REVIEW 2 major objections 4 minor
Reformulating the Lanczos recursion as Hamiltonian polynomials yields an orthogonal quantum Krylov method that needs no overlap-matrix regularization.
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 →
T0 review · grok-4.5
2026-07-13 02:43 UTC pith:DWEPLW3G
load-bearing objection Clean operator-level Lanczos lift that actually kills the overlap-matrix problem; the GQSP success-probability issue is real but already treated as practical, not fatal. the 2 major comments →
Orthogonal Quantum Krylov Diagonalisation
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
By expressing Lanczos vectors as polynomial transformations of the Hamiltonian and implementing them with block encoding plus generalized quantum signal processing, OQKD generates an orthonormal Krylov basis by construction. It therefore reproduces the classical Lanczos orthogonality, tridiagonal structure, and convergence behavior, completely eliminating the overlap-matrix regularization required by prior quantum Krylov methods.
What carries the argument
Operator-level Lanczos polynomials Pn(H̃) defined by the rescaled three-term recurrence, expanded in the Chebyshev basis, and applied via GQSP on a block-encoded walk operator; the coefficients α̃ and β̃ are obtained only from diagonal expectation values, closing the recursion without storing prior vectors.
Load-bearing premise
The success probability of high-degree Lanczos polynomials either stays usable or can be kept nearly constant by restarting with fixed low-degree polynomials without spoiling the exponential convergence rate.
What would settle it
On a larger Heisenberg or Hubbard instance, check whether restarted OQKD energy error continues to track classical Lanczos while the GQSP success probability for the accumulated restart polynomial stays of order 0.1 or higher rather than decaying exponentially with restart count or system size.
If this is right
- Quantum Krylov solvers can reach larger stable subspace dimensions without thresholding the overlap matrix.
- Measurement cost loses the exponential dependence on initial-state overlap that appears in regularized Chebyshev QKD.
- Only diagonal elements need measuring, avoiding Hadamard tests and cutting the number of terms thanks to the tridiagonal form.
- Approximate ground states from the restarted protocol can serve as high-overlap inputs to quantum phase estimation.
- Asymptotic query complexity remains the same as Chebyshev-based quantum Krylov methods.
Where Pith is reading between the lines
- If the restarted protocol’s success probability stays nearly constant with system size, OQKD-style preparation could become a standard front-end for fault-tolerant phase estimation on strongly correlated molecules.
- The same operator-level orthogonalization idea could extend to other classical three-term recurrences once suitable polynomial transforms exist.
- Growth of the GQSP normalization factor remains the practical bottleneck; any advance in synthesizing unbounded polynomials would immediately raise OQKD’s reach.
- Shot-budget comparisons between restarting and amplitude-amplified high-degree OQKD on larger lattices would quantify when each strategy wins.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript introduces Orthogonal Quantum Krylov Diagonalization (OQKD), which reformulates the classical Lanczos three-term recurrence at the operator level so that each Krylov vector is realized as a polynomial of the rescaled Hamiltonian acting on a fixed initial state. The polynomials are expanded in the Chebyshev basis and implemented via block encoding plus Generalized Quantum Signal Processing (GQSP). Because the basis is orthonormal by construction, the projected Hamiltonian is tridiagonal and no overlap-matrix regularization is required. The authors derive a recursive formula for the Lanczos coefficients from diagonal moments only (Eqs. 13–14, Appendix A), give the Chebyshev-coefficient recursion (Appendix B), and show that the GQSP query complexity matches that of Chebyshev-based QKD. Classical state-vector simulations on the 4 imes4 J1–J2 Heisenberg model reproduce classical Lanczos convergence to machine precision while keeping the overlap condition number near unity. A restarted protocol (ROQKD) that replaces a single high-degree polynomial by a sequence of fixed low-degree transformations is proposed for state preparation, with numerical evidence that success probability remains nearly constant while infidelity decays exponentially.
Significance. If the construction holds under realistic sampling noise and larger system sizes, OQKD closes a long-standing gap between quantum Krylov methods and the classical Lanczos algorithm: it supplies an orthogonal, tridiagonal quantum subspace method that eliminates the ill-conditioning and thresholding overhead of existing QKD/CQKD approaches. The measurement-complexity analysis (Sec. III D) cleanly separates the overlap-dependent exponential cost of CQKD from an implementation-dependent GQSP success probability, and the restarted protocol offers a concrete route to high-overlap initial states for Quantum Phase Estimation. The algebraic derivations (Appendices A–C) are self-contained and the numerics are parameter-free comparisons against the classical algorithm, giving the work a solid technical foundation.
major comments (2)
- [Sec. III C–D, Fig. 3, Eq. 14] Sec. III C–D and Fig. 3: the exponential decay of the GQSP success probability p_succ = 1/Λ_n^{2} with Lanczos dimension is acknowledged but never quantified under statistical sampling noise. Because each α_i and ⟨H^{2}⟩ must be estimated on the prepared state, finite-shot fluctuations will feed into the recursive coefficient construction (Eq. 14 and Appendix B). A short analysis or numerical experiment showing how shot noise propagates into the subsequent polynomials and into the final Ritz values is needed to substantiate the claim that OQKD’s measurement complexity is free of the |γ0| dependence that plagues CQKD.
- [Sec. IV, Appendix D, Fig. 4] Sec. IV and Appendix D: the restarted protocol is presented as preserving exponential Lanczos convergence while keeping success probability nearly constant (Fig. 4). The only numerical evidence is for a single 4 imes4 instance and small n_max. The Kaniel–Paige scaling argument in Appendix D predicts D ~ O(N^{2}) for random initial states; it is not shown whether a fixed n_max independent of N still yields a useful overlap after a polynomial number of restarts, or whether the accumulated polynomial Q^(R) eventually reintroduces an exponential cost. A scaling plot or analytic bound for the restart depth versus system size would strengthen the state-preparation claim.
minor comments (4)
- [Fig. 2] Fig. 2 caption and main text: the threshold value 10^{-9} used for QKD/CQKD is stated only in the caption; it should also appear in the methods paragraph so that the comparison is fully reproducible.
- [Eq. 22, Appendix C] Eq. (22) versus the tighter bound (C5): the paper uses the triangular-inequality estimate for Λ_n; a brief remark on how much the success probability could improve with the optimal Λ_n would be useful.
- [throughout] Typographical inconsistencies: “OKQD” appears in several places (e.g., Sec. III C, Algorithm 2) instead of “OQKD”; “staightworward” (p. 5) should be “straightforward”.
- [Algorithm 1] Algorithm 1 step 4: “recover the normalization factors associated with the Lanczos vectors” is slightly ambiguous; a one-sentence clarification that the vectors are already normalized by construction would help.
Circularity Check
No significant circularity: OQKD is a direct constructive operator-level transcription of the classical Lanczos three-term recurrence, verified by independent numerics.
full rationale
The paper defines Lanczos vectors as polynomials Pn(H̃)|Φ0⟩ obeying the classical three-term recurrence (Eqs. 9–10), obtains the coefficients α̃n, β̃n from measurable moments via the tridiagonal identity (Eq. 14 and Appendix A), expands them in the Chebyshev basis (Appendix B), and implements the resulting polynomials by block encoding + GQSP (Eqs. 18–25, Appendix C). Orthogonality, the tridiagonal projected Hamiltonian, and the classical convergence profile therefore hold by construction of the recurrence, not by fitting or by a self-citation chain. Numerical state-vector simulations on the independent 4 imes4 J1–J2 Heisenberg model simply confirm that the quantum construction reproduces the classical Lanczos curves (Fig. 2). Measurement-complexity scalings follow from standard sampling arguments once the overlap matrix is eliminated (Sec. III D). The restarted protocol (Sec. IV) is an engineering device that re-uses the same polynomials at fixed low degree; it introduces no fitted parameters that are later called predictions. No uniqueness theorems, ansatzes, or load-bearing self-citations appear. The derivation is therefore self-contained and non-circular.
Axiom & Free-Parameter Ledger
free parameters (2)
- n_max (restart depth)
- Λ_n (GQSP normalization)
axioms (4)
- domain assumption A block encoding of the rescaled Hamiltonian H̃ exists and can be qubitized into a walk operator W (standard LCU or PREP-SELECT construction).
- domain assumption GQSP can realize any polynomial whose Laurent form satisfies the unit-circle bound after normalization by Λ_n (Motlagh & Wiebe).
- standard math The classical three-term Lanczos recurrence produces an orthonormal basis of the Krylov subspace in exact arithmetic.
- domain assumption Diagonal moments ⟨ψ|H̃|ψ⟩ and ⟨ψ|H̃²|ψ⟩ can be estimated to sufficient precision by standard quantum measurements.
invented entities (2)
-
OQKD (Orthogonal Quantum Krylov Diagonalization)
no independent evidence
-
ROQKD (Restarted OQKD)
no independent evidence
read the original abstract
Quantum subspace-diagonalization methods, particularly Quantum Krylov Diagonalization (QKD), provide a promising route for computing low-energy spectra of quantum many-body Hamiltonians. However, existing quantum Krylov approaches rely on non-orthogonal Krylov bases, requiring overlap-matrix regularization that limits numerical stability and accuracy. In this work, we introduce an Orthogonal Quantum Krylov Diagonalization (OQKD) framework that reformulates the classical Lanczos recursion at the operator level, enabling an orthogonal quantum implementation of Krylov-subspace diagonalization. By expressing Lanczos vectors as polynomial transformations of the Hamiltonian, OQKD reproduces the orthogonality, tridiagonal structure, and convergence behavior of the classical Lanczos algorithm thus eliminating the need for overlap-matrix regularization. We further show that the required Lanczos polynomials can be implemented using block encoding and Generalized Quantum Signal Processing with the same asymptotic query complexity as Chebyshev-based QKD methods. Numerical simulations of the $J_1$--$J_2$ Heisenberg model confirm the classical Lanczos convergence and numerical stability of the proposed method, while the measurement-complexity scaling is established analytically. Building upon the OQKD framework, we then introduce a restarted state-preparation protocol that replaces a single high-degree polynomial transformation with a sequence of fixed low-degree transformations, maintaining an affordable block encoding success probability while retaining comparable convergence. These results establish OQKD as an orthogonal quantum analog of the classical Lanczos algorithm and identify the restarted protocol as a promising state-preparation strategy for Quantum Phase Estimation.
Figures
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.