system until it settles into its most stable state.

For years, scientists have judged these machines by one simple question: "Did they find the right answer?"

The authors of this paper argue that this is a bit like judging a chef only by whether the customer finished their meal, without tasting the food to see if it was cooked correctly. If the customer is hungry enough, they might eat a burnt steak and still be satisfied. Similarly, a quantum computer might get the right answer by luck, or it might fail in a way that looks like a right answer but is actually wrong.

The New Metric: Checking the "Recipe" Instead of the "Plate"

The authors propose a new way to test these machines. Instead of just looking at the final answer, they want to measure how well the machine followed the rules of physics during the process.

They call this metric the "Equation of State Accuracy."

Think of it this way:

The Old Way: You ask the quantum computer, "What is the best schedule for the trains?" It gives you an answer. You check if it works.
The New Way: You ask the computer, "As you were slowly changing the schedule, what was the energy level at every single step?"

If the computer is truly doing "quantum annealing," its energy levels should follow a very specific, smooth curve (like a perfect slide down a hill). If the machine is noisy, broken, or just acting like a regular classical computer, that curve will be jagged or wrong. By measuring how closely the machine's energy curve matches the perfect theoretical curve, the authors can tell if the machine is actually doing quantum magic or just faking it.

The "Thermal" vs. "Quantum" Battle

A big question in this field is: Is the quantum computer actually using quantum mechanics (like tunneling through walls), or is it just acting like a regular computer that uses heat (like shaking a box of marbles) to find the solution?

To answer this, the authors built a "super-classical" simulator. They created a digital model that acts like a thermal annealer (a heat-based solver) but is tuned so perfectly that it tries to mimic the quantum machine.

The Analogy: Imagine you are trying to guess the secret recipe of a famous chef. Instead of just tasting the dish, you try to recreate the dish using only a standard kitchen. If your standard kitchen can make a dish that tastes exactly like the chef's (within a tiny margin of error), then maybe the chef isn't using any secret "quantum" ingredients after all.

What They Found

The authors ran these simulations on two types of systems:

Rydberg Atoms: These are atoms excited to a high energy state, used in some of the newest quantum computers.
A Real-World Financial Problem: A specific puzzle about predicting "fallen angel" loans (loans that are likely to default).

The Results:

The Scale: They simulated systems with up to 100 million atoms on a single standard computer chip. This is massive.
The Precision: They found that to beat their classical "thermal" simulator, a real quantum annealer needs to be incredibly precise.
- If the machine's error is around 1% to 0.1% (10⁻² to 10⁻³), it is indistinguishable from a classical computer using heat. It's not doing anything special.
- To show a real advantage, the machine needs to be precise to 0.01% (10⁻⁴) or better.
The Reality Check: The authors compared their results to current experimental machines (like those using Rydberg atoms). They found that current machines are orders of magnitude less precise than what is needed to prove they are doing something truly quantum. The "noise" in current machines is too high; they are essentially just acting like very expensive, very slow classical computers.

The "Gap" Problem

The paper also highlights a specific moment in the process called the "Gap Closing."
Imagine the mountain range has a very narrow, deep canyon. To get to the bottom, you have to cross it.

In a perfect quantum world, the machine can "tunnel" through the canyon.
In the real world, if the machine is too slow or too noisy, it gets stuck on the edge.

The authors found that their simulation is most sensitive to errors right at this canyon. If a real machine can't measure the energy curve accurately at this specific point, it means it's failing to solve the hardest part of the problem.

Summary

In simple terms, this paper says:

Stop just asking "Did you get the answer?"
Start asking "Did you follow the quantum rules correctly while getting there?"
They created a new test (measuring the energy curve) that acts like a lie detector for quantum computers.
Their tests show that current quantum annealers are still too "noisy" and imprecise to prove they are doing anything better than a very smart classical computer. They need to get much more accurate before they can claim a true advantage.

Technical Summary: A Fidelity Metric for Quantum Annealing Benchmarked by Extreme Scaling Quantum Monte-Carlo Simulations

Problem Statement
Current benchmarks for Quantum Annealing (QA) are predominantly pragmatic, measuring performance solely by the speed or success rate of solving specific optimization problems (typically formulated as Quadratic Unconstrained Binary Optimization, or QUBO). The authors argue this approach is "inherently brittle" because results depend heavily on the specific problem instance and the classical algorithm used as a reference. Furthermore, this methodology fails to distinguish between intrinsic limitations of the annealing process (e.g., exponentially vanishing energy gaps leading to Landau-Zener transitions) and extrinsic hardware limitations (e.g., noise, decoherence, finite temperature). Consequently, there is a lack of a robust metric to quantify the fidelity of the quantum annealing process itself, independent of the final solution's quality.

Methodology
The paper proposes a new fidelity metric: the accuracy ( $\epsilon$ ) of the equation of state $E(h_x) \equiv \langle \hat{H} \rangle$ during the annealing process. This metric is analogous to "fidelity-per-gate" in gate-based quantum computing. To obtain the equation of state experimentally, one would measure both the longitudinal energy ( $\langle \hat{H}_z \rangle$ ) and the transverse field energy ( $\langle \hat{H}_x \rangle$ ) at various points in the annealing schedule.

To establish benchmark values for $\epsilon$ and test the limits of classical simulation, the authors employ two variants of Simulated Quantum Annealing (SQA) using Quantum Monte-Carlo (QMC) techniques on Rydberg atom systems:

Variational Quantum Monte-Carlo (VMC):
- Ansatz: The authors use a "thermal ansatz" inspired by simulated thermal annealing. The wavefunction is modeled as $\psi(n) \propto e^{-E_{\text{eff}}(n)/2}$ , where the effective energy $E_{\text{eff}}(n)$ is a modified classical energy functional.
- Variants: They test a "simple thermal ansatz" ( $E_{\text{eff}} = \beta_{\text{eff}} F(n)$ ) and a "staggered thermal ansatz" ( $E_{\text{eff}} = \beta_{\text{eff}} F(n) + \sum \beta_k N_k(n)$ ) to capture symmetry breaking.
- Scale: This approach is computationally lightweight, allowing simulations on a single CPU up to 100,000,000 atoms ( $N=10^8$ ).
- Accuracy: Achieves an estimated accuracy of $\epsilon \sim 10^{-2} - 10^{-3}$ .
Green Function Monte-Carlo (GFMC):
- Method: A post-VMC method that evolves the variational ansatz in imaginary time to project onto the true ground state. It is unbiased and free of the sign problem for transverse field Ising models.
- Scale: Simulations performed up to 100,000 atoms ( $N=10^5$ ).
- Accuracy: Achieves higher precision with $\epsilon \sim 10^{-4}$ .

The authors utilize these methods to calculate the equation of state for Rydberg atoms on square and triangular lattices, as well as for a real-world QUBO instance ("fallen angel problem" related to financial risk management).

Key Results

Scalability and Precision: The VMC approach with the thermal ansatz successfully simulated systems up to $10^8$ atoms with a relative error $\delta E/E_0 < 10^{-2}$ . The GFMC approach reached $\epsilon \sim 10^{-4}$ for $N=10^5$ .
Error Localization: The relative error in the equation of state peaks precisely at the location of the energy gap closing (phase transitions), which corresponds to the regime where Landau-Zener transitions occur. Away from these critical points, the error is significantly lower.
Comparison to Experiments: The authors estimate that current Rydberg atom experimental platforms operate with an accuracy significantly lower than these classical benchmarks. The classical simulations outperform current experimental platforms in both system size and precision by orders of magnitude.
QUBO Benchmarking: For the "fallen angel" QUBO instance, the simple thermal ansatz yielded a relative error of $\approx 10^{-3}$ . The authors note that while the final solution error ( $\delta F/F$ ) can be arbitrarily reduced by increasing the number of shots, the equation of state error ( $\delta E/E_0$ ) provides a more intrinsic measure of the hardware's or algorithm's capability.
Thermal vs. Quantum: The study suggests that within an accuracy of $\epsilon \sim 10^{-2} - 10^{-3}$ , a quantum annealer is effectively indistinguishable from its thermal classical counterpart, implying that current experimental noise levels may mask genuine quantum advantages.

Significance and Claims
The paper claims that measuring the accuracy of the equation of state is a crucial step forward for QA benchmarking because:

It provides a direct diagnostic of the dominant mechanisms causing deviations from ideal adiabatic evolution (distinguishing intrinsic gap issues from extrinsic noise).
It offers a hardware-agnostic metric that does not rely on the specific difficulty of a chosen QUBO instance.
It sets a severe constraint for future hardware: to demonstrate a quantum advantage, experimental platforms must achieve an equation of state accuracy that surpasses these classical benchmarks (specifically $\epsilon < 10^{-4}$ for large systems).

The authors conclude that the "deceptively simple" thermal variational ansatz is a powerful tool for simulating QA, suggesting that the path to understanding quantum advantages may lie in identifying regimes where quantum fluctuations provide a distinct advantage over even the most sophisticated classical thermal simulations. They do not claim that QA is currently superior to classical heuristics; rather, they provide the tools to rigorously test that hypothesis in the future.