🔬 physics

Benchmarking Open-Access Quantum Hardware Through Many-Body Dynamics in the 1D Tilted-Field Ising Model

This study benchmarks the IBM \texttt{ibm\_marrakesh} quantum hardware by simulating a 15-qubit tilted-field Ising model, revealing that while the device captures early-time magnetization trends, noise and decoherence significantly degrade long-range correlation spreading and local magnetization fidelity compared to ideal simulations.

Original authors: Samarth Lamba, Aleksandra A. Ziolkowska

Published 2026-07-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Samarth Lamba, Aleksandra A. Ziolkowska

Original paper licensed under CC BY 4.0 (https://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: Testing a Noisy Quantum Computer

Imagine you have a brand-new, high-tech robot designed to solve complex puzzles. You want to know if it actually works as advertised or if it just makes a lot of noise and guesses.

This paper is a "stress test" for a real, publicly available quantum computer (specifically, one called IBM Marrakesh). The researchers wanted to see if this machine could accurately simulate a specific physics problem: a chain of 15 tiny magnets (qubits) interacting with each other and an external magnetic field.

They compared the robot's performance against a perfect, "ideal" simulation running on a powerful classical computer. The goal wasn't to build a new product, but to see how much of the real physics the noisy machine could actually preserve.

The Setup: A Line of 15 Spinners

Think of the system as a line of 15 children standing in a row, each holding a spinning top.

The Rules (The Hamiltonian): The children are told to follow specific rules. They want to face the same direction as their neighbors (like magnets attracting), but there is also a wind blowing them sideways (the magnetic field).
The Disturbance: At the start, everyone is facing "North" (spin-up), except for the child in the exact middle (the 8th child), who is forced to face "South" (spin-down).
The Goal: Watch how the "South" facing child influences the others over time. Does the "South" feeling spread down the line? Do the tops start spinning in sync?

The researchers ran this scenario in two ways:

The Ideal Simulation: A perfect computer calculation where nothing goes wrong.
The Real Hardware: The actual IBM quantum computer, which is like a robot trying to do the same dance but has shaky hands and gets tired easily.

The Experiment: Two Different Scenarios

The researchers tested two main versions of the "wind" (magnetic field):

The "Straight" Wind: The wind only blows from the side. This is a predictable, orderly system (like a well-rehearsed dance).
The "Tilted" Wind: The wind blows from the side and pushes from the front. This makes the system chaotic and much harder to predict (like a dance where the music changes tempo randomly).

They also tested three different strengths of "friendship" between the neighbors (how strongly the children influence each other).

The Results: What Worked and What Didn't

1. The Early Moments (The First Few Seconds)

The Good News: The real quantum computer was surprisingly good at the very beginning. Just like the ideal simulation, the real machine showed that the "South" facing child in the middle started to influence the neighbors. The general trend of the magnets changing direction was visible.

Analogy: If you shout "Hello" in a quiet room, the real robot heard it and shouted back correctly.

2. The Details (The Specifics)

The Bad News: As time went on, the real machine's performance got "squished."

Compressed Values: In the perfect simulation, the magnets showed a wide range of behaviors (some very North, some very South, some in between). On the real machine, everything looked "meh"—the values were all clumped together in the middle.
Lost Separation: The perfect simulation showed clear differences between the child in the middle and the child at the end of the line. The real machine couldn't tell them apart; it looked like everyone was doing the same thing.
Analogy: Imagine a high-definition movie (the ideal simulation) vs. a blurry, low-resolution video (the real hardware). You can see the actors moving in both, but the blurry video loses all the fine details and facial expressions.

3. The Spread of Information (The "Light Cone")

This was the biggest gap. In the perfect simulation, you could clearly see a "wave" of influence spreading out from the center child to the edges, like a ripple in a pond.

On the Real Machine: The ripple started at the center, but it died out very quickly. By the time it should have reached the edges, the signal was just random static.
Analogy: It's like trying to pass a secret message down a line of people in a noisy stadium. The first few people get it right, but by the time it reaches the back of the line, the message is garbled and lost because of the noise.

Why Did the Real Machine Fail?

The paper explains that the real quantum computer is "noisy." It suffers from three main problems:

Shaky Hands (Gate Errors): The machine tries to rotate the magnets, but sometimes it rotates them slightly too far or not far enough.
Bad Hearing (Readout Errors): When the machine tries to "read" the final direction of the magnets, it sometimes mistakes "North" for "South."
Getting Tired (Decoherence): Quantum states are fragile. As the machine runs more steps (time passes), the environment (heat, vibration) causes the quantum information to leak away, turning the precise quantum dance into a messy classical shuffle.

The Conclusion

The study concludes that current open-access quantum computers are good enough to see the "big picture" of what is happening in the very short term, but they cannot yet reliably track complex, long-term interactions or the spread of information across a system.

The machine preserves the idea of the physics, but the details get washed out by noise. To get the full picture in the future, we will need better error correction (like noise-canceling headphones for quantum computers), but right now, those tools are too heavy and complex for these small machines to carry.

In short: The quantum computer passed the "first grade" of physics simulation but is still struggling with the "advanced math" of complex, long-term dynamics.

Technical Summary: Benchmarking Open-Access Quantum Hardware Through Many-Body Dynamics in the 1D Tilted-Field Ising Model

Problem Statement
Hamiltonian simulation is a primary proposed application for near-term quantum computers, yet its practical utility depends on whether physical hardware can preserve intended many-body dynamics amidst sampling limits, gate noise, decoherence, and circuit depth constraints. While the transverse-field Ising model is a standard integrable system, the addition of a longitudinal field creates the tilted-field Ising model, which breaks integrability and introduces complex, non-trivial dynamics including quantum chaos and thermalization. The central problem addressed is determining the extent to which present-day open-access quantum hardware can reproduce these physically meaningful many-body observables compared to ideal, noiseless simulations.

Methodology
The study utilizes a 15-qubit one-dimensional tilted-field Ising chain as a benchmark system, executed on the IBM Quantum backend ibm_marrakesh.

System Initialization: The system is initialized with all qubits in the spin-up state ( $|0\rangle$ ), except for the central qubit ( $q8$ ), which is flipped to spin-down ( $|1\rangle$ ) using an X gate to create a localized excitation.
Hamiltonian and Parameters: The dynamics are governed by the Hamiltonian $H = -J \sum Z_i Z_{i+1} - h_x \sum X_i - h_z \sum Z_i$ . The study compares an integrable reference case ( $h_z = 0$ ) against a non-integrable tilted-field case ( $h_z = h_x = 0.35625$ ) across three interaction regimes: $J < h_x$ , $J = h_x$ , and $J > h_x$ .
Simulation Technique: Time evolution is approximated using a second-order symmetric Trotter decomposition with a timestep $\Delta t = 0.1$ . This choice balances accuracy (local error scales as $O(\Delta t^3)$ ) with circuit depth to remain feasible on open-access hardware.
Observables: Three key observables are tracked to evaluate dynamics:
1. Local Magnetization: $\langle Z_i(t) \rangle$ to track individual spin orientation.
2. Total Magnetization: $M_z(t) = \sum \langle Z_i(t) \rangle$ to measure global spin polarization memory.
3. Connected Spin Correlations: $C_{ij}^{conn}(t) = \langle Z_i Z_j \rangle - \langle Z_i \rangle \langle Z_j \rangle$ to analyze correlation spreading and light-cone structures.
Experimental Execution: Circuits were run on ibm_marrakesh with 2048 shots per circuit. The study employed resilience level 1 with dynamical decoupling (XpXm sequence) but did not use user-implemented zero-noise extrapolation or full quantum error correction. Hardware results were sampled every 0.5 time units up to $t=35$ .

Key Results
The study compares ideal classical Trotterized simulations with hardware execution results:

Local Magnetization: The hardware reproduces the broad early-time trend of magnetization decay. However, hardware values are "compressed," showing reduced separation between qubits compared to the ideal simulation. In the tilted-field cases ( $h_z = h_x$ ), the ideal simulation retains higher local magnetization values and oscillatory structure, whereas hardware data show a faster collapse toward zero and reduced site-to-site variation.
Total Magnetization: Similar to local magnetization, hardware data show a rapid initial drop followed by a lower steady state compared to the ideal simulation. In the tilted-field regime, the ideal simulation maintains significant positive magnetization (memory of initial state), while hardware data show a much stronger reduction, indicating a loss of global magnetization memory.
Connected Correlations: The ideal simulation exhibits a clear "light-cone" structure where correlations spread outward from the central excitation ( $q8$ ) in a structured manner. In contrast, the hardware results display a significantly weaker and noisier pattern. The distinct spreading front is largely obscured, replaced by scattered low-intensity fluctuations, particularly at longer times and further distances from the center.

Key Contributions

Benchmarking Protocol: The paper establishes a reproducible, observable-level benchmark for correlation-spreading fidelity in a non-integrable spin-chain system on Heron-class IBM hardware.
Quantitative Comparison: It provides a direct, quantitative comparison of ideal versus noisy dynamics for a 15-qubit system, specifically isolating the effects of breaking integrability on hardware performance.
Noise Characterization: The study demonstrates that while open-access hardware can capture coarse, early-time features of intended dynamics, longer-time correlation spreading and detailed oscillatory structures are increasingly obscured by finite sampling, two-qubit gate errors, readout errors, and decoherence.

Significance and Claims
The paper modestly claims that the study serves as a near-term benchmark rather than a fully error-corrected simulation. Its primary significance lies in demonstrating the current limits of open-access quantum hardware in preserving physically meaningful many-body behavior.

Preservation of Dynamics: The hardware preserves some early-time features (general magnetization trends) but fails to reliably preserve long-time correlation spreading.
Limitations: The results highlight that without error mitigation or correction, the detailed structure of non-integrable dynamics (such as specific correlation patterns and oscillatory memory) is difficult to extract from noisy data.
Future Directions: The authors suggest that while error mitigation (e.g., readout error mitigation, zero-noise extrapolation) could improve accuracy, full quantum error correction remains a significant overhead for near-term devices. They propose future work involving cross-backend comparisons, smaller spin chains to isolate noise, and the inclusion of more complex observables like out-of-time-order correlators (OTOCs) and entanglement entropy to further probe quantum chaos and information spreading.