Pith. sign in

REVIEW 4 major objections 5 minor 73 references

Quantum Krylov diagonalization works on today's hardware for the Hubbard model when evolution time, subspace size, and truncation are balanced against the low-energy gap.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-30 17:46 UTC pith:SAYTJ2EH

load-bearing objection Solid empirical parameter map of QKD on 1D Hubbard with a small IBM demo; useful practice notes, limited novelty, and one underplayed Trotter/Toeplitz caveat. the 4 major comments →

arxiv 2607.23604 v1 pith:SAYTJ2EH submitted 2026-07-26 quant-ph

Performance and Stability of Quantum Krylov Diagonalization for the Hubbard Model

Mohammad Mirzakhani , Hyukgun Kwon , Kyungsun Moon This is my paper
classification quant-ph
keywords quantum Krylov diagonalizationHubbard modelNISQJordan–Wignersingular-value truncationTrotterizationground-state energyHadamard test
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper asks whether quantum Krylov diagonalization (QKD) can reliably estimate ground-state energies of the half-filled one-dimensional Hubbard model on near-term quantum computers. QKD builds a small subspace by repeatedly applying real-time evolution to a reference state, measures Hamiltonian and overlap matrix elements on the quantum device, and finishes with a classical generalized eigenvalue solve. The authors show that success hinges on the Hamiltonian's low-energy spectrum and on numerical stability: systems with nearly closing gaps need longer evolution times to separate nearby states, while evolution time, Krylov dimension, Trotter steps, and singular-value truncation must be chosen together so that discretization error and ill-conditioning do not ruin the solve. Using a shallower Jordan–Wigner time-evolution circuit, they map how these knobs interact across system size and interaction strength, and they run the method on IBM hardware with only light readout mitigation and modest shots. The hardware runs track the ideal convergence trends, arguing that QKD is already a practical route to strongly correlated fermionic ground states on NISQ devices.

Core claim

For the half-filled 1D Hubbard model with periodic boundaries, QKD accuracy is controlled by a trade-off between low-energy spectral structure and numerical stability: near-closing gaps demand longer evolution times to resolve nearby eigenstates, while evolution time, Krylov dimension, Trotter number, and singular-value truncation threshold must be balanced to avoid instabilities and accumulated Trotter error; with that balance, IBM hardware reproduces the ideal convergence trends under lightweight readout mitigation and a modest shot budget.

What carries the argument

Unitary Krylov subspace from real-time evolution: basis states |ψ_j⟩ = e^{-i H j t} |ψ_0⟩, with Hamiltonian and overlap matrices measured by a Hadamard test, regularized by singular-value truncation, then solved as the generalized eigenvalue problem H̃ c = E S̃ c.

Load-bearing premise

The non-interacting Slater determinant stays informative enough as a starting state, and first-order Trotter plus a fixed truncation cutoff keep discretization and conditioning errors under control for the modest sizes and interaction strengths studied.

What would settle it

On the same L=4 or L=8 half-filled Hubbard chains (or larger), show that even with longer evolution, higher Trotter number, and tuned SVT the energy error fails to improve with Krylov dimension, or that IBM runs with ~10^4 shots and TREX no longer track the ideal convergence curve.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Parameter selection for QKD can follow explicit guidelines: longer t_evol for small-gap spectra, intermediate SVT (~0.1), and enough Trotter steps to delay breakdown at large D.
  • QKD can beat iterative phase estimation on accuracy while using shallower, fixed-depth circuits whose cost grows mainly in classical post-processing.
  • Finite-shot noise sets an accuracy floor; beyond modest D, more Krylov states help only if the measurement budget rises roughly as 1/√N_shots.
  • With only lightweight readout mitigation, current superconducting processors already show the same QKD convergence trends as noiseless simulations for small Hubbard chains.

Where Pith is reading between the lines

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

  • If reference-state fidelity collapses at stronger U or larger L, hybrid preparation (e.g., a short variational warm-start) may become necessary before QKD remains efficient.
  • The same gap-vs-evolution-time rule should apply to other fermionic lattice models with accidental near-degeneracies, not only 1D Hubbard.
  • Toeplitz structure plus the simplified JW string together suggest measurement and depth costs that scale more gently than naive subspace methods, inviting resource estimates for 2D clusters.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript presents a systematic numerical and experimental study of quantum Krylov diagonalization (QKD) applied to the ground-state energy of the half-filled 1D Hubbard model with periodic boundary conditions, for system sizes L=2–10 and interaction strengths U0=2–8. Using exact statevector simulations, finite-shot simulations with bootstrap error estimates, and runs on IBM's ibm_yonsei processor (L=3,6; TREX readout mitigation; 10^4 shots), the authors map out how the Krylov dimension D, evolution time t_evol, Trotter number N_trot, and SVT threshold δ_SVT jointly control convergence and numerical stability. The main qualitative findings are: (i) systems with near-closing spectral gaps (L=4,8) converge slowly and require longer evolution times; (ii) at large D a breakdown occurs whose onset is delayed by increasing N_trot, attributed to accumulated first-order Trotter error; (iii) an intermediate δ_SVT is optimal; (iv) hardware data reproduce the ideal convergence trends qualitatively. A comparison with IQPE (Table II, Fig. 7) argues QKD reaches higher accuracy at much shallower circuit depth. Exact benchmarks come from independent QuSpin diagonalization.

Significance. If the results hold, this is a useful, practice-oriented contribution: it is one of the more complete empirical parameter studies of QKD for a correlated fermionic model, and the guidelines (t_evol/Δt_o trade-off, interplay of gap structure with convergence, SVT balancing) are stated concretely enough to be used and falsified by other groups. Strengths worth naming: energies are benchmarked against independent exact diagonalization (QuSpin), so the central accuracy claims are not self-referential; statistical claims in the shot-based and hardware sections carry bootstrap or multi-run error bars; the per-size simulation parameters are disclosed (Table I); and the hardware demonstration (L=6, 12 qubits, ~10^4 shots, TREX only) is a genuine, if modest, data point on current devices. The gap–convergence connection (Appendix C, Figs. 11–12) is a nice mechanistic check rather than a post-hoc rationalization. The study is confirmatory rather than conceptually new — the Toeplitz Hadamard-test methodology follows Ref. [49] and the circuit construction follows the authors' Ref. [54] — but the systematic character of the scan is where the value lies.

major comments (4)
  1. [Sec. II B–C, Eqs. (8)–(11)] Sec. II B, Eqs. (8)–(9), and Sec. II C: the compact Toeplitz form S̃_jk = ⟨ψ0|U_{k−j}|ψ0⟩ is exact only for exact time evolution, since it relies on the group property U^{j†}U^k = U_{k−j}. The authors acknowledge this ('when the time evolution is approximated via Trotterization, the two expressions are no longer equivalent'), but Sec. II C then asserts the Toeplitz-Hermitian structure and the resulting linear-in-D measurement scaling unconditionally, and the Hadamard-test circuits of Fig. 2 measure only the compact form for all Trotterized simulations and hardware runs. Under first-order Trotter, Û^{j†}Û^k ≠ Û_{k−j}, so the assembled S̃, H̃ are not literally the Gram/projected matrices of the Krylov subspace actually generated, and the Rayleigh–Ritz justification of Eq. (10) holds only up to this structural inconsistency. This matters for the paper's own narrative: each matrix element's'
  2. [Sec. III A 1, Fig. 3] Relatedly, the interpretation of Fig. 3 attributes the large-D breakdown entirely to 'accumulated Trotter discretization errors.' An alternative contribution is the structural mismatch described above, which grows precisely as the smallest singular values of S̃ approach δ_SVT — the same onset the authors observe. These two mechanisms are distinguishable at essentially zero cost: in statevector simulation, compare (a) the compact-form matrices with (b) matrices built elementwise from ⟨ψ0|Û^{j†} H Û^k|ψ0⟩ using the Trotterized Û, at the same N_trot. If (b) delays or removes the breakdown relative to (a), the attribution in Sec. III A 1 must be revised; if not, the compact-form practice is empirically vindicated and the paper is strengthened. Either outcome is publishable, but the load-bearing interpretive claim of Fig. 3 currently rests on an untested assumption.
  3. [Sec. III A 2, Table I] Table I and Sec. III A 2: the system-size comparison tunes t_evol (from Δt_o/6 to 2Δt_o, non-monotonically) and N_trot (100 or 200) separately and by hand for each L. The paper is transparent about this, but it means the L-dependence in Fig. 4(a) conflates genuine size scaling with per-size parameter choices — e.g., L=4 is run at t_evol=Δt_o/4 while L=5 runs at Δt_o. Since 'system size' is one of the axes of the claimed systematic study and feeds the scalability discussion, at least one fixed-parameter scan (same t_evol/Δt_o and N_trot across all L) should be shown, or the text should state explicitly that Fig. 4(a) demonstrates achievable accuracy under per-size tuning rather than intrinsic size dependence.
  4. [Sec. III A 5, Table II] Sec. III A 5, Table II, Fig. 7: the QKD–IQPE comparison concludes QKD has 'significantly lower quantum resource requirements,' but the accounting is limited to single-circuit depth and width. IQPE is run once at depth ~2×10^5, while QKD requires O(D × N_Pauli) Hadamard-test circuit executions at depth ~6×10^3 each, with a shot budget per circuit; the total gate/shot cost is not compared, and the IQPE parameters (m=5, N_trot=10) appear chosen to be unfavorable without justification. The depth-per-circuit claim is true as stated and worth keeping, but the broader resource claim should either be supported by a total-cost comparison or narrowed explicitly to circuit depth.
minor comments (5)
  1. [Figs. 5, 6, 8 captions vs. Table I] Parameter inconsistency in captions: Table I gives Δt_o=0.17 for L=10, so t_evol=2Δt_o should be 0.34, but Figs. 5, 6, and 8(b) state t_evol=2Δt_o=0.28 (0.28 is the L=6 value of Δt_o). Please reconcile — either the Δt_o values used for L=10 differ from Table I, or the captions are wrong.
  2. [Sec. III B 2] The hardware section (Sec. III B 2) does not report device calibration data, run dates, or qubit layout for the ibm_yonsei experiments; even approximate two-qubit error rates would help the reader judge the L=6 results. A statement on code/data availability is also absent.
  3. [Sec. II C, Eq. (12)] Only first-order Trotter is used throughout. Since N_trot=500 is already being simulated, a brief remark on why second-order formulas were not considered (they would change the depth/error trade-off underlying several conclusions) would be useful.
  4. [various] Typos and small items: 'NUMERICAL RESUL TS' (Sec. III heading); 'ans¨atze' (Sec. I); 'eigenstates evolves' (Sec. III A 2); 'Trotter–Suzuki' vs 'Trotter-Suzuki' hyphenation inconsistent; in Fig. 12 the SVT threshold δ_SVT=0.9 departs from the stated default without comment in the main text.
  5. [Sec. III A 3, Fig. 5] The choice of the noninteracting Slater determinant as reference is reasonable here (fidelities ≳0.8 except L=4,8), but the paper would benefit from one sentence on expectations when this overlap collapses at stronger U0 or larger L, since the guidelines depend on it.

Circularity Check

0 steps flagged

No circularity: QKD energies are benchmarked against independent exact diagonalization; parameter studies and hardware trends are not forced by fitted inputs or self-citation.

full rationale

The paper is a systematic numerical and hardware study of Quantum Krylov diagonalization on the half-filled 1D Hubbard model. Ground-state energies from QKD are compared throughout to independent exact diagonalization (QuSpin), not to quantities defined from the QKD outputs themselves. The reference state is the non-interacting Slater determinant; t_evol is scaled to the spectral norm Δt_o = π/∥H∥; δ_SVT defaults to 0.1; and convergence is reported as absolute error |E_QKD − E_exact|. None of these choices makes the reported GSE or the claimed interplay between spectral gaps and numerical stability true by construction. The self-citation to the authors’ prior low-depth Jordan–Wigner circuit work [54] supplies an implementation detail (CNOT reduction under PBC) quantified in Appendix A; it is not a uniqueness theorem, fitted constant, or premise that forces the convergence guidelines or hardware trends. The Toeplitz/Hadamard-test measurement structure and the acknowledged inequivalence of compact vs. full matrix elements under Trotterization are methodological choices that may affect correctness or interpretation of the large-D breakdown, but they do not render any claimed prediction equivalent to its inputs. The derivation chain is therefore self-contained against external benchmarks, with no self-definitional loop, fitted-input-as-prediction, or load-bearing self-citation circularity.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

Central claims rest on standard quantum-simulation assumptions (JW mapping, first-order Trotter, Hadamard test, SVT-regularized generalized eigenproblem) plus the authors’ previously published circuit simplification. Free parameters are the hand-chosen algorithmic knobs (t_evol relative to π/‖H‖, N_trot, δ_SVT, reference state) that are varied to produce the guidelines; no physical constants are fitted to the target energies.

free parameters (4)
  • t_evol / Δt_o ratio = varies (e.g. 1/6 … 2, or 4 for gap test)
    Chosen per lattice size and figure (Table I, Fig. 3–6) to balance Krylov progress against Trotter error; not derived from a uniqueness condition.
  • N_trot = 100–600 depending on panel
    Hand-selected Trotter steps (100–600) to keep discretization below the observed error floor.
  • δ_SVT = 0.1 (default); 10^{-4}–0.9 swept
    Singular-value truncation threshold; default 0.1, swept in Fig. 6; controls stability versus information loss.
  • reference state |ψ0⟩ = noninteracting ground state
    Fixed as the U_0=0 Slater determinant; fidelity to the true ground state is an uncontrolled input that strongly affects convergence (Fig. 4b, 5 inset).
axioms (5)
  • standard math First-order Trotter–Suzuki product formula approximates e^{-iHt} with error controlled by N_trot
    Sec. II C, Eq. (12); used for all noisy and hardware circuits.
  • domain assumption Jordan–Wigner mapping plus the authors’ PBC string reduction yields a valid qubit Hamiltonian and shallower circuits
    Sec. II A and Appendix A; relies on prior work [54].
  • domain assumption SVT at threshold δ_SVT yields a numerically stable truncated generalized eigenproblem whose lowest Ritz value approximates the GSE
    Sec. II B; standard in the QKD literature the paper cites.
  • standard math Hadamard-test circuits with known vacuum phase ϕ recover the complex matrix elements ⟨ψ0|P U^{k-j}|ψ0⟩
    Sec. II C, Appendix B, following Yoshioka et al.
  • ad hoc to paper Lightweight TREX readout mitigation plus 10^4 shots suffice to reveal the ideal convergence trend on ibm_yonsei
    Sec. III B 2; empirical claim for the reported device and sizes only.

pith-pipeline@v1.2.0-grok45-kimik3 · 23646 in / 3213 out tokens · 55462 ms · 2026-07-30T17:46:28.596049+00:00 · methodology

0 comments
read the original abstract

Quantum Krylov diagonalization (QKD) has emerged as a promising hybrid quantum-classical approach for estimating ground-state properties of many-body systems on near-term quantum devices. In this work, we investigate the convergence, stability, and hardware performance of QKD for the one-dimensional Hubbard model with periodic boundary conditions. Building upon our previously developed low-depth Jordan--Wigner implementation, which reduces the number of two-qubit (CNOT) gates required for quantum time evolution, we perform a systematic study of the influence of the Krylov dimension, Hamiltonian evolution parameters, system size, interaction strength, and singular-value truncation (SVT) on the convergence of the method. Our results show that the performance of QKD is governed by a delicate interplay between the low-energy spectral structure of the Hamiltonian and numerical stability. In particular, systems with near-closing energy gaps require longer evolution times to efficiently resolve nearby eigenstates, while the evolution time, Krylov dimension, Trotter number, and SVT threshold must be carefully balanced to avoid numerical instabilities and accumulated time-discretization errors. This analysis provides practical guidelines for selecting algorithmic parameters in QKD. Finally, we demonstrate the algorithm on IBM quantum hardware, where the experimental results reproduce the convergence trends predicted by ideal simulations using only lightweight readout-error mitigation and a modest measurement budget. Together, these results demonstrate that QKD is a practical and hardware-efficient approach for studying strongly correlated fermionic systems on current NISQ quantum processors.

Figures

Figures reproduced from arXiv: 2607.23604 by Hyukgun Kwon, Kyungsun Moon, Mohammad Mirzakhani.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) 1D Hubbard chain with PBC, illustrated for the [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Hadamard-test circuit used to evaluate the real (via [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (a) Absolute GSE error as a function of the Krylov dimension [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. (a) Convergence of the GSE error as a function of [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Absolute GSE error as a function of the Krylov di [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Absolute GSE error as a function of the Krylov di [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison of the absolute GSE error obtained with [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Absolute GSE error as a function of the Krylov dimension [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Absolute GSE error as a function of the Krylov dimension [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Reduction in the number of CNOT gates as a [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12. GSE error as a function of the Krylov dimension [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

73 extracted references · 3 linked inside Pith

  1. [1]

    Figure 3(a) shows the convergence of the QKD GSE for the half-filled six-site Hubbard model as a function of the Krylov dimensionDfor several choices oft evol

    Effect of the time step on Krylov convergence The performance of the QKD algorithm depends criti- cally on the choice of the time-evolution parametert evol, which determines the structure of the Krylov subspace. Figure 3(a) shows the convergence of the QKD GSE for the half-filled six-site Hubbard model as a function of the Krylov dimensionDfor several cho...

  2. [2]

    System size To assess the scalability of the QKD approach, we in- vestigate the dependence of the algorithmic performance (a) (b) FIG. 4. (a) Convergence of the GSE error as a function of the Krylov dimensionDfor different lattice sizesL. For each system size, the time-evolution parametert evol and the num- ber of Trotter stepsN trot are selected relative...

  3. [3]

    Throughout this analysis, the system size is fixed atL= 10, while the time evolution is performed witht evol = 2∆to usingN trot = 200 Trotter steps

    Coulomb interaction We next investigate the influence of the on-site Coulomb interaction strength on the convergence of QKD by varyingU 0 from 2 to 8. Throughout this analysis, the system size is fixed atL= 10, while the time evolution is performed witht evol = 2∆to usingN trot = 200 Trotter steps. This allows us to isolate the effect of increasing electr...

  4. [4]

    As discussed in Sec

    Singular-value truncation We finally investigate the role of the SVT threshold, which regularizes the generalized eigenvalue problem by discarding Krylov basis vectors associated with small singular values of the overlap matrix. As discussed in Sec. II B, this regularization improves the conditioning of the overlap matrix but may also remove physically re...

  5. [5]

    (b) Quantum circuit implementing the time-evolution op- erator of the Hubbard Hamiltonian [Eq. (1)]. The firstL qubits encode the spin-up orbitals, while the remainingL qubits encode the spin-down orbitals. The red blocks,U Kin(t), implement the hopping evolution between neighboring lat- tice sites, whereas the blue blocks,U U (t), implement the on- site ...

  6. [6]

    The simulation parameters for both methods are summarized in Table II

    Comparison with Iterative Quantum Phase Estimation To further assess the practical performance of QKD, we compare it with IQPE [54] for theL= 6 Hubbard model atU 0 = 3 under noiseless conditions. The simulation parameters for both methods are summarized in Table II. Figure 7 compares the absolute GSE errors of the two approaches. IQPE yields an error of a...

  7. [7]

    Finite-shot sampling effects The noiseless simulations presented in the previous sec- tion isolate the intrinsic convergence properties of the QKD algorithm. In practical quantum processors, how- ever, expectation values are estimated from a finite num- ber of circuit executions, introducing statistical fluctu- ations in the measured Hamiltonian and overl...

  8. [8]

    Realistic device-noise simulations Finally, we assess the performance of the QKD al- gorithm under realistic hardware conditions using the IBM quantum processoribm yonsei, based on IBM’s 10 (a) L = 6 (b) L = 10 FIG. 8. Absolute GSE error as a function of the Krylov dimensionDfor the Hubbard model with (a)L= 6 and (b)L= 10 lattice sites atU 0 = 3. The red ...

    2025

  9. [9]

    R. P. Feynman, Int. J. Theor. Phys.21, 467 (1982)

    1982

  10. [10]

    Lloyd, Science273, 1073 (1996)

    S. Lloyd, Science273, 1073 (1996)

    1996

  11. [11]

    Bloch, J

    I. Bloch, J. Dalibard, and W. Zwerger, Rev. Mod. Phys. 80, 885 (2008)

    2008

  12. [12]

    Hubbard, Proc

    J. Hubbard, Proc. R. Soc. A276, 238 (1963)

    1963

  13. [13]

    E. H. Lieb and F. Y. Wu, Phys. Rev. Lett.20, 1445 (1968)

    1968

  14. [14]

    F. H. Essler, H. Frahm, F. G¨ ohmann, A. Kl¨ umper, and V. E. Korepin,The one-dimensional Hubbard model (Cambridge University Press, 2005)

    2005

  15. [15]

    M. Qin, T. Sch¨ afer, S. Andergassen, P. Corboz, and E. Gull, Annu. Rev. Condens. Matter Phys.13, 275 (2022)

    2022

  16. [16]

    Or´ us, Ann

    R. Or´ us, Ann. Phys.349, 117 (2014)

    2014

  17. [17]

    I. M. Georgescu, S. Ashhab, and F. Nori, Rev. Mod. Phys.86, 153 (2014)

    2014

  18. [18]

    Bauer, S

    B. Bauer, S. Bravyi, M. Motta, and G. K.-L. Chan, Chem. Rev.120, 12685 (2020)

    2020

  19. [19]

    Aspuru-Guzik, A

    A. Aspuru-Guzik, A. D. Dutoi, P. J. Love, and M. Head- Gordon, Science309, 1704 (2005)

    2005

  20. [20]

    A. Y. Kitaev, arXiv: 10.48550/arXiv.quant-ph/9511026 (1995)

    Pith/arXiv arXiv doi:10.48550/arxiv.quant-ph/9511026 1995

  21. [21]

    D. S. Abrams and S. Lloyd, Phys. Rev. Lett.79, 2586 (1997)

    1997

  22. [22]

    M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information(Cambridge University Press, 2000)

    2000

  23. [23]

    Ortiz, J

    G. Ortiz, J. E. Gubernatis, E. Knill, and R. Laflamme, Phys. Rev. A64, 022319 (2001)

    2001

  24. [24]

    J. D. Whitfield, J. Biamonte, and A. Aspuru-Guzik, Mol. Phys.109, 735 (2011)

    2011

  25. [25]

    Wecker, M

    D. Wecker, M. B. Hastings, N. Wiebe, B. K. Clark, C. Nayak, and M. Troyer, Phys. Rev. A92, 062318 (2015)

    2015

  26. [26]

    Babbush, N

    R. Babbush, N. Wiebe, J. McClean, J. McClain, H. Neven, and G. K.-L. Chan, Phys. Rev. X8, 011044 (2018)

    2018

  27. [27]

    Preskill, Quantum2, 79 (2018)

    J. Preskill, Quantum2, 79 (2018)

    2018

  28. [28]

    McArdle, S

    S. McArdle, S. Endo, A. Aspuru-Guzik, S. C. Benjamin, and X. Yuan, Rev. Mod. Phys.92, 015003 (2020)

    2020

  29. [29]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’brien, Nat. Commun.5, 4213 (2014)

    2014

  30. [30]

    J. R. McClean, J. Romero, R. Babbush, and A. Aspuru- Guzik, New J. Phys.18, 023023 (2016)

    2016

  31. [31]

    Kandala, A

    A. Kandala, A. Mezzacapo, K. Temme, M. Takita, M. Brink, J. M. Chow, and J. M. Gambetta, Nature549, 242 (2017)

    2017

  32. [32]

    Cerezo, A

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio,et al., Nat. Rev. Phys.3, 625 (2021)

    2021

  33. [33]

    Stanisic, J

    S. Stanisic, J. L. Bosse, F. M. Gambetta, R. A. Santos, W. Mruczkiewicz, T. E. O’Brien, E. Ostby, and A. Mon- tanaro, Nat. Commun.13, 5743 (2022)

    2022

  34. [34]

    P. G. Szalay, M. Nooijen, and R. J. Bartlett, J. Chem. Phys.103, 281 (1995)

    1995

  35. [35]

    Harsha, T

    G. Harsha, T. Shiozaki, and G. E. Scuseria, J. Chem. Phys.148, 044107 (2018)

    2018

  36. [36]

    Dallaire-Demers, J

    P.-L. Dallaire-Demers, J. Romero, L. Veis, S. Sim, and A. Aspuru-Guzik, Quantum Sci. Technol.4, 045005 (2019)

    2019

  37. [37]

    H. R. Grimsley, S. E. Economou, E. Barnes, and N. J. Mayhall, Nat. Commun.10, 3007 (2019)

    2019

  38. [38]

    J. R. McClean, M. E. Kimchi-Schwartz, J. Carter, and W. A. de Jong, Phys. Rev. A95, 042308 (2017)

    2017

  39. [39]

    Higgott, D

    O. Higgott, D. Wang, and S. Brierley, Quantum3, 156 (2019)

    2019

  40. [40]

    Santagati, J

    R. Santagati, J. Wang, A. A. Gentile, S. Paesani, 15 N. Wiebe, J. R. McClean, S. Morley-Short, P. J. Shad- bolt, D. Bonneau, J. W. Silverstone,et al., Sci. Adv.4, eaap9646 (2018)

    2018

  41. [41]

    D. Wang, O. Higgott, and S. Brierley, Phys. Rev. Lett. 122, 140504 (2019)

    2019

  42. [42]

    Wecker, M

    D. Wecker, M. B. Hastings, and M. Troyer, Phys. Rev. A92, 042303 (2015)

    2015

  43. [43]

    J. F. Gonthier, M. D. Radin, C. Buda, E. J. Doskocil, C. M. Abuan, and J. Romero, Phys. Rev. Res.4, 033154 (2022)

    2022

  44. [44]

    R. M. Parrish and P. L. McMahon, arXiv: 10.48550/arXiv.1909.08925 (2019)

    Pith/arXiv arXiv doi:10.48550/arxiv.1909.08925 1909

  45. [45]

    Motta, C

    M. Motta, C. Sun, A. T. Tan, M. J. O’Rourke, E. Ye, A. J. Minnich, F. G. Brandao, and G. K.-L. Chan, Nat. Phys.16, 205 (2020)

    2020

  46. [46]

    Saad,Numerical methods for large eigenvalue prob- lems: revised edition(SIAM, 2011)

    Y. Saad,Numerical methods for large eigenvalue prob- lems: revised edition(SIAM, 2011)

    2011

  47. [47]

    N. H. Stair, R. Huang, and F. A. Evangelista, J. Chem. Theory Comput.16, 2236 (2020)

    2020

  48. [48]

    J. Cohn, M. Motta, and R. M. Parrish, PRX Quantum 2, 040352 (2021)

    2021

  49. [49]

    Seki and S

    K. Seki and S. Yunoki, PRX Quantum2, 010333 (2021)

    2021

  50. [50]

    E. N. Epperly, L. Lin, and Y. Nakatsukasa, SIAM J. Ma- trix Anal. Appl.43, 1263 (2022)

    2022

  51. [51]

    Klymko, C

    K. Klymko, C. Mejuto-Zaera, S. J. Cotton, F. Wudarski, M. Urbanek, D. Hait, M. Head-Gordon, K. B. Whaley, J. Moussa, N. Wiebe, W. A. de Jong, and N. M. Tubman, PRX Quantum3, 020323 (2022)

    2022

  52. [52]

    C. L. Cortes and S. K. Gray, Phys. Rev. A105, 022417 (2022)

    2022

  53. [53]

    Kirby, M

    W. Kirby, M. Motta, and A. Mezzacapo, Quantum7, 1018 (2023)

    2023

  54. [54]

    G. Lee, D. Lee, and J. Huh, Quantum8, 1477 (2024)

    2024

  55. [55]

    T. E. Baker, Phys. Rev. A110, 012420 (2024)

    2024

  56. [56]

    Kirby, Quantum8, 1457 (2024)

    W. Kirby, Quantum8, 1457 (2024)

    2024

  57. [57]

    Yoshioka, M

    N. Yoshioka, M. Amico, W. Kirby, P. Jurcevic, A. Dutt, B. Fuller, S. Garion, H. Haas, I. Hamamura, A. Ivrii, et al., Nat. Commun.16, 5014 (2025)

    2025

  58. [58]

    M. G. J. a. Oliveira and N. Glaser, Phys. Rev. A112, 052442 (2025)

    2025

  59. [59]

    Jordan and E

    P. Jordan and E. Wigner, Zeitschrift f¨ ur Physik47, 631 (1928)

    1928

  60. [60]

    S. B. Bravyi and A. Y. Kitaev, Ann. Phys.298, 210 (2002)

    2002

  61. [61]

    J. T. Seeley, M. J. Richard, and P. J. Love, J. Chem. Phys.137, 224109 (2012)

    2012

  62. [62]

    Mirzakhani and K

    M. Mirzakhani and K. Moon, EPJ Quantum Technol.12, 134 (2025)

    2025

  63. [63]

    N. F. Mott, Proc. Phys. Soc. A62, 416 (1949)

    1949

  64. [64]

    Kaniel, Math

    S. Kaniel, Math. Comput.20, 369 (1966)

    1966

  65. [65]

    Saad, SIAM J

    Y. Saad, SIAM J. Numer. Anal.17, 687 (1980)

    1980

  66. [66]

    Hatano and M

    N. Hatano and M. Suzuki, inQuantum annealing and other optimization methods(Springer, 2005) pp. 37–68

    2005

  67. [67]

    Javadi-Abhari, M

    A. Javadi-Abhari, M. Treinish, K. Krsulich, C. J. Wood, J. Lishman, J. Gacon, S. Martiel, P. D. Nation, L. S. Bishop, A. W. Cross, B. R. Johnson, and J. M. Gam- betta, arXiv: 10.48550/arXiv.2405.08810 (2024)

    Pith/arXiv arXiv doi:10.48550/arxiv.2405.08810 2024

  68. [68]

    Weinberg and M

    P. Weinberg and M. Bukov, SciPost Phys.7, 020 (2019)

    2019

  69. [69]

    van den Berg, Z

    E. van den Berg, Z. K. Minev, and K. Temme, Phys. Rev. A105, 032620 (2022)

    2022

  70. [70]

    Temme, S

    K. Temme, S. Bravyi, and J. M. Gambetta, Phys. Rev. Lett.119, 180509 (2017)

    2017

  71. [71]

    Li and S

    Y. Li and S. C. Benjamin, Phys. Rev. X7, 021050 (2017)

    2017

  72. [72]

    S. Endo, S. C. Benjamin, and Y. Li, Phys. Rev. X8, 031027 (2018)

    2018

  73. [73]

    Takagi, S

    R. Takagi, S. Endo, S. Minagawa, and M. Gu, npj Quan- tum Inf.8, 114 (2022)

    2022