Encoding Choices and Fault-Tolerant Resource Estimates for Digital Quantum Hamiltonian Descent

Chenxu Liu; Meng Wang; Mingze Li; Muqing Zheng; Samuel Stein; Yousu Chen

REVIEW 3 major objections 5 minor 70 references

Encoding Choices and Fault-Tolerant Resource Estimates for Digital Quantum Hamiltonian Descent

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper argues that, for fault-tolerant implementations of quantum Hamiltonian descent, binary amplitude encoding should replace one-hot encoding: it shrinks the data register from O(dN) to O(d log N) qubits and, across every benchmark s

desk verdict A credible first systematic encoding comparison for fault-tolerant QHD; the binary preference is real under the paper's cost model, but the conditionality should be front and center. read the letter →

arxiv 2607.16996 v1 pith:MBK5ZVLQ submitted 2026-07-18 quant-ph

Chenxu Liu , Meng Wang , Mingze Li , Muqing Zheng , Samuel Stein , Yousu Chen This is my paper · ORCID

classification quant-ph PACS 03.67.Lx03.67.Pp

keywords quantumHamiltoniandescentbinaryamplitudeencodingone-hotfault-tolerantcomputingClifford+TsynthesisFouriertransformTrottererrorresourceestimation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Quantum Hamiltonian descent (QHD) solves continuous optimization by evolving a wavefunction whose potential term encodes the objective, and a digital implementation must first decide how to map the optimization grid onto qubits. This paper argues that binary amplitude encoding is the better choice for fault-tolerant implementations: it shrinks the data register from O(dN) to O(d log N) qubits, matches or beats one-hot encoding in asymptotic rotation cost, and uses fewer arbitrary R_z rotations on every benchmark problem studied. The paper also shows that the binary kinetic evolution can be approximated by a low-momentum quadratic spectrum together with an approximate quantum Fourier transform, cutting kinetic cost to polylogarithmic scaling in the grid resolution. It finds, however, that when the objective function is dense and transcendental, like Ackley, potential synthesis dominates and erases much of that kinetic saving. The work thereby gives concrete circuit constructions and resource estimates for planning QHD on future fault-tolerant hardware, and it redirects attention to structure-exploiting compilation of the potential term.

What carries the argument

Two complementary representations of the same discretized Hamiltonian carry the comparison. One-hot encoding places a single excited qubit at each grid point, turning the finite-difference Laplacian into local XX+YY hopping and the potential into low-weight occupation-projector Pauli strings. Binary amplitude encoding writes the grid index directly as a bit string; the Laplacian is diagonalized by the quantum Fourier transform, and the diagonal potential is expanded into Pauli-Z strings by Walsh-Hadamard transform. The mechanism that enables the polylogarithmic kinetic saving is the quadratic low-momentum identity k^2 = sum_ℓ 2^{2ℓ} k_ℓ + 2 sum_{ℓ<m} 2^{ℓ+m} k_ℓ k_m, which reduces the Fourie

What would settle it

Implement the same Ackley-2D benchmark at N=64 on a surface-code model that explicitly charges for routing and lattice surgery of the binary Pauli strings (weight up to 2 log N), and compare total logical space-time volume against the one-hot circuit; if one-hot has smaller volume at equal target error, the paper's binary-preference claim fails under that cost model.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is an encoding-aware cost comparison with a clear winner: for d variables and N grid points per variable, binary amplitude encoding uses d log N qubits instead of dN, with comparable or better asymptotic rotation scaling and fewer R_z rotations in every benchmark tested. The binary advantage comes from diagonalizing the finite-difference kinetic operator through the quantum Fourier transform and expanding the potential in Walsh-Hadamard Pauli-Z strings; a low-momentum k^2 approximation plus approximate QFT pushes kinetic cost to polylogarithmic scaling. The paper concludes binary is the preferred encoding for fault-tolerant QHD under the stated

Load-bearing premise

The binary-preference conclusion stands only if arbitrary non-Clifford rotations truly dominate, with Clifford gates, high-weight Pauli measurements, and long-range routing overhead cheap enough to ignore; on connectivity-limited hardware that assumption can fail.

Editorial extensions

If this is right

Fault-tolerant QHD resource planning can adopt binary encoding as a default, reducing logical data qubits by a factor of N / log N at equal grid resolution.
T-gate counts drop in every benchmark, and the reduction compounds with qubit savings to lower the required surface-code distance and physical-qubit footprint by 85–91% in the studied cases.
The kinetic cost of binary QHD can be made polylogarithmic in N through a low-momentum k^2 phase and approximate QFT, but only when potential synthesis does not dominate; for dense transcendental objectives such as Ackley the kinetic savings become marginal.
For polynomial or otherwise structured objectives, binary Walsh-Hadamard decomposition is sparse and yields smaller, shallower circuits; for dense coupled objectives, one-hot retains advantages in parallelism and depth despite its much larger qubit count.
Because the number of registers still grows linearly with the optimization dimension d, high-dimensional applications require iterative refinement or further structural decomposition.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper: the same low-momentum idea could be applied on the potential side—approximating a smooth objective by a low-degree polynomial before Walsh-Hadamard decomposition would sparsify the Pauli expansion, directly attacking the Ackley-type bottleneck the paper identifies.
Beyond the paper: the binary advantage is not hardware-independent; on modular or low-connectivity architectures, high-weight Pauli-Z strings (weight up to s_max log N) may incur routing and lattice-surgery costs that flip the comparison, so a Gray-code or hybrid encoding is a natural testable alternative.
Beyond the paper: schedule optimization could be used deliberately to keep the wavefunction in the low-momentum sector, making the k^2 kinetic approximation a controllable resource tradeoff rather than a heuristic that may or may not hold.
Beyond the paper: the benchmarks suggest a practical selection rule—binary encoding for separable or polynomial objectives, one-hot encoding for dense coupled landscapes—that could be validated on a wider objective suite with end-to-end physical-resource counts.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Worth a look. This is the first systematic encoding comparison for digital QHD, and it's a solid one. The main result—binary amplitude encoding cuts the data register to O(d log N) and uses fewer R_z rotations than one-hot across the tested benchmarks—holds up under the paper's stated cost model. The caveats are mostly in the text, which is more than many resource papers do.

What's genuinely new: they derive explicit binary-encoding circuits for the QHD kinetic operator, including the low-momentum k^2 approximation and AQFT variants, validate them against a split-operator Schrödinger solver, and connect the per-step scalings to Clifford+T resource counts. Table I is internally consistent, and the numerical validation in Figs. 2–5 is convincing for what it covers. The observation that Ackley's dense potential erases the kinetic savings is also a useful counterpoint to naive optimization.

Soft spots, in rough order of weight. First, the claim that binary is 'preferred for fault-tolerant implementations' is explicitly conditional on the Sec. III A cost model: arbitrary rotations dominate, while Clifford operations, high-weight Pauli measurements, and routing are treated as cheap. The paper's own Ackley numbers at N=64 show how load-bearing that assumption is—binary has 4.6x more CNOTs and roughly 30x more depth than one-hot—and Sec. IV E acknowledges that connectivity overheads could erase the advantage. So the body is honest, but the abstract's phrasing glides over the condition; it should carry the same qualification as the conclusion. Second, the FT tables use 100 Trotter steps while the validation uses 10^4. The per-step circuit structure is identical, so this is defensible for resource estimation, but it means the T-counts do not reflect the actual Trotter error or step count needed for a fair algorithmic comparison. Third, no code or data are released, and the analysis relies on unreleased tools; that limits independent checking and should be fixed before publication.

Who this is for: people doing early-FT resource estimation for QHD or first-quantized simulation, and groups choosing encodings for continuous optimization on future hardware. It deserves a serious referee. With the code released and the cost-model caveat moved into the abstract, I'd be comfortable citing it as the default reference for encoding tradeoffs in digital QHD.

Referee Report

3 major / 5 minor

Summary. The paper analyzes how the choice of data encoding (one-hot vs. binary amplitude) affects the gate counts and fault-tolerant (FT) resource requirements of digital quantum Hamiltonian descent (QHD). It derives asymptotic scalings for kinetic and potential evolution operators, constructs explicit Clifford+Rz circuits, validates them against a classical split-operator Schrödinger solver on one- and two-dimensional benchmarks, and reports Clifford+T resource estimates using a surface-code cost model. The central claim is that binary encoding reduces the data register from O(dN) to O(d log N) qubits and generally uses fewer Rz rotations than one-hot encoding, making it the preferred choice when arbitrary rotations dominate the FT cost. The paper also examines approximate kinetic implementations (AQFT and low-momentum k^2 approximation) and shows that these reduce kinetic cost to polylogarithmic scaling, though potential synthesis, as in the Ackley benchmark, can dominate and erase those savings.

Significance. If the claims hold, this is a useful contribution: it provides a systematic, encoding-aware resource comparison for a specific continuous-optimization quantum algorithm, with analytic scalings that are internally consistent and supported by an independent classical solver validation. The paper gives concrete circuit constructions for both encodings, and its numerical data on benchmark objectives provide a starting point for practical FT resource planning. The main strengths are the explicit derivation of rotation counts in Eqs. (14), (24), (72), (77), the validation of circuit dynamics against a split-operator solver (Figs. 2–5), and the recognition that objective structure, not just encoding, determines the dominant cost. However, the quantitative FT resource estimates rest on assumptions that are acknowledged but not stress-tested, and the headline preference for binary encoding is conditional on a cost model that may not hold in realistic surface-code layouts.

major comments (3)

[Sec. VI B, Table III, and Sec. VII] The FT resource estimates use 100 Trotter steps, while the accuracy validation in Sec. V B uses 10^4 Trotter steps (and Fig. 5 uses up to 2×10^5). The paper's own Trotter-error discussion in Sec. VII gives r=O(N) for a conservative operator-norm bound at fixed target error. If that scaling applies, the total T-counts in Table III would be multiplied by a resolution-dependent factor reflecting the actual number of steps needed for the chosen accuracy, and the 100-step resource estimates would substantially understate the total FT cost. The claim that the per-step structure is repeated does not justify the use of 100 steps when the validation uses orders of magnitude more. The manuscript should either include the step-count dependence in the FT resource model or justify why 100 steps is sufficient for the specific benchmarks, e.g., by reporting solution-quality versus step-count data.
[Sec. III A, Sec. IV E, Table II, Table III] The conclusion that binary encoding is preferred for FT implementations rests on the cost model of Sec. III A, where Clifford/CNOT operations and high-weight Pauli measurements are comparatively cheap. The paper's own data in Table II show that for Ackley at N=64, binary encoding uses 41,242 CX/step vs. 8,960 for one-hot and has depth 7.86e6 vs. 2.71e5. Sec. IV E acknowledges that connectivity-dependent overheads (routing, lattice-surgery distance, high-weight Pauli strings up to s_max log N) are not included, but no quantitative sensitivity analysis is provided. Since surface-code lattice-surgery cost grows with Pauli weight and patch distance, a modest per-CNOT or per-weight penalty could erase the observed 2.9× Rz advantage. The manuscript should either provide a sensitivity analysis with a plausible connectivity/weight cost model or explicitly state that the preference for binary enc
[Sec. V C, Sec. VII, Table I] The AQFT-k^2 kinetic approximation is validated only for one-dimensional benchmarks and for the specific QHD-C schedule with observables <x> and sigma_x. The claim that this approximation reduces binary kinetic cost to polylogarithmic scaling, and the associated resource estimates in Table I, assume that the wavefunction remains concentrated in low-momentum modes for the entire evolution. The paper does not provide an error bound or a quantitative condition for when this assumption holds. In higher-dimensional problems or with different schedules, diabatic excitation may populate high-momentum modes, making the k^2 approximation unreliable. Please add a formal error estimate or a numerical study that maps the range of validity of the low-momentum approximation in the resource-relevant regime.

minor comments (5)

[Data and Code Availability] The statement 'available from the corresponding author upon reasonable request' falls short of modern reproducibility expectations for a resource-analysis paper. Consider releasing the circuit-generation code and the benchmark data, or at least the exact gate-count extraction scripts.
[Sec. VI B refs [59,60,64]] The FT resource estimates depend on NWQEC and NWQRE, which are described as in-preparation or unreleased tools. This limits independent verifiability; please provide a detailed description of the transpilation and resource-estimation algorithms in an appendix or cite a peer-reviewed version.
[Fig. 5 caption] The caption states that for N=16..128, 10^5 Trotter steps are used, while the main text Sec. V B states that the default is delta_t=10^-3, i.e., 10^4 steps. Please reconcile or clarify which step count is used for which figure.
[Sec. VI A] Minor typographical issues: 'optimization level0' should be 'optimization level 0'; and the N^2/N scaling guides in Fig. 6 are described as visual guides, but the caption should also state that they are not fits (which it does), to avoid overinterpretation by readers.
[Eq. (36)] The uniform error allocation in Eq. (36) is a valid worst-case assumption, but for a paper emphasizing resource estimation it would be helpful to note that non-uniform allocations can reduce T-count; this is mentioned in Sec. VII but could be stated earlier.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: gate-count scalings are derived by explicit counting and validated against an independent classical solver; the binary preference is conditional on the stated FT cost model, not by construction.

full rationale

The paper's central scalings are obtained from explicit circuit-level counting rather than from fitting or from self-referential definitions. The one-hot kinetic count M_K=2dN follows from the nearest-neighbor XX+YY bond decomposition of the finite-difference Laplacian, and the potential count M_V=O(G N^{s_max}) follows from expanding occupation projectors into Pauli-Z strings. The binary potential count uses the Walsh–Hadamard coefficient expansion of the sampled diagonal operator, and the binary kinetic count follows from the exact-QFT diagonalization of the finite-difference Laplacian. None of these steps assumes the conclusion that binary encoding is preferred; they are direct counting arguments for circuit implementations that are then numerically validated against a classical split-operator Schrödinger solver (Figs. 2–5). The statement that binary encoding is preferred for fault-tolerant implementations is explicitly conditioned on the resource model of Sec. III A, in which non-Clifford rotations dominate and Clifford/connectivity overheads are comparatively cheap. The paper itself repeatedly acknowledges that this condition is substantive: Sec. IV E states that 'connectivity-dependent overheads are not included in the asymptotic estimates above and could shift the practical preference,' and Sec. VII lists lattice-surgery layout optimization as an omitted consideration. Thus the main claim is a conditional resource comparison, not a circularly defined prediction. The self-citations to the authors' earlier QHD resource work (Ref. [28]) and to their in-house tools NWQEC/NWQRE (Refs. [59–61,64]) supply framework and tooling, but the central gate-count scalings and T-count reductions are supported by independent counting and by standard, externally published synthesis and surface-code methods (GridSynth, Litinski's lattice-surgery model). Even if the unreleased tools introduce reproducibility risk, there is no exhibited reduction of the paper's results to its own assumptions by construction. No circular step satisfying the quoted-evidence standard was found.

Assumptions & free parameters 5 free parameters · 8 assumptions · 0 invented entities

The central claims rest on standard linear-algebraic facts (Walsh-Hadamard decomposition, QFT diagonalization) and on a fault-tolerant cost model that is an assumption about future hardware, not a derived result. The hand-chosen schedule, error budgets, and unspecified angle threshold affect the numerical resource counts but not the asymptotic comparisons.

free parameters (5)

QHD-C schedule parameters (s, T, δt) = s=1, T=10, δt=10^-3
Chosen by hand for benchmarks; determine dynamics and Trotter step count, not fitted.
Angle threshold δθ = unspecified
Used in §VI A to discard small rotations and merge into Clifford gates; no value given, so R_z and T counts depend on an unknown parameter.
AQFT truncation order = 3 (Fig. 8, main text)
Set by hand; controls how many controlled rotations are omitted and thus the accuracy/cost tradeoff.
Trotter step count for FT estimates = 100
Chosen in §VI B for tractability of transpilation; Table III totals are for 100 steps, not the 10^4 used in accuracy verification.
Error budgets = ϵ_syn=1e-4, ϵ_T=5e-5, ϵ_L=5e-5
Allocated by hand in §VI B; affect code distance and T-count synthesis precision.

assumptions (8)

standard math Walsh-Hadamard expansion represents any diagonal operator as up to 2^n-1 Pauli-Z strings (Eq. 53-54)
Used for binary potential and Fourier-basis kinetic phase synthesis.
standard math Quantum Fourier transform diagonalizes the periodic finite-difference Laplacian (Eq. 63-65)
Basis of the binary kinetic construction.
standard math Trotter-Suzuki product formula approximates time-ordered evolution (Eq. 6)
Time discretization; error analysis follows Refs [63,65].
domain assumption Surface-code cost model: Clifford/CNOT operations are comparatively cheap; arbitrary rotations dominate via Clifford+T synthesis (Eastin-Knill)
Frames the whole comparison in §III A; if false, encoding preference could change.
domain assumption Low-momentum approximation λ_k ≈ -4π^2 k^2/(h^2 N^2) is accurate for QHD dynamics (Eq. 89)
Validated only on 1D/2D benchmarks; may fail if potential excites high momenta.
domain assumption Objective function decomposes into support-local terms f(x)=Σ f_m(x_{S_m}) with small s_max (Eq. 19, 50)
Determines potential rotation count O(G N^{s_max}); dense d-variable terms would change scaling.
ad hoc to paper QHD-C schedule e^{φ}=2/(s+t^3), e^{χ}=2t^3 (Eq. 92)
A specific schedule chosen for benchmarks; results may be schedule-dependent.
ad hoc to paper Uniform error budget allocation over rotations (Eq. 36)
Simplifies synthesis error; non-uniform allocation could change T counts.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Encoding Choices and Fault-Tolerant Resource Estimates for Digital Quantum Hamiltonian Descent." pith.science (2026). https://pith.science/paper/MBK5ZVLQ

@misc{pith2026260716996,
  author       = {Pith},
  title        = {Pith review of: Encoding Choices and Fault-Tolerant Resource Estimates for Digital Quantum Hamiltonian Descent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MBK5ZVLQ}},
  note         = {Machine review of arXiv:2607.16996}
}

abstract

Quantum Hamiltonian descent (QHD) formulates continuous optimization as time-dependent quantum dynamics, where a kinetic term drives exploration and a potential term encodes the objective function. Digital implementations of QHD require encoding the search space into qubits, and this choice can shift the dominant cost among logical qubits, circuit depth, non-Clifford rotations, and potential synthesis. In this work, we present an encoding-aware resource analysis comparing one-hot and binary amplitude encodings for QHD. We derive gate-count scalings, construct and validate circuits against classical \Sch-equation solvers, and estimate Clifford+$R_z$ and fault-tolerant Clifford+$T$ resources on benchmark optimization problems. Binary encoding reduces the data register from $O(dN)$ to $O(d\log N)$ qubits and gives comparable asymptotic scaling for both kinetic and potential evolutions. Across all benchmark problems studied, binary encoding also uses fewer $R_z$ rotations than one-hot encoding, making it the preferred option for fault-tolerant implementations where arbitrary rotations dominate the cost. Kinetic approximations based on low-momentum spectra and approximate QFTs can further reduce the binary kinetic cost to polylogarithmic scaling. However, for targets such as Ackley, potential synthesis can dominate the total cost and reduce the benefit of kinetic approximations. These results suggest that exploiting the analytic structure of the target function to compile the potential evolution in QHD more efficiently is needed for further resource reductions.

Figures

Figures reproduced from arXiv: 2607.16996 by the authors.

**Figure 2.** FIG. 2. Comparison of the QHD dynamics obtained from the classical split-operator solver and from the one-hot and binary [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗

**Figure 3.** FIG. 3. Comparison of the [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗

Figures from the paper (10 more)

**Figure 4.** Figure 4: FIG. 4. Comparison of the [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

**Figure 5.** Figure 5: FIG. 5. Effect of approximate-QFT truncation on the two [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]

**Figure 6.** Figure 6: FIG. 6. Total CX and [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

**Figure 8.** Figure 8: FIG. 8. Gate-count scaling for different kinetic-evolution ap [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]

**Figure 9.** Figure 9: FIG. 9. Total [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]

**Figure 10.** Figure 10: FIG. 10 [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]

**Figure 11.** Figure 11: FIG. 11. The four two-dimensional benchmark target func [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]

**Figure 12.** Figure 12: FIG. 12. The CX and [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]

**Figure 13.** Figure 13: FIG. 13. Gate-count scaling for different kinetic-evolution approximations for (a) Three-Hump Camel, (b) Alpine 1, and (c) [PITH_FULL_IMAGE:figures/full_fig_p026_13.png]

**Figure 14.** Figure 14: FIG. 14 [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 1 canonical work pages

[1]

One-hot realization In the one-hot construction, each coordinate register contains one qubit per grid point, as in Eq. (8). The cir- cuit acts on the full qubit Hilbert space, but the intended physical subspace is the tensor product of Hamming- weight-one subspaces, one for each coordinate. The potential circuit follows directly from the projec- tor repre...
[2]

A grid point is encoded directly as a computational-basis bit string, so no one-hot subspace constraint is required

Binary realization The binary construction usesb= log 2 Nqubits per coordinate register. A grid point is encoded directly as a computational-basis bit string, so no one-hot subspace constraint is required. The potential term remains diago- nal and is compiled by the Pauli-Zexpansion described in Sec. IV. In particular, each support-local diagonal block is...
[3]

These functions are plotted in Fig

One-Dimensional T argets We use four one-dimensional target functions as bench- mark problems. These functions are plotted in Fig. 2e to Fig. 2h. a. Centered Quadratic: f(u) = u− 1 2 2 .(A1) The unique global minimizer is u⋆ = 1 2 , f(u ⋆) = 0.(A2) b. Double-Well: f(u) = 4(u−0.3) 2(u−0.7) 2 + 0.02(u−0.55).(A3) The global minimizer is u⋆ ≈0.285901, f(u ⋆)≈...
[4]

These four functions are plotted in Fig

Two-Dimensional T argets We mainly focus on four two-dimensional target func- tions for resource analysis and verification of the perfor- mance of the QHD algorithm. These four functions are plotted in Fig. 11. a. Coupled Quadratic 2D. f(x, y) = (x−0.25) 2 + 1.4(y−0.65) 2 + 0.2(x−0.25)(y−0.65). (A9) The unique global minimizer is (x⋆, y⋆) = (0.25,0.65), f...
[5]

P. W. Shor, Polynomial-time algorithms for prime factor- ization and discrete logarithms on a quantum computer, SIAM review41, 303 (1999)

1999
[6]

P. W. Shor, Algorithms for quantum computation: dis- crete logarithms and factoring, inProceedings 35th An- nual Symposium on Foundations of Computer Science (IEEE, 1994) pp. 124–134

1994
[7]

Zhou, S.-T

L. Zhou, S.-T. Wang, S. Choi, H. Pichler, and M. D. 26 103 CX gates / Trotter step N log2 N Exact k2 AQFT AQFT-k2 16 32 64 128 256 512 Resolution (N) 102 103 Rz gates / Trotter step N log2 N (a) Three-Hump Camel 103 104 CX gates / Trotter step N log2 N Exact k2 AQFT AQFT-k2 16 32 64 128 256 512 Resolution (N) 102 103 Rz gates / Trotter step N log2 N (b) A...

2020
[8]

I. M. Georgescu, S. Ashhab, and F. Nori, Quantum sim- ulation, Reviews of Modern Physics86, 153 (2014)

2014

Show all 70 references

[9]

Kandala, A

A. Kandala, A. Mezzacapo, K. Temme, M. Takita, M. Brink, J. M. Chow, and J. M. Gambetta, Hardware- efficient variational quantum eigensolver for small molecules and quantum magnets, Nature549, 242 (2017)

2017
[10]

Rebentrost, B

P. Rebentrost, B. Gupt, and T. R. Bromley, Quantum computational finance: Monte carlo pricing of financial derivatives, Physical Review A98, 022321 (2018)

2018
[11]

Woerner and D

S. Woerner and D. J. Egger, Quantum risk analysis, npj Quantum Information5, 1 (2019)

2019
[12]

Preskill, Quantum Computing in the NISQ era and beyond, Quantum2, 79 (2018)

J. Preskill, Quantum Computing in the NISQ era and beyond, Quantum2, 79 (2018)

2018
[13]

Preskill, Beyond nisq: The megaquop ma- chine, ACM Transactions on Quantum Computing6, 10.1145/3723153 (2025)

J. Preskill, Beyond nisq: The megaquop ma- chine, ACM Transactions on Quantum Computing6, 10.1145/3723153 (2025)

2025 doi
[14]

S. M. LaValle,Planning Algorithms(Cambridge Univer- sity Press, 2006)

2006
[15]

J. T. Betts,Practical Methods for Optimal Control and Estimation Using Nonlinear Programming, 2nd ed. (SIAM, 2010)

2010
[16]

Nocedal and S

J. Nocedal and S. J. Wright,Numerical Optimization (Springer New York, NY, 2006)

2006
[17]

Markowitz, Portfolio selection, The Journal of Finance 7, 77 (1952)

H. Markowitz, Portfolio selection, The Journal of Finance 7, 77 (1952)

1952
[18]

Cornu´ ejols and R

G. Cornu´ ejols and R. T¨ ut¨ unc¨ u,Optimization Methods in Finance(Cambridge University Press, 2007)

2007
[19]

M. B. Zafar, I. Valera, M. G. Rodriguez, and K. P. Gum- madi, Fairness constraints: Mechanisms for fair classifica- tion, inProceedings of the 20th International Conference 27 on Artificial Intelligence and Statistics, Proceedings of Machine Learning Research, Vol. 54 (PMLR, 2...

2017
[20]

Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society: Series B (Methodological)58, 267 (1996)

R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society: Series B (Methodological)58, 267 (1996)

1996
[21]

H. Xu, C. Caramanis, and S. Mannor, Robustness and regularization of support vector machines, Journal of Ma- chine Learning Research10, 1485 (2009)

2009
[22]

S. H. Low, Convex relaxation of optimal power flow— part I: Formulations and equivalence, IEEE Transactions on Control of Network Systems1, 15 (2014)

2014
[23]

Aravena, D

I. Aravena, D. K. Molzahn, S. Zhang, C. G. Petra, F. E. Curtis, S. Tu, A. W¨ achter, E. Wei, E. Wong, A. Gho- lami, K. Sun, X. A. Sun, S. T. Elbert, J. T. Holzer, and A. Veeramany, Recent developments in security- constrained AC optimal power flow: Overview of chal- lenge 1 in...

1997
[24]

Holzer, C

J. Holzer, C. Coffrin, C. DeMarco, R. Duthu, S. Elbert, B. Eldridge, T. Elgindy, M. Garcia, S. Greene, N. Guo, E. Hale, B. Lesieutre, T. Mak, C. McMillan, H. Mit- telmann, H. Oh, R. O’Neill, T. Overbye, B. Palmintier, R. Parker, F. Safdarian, A. Tbaileh, P. Van Henten- ryck, A...

2024
[25]

Z. Wu, P. Sampourmahani, M. Mohammadisiahroudi, and T. Terlaky, A quantum dual logarithmic barrier method for linear optimization, INFORMS Journal on Optimization (2025)

2025
[26]

Das and B

A. Das and B. K. Chakrabarti, Colloquium: Quantum annealing and analog quantum computation, Rev. Mod. Phys.80, 1061 (2008)

2008
[27]

Abbas, A

A. Abbas, A. Ambainis, B. Augustino, A. B¨ artschi, H. Buhrman, C. Coffrin, G. Cortiana, V. Dunjko, D. J. Egger, B. G. Elmegreen,et al., Challenges and opportu- nities in quantum optimization, Nature Reviews Physics 6, 718 (2024)

2024
[28]

Tasseff, T

B. Tasseff, T. Albash, Z. Morrell, M. Vuffray, A. Y. Lokhov, S. Misra, and C. Coffrin, On the emerging po- tential of quantum annealing hardware for combinatorial optimization, Journal of Heuristics30, 325–358 (2024)

2024
[29]

Kim, S.-W

S. Kim, S.-W. Ahn, I.-S. Suh, A. W. Dowling, E. Lee, and T. Luo, Quantum annealing for combinatorial optimiza- tion: a benchmarking study, npj Quantum Information 11, 10.1038/s41534-025-01020-1 (2025)

2025 doi
[30]

J. Leng, E. Hickman, J. Li, and X. Wu, Quantum hamil- tonian descent, arXiv preprint arXiv:2303.01471 (2023)

2023 arXiv
[31]

Kushnir, J

S. Kushnir, J. Leng, Y. Peng, L. Fan, and X. Wu, Qh- dopt: A software for nonlinear optimization with quan- tum hamiltonian descent, INFORMS Journal on Com- puting37, 107 (2025)

2025
[32]

Z. Wu, M. Li, M. Zheng, M. Wang, J. Liu, S. Stein, A. Li, Y. Chen, and C. Liu, Benchmarking and resource analysis for augmented-lagrangian quantum hamiltonian descent (2026), arXiv:2605.12066 [quant-ph]

2026 arXiv
[33]

M. Li, L. Fan, and Z. Han, Quantum hamiltonian descent based augmented lagrangian method for constrained non- convex nonlinear optimization (2025), arXiv:2508.02969 [math.OC]

2025 arXiv
[34]

J. Leng, J. Li, Y. Peng, and X. Wu, Expanding hardware- efficiently manipulable hilbert space via hamiltonian em- bedding, Quantum9, 1857 (2025)

2025
[35]

Kushnir, J

S. Kushnir, J. Leng, Y. Peng, L. Fan, and X. Wu, Qhdopt: A software for nonlinear optimization with quantum hamiltonian descent (2024), arXiv:2409.03121 [quant-ph]

2024 arXiv
[36]

Eastin and E

B. Eastin and E. Knill, Restrictions on transversal en- coded quantum gate sets, Phys. Rev. Lett.102, 110502 (2009)

2009
[37]

Bombin and M

H. Bombin and M. A. Martin-Delgado, Quantum mea- surements and gates by code deformation, Journal of Physics A: Mathematical and Theoretical42, 095302 (2009)

2009
[38]

Horsman, A

D. Horsman, A. G. Fowler, S. Devitt, and R. V. Meter, Surface code quantum computing by lattice surgery, New Journal of Physics14, 123011 (2012)

2012
[39]

B. J. Brown, K. Laubscher, M. S. Kesselring, and J. R. Wootton, Poking holes and cutting corners to achieve clif- ford gates with the surface code, Phys. Rev. X7, 021029 (2017)

2017
[40]

Litinski, A Game of Surface Codes: Large-Scale Quan- tum Computing with Lattice Surgery, Quantum3, 128 (2019)

D. Litinski, A Game of Surface Codes: Large-Scale Quan- tum Computing with Lattice Surgery, Quantum3, 128 (2019)

2019
[41]

Litinski, Magic state distillation: Not as costly as you think, Quantum3, 205 (2019)

D. Litinski, Magic state distillation: Not as costly as you think, Quantum3, 205 (2019)

2019
[42]

Gidney, N

C. Gidney, N. Shutty, and C. Jones, Magic state culti- vation: growing t states as cheap as cnot gates (2024), arXiv:2409.17595 [quant-ph]

2024 arXiv
[43]

Rosenfeld, C

E. Rosenfeld, C. Gidney, G. Roberts, A. Morvan, N. Lacroix, D. Kafri, J. Marshall, M. Li, V. Sivak, D. Abanin, A. Abbas, R. Acharya, L. A. Beni, G. Aigeldinger, R. Alcaraz, S. Alcaraz, T. I. Andersen, M. Ansmann, F. Arute, K. Arya, W. Askew, N. As- trakhantsev, J. Atalaya, R. ...

2025
[44]

Jochym-O’Connor and R

T. Jochym-O’Connor and R. Laflamme, Using concate- nated quantum codes for universal fault-tolerant quan- tum gates, Phys. Rev. Lett.112, 010505 (2014)

2014
[45]

J. T. Anderson, G. Duclos-Cianci, and D. Poulin, Fault- tolerant conversion between the steane and reed-muller quantum codes, Phys. Rev. Lett.113, 080501 (2014)

2014
[46]

Bravyi, G

S. Bravyi, G. Smith, and J. A. Smolin, Trading classical and quantum computational resources, Phys. Rev. X6, 021043 (2016)

2016
[47]

M. Amy, P. Azimzadeh, and M. Mosca, On the controlled-not complexity of controlled-not–phase cir- cuits, Quantum Science and Technology4, 015002 (2018)

2018
[48]

Cowtan, S

A. Cowtan, S. Dilkes, R. Duncan, W. Simmons, and S. Sivarajah, Phase gadget synthesis for shallow circuits (2019), arXiv:1906.01734 [quant-ph]

2019 arXiv
[49]

Vandaele, S

V. Vandaele, S. Martiel, and T. Goubault de Brugi` ere, Phase polynomials synthesis algorithms for NISQ archi- tectures and beyond (2021), arXiv:2104.00934 [quant- ph]

2021 arXiv
[50]

N. J. Ross and P. Selinger, Optimal ancilla-free clif- ford+ t approximation of z-rotations, arXiv preprint arXiv:1403.2975 (2014)

2014 arXiv
[51]

Hadfield, On the representation of boolean and real functions as hamiltonians for quantum comput- ing, ACM Transactions on Quantum Computing2, 10.1145/3478519 (2021)

S. Hadfield, On the representation of boolean and real functions as hamiltonians for quantum comput- ing, ACM Transactions on Quantum Computing2, 10.1145/3478519 (2021)

2021 doi
[52]

Welch, D

J. Welch, D. Greenbaum, S. Mostame, and A. Aspuru- Guzik, Efficient quantum circuits for diagonal unitaries without ancillas, New Journal of Physics16, 033040 (2014)

2014
[53]

Zhang, K

S. Zhang, K. Huang, and L. Li, Depth-optimized quan- tum circuit synthesis for diagonal unitary operators with asymptotically optimal gate count, Phys. Rev. A109, 042601 (2024)

2024
[54]

C. C. Chang, K. S. McElvain, E. Rrapaj, and Y. Wu, Im- proving Schr¨ odinger equation implementations with gray code for adiabatic quantum computers, PRX Quantum 3, 020356 (2022)

2022
[55]

J. Hu, S. Jin, N. Liu, and L. Zhang, Quantum circuits for partial differential equations via Schr¨ odingerisation, Quantum8, 1563 (2024)

2024
[56]

Klappenecker and M

A. Klappenecker and M. Roetteler, On the irresistible efficiency of signal processing methods in quantum com- puting (2001), arXiv:quant-ph/0111039 [quant-ph]

2001 arXiv
[57]

M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information(Cambridge University Press, 2010)

2010
[58]

Y. Nam, Y. Su, and D. Maslov, Approximate quantum fourier transform with O(n log(n)) T gates, npj Quantum Information6, 26 (2020)

2020
[59]

Ahmadkhaniha, L

A. Ahmadkhaniha, L. Chen, J. Doliskani, and Z. Sun, Qrtlib: A library for fast quantum real transforms (2025), arXiv:2510.16625 [quant-ph]

2025 arXiv
[60]

D. C. McKay, C. J. Wood, S. Sheldon, J. M. Chow, and J. M. Gambetta, Efficientzgates for quantum comput- ing, Phys. Rev. A96, 022330 (2017)

2017
[61]

Gottesman, The heisenberg representation of quan- tum computers (1998), arXiv:quant-ph/9807006 [quant- ph]

D. Gottesman, The heisenberg representation of quan- tum computers (1998), arXiv:quant-ph/9807006 [quant- ph]

1998 arXiv
[62]

Aaronson and D

S. Aaronson and D. Gottesman, Improved simulation of stabilizer circuits, Phys. Rev. A70, 052328 (2004)

2004
[63]

M. Wang, C. Liu, A. Li, and A. Aniceto, Nwqec: A toolkit for fault-tolerant quantum circuit transpilation and t-count optimization (2025)

2025
[64]

M. Wang, C. Liu, S. Stein, Y. Ding, P. Das, P. J. Nair, and A. Li, Optimizing ftqc programs through qec transpiler and architecture codesign, arXiv preprint arXiv:2412.15434 (2024)

2024 arXiv
[65]

M. Wang, C. Liu, S. Garner, S. Stein, Y. Ding, P. J. Nair, and A. Li, Tableau-based framework for efficient logical quantum compilation, arXiv preprint arXiv:2509.02721 (2025)

2025 arXiv
[66]

This threshold is motivated by our numerical observa- tions: when the average fidelity relative to the classical solver is above 0.98, the success probability, defined as the probability of measuring a point within 0.1 of the optimal solutionx 0, remains nearly unchanged, with...
[67]

D. An, D. Fang, and L. Lin, Time-dependent unbounded Hamiltonian simulation with vector norm scaling, Quan- tum5, 459 (2021)

2021
[68]

Z. Song, M. Wang, M. Zheng, S. Stein, S. H. Bryngelson, J. M¨ ulmenst¨ adt, X. Li, A. Li, and C. Liu, NWQRE: A modular quantum resource estimation workflow for fault- tolerant algorithms (2026), manuscript in preparation

2026
[69]

A. M. Childs, Y. Su, M. C. Tran, N. Wiebe, and S. Zhu, Theory of Trotter error with commutator scaling, Phys- ical Review X11, 011020 (2021)

2021
[70]

Leng, Jiaqi and Hickman, Ethan and Li, Joseph and Wu, Xiaodi, Quantum Hamiltonian Descent: numerical simulation, real-machine deployment, and benchmarking,https://github.com/jiaqileng/ quantum-hamiltonian-descent, accessed: June 14, 2026

2026

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

One-hot realization In the one-hot construction, each coordinate register contains one qubit per grid point, as in Eq. (8). The cir- cuit acts on the full qubit Hilbert space, but the intended physical subspace is the tensor product of Hamming- weight-one subspaces, one for each coordinate. The potential circuit follows directly from the projec- tor repre...

[2] [2]

A grid point is encoded directly as a computational-basis bit string, so no one-hot subspace constraint is required

Binary realization The binary construction usesb= log 2 Nqubits per coordinate register. A grid point is encoded directly as a computational-basis bit string, so no one-hot subspace constraint is required. The potential term remains diago- nal and is compiled by the Pauli-Zexpansion described in Sec. IV. In particular, each support-local diagonal block is...

[3] [3]

These functions are plotted in Fig

One-Dimensional T argets We use four one-dimensional target functions as bench- mark problems. These functions are plotted in Fig. 2e to Fig. 2h. a. Centered Quadratic: f(u) = u− 1 2 2 .(A1) The unique global minimizer is u⋆ = 1 2 , f(u ⋆) = 0.(A2) b. Double-Well: f(u) = 4(u−0.3) 2(u−0.7) 2 + 0.02(u−0.55).(A3) The global minimizer is u⋆ ≈0.285901, f(u ⋆)≈...

[4] [4]

These four functions are plotted in Fig

Two-Dimensional T argets We mainly focus on four two-dimensional target func- tions for resource analysis and verification of the perfor- mance of the QHD algorithm. These four functions are plotted in Fig. 11. a. Coupled Quadratic 2D. f(x, y) = (x−0.25) 2 + 1.4(y−0.65) 2 + 0.2(x−0.25)(y−0.65). (A9) The unique global minimizer is (x⋆, y⋆) = (0.25,0.65), f...

[5] [5]

P. W. Shor, Polynomial-time algorithms for prime factor- ization and discrete logarithms on a quantum computer, SIAM review41, 303 (1999)

1999

[6] [6]

P. W. Shor, Algorithms for quantum computation: dis- crete logarithms and factoring, inProceedings 35th An- nual Symposium on Foundations of Computer Science (IEEE, 1994) pp. 124–134

1994

[7] [7]

Zhou, S.-T

L. Zhou, S.-T. Wang, S. Choi, H. Pichler, and M. D. 26 103 CX gates / Trotter step N log2 N Exact k2 AQFT AQFT-k2 16 32 64 128 256 512 Resolution (N) 102 103 Rz gates / Trotter step N log2 N (a) Three-Hump Camel 103 104 CX gates / Trotter step N log2 N Exact k2 AQFT AQFT-k2 16 32 64 128 256 512 Resolution (N) 102 103 Rz gates / Trotter step N log2 N (b) A...

2020

[8] [8]

I. M. Georgescu, S. Ashhab, and F. Nori, Quantum sim- ulation, Reviews of Modern Physics86, 153 (2014)

2014