arxiv: 2604.27838 · v1 · submitted 2026-04-30 · 🪐 quant-ph · cs.DS

Recognition: unknown

Heisenberg-limited Hamiltonian learning without short-time control

Myeongjin Shin , Junseo Lee , Changhun Oh

Authors on Pith no claims yet

Pith reviewed 2026-05-07 05:52 UTC · model grok-4.3

classification 🪐 quant-ph cs.DS

keywords Hamiltonian learningHeisenberg limitquantum controlsparse Hamiltonianspure-state tomographyquantum dynamicscontinuous control emulation

0 comments

The pith

Heisenberg-limited Hamiltonian learning is possible using only evolution times bounded below by a fixed minimum T.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that optimal-precision learning of unknown quantum Hamiltonians does not require access to arbitrarily short-time dynamics. It supplies a concrete method to emulate the continuous control steps of prior iterative algorithms by using only longer evolutions of duration at least T. The approach reduces the original problem to sparse pure-state tomography. For Hamiltonians that are logarithmically sparse the total evolution time needed to reach precision ε still scales as 1/ε, which is information-theoretically optimal and independent of the value of T. For polynomially sparse many-body Hamiltonians the minimum allowed evolution time can be relaxed at only polynomial extra cost in total time.

Core claim

We introduce a framework in which every query to the unknown dynamics has duration at least a prescribed minimum time T, and show that this restriction does not preclude Heisenberg-limited scaling. The key ingredient is a method for emulating the continuous quantum control required by iterative learning algorithms using only such lower-bounded evolution times. This reduces the learning task to sparse pure-state tomography. Notably, for logarithmically sparse Hamiltonians, our algorithm achieves the information-theoretically optimal 1/ε scaling in total evolution time for any arbitrary constant minimum evolution time T.

What carries the argument

Emulation of continuous quantum control operations via sequences of evolution times each at least T, which converts the Hamiltonian-learning problem into one of sparse pure-state tomography.

If this is right

For logarithmically sparse Hamiltonians the total evolution time achieves the optimal 1/ε Heisenberg scaling for any fixed minimum query duration T.
For polynomially sparse many-body Hamiltonians a quantitative tradeoff appears: the minimum required evolution time can be relaxed from the usual limit at only polynomial cost in total evolution time.
High-bandwidth ultra-short pulses are not fundamentally required to reach information-theoretically optimal quantum learning performance.
The learning procedure can be implemented on hardware whose control bandwidth is finite and whose transient pulse errors prevent arbitrarily short gates.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Experimental platforms that cannot produce fast pulses may still reach optimal Hamiltonian learning rates by adopting the emulation sequence.
The same control-emulation idea could be applied to other iterative quantum algorithms that currently assume continuous or short-time access.
One could test the polynomial tradeoff numerically on small many-body Hamiltonians to see how the constant factors behave in practice.
If the emulation step incurs an overhead that grows with system size, the overall scaling for polynomially sparse systems may become worse than the paper states.

Load-bearing premise

The proposed emulation of continuous control can be realized using only evolutions of length at least T without introducing errors beyond those already accounted for by the sparsity assumption.

What would settle it

Run the algorithm on a logarithmically sparse Hamiltonian with known ground truth; measure whether the observed total evolution time to reach precision ε scales linearly as 1/ε when the minimum allowed query time is held fixed at any chosen constant T.

Figures

Figures reproduced from arXiv: 2604.27838 by Changhun Oh, Junseo Lee, Myeongjin Shin.

**Figure 1.** Figure 1: Comparison between previous short-time control approaches and our framework. Previous Heisenberg view at source ↗

**Figure 2.** Figure 2: Schematic overview of our iterative learning protocol. At iteration view at source ↗

read the original abstract

Characterizing quantum systems by learning their underlying Hamiltonians is a central task in quantum information science. While recent algorithmic advances have achieved near-optimal efficiency in this task, they critically rely on accessing arbitrarily short-time dynamics. This reliance poses severe experimental challenges due to finite control bandwidth and transient pulse errors. In this work, we demonstrate that Heisenberg-limited Hamiltonian learning can be achieved without short-time control. We introduce a framework in which every query to the unknown dynamics has duration at least a prescribed minimum time $T$, and show that this restriction does not preclude Heisenberg-limited scaling. The key ingredient is a method for emulating the continuous quantum control required by iterative learning algorithms using only such lower-bounded evolution times. This reduces the learning task to sparse pure-state tomography. Notably, for logarithmically sparse Hamiltonians, our algorithm achieves the information-theoretically optimal $1/\varepsilon$ scaling in total evolution time for any arbitrary constant minimum evolution time $T$. For many-body (polynomially sparse) systems, we uncover a rigorous quantitative tradeoff, showing that the minimum required evolution time can be significantly relaxed from the standard limit at a polynomial cost in total evolution time. Our results affirmatively resolve a prominent open problem in the field and reveal that high-bandwidth, ultra-short pulses are not fundamentally necessary for optimal quantum learning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a workable way to hit Heisenberg-limited Hamiltonian learning with any fixed minimum query time T by emulating short controls through sequences of longer evolutions and reducing the task to sparse pure-state tomography.

read the letter

The central advance is the emulation technique that lets iterative learning algorithms run without ever needing evolution times shorter than a prescribed T. For Hamiltonians that are logarithmically sparse the total evolution time still scales as 1/ε, which matches the information-theoretic optimum and removes the experimental requirement for ultra-short pulses. The reduction to sparse tomography is clean and the quantitative tradeoff they derive for polynomially sparse many-body Hamiltonians is useful: you can relax the minimum time at a polynomial cost in total time. That tradeoff is stated explicitly and looks reproducible from the abstract and the claimed lemmas. The proofs appear to track additive errors carefully enough that the sum over the 1/ε steps stays inside the target precision, at least for the log-sparse case. The citation pattern is standard and the prior work on short-time methods is acknowledged without over-claiming. The main soft spot is that the emulation construction must be checked in detail for whether it introduces any hidden dependence on the unknown spectrum or requires extra assumptions beyond the stated sparsity; if the error per emulation step is not strictly o(ε) independent of T, the claimed scaling would degrade. The paper does not appear to rely on circular fitting or invented structure. This is the kind of result experimental groups working on device characterization will want to read once the proofs are verified. It is worth sending to a serious referee who can examine the error lemmas and the tomography subroutine; the core idea is technically substantive and directly addresses a practical bottleneck even if the constants turn out larger than hoped.

Referee Report

2 major / 2 minor

Summary. The manuscript claims that Heisenberg-limited Hamiltonian learning is possible without access to arbitrarily short-time dynamics. By emulating the continuous control needed for iterative learning algorithms using only black-box queries to U(τ) with τ ≥ T (T any fixed constant), the problem reduces to sparse pure-state tomography. For logarithmically sparse Hamiltonians this yields the information-theoretically optimal 1/ε total-evolution-time scaling; for polynomially sparse (many-body) Hamiltonians a quantitative tradeoff is derived between the minimum allowed evolution time and the total evolution time.

Significance. If the central claims hold, the work affirmatively resolves a prominent open problem by showing that high-bandwidth, ultra-short pulses are not required for optimal quantum Hamiltonian learning. The reduction to sparse tomography and the explicit tradeoff for polynomially sparse systems are technically notable; the result would have direct experimental implications for platforms with finite control bandwidth.

major comments (2)

[Emulation framework (the section that reduces iterative learning to sparse tomography)] The emulation of continuous/short-time control via sequences of ≥T evolutions is load-bearing for the 1/ε claim. The manuscript must supply an explicit error lemma (presumably in the section detailing the emulation procedure) that bounds the total additive error accumulated over Θ(1/ε) iterative steps by o(ε), without introducing factors of 1/T or poly(1/ε) that would push the total evolution time to ω(1/ε). Because the Hamiltonian is known only up to log-sparsity, the construction cannot rely on eigenvalue knowledge or T-dependent cancellations; the current abstract provides no such accounting.
[Theorem stating the 1/ε scaling for log-sparse Hamiltonians] For the logarithmically sparse case, the information-theoretic optimality statement requires that every emulation step contributes an error small enough that the sum remains o(ε). If the number of long-time queries per emulation step scales with 1/T or injects an error linear in T that cannot be absorbed into the constant, the claimed 1/ε bound fails even for constant T. A concrete query-complexity accounting (total evolution time versus ε and T) is needed.

minor comments (2)

[Abstract and the corresponding theorem for many-body systems] The abstract asserts a 'rigorous quantitative tradeoff' for polynomially sparse systems; the main text should state the precise polynomial degree in the total-evolution-time cost as a function of the minimum allowed T.
[Introduction and preliminaries] Notation for the sparsity parameter (logarithmic vs. polynomial) and the precise definition of 'sparse pure-state tomography' should be introduced early and used consistently.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript and for identifying the need for greater explicitness in the error analysis of the emulation framework. We agree that these points require clarification to fully substantiate the 1/ε claim. We will revise the manuscript to extract and strengthen the relevant error bounds into a dedicated lemma and to include an explicit query-complexity accounting. Our responses to the major comments are provided below.

read point-by-point responses

Referee: [Emulation framework (the section that reduces iterative learning to sparse tomography)] The emulation of continuous/short-time control via sequences of ≥T evolutions is load-bearing for the 1/ε claim. The manuscript must supply an explicit error lemma (presumably in the section detailing the emulation procedure) that bounds the total additive error accumulated over Θ(1/ε) iterative steps by o(ε), without introducing factors of 1/T or poly(1/ε) that would push the total evolution time to ω(1/ε). Because the Hamiltonian is known only up to log-sparsity, the construction cannot rely on eigenvalue knowledge or T-dependent cancellations; the current abstract provides no such accounting.

Authors: We agree that an explicit error lemma is required for a self-contained proof of the claimed scaling. While the original manuscript derives the necessary bounds inside the proof of the main theorem (Section 4), we acknowledge that presenting them as a standalone lemma improves readability and rigor. In the revised manuscript we will add Lemma 3.2 in the emulation section. The lemma states that, for any fixed T > 0 and target error δ, there exists a sequence of O(1) queries each of duration at least T that emulates a short-time evolution step with additive error at most δ in operator norm. The construction uses only the logarithmic sparsity assumption to bound the difference between the true and emulated unitaries via a black-box approximation argument that does not invoke eigenvalue knowledge or T-dependent cancellations. By setting δ = Θ(ε²) per step, the accumulated error over the Θ(1/ε) iterations of the outer learning algorithm remains o(ε). No 1/T or poly(1/ε) factors appear in the total evolution time because the number of long-time queries per emulation step is independent of both T (for constant T) and ε. We will also update the abstract to reference this lemma. revision: yes
Referee: [Theorem stating the 1/ε scaling for log-sparse Hamiltonians] For the logarithmically sparse case, the information-theoretic optimality statement requires that every emulation step contributes an error small enough that the sum remains o(ε). If the number of long-time queries per emulation step scales with 1/T or injects an error linear in T that cannot be absorbed into the constant, the claimed 1/ε bound fails even for constant T. A concrete query-complexity accounting (total evolution time versus ε and T) is needed.

Authors: We thank the referee for this observation. The emulation procedure is constructed so that the per-step query count is O(1) and independent of T for any fixed positive constant T; the error introduced by each emulation is controlled solely by the sparsity parameter and the chosen δ, without a linear-in-T term that cannot be absorbed. Consequently the total evolution time remains Θ(1/ε) with a T-dependent prefactor that stays finite for each fixed T. In the revised manuscript we will augment the statement of Theorem 1 with an explicit complexity corollary (Corollary 1.1) that reads: for any fixed T > 0 and log-sparse Hamiltonian, the algorithm uses at most C(T)/ε total evolution time, where C(T) is a constant independent of ε. The proof will tabulate the contribution of each emulation step, confirming that the sum of emulation errors is o(ε) and that no additional poly(1/ε) or 1/T blow-up occurs. This accounting will also be summarized in a new table in Section 4. revision: yes

Circularity Check

0 steps flagged

No significant circularity: reduction to independent sparse tomography

full rationale

The paper's central derivation introduces an emulation technique for continuous control using only evolutions of duration at least T and reduces the Hamiltonian learning task to the independent problem of sparse pure-state tomography. No self-definitional steps, fitted inputs renamed as predictions, or load-bearing self-citations appear in the provided abstract or described framework. The 1/ε scaling claim for log-sparse Hamiltonians follows from applying known tomography bounds to the emulated queries rather than re-deriving them from the target result itself. The approach is self-contained against external benchmarks for sparse tomography and does not reduce any prediction to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard quantum mechanics (Schrödinger evolution) and the ability to perform queries of duration at least T. The sparsity assumptions on the Hamiltonian are domain-specific but not invented ad hoc. No free parameters or new physical entities are introduced in the abstract.

axioms (2)

domain assumption The unknown system evolves unitarily under a time-independent Hamiltonian according to the Schrödinger equation.
Standard background assumption for all Hamiltonian learning tasks.
domain assumption The Hamiltonian is either logarithmically sparse or polynomially sparse.
Required for the stated optimal scaling and tradeoff results.

pith-pipeline@v0.9.0 · 5531 in / 1389 out tokens · 28655 ms · 2026-05-07T05:52:01.352037+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

64 extracted references · 11 canonical work pages · 1 internal anchor

[1]

Gentile, Nathan Wiebe, Francesco Petruccione, Jeremy L

Jian Wang, Stefano Paesani, Raffaele Santagati, Sebastian Knauer, Antonio A. Gentile, Nathan Wiebe, Francesco Petruccione, Jeremy L. O’Brien, John G. Rarity, and Anthony Laing. Experimental quantum hamiltonian learning.Nature Physics, 13(6):551–555, 2017

2017
[2]

Shengshi Pang and Todd A. Brun. Quantum metrology for a general hamiltonian parameter.Physical Review 17 A, 90(2):022117, 2014

2014
[3]

Valdez, Giuseppe Mussardo, et al

Paul Shaffer, Angelica C. Valdez, Giuseppe Mussardo, et al. Practical verification protocols for analog quantum simulators.npj Quantum Information, 7(1):1–9, 2021

2021
[4]

Practical hamiltonian learning with unitary dynamics and gibbs states.Nature Communications, 15(1):312, 2024

Andi Gu, Lukasz Cincio, and Patrick J Coles. Practical hamiltonian learning with unitary dynamics and gibbs states.Nature Communications, 15(1):312, 2024

2024
[5]

Learning many-body hamiltonians with heisenberg- limited scaling.Physical Review Letters, 130(20):200403, 2023

Hsin-Yuan Huang, Yu Tong, Di Fang, and Yuan Su. Learning many-body hamiltonians with heisenberg- limited scaling.Physical Review Letters, 130(20):200403, 2023

2023
[6]

Robust and efficient hamiltonian learning.Quantum, 7:1045, 2023

Wenjun Yu, Jinzhao Sun, Zeyao Han, and Xiao Yuan. Robust and efficient hamiltonian learning.Quantum, 7:1045, 2023

2023
[7]

Learning𝑘-body hamiltonians via compressed sensing.arXiv:2410.18928, 2024

Muzhou Ma, Steven T Flammia, John Preskill, and Yu Tong. Learning𝑘-body hamiltonians via compressed sensing.arXiv:2410.18928, 2024

work page arXiv 2024
[8]

Flammia, and Susanne F

Hong-Ye Hu, Muzhou Ma, Weiyuan Gong, Qi Ye, Yu Tong, Steven T. Flammia, and Susanne F. Yelin. Ansatz-free hamiltonian learning with heisenberg-limited scaling.PRX Quantum, 6:040315, Oct 2025

2025
[9]

Nearly optimal algorithms to learn sparse quantum hamiltonians in physically motivated distances.arXiv:2509.09813, 2025

Amira Abbas, Nunzia Cerrato, Francisco Escudero Gutiérrez, Dmitry Grinko, Francesco Anna Mele, and Pulkit Sinha. Nearly optimal algorithms to learn sparse quantum hamiltonians in physically motivated distances.arXiv:2509.09813, 2025

work page arXiv 2025
[10]

Heisenberg-limited hamiltonian learning for interacting bosons.npj Quantum Information, 10(1):83, 2024

Haoya Li, Yu Tong, Tuvia Gefen, Hongkang Ni, and Lexing Ying. Heisenberg-limited hamiltonian learning for interacting bosons.npj Quantum Information, 10(1):83, 2024

2024
[11]

Learning the structure of any hamiltonian from minimal assumptions

Andrew Zhao. Learning the structure of any hamiltonian from minimal assumptions. InProceedings of the 57th Annual ACM Symposium on Theory of Computing, pages 1201–1211, 2025

2025
[12]

Practical characterization of quantum devices without tomography.Physical Review Letters, 107(21):210404, 2011

Marcus P da Silva, Olivier Landon-Cardinal, and David Poulin. Practical characterization of quantum devices without tomography.Physical Review Letters, 107(21):210404, 2011

2011
[13]

Hamiltonian learning and certification using quantum resources.Physical Review Letters, 112(19):190501, 2014

Nathan Wiebe, Christopher Granade, Christopher Ferrie, and David G Cory. Hamiltonian learning and certification using quantum resources.Physical Review Letters, 112(19):190501, 2014

2014
[14]

The advantage of quantum control in many- body hamiltonian learning.Quantum, 8:1537, 2024

Alicja Dutkiewicz, Thomas E O’Brien, and Thomas Schuster. The advantage of quantum control in many- body hamiltonian learning.Quantum, 8:1537, 2024

2024
[15]

Optimal short-time measurements for hamiltonian learning.arXiv:2108.08824, 2021

Assaf Zubida, Elad Yitzhaki, Netanel H Lindner, and Eyal Bairey. Optimal short-time measurements for hamiltonian learning.arXiv:2108.08824, 2021

work page arXiv 2021
[16]

Optimal learning of quantum hamiltonians from high- temperature gibbs states

Jeongwan Haah, Robin Kothari, and Ewin Tang. Optimal learning of quantum hamiltonians from high- temperature gibbs states. In2022 IEEE 63rd Annual Symposium on Foundations of Computer Science (FOCS), pages 135–146. IEEE, 2022

2022
[17]

Structure learning of hamiltonians from real- time evolution

Ainesh Bakshi, Allen Liu, Ankur Moitra, and Ewin Tang. Structure learning of hamiltonians from real- time evolution. In2024 IEEE 65th Annual Symposium on Foundations of Computer Science (FOCS), pages 1037–1050. IEEE, 2024

2024
[18]

Theory of trotter error with commutator scaling.Physical Review X, 11(1):011020, 2021

Andrew M Childs, Yuan Su, Minh C Tran, Nathan Wiebe, and Shuchen Zhu. Theory of trotter error with commutator scaling.Physical Review X, 11(1):011020, 2021

2021
[19]

CRC press, 2025

Sergio Blanes and Fernando Casas.A concise introduction to geometric numerical integration. CRC press, 2025

2025
[20]

Efficient quantum algorithms for simulating sparse hamiltonians.Communications in Mathematical Physics, 270(2):359–371, 2007

Dominic W Berry, Graeme Ahokas, Richard Cleve, and Barry C Sanders. Efficient quantum algorithms for simulating sparse hamiltonians.Communications in Mathematical Physics, 270(2):359–371, 2007

2007
[21]

Errors due to finite rise and fall times of pulses in superconducting charge qubits.Physical Review B, 65(14):144526, 2002

Sangchul Oh. Errors due to finite rise and fall times of pulses in superconducting charge qubits.Physical Review B, 65(14):144526, 2002

2002
[22]

Orlando, and William D

Simon Gustavsson, Olger Zwier, Jonas Bylander, Fei Yan, Fumiki Yoshihara, Yasunobu Nakamura, Terry P. Orlando, and William D. Oliver. Improving quantum gate fidelities by using a qubit to measure microwave pulse distortions.Physical Review Letters, 110(4):040502, 2013

2013
[23]

In situ characterization of qubit control lines: A qubit as a vector network analyzer.Physical Review Letters, 123(15):150501, 2019

Markus Jerger, Anatoly Kulikov, Zénon Vasselin, and Arkady Fedorov. In situ characterization of qubit control lines: A qubit as a vector network analyzer.Physical Review Letters, 123(15):150501, 2019

2019
[24]

M. A. Rol, L. Ciorciaro, F. K. Malinowski, B. M. Tarasinski, R. E. Sagastizabal, C. C. Bultink, Y. Salathe, N. Haandbaek, J. Sedivy, and L. DiCarlo. Time-domain characterization and correction of on-chip distortion of control pulses in a quantum processor.Applied Physics Letters, 116(5):054001, 2020

2020
[25]

Guang Hao Low and Isaac L. Chuang. Optimal hamiltonian simulation by quantum signal processing. Physical Review Letters, 118(1):010501, 2017

2017
[26]

Guang Hao Low and Isaac L. Chuang. Hamiltonian simulation by qubitization.Quantum, 3:163, 2019

2019
[27]

Quantum singular value transformation and beyond: Exponential improvements for quantum matrix arithmetics

András Gilyén, Yuan Su, Guang Hao Low, and Nathan Wiebe. Quantum singular value transformation and beyond: Exponential improvements for quantum matrix arithmetics. InProceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing, pages 193–204, 2019

2019
[28]

Childs and Nathan Wiebe

Andrew M. Childs and Nathan Wiebe. Hamiltonian simulation using linear combinations of unitary opera- tions.Quantum Information and Computation, 12(11-12):901–924, 2012

2012
[29]

Berry, Andrew M

Dominic W. Berry, Andrew M. Childs, Richard Cleve, Robin Kothari, and Rolando D. Somma. Simulating hamiltonian dynamics with a truncated taylor series.Physical Review Letters, 114(9):090502, 2015

2015
[30]

Circuit complexity of quantum access models for encoding classical data

Xiao-Ming Zhang and Xiao Yuan. Circuit complexity of quantum access models for encoding classical data. npj Quantum Information, 10:42, 2024

2024
[31]

Query-optimal estimation of unitary channels in diamond distance

Jeongwan Haah, Robin Kothari, Ryan O’Donnell, and Ewin Tang. Query-optimal estimation of unitary channels in diamond distance. In2023 IEEE 64th Annual Symposium on Foundations of Computer Science (FOCS), pages 363–390. IEEE, 2023

2023
[32]

Learning quantum hamiltonians at any temper- 18 ature in polynomial time

Ainesh Bakshi, Allen Liu, Ankur Moitra, and Ewin Tang. Learning quantum hamiltonians at any temper- 18 ature in polynomial time. InProceedings of the 56th Annual ACM Symposium on Theory of Computing, pages 1470–1477, 2024

2024
[33]

Learning hamiltonians in the heisenberg limit with static single-qubit fields.arXiv preprint arXiv:2601.10380, 2026

Shrigyan Brahmachari, Shuchen Zhu, Iman Marvian, and Yu Tong. Learning hamiltonians in the heisenberg limit with static single-qubit fields.arXiv preprint arXiv:2601.10380, 2026

work page arXiv 2026
[34]

Quantum probe tomography.arXiv:2510.08499, 2025

Sitan Chen, Jordan Cotler, and Hsin-Yuan Huang. Quantum probe tomography.arXiv:2510.08499, 2025

work page arXiv 2025
[35]

Quantum algorithmic measurement.Nature Communi- cations, 13(1):1–9, 2022

Dorit Aharonov, Jordan Cotler, and Xiao-Liang Qi. Quantum algorithmic measurement.Nature Communi- cations, 13(1):1–9, 2022

2022
[36]

Quantum algorithms for testing Hamiltonian symmetry.Physical Review Letters, 129(16):160503, 2022

Margarite L LaBorde and Mark M Wilde. Quantum algorithms for testing Hamiltonian symmetry.Physical Review Letters, 129(16):160503, 2022

2022
[37]

Unitary property testing lower bounds by polynomials

Adrian She and Henry Yuen. Unitary property testing lower bounds by polynomials. In14th Innovations in Theoretical Computer Science Conference (ITCS 2023), volume 251, pages 96:1–96:17, 2023

2023
[38]

Caro, and Aadil Oufkir

Andreas Bluhm, Matthias C. Caro, and Aadil Oufkir. Hamiltonian property testing.arXiv:2403.02968, 2024

work page arXiv 2024
[39]

Testing and learning structured quantum hamiltonians

Srinivasan Arunachalam, Arkopal Dutt, and Francisco Escudero Gutiérrez. Testing and learning structured quantum hamiltonians. InProceedings of the 57th Annual ACM Symposium on Theory of Computing, pages 1263–1270, 2025

2025
[40]

Hamiltonian locality testing via trotterized postselection.arXiv, 2025

John Kallaugher and Daniel Liang. Hamiltonian locality testing via trotterized postselection.arXiv, 2025

2025
[41]

Quantum Hamiltonian Certification

Minbo Gao, Zhengfeng Ji, Qisheng Wang, Wenjun Yu, and Qi Zhao. Quantum hamiltonian certification. arXiv:2505.13217, 2025

work page internal anchor Pith review arXiv 2025
[42]

Certifying and learning quantum ising hamiltonians.arXiv preprint arXiv:2509.10239, 2025

Andreas Bluhm, Matthias C Caro, Francisco Escudero Gutiérrez, Aadil Oufkir, and Cambyse Rouzé. Cer- tifying and learning quantum ising hamiltonians.arXiv:2509.10239, 2025

work page arXiv 2025
[43]

Optimal certification of constant-local hamiltonians.arXiv preprint arXiv:2512.09778, 2025

Junseo Lee and Myeongjin Shin. Optimal certification of constant-local hamiltonians.arXiv:2512.09778, 2025

work page arXiv 2025
[44]

Certifying and learning local quantum hamiltonians.arXiv:2603.29809, 2026

Andreas Bluhm, Matthias C Caro, Francisco Escudero Gutiérrez, Junseo Lee, Aadil Oufkir, Cambyse Rouzé, and Myeongjin Shin. Certifying and learning local quantum hamiltonians.arXiv:2603.29809, 2026

work page arXiv 2026
[45]

Lower bounds for learning hamiltonians from time evolution.arXiv e-prints, pages arXiv–2509, 2025

Ziyun Chen and Jerry Li. Lower bounds for learning hamiltonians from time evolution.arXiv e-prints, pages arXiv–2509, 2025

2025
[46]

Learning hamiltonians using quantum equilibration

Constantin Cedillo Vayson de Pradenne, Jordan Cotler, and Hsin-Yuan Huang. Learning hamiltonians using quantum equilibration. To appear
[47]

Learning stabilizer states by Bell sampling

Ashley Montanaro. Learning stabilizer states by bell sampling.arXiv:1707.04012, 2017

work page Pith review arXiv 2017
[48]

Dominik Hangleiter and Michael J. Gullans. Bell sampling from quantum circuits.Physical Review Letters, 133:020601, 2024

2024
[49]

Evaluating the fréchet derivative of the matrix exponential.Numerische Mathematik, 63:213– 226, 1992

Roy Mathias. Evaluating the fréchet derivative of the matrix exponential.Numerische Mathematik, 63:213– 226, 1992

1992
[50]

E. B. Dynkin. On the representation by means of commutators of the serieslog(𝑒𝑥𝑒𝑦)for noncommutative 𝑥and𝑦.Matematicheskii Sbornik (N.S.), 25(67):155–162, 1949. In Russian. 19 Appendices CONTENTS I. Introduction 1 II. Problem setup and main results 3 III. Algorithm overview 6 IV. Long-time emulation of residual dynamics 6 V. Coefficient extraction by pure...

1949
[51]

Pure-state tomography 22
[52]

Reduction to pure-state tomography 27

Technical tools 24 C. Reduction to pure-state tomography 27
[53]

Correctness of Algorithm 2 27
[54]

Approximating continuous quantum control using long-time evolution 29

Correctness of Algorithm 1 28 D. Approximating continuous quantum control using long-time evolution 29
[55]

Approximating continuous quantum control with arbitrary long time evolution 29
[56]

The main algorithm 31 F

Approximating continuous quantum control for many-body Hamiltonians 30 E. The main algorithm 31 F. Hamiltonian learning with long time dynamics in the large error regime 33 Appendix A: Related work The interactions of a quantum system are encoded in its Hamiltonian, and characterizing this Hamilto- nian from real-time dynamics is a fundamental problem in ...
[57]

For two qubits, the Bell states are |𝜎00⟩= 1√ 2(|00⟩+|11⟩),|𝜎 01⟩= 1√ 2(|00⟩ − |11⟩),|𝜎 10⟩= 1√ 2(|01⟩+|10⟩),|𝜎 11⟩= 1√ 2(|01⟩ − |10⟩)

Bell sampling We recall the Bell basis and the Bell-sampling primitive used throughout this work. For two qubits, the Bell states are |𝜎00⟩= 1√ 2(|00⟩+|11⟩),|𝜎 01⟩= 1√ 2(|00⟩ − |11⟩),|𝜎 10⟩= 1√ 2(|01⟩+|10⟩),|𝜎 11⟩= 1√ 2(|01⟩ − |10⟩). (B10) 22 The2 𝑛-qubit Bell basis is given by all tensor products|𝜎s⟩=|𝜎 𝑠1 ⟩⊗· · ·⊗|𝜎 𝑠𝑛 ⟩, where𝑠𝑖 ∈ {00,01,10,11}. Equiva...
[58]

We there- fore analyze the sample complexity of reconstructing a sparse pure quantum state under two different norms

Pure-state tomography Our Hamiltonian learning algorithm reduces the problem to sparse pure-state tomography. We there- fore analyze the sample complexity of reconstructing a sparse pure quantum state under two different norms. In particular, we provide guarantees for recovery in theℓ∞ andℓ 2 norms of the coefficient vector. Lemma 4(Sparse pure-state tomo...
[59]

We begin with Duhamel’s principle, which provides a convenient way to express and bound𝑒−𝑖𝑋𝑡 𝑒𝑖𝑌 𝑡

T echnical tools In this section, we collect several useful mathematical lemmas that will be used throughout the proofs. We begin with Duhamel’s principle, which provides a convenient way to express and bound𝑒−𝑖𝑋𝑡 𝑒𝑖𝑌 𝑡. Lemma 6(Duhamel’s principle [49]).Let𝑋, 𝑌be Hermitian matrices, and let𝑉(𝑡) :=𝑒 −𝑖𝑋𝑡 𝑒𝑖𝑌 𝑡 and ∆ :=𝑋−𝑌. Then, for every𝑡≥0, 𝑉(𝑡) =𝐼−𝑖 ∫︁...
[60]

For convenience, define a unitary𝑉satisfying𝑉(𝑃 𝑥 ⊗𝐼)|Φ⟩=|𝑥⟩for all𝑥

Correctness of Algorithm 2 In this section, we reduce Hamiltonian learning to pure-state tomography. For convenience, define a unitary𝑉satisfying𝑉(𝑃 𝑥 ⊗𝐼)|Φ⟩=|𝑥⟩for all𝑥. Such a unitary exists since the states{(𝑃 𝑥 ⊗𝐼)|Φ⟩: 𝑥∈ {0,1} 2𝑛}form an orthonormal basis. Theorem 3(Sparse Hamiltonian learning via pure-state tomography).Let𝐴= ∑︀ 𝑥∈{0,1}2𝑛 𝛼𝑥𝑃𝑥 be an ...
[61]

Correctness of Algorithm 1 The correctness of Algorithm 1 is guaranteed provided that𝜀≤𝑐/(10𝑐𝐹 𝑐∞). Theorem 4(Sparse Hamiltonian learning in Frobenius norm via pure-state tomography).Let𝐴=∑︀ 𝑥∈{0,1}2𝑛 𝛼𝑥𝑃𝑥 be an𝑛-qubit traceless Hermitian operator, where{𝑃𝑥}is the Pauli basis and|supp𝑃 (𝐴, 𝑐𝜀/8)| ≤ 𝑠. Assume that‖𝐴‖ ∞ ≤𝑐 ∞𝜀and‖𝐴‖ 𝐹 ≤𝑐 𝐹 𝜀. Then, using que...
[62]

We assume that𝐻and𝐻 𝑗 are𝑚-sparse traceless Hamiltonians satisfying‖𝐻−𝐻 𝑗‖ℓ∞ ≤𝜀. Our goal is to learn a Hermitian matrix̃︁𝑊such that‖𝑊− ̃︁𝑊‖ 𝐹 ≤𝑐𝜀, for a desired constant𝑐, using queries to 𝑒−𝑖𝐻𝑇 and with total evolution time achieving Heisenberg-limited scaling. This procedure will serve as a subroutine for Hamiltonian learning
[63]

Using Lemma 6 and Lemma 7, we have ‖𝑊𝑗‖𝐹 ≤𝜋‖𝐼−𝐶 𝑗‖𝐹 ≤𝜋𝑇‖𝐻−𝐻 𝑗‖𝐹 ≤2𝜋𝑇 √𝑚 𝜀,(D1) and ‖𝑊𝑗‖∞ ≤𝜋‖𝐼−𝐶 𝑗‖∞ ≤𝜋𝑇‖𝐻−𝐻 𝑗‖∞ ≤2𝜋𝑇 𝑚𝜀,(D2) since‖𝐻−𝐻 𝑗‖ℓ∞ ≤𝜀and both𝐻, 𝐻 𝑗 are𝑚-sparse

Approximating continuous quantum control with arbitrary long time evolution To efficiently learn the Hamiltonian𝑊𝑗, we first bound its norm and sparsity. Using Lemma 6 and Lemma 7, we have ‖𝑊𝑗‖𝐹 ≤𝜋‖𝐼−𝐶 𝑗‖𝐹 ≤𝜋𝑇‖𝐻−𝐻 𝑗‖𝐹 ≤2𝜋𝑇 √𝑚 𝜀,(D1) and ‖𝑊𝑗‖∞ ≤𝜋‖𝐼−𝐶 𝑗‖∞ ≤𝜋𝑇‖𝐻−𝐻 𝑗‖∞ ≤2𝜋𝑇 𝑚𝜀,(D2) since‖𝐻−𝐻 𝑗‖ℓ∞ ≤𝜀and both𝐻, 𝐻 𝑗 are𝑚-sparse. We now bound the sparsity of𝑊𝑗. L...
[64]

From the previous section, we have|supp 𝑃 (𝑊𝑗)| ≤4 𝑚, which becomes exponential in𝑛

Approximating continuous quantum control for many-body Hamiltonians When𝐻is a many-body Hamiltonian, its sparsity scales as𝑚= poly(𝑛). From the previous section, we have|supp 𝑃 (𝑊𝑗)| ≤4 𝑚, which becomes exponential in𝑛. Thus, directly learning𝑊𝑗 is infeasible. To address this, we restrict the evolution time to scale with the sparsity, setting𝑇=𝒪(𝑚−1/𝐾)for...