Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Paper 13 - Cross-Device and Cross-Temperature Noise Structure in Variational Gibbs-State Preparation: A Full-Tomography Held-Out Reanalysis

Roshankumar chandaliya

Published Oct 6, 2026• 0 citations
Publisher

Abstract

Overview This study investigates how quantum noise affects variational Gibbs-state preparation on trapped-ion quantum computers. What Was Analyzed The analysis uses full two-qubit tomography instead of relying only on a single quantity such as fidelity or effective temperature. It compares complete quantum states across different field settings, three trapped-ion devices, and two temperature regimes. Noise Models Compared Several noise descriptions are tested, including scalar depolarization, Pauli-channel models, coherent-error corrections, and generalized amplitude damping (GAD). Predictive Testing The models are trained on selected conditions and then tested on unseen conditions without refitting. This allows the analysis to measure whether a noise model remains useful outside the data used to fit it. Main Result A shared GAD-based model shows the most stable predictive behaviour across the tested devices and temperatures. In the strongest frozen cross-temperature test, parameters fitted at beta = 5 are applied directly at beta = 1 without refitting. The model: improves all 9 of 9 QPU cases reduces mean Hilbert-Schmidt distance from 0.08644669 to 0.07277861 gives a 15.81% reduction in mean error Interpretation The results support a shared damping-like noise-model family within the tested system. However, the fitted parameters are not universal, and the study does not claim a unique microscopic noise mechanism, universal hardware-noise parameters, retrocausality, or a universal physical law. Reproducibility The calculations are also implemented in QDL Research Suite v1.9.0, allowing the main numerical results to be reproduced and independently checked. Software DOI: 10.5281/zenodo.23181558

DOI
10.5281/zenodo.23183457

Related papers

Variational Gibbs state preparation on trapped-ion devices

Reece Robertson, Mirko Consiglio, Josey Stevens +3 more

We implement a variational quantum algorithm for Gibbs state preparation of a transverse-field Ising model on IonQ’s quantum computers. To this end, we train the variational parameters via classical simulation and perform state tomography on the quantum devices to evaluate the fidelity of the prepared Gibbs state. As a main result, we find that fidelity decreases (non-monotonically) as a function of the inverse temperature β of the system. Fidelity also decreases as a function of the size of the system. Interestingly, we find that a Gibbs state prepared for a specified β is a better representative of a Gibbs state prepared for a lower β ; or in other words, thermal fluctuations in the quantum hardware lead to digital heating, that is, an increase in the temperature of the prepared Gibbs state above what was intended.

npj Quantum Information•Jul 11, 2026•0 citations

Variational Gibbs State Preparation on Trapped-Ion Devices

Reece Robertson, Mirko Consiglio, Josey Stevens +3 more

We implement a variational quantum algorithm for Gibbs state preparation of a transverse-field Ising model on IonQ's quantum computers. To this end, we train the variational parameters via classical simulation and perform state tomography on the quantum devices to evaluate the fidelity of the prepared Gibbs state. As a main result, we find that fidelity decreases (non-monotonically) as a function of the inverse temperature $β$ of the system. Fidelity also decreases as a function of the size of the system. Interestingly, we find that a Gibbs state prepared for a specified $β$ is a better representative of a Gibbs state prepared for a $\textit{lower}$ $β$; or in other words, thermal fluctuations in the quantum hardware lead to digital heating, that is, an increase in the temperature of the prepared Gibbs state above what was intended.

quant-pharXiv:2603.03801•Mar 5, 2026•6 citations

A pre-registered entanglement-forged H₄ benchmark on trapped-ion hardware: accurate noise modelling, incomplete error correction

Venkata Rao Allu

We report a pre-registered benchmark of an entanglement-forged H4 ground-state calculation on IonQ Forte Enterprise 1, a live trapped-ion processor. The analysis was frozen and publicly committed before any hardware data existed. The experiment comprised one submission of 81 new circuits at 1,100 shots per circuit, with device debiasing disabled. A three-parameter analytic noise model, fixed in advance, predicted the uncorrected energy error to within approximately 1%: 23.84 kcal/mol measured versus 24.06 kcal/mol predicted. However, the frame-free corrected estimator returned an error of 3.46 kcal/mol, compared with 0.83 kcal/mol predicted. This missed the conventional chemical-accuracy threshold of 1 kcal/mol. The shot-noise-weighted chi-square per degree of freedom increased from approximately 2 in simulator rehearsals to 5.93 on hardware. Analysis of 732 residuals identified elevated discrepancies in strongly corrected, high-weight XY-type observables and a systematic offset in one measurement group. A subsequent correction-magnitude refit showed that improving the goodness-of-fit statistic did not necessarily improve energy accuracy. These results demonstrate that a noise model that predicts uncorrected hardware behaviour accurately may still fail to predict the performance of the correction built from it. The paper reports the pre-registered experimental design, hardware results, residual diagnostics, follow-up tests, and methodological limitations. This work is not a blind variational ground-state search. The exact-state dependencies of the shared-frame estimator are disclosed, and the results do not establish reliable chemical accuracy for the frame-free estimator. Code and data repository: https://github.com/Venkatallu11/quantum-chemistry-vqe

•Oct 16, 2026•0 citations

A context-aware gate set tomography: Improving the self-consistent characterization of trapped-ion universal gate sets by leveraging non-Markovianity

Pablo Viñas, A. Bermudez

Abstract To progress in the characterization of noise for quantum computers, gate set tomography (GST) has emerged as a self-consistent protocol that accurately estimates noisy gates, state preparations, and measurements. In its original incarnation, GST improves estimation precision by applying gates sequentially, assuming the noise yields fixed completely-positive and trace-preserving (CPTP) maps independent of prior gate history. This ‘Markovian’ assumption can conflict with experiments, where time-correlated noise may induce non-Markovian dynamics, or slow drifts and cumulative calibration errors introduce context dependence, causing CP-divisible maps to vary with circuit depth. In this work, we address this issue for trapped-ion devices with phonon-mediated two-qubit gates. Using detailed microscopic modeling of high-fidelity light-shift gates, we tailor GST to capture the main source of context dependence: motional degrees of freedom. Rather than invalidating GST, context dependence can be incorporated into the gate-set parametrization, reducing sampling cost. Our results identify a promising research avenue that might be applicable to other platforms where microscopic modeling can be incorporated: the development of a context-aware GST.

arXiv:2507.02542•npj Quantum Information•Jun 26, 2026•2 citations

Adiabatic preparation of thermal states and entropy-noise relation on noisy quantum computers

Etienne Granet, Henrik Dreyer

We consider the problem of preparing thermal equilibrium states at finite temperature on quantum computers. Assuming thermalization, we show that states that are locally at thermal equilibrium can be prepared by evolving adiabatically an initial thermal Gibbs state of a simple Hamiltonian with an interpolating time-dependent Hamiltonian, identically to adiabatic ground state preparation. We argue that the entropy density of local density matrices is conserved during the adiabatic evolution in the thermodynamic limit, so that both the entropy and energy of the final state can be computed, and thus the final temperature too. We show that in the presence of hardware noise, the entropy created by the noisy evolution can be precisely benchmarked with mirror circuits. We give numerical evidence that the resulting thermal state preparation protocol is noise-resilient for depolarizing noise, in the sense that the energy-temperature curve measured on a noisy quantum computer is remarkably insensitive to the amplitude of depolarizing noise in the state preparation. We finally propose a protocol to estimate the lack of adiabaticity in a given actual Trotter implementation of the dynamics. We test our protocol on Quantinuum’s H1-1 ion-trap device. We measure that a circuit with 640 two-qubit gates implemented on hardware generates an entropy per site of 0.166 ± 0.0045, giving a benchmark metric for this state preparation. We report the preparation of a thermal state with temperature 2.56 ± 0.26 of the Ising model in size 5 × 4.

arXiv:2509.05206•npj Quantum Information•Jul 10, 2026•9 citations

A Diagnostic Framework for Quantum Noise Channels in Single-Qubit Cyber-Physical Systems

Polina Kravchenko, Yevhen S. Darnapuk, Gennady P. Chuiko

Single-qubit channels provide a controlled setting for comparing how different noise mechanisms change a quantum state. Using density matrices and Kraus operators, this paper compares bit-flip, phase-flip, depolarising, amplitudedamping, and phase-damping channels under the same set of input states and diagnostic metrics. The channels are tested on selected initial states to compare their effects on computational basis and superposition states. For each noisy state, purity, fidelity, von Neumann entropy, Bloch-vector norm, and quantum Fisher information (QFI) for$\boldsymbol{Z}$-axis phase encoding are calculated. The procedure is implemented in the Maple computing environment and used to perform parameter sweeps, generate two- and three-dimensional visualisations, and identify changes in metric behaviour. An engineering degradation index is introduced to combine the individual metrics and distinguish working, degraded, and critical parameter ranges. The calculations show that the$\vert+\rangle$state is unchanged by bit-flip noise, whereas phase-flip noise produces maximum mixedness at$\boldsymbol{p}=\mathbf{0. 5}$, and phase damping progressively removes coherence and$\boldsymbol{Z}$-axis phase sensitivity. The Maple implementation returns metric curves and degradation-score crossings for all channels considered, along with state-angle surfaces for amplitude and phase damping.

International Conference on Computer, Control, Informatics and its Applications•Sep 1, 2026•0 citations
Overview This study investigates how quantum noise affects variational Gibbs-state preparation on trapped-ion quantum computers. What Was Analyzed The analysis uses full two-qubit tomography instead of relying only on a single quantity such as fidelity or effective temperature. It compares complete quantum states across different field settings, three trapped-ion devices, and two temperature regimes. Noise Models Compared Several noise descriptions are tested, including scalar depolarization, Pauli-channel models, coherent-error corrections, and generalized amplitude damping (GAD). Predictive Testing The models are trained on selected conditions and then tested on unseen conditions without refitting. This allows the analysis to measure whether a noise model remains useful outside the data used to fit it. Main Result A shared GAD-based model shows the most stable predictive behaviour across the tested devices and temperatures. In the strongest frozen cross-temperature test, parameters fitted at beta = 5 are applied directly at beta = 1 without refitting. The model: improves all 9 of 9 QPU cases reduces mean Hilbert-Schmidt distance from 0.08644669 to 0.07277861 gives a 15.81% reduction in mean error Interpretation The results support a shared damping-like noise-model family within the tested system. However, the fitted parameters are not universal, and the study does not claim a unique microscopic noise mechanism, universal hardware-noise parameters, retrocausality, or a universal physical law. Reproducibility The calculations are also implemented in QDL Research Suite v1.9.0, allowing the main numerical results to be reproduced and independently checked. Software DOI: 10.5281/zenodo.23181558
We implement a variational quantum algorithm for Gibbs state preparation of a transverse-field Ising model on IonQ’s quantum computers. To this end, we train the variational parameters via classical simulation and perform state tomography on the quantum devices to evaluate the fidelity of the prepared Gibbs state. As a main result, we find that fidelity decreases (non-monotonically) as a function of the inverse temperature β of the system. Fidelity also decreases as a function of the size of the system. Interestingly, we find that a Gibbs state prepared for a specified β is a better representative of a Gibbs state prepared for a lower β ; or in other words, thermal fluctuations in the quantum hardware lead to digital heating, that is, an increase in the temperature of the prepared Gibbs state above what was intended.
We implement a variational quantum algorithm for Gibbs state preparation of a transverse-field Ising model on IonQ's quantum computers. To this end, we train the variational parameters via classical simulation and perform state tomography on the quantum devices to evaluate the fidelity of the prepared Gibbs state. As a main result, we find that fidelity decreases (non-monotonically) as a function of the inverse temperature βββ of the system. Fidelity also decreases as a function of the size of the system. Interestingly, we find that a Gibbs state prepared for a specified βββ is a better representative of a Gibbs state prepared for a lower\textit{lower}lower βββ; or in other words, thermal fluctuations in the quantum hardware lead to digital heating, that is, an increase in the temperature of the prepared Gibbs state above what was intended.
We report a pre-registered benchmark of an entanglement-forged H4 ground-state calculation on IonQ Forte Enterprise 1, a live trapped-ion processor. The analysis was frozen and publicly committed before any hardware data existed. The experiment comprised one submission of 81 new circuits at 1,100 shots per circuit, with device debiasing disabled. A three-parameter analytic noise model, fixed in advance, predicted the uncorrected energy error to within approximately 1%: 23.84 kcal/mol measured versus 24.06 kcal/mol predicted. However, the frame-free corrected estimator returned an error of 3.46 kcal/mol, compared with 0.83 kcal/mol predicted. This missed the conventional chemical-accuracy threshold of 1 kcal/mol. The shot-noise-weighted chi-square per degree of freedom increased from approximately 2 in simulator rehearsals to 5.93 on hardware. Analysis of 732 residuals identified elevated discrepancies in strongly corrected, high-weight XY-type observables and a systematic offset in one measurement group. A subsequent correction-magnitude refit showed that improving the goodness-of-fit statistic did not necessarily improve energy accuracy. These results demonstrate that a noise model that predicts uncorrected hardware behaviour accurately may still fail to predict the performance of the correction built from it. The paper reports the pre-registered experimental design, hardware results, residual diagnostics, follow-up tests, and methodological limitations. This work is not a blind variational ground-state search. The exact-state dependencies of the shared-frame estimator are disclosed, and the results do not establish reliable chemical accuracy for the frame-free estimator. Code and data repository: https://github.com/Venkatallu11/quantum-chemistry-vqe
Abstract To progress in the characterization of noise for quantum computers, gate set tomography (GST) has emerged as a self-consistent protocol that accurately estimates noisy gates, state preparations, and measurements. In its original incarnation, GST improves estimation precision by applying gates sequentially, assuming the noise yields fixed completely-positive and trace-preserving (CPTP) maps independent of prior gate history. This ‘Markovian’ assumption can conflict with experiments, where time-correlated noise may induce non-Markovian dynamics, or slow drifts and cumulative calibration errors introduce context dependence, causing CP-divisible maps to vary with circuit depth. In this work, we address this issue for trapped-ion devices with phonon-mediated two-qubit gates. Using detailed microscopic modeling of high-fidelity light-shift gates, we tailor GST to capture the main source of context dependence: motional degrees of freedom. Rather than invalidating GST, context dependence can be incorporated into the gate-set parametrization, reducing sampling cost. Our results identify a promising research avenue that might be applicable to other platforms where microscopic modeling can be incorporated: the development of a context-aware GST.
We consider the problem of preparing thermal equilibrium states at finite temperature on quantum computers. Assuming thermalization, we show that states that are locally at thermal equilibrium can be prepared by evolving adiabatically an initial thermal Gibbs state of a simple Hamiltonian with an interpolating time-dependent Hamiltonian, identically to adiabatic ground state preparation. We argue that the entropy density of local density matrices is conserved during the adiabatic evolution in the thermodynamic limit, so that both the entropy and energy of the final state can be computed, and thus the final temperature too. We show that in the presence of hardware noise, the entropy created by the noisy evolution can be precisely benchmarked with mirror circuits. We give numerical evidence that the resulting thermal state preparation protocol is noise-resilient for depolarizing noise, in the sense that the energy-temperature curve measured on a noisy quantum computer is remarkably insensitive to the amplitude of depolarizing noise in the state preparation. We finally propose a protocol to estimate the lack of adiabaticity in a given actual Trotter implementation of the dynamics. We test our protocol on Quantinuum’s H1-1 ion-trap device. We measure that a circuit with 640 two-qubit gates implemented on hardware generates an entropy per site of 0.166 ± 0.0045, giving a benchmark metric for this state preparation. We report the preparation of a thermal state with temperature 2.56 ± 0.26 of the Ising model in size 5 × 4.
Single-qubit channels provide a controlled setting for comparing how different noise mechanisms change a quantum state. Using density matrices and Kraus operators, this paper compares bit-flip, phase-flip, depolarising, amplitudedamping, and phase-damping channels under the same set of input states and diagnostic metrics. The channels are tested on selected initial states to compare their effects on computational basis and superposition states. For each noisy state, purity, fidelity, von Neumann entropy, Bloch-vector norm, and quantum Fisher information (QFI) forZ\boldsymbol{Z}Z-axis phase encoding are calculated. The procedure is implemented in the Maple computing environment and used to perform parameter sweeps, generate two- and three-dimensional visualisations, and identify changes in metric behaviour. An engineering degradation index is introduced to combine the individual metrics and distinguish working, degraded, and critical parameter ranges. The calculations show that the∣+⟩\vert+\rangle∣+⟩state is unchanged by bit-flip noise, whereas phase-flip noise produces maximum mixedness atp=0.5\boldsymbol{p}=\mathbf{0. 5}p=0.5, and phase damping progressively removes coherence andZ\boldsymbol{Z}Z-axis phase sensitivity. The Maple implementation returns metric curves and degradation-score crossings for all channels considered, along with state-angle surfaces for amplitude and phase damping.