Pith

open record

sign in

arxiv: 2607.05873 · v1 · pith:55DPFXQZ · submitted 2026-07-07 · quant-ph · math-ph· math.MP

Fixing Divergence in Carleman Linearization via Analytical Continuation

Mingshuo Zhu , Hayato Higuchi , Hokuto Iwakiri , Kouki Nakamura , Naohisa Sueishi , Shih-Yen Tseng , Shoichiro Tsutsui This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 reserved 2026-07-08 21:45 UTCglm-5.2pith:55DPFXQZrecord.jsonopen to challenge →

Figure 1
Figure 1. Figure 1: FIG. 1: Comparison of Carleman linearization with the exact solution of the logistic equation, a [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] reproduced from arXiv: 2607.05873
classification quant-ph math-phmath.MP
keywords equationsdifferentialdivergencemethodquantumanalyticalcarlemancontinuation
0
0 comments X

The pith

Conformal map cures Carleman linearization's long-time blowup

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

Carleman linearization is a technique that rewrites a nonlinear differential equation as an infinite-dimensional linear one, making it attractive for quantum algorithms. The method has a known defect: the resulting series solution diverges after a finite time. This paper pinpoints the cause: the spectral representation of the solution is a Laurent series (a power series that can include negative powers) with a finite radius of convergence, and time evolution eventually pushes the evaluation point outside that radius. The authors fix this by applying a Mobius conformal map, which is a simple rational transformation of the complex variable that sends the positive real time axis into the interior of the unit disk. Under this map, the divergent spectral series is replaced by a convergent one. The correction takes the concrete form of inserting a regularized function f_{M,c}(k,t) into each eigenmode of the Carleman solution. For positive integer eigenvalues, this function is a regularized incomplete beta function I_{1-omega}(k, M-k+1), which decays to zero as time grows, counteracting the exponentially growing factor e^{kt} that caused the original divergence. The authors prove a convergence condition c < (1-x_0)/x_0 for the logistic equation, where c is the conformal-map scale parameter and x_0 is the initial condition. They extend the method to KPP-Fisher reaction-diffusion equations (handling both integer and non-integer eigenvalue spectra) and to cubic nonlinearities in phase-field models, where a modified conformal map using the variable eta = e^{2t} is required because the standard map fails to move the relevant branch points outside the unit disk. Finally, they show the regularized solution can be implemented on a quantum computer via block encoding of the Carleman matrix and a Linear Combination of Unitaries (LCU) eigenvalue transformation, and provide a gate-complexity and error budget for the logistic case.

Core claim

The long-time divergence of Carleman linearization is caused by evaluating a Laurent series outside its radius of convergence. Inserting a regularized incomplete beta function f_{M,c}(k,t) = I_{1-omega}(k, M-k+1) into the spectral solution of the Carleman system, guided by a Mobius conformal map that sends the time axis into the unit disk, replaces the divergent factor e^{kt} with a bounded factor e^{kt} f_{M,c}(k,t) and restores convergence for all positive time, subject to an explicit condition relating the map parameter to the initial data.

What carries the argument

The regularized function f_{M,c}(k,t), defined as a regularized incomplete beta function I_{1-omega}(k, M-k+1) for positive eigenvalues k (and modified to k/2 for cubic nonlinearities), which is inserted into the eigenvalue decomposition of the Carleman matrix to replace the divergent exponential growth factor.

If this is right

  • Quantum algorithms for nonlinear ODEs and PDEs based on Carleman linearization can potentially run for arbitrarily long times without the exponential divergence that previously limited them, provided the convergence condition on the conformal-map parameter is met.
  • The regularization is an eigenvalue transformation of the Carleman matrix, so it can be implemented on a quantum computer using standard block-encoding and LCU or QSVT primitives, turning a classical analytic-continuation trick into a quantum circuit.
  • The need for a different conformal map (using eta = e^{2t}) for cubic nonlinearities versus quadratic ones suggests a systematic recipe: the map must be tailored to the singularity structure of the exact solution, and the paper's approach gives a template for identifying the right map for other nonlinearity classes.
  • If the same divergence mechanism and fix apply to lattice Boltzmann and Navier-Stokes simulations, as the authors suggest, the method could stabilize Carleman-based computational fluid dynamics on both classical and quantum hardware.

Where Pith is reading between the lines

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

  • The convergence condition c < (1-x_0)/x_0 ties the method's validity to the initial condition: for x_0 close to 1, the allowable range of c shrinks toward zero, meaning the regularization may become impractical for initial states near the nonlinear saturation point. This suggests a fundamental limitation for certain classes of initial data that is not explored in depth.
  • The error bound for the quantum implementation includes the condition number kappa(P_K) of the eigenvector matrix of the truncated Carleman matrix (Eq. 64), which is never quantified. Since Carleman matrices are non-normal, kappa could grow rapidly with truncation order K, potentially offsetting the exponential convergence gained from the conformal map. The practical quantum complexity may therefo
  • The ad hoc perturbation (epsilon_i, with magnitude up to 10^{-4}) introduced to break eigenvalue degeneracies in the KPP-Fisher L=3 case suggests that spectral degeneracies are a practical bottleneck for the method on systems with spatial structure. Whether this perturbation introduces systematic bias at long times is not analyzed.
  • The choice M = K (mapping series cutoff equal to Carleman truncation order) is advocated as the practical default, but the paper also shows that M > K can accelerate convergence or cause instability depending on the regime. This trade-off resembles a resolution-of-identity problem and may admit an optimal scaling relation M ~ alpha * K that the paper does not derive.

Load-bearing premise

The method assumes the eigenvalues and eigenvectors of the truncated Carleman matrix can be computed accurately and that the spectral decomposition is well-conditioned. For the KPP-Fisher system, the authors explicitly note that eigenvector computation becomes unstable for larger truncation orders due to spectral degeneracy, requiring an ad hoc perturbation to break the degeneracy, and the condition number of the eigenvector matrix appears in the error bound but is neverquant

What would settle it

If the condition number kappa(P_K) of the truncated Carleman eigenvector matrix grows superpolynomially with K for a non-trivial class of nonlinear ODEs, then the LCU normalization factor Lambda would grow accordingly, and the quantum gate complexity T_total = O(Lambda * K^2 * log K / epsilon_meas) would lose any quantum advantage despite the conformal-map regularization eliminating the series divergence.

Figures

Figures reproduced from arXiv: 2607.05873 by Hayato Higuchi, Hokuto Iwakiri, Kouki Nakamura, Mingshuo Zhu, Naohisa Sueishi, Shih-Yen Tseng, Shoichiro Tsutsui.

Figure 2
Figure 2. Figure 2: FIG. 2: Conformal mapping between the [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Comparison of heatmaps for the regularized incomplete beta function I [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Time evolution of the logistic solution after conformal mapping for [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: The Carleman solution to the logistic equation with di [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Phase diagram in the ( [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: Phase diagram in the ( [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: First component of the Carleman solution to the [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: Comparision among the Euler’s method, the original Carleman solution and the [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10: First component of the Carleman solution to the [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11: Comparision among the Euler’s method, the original Carleman solution and the [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12: Time evolution of the cubic logistic solution after applying the new conformal map, for [PITH_FULL_IMAGE:figures/full_fig_p019_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13: First component of the Carleman solution to the [PITH_FULL_IMAGE:figures/full_fig_p020_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14: Comparision among the Euler’s method, the original Carleman solution and the [PITH_FULL_IMAGE:figures/full_fig_p021_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15: First component of the Carleman solution to the [PITH_FULL_IMAGE:figures/full_fig_p021_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16: Comparision among the Euler’s method, the original Carleman solution and the [PITH_FULL_IMAGE:figures/full_fig_p021_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17: Phase diagram in the ( [PITH_FULL_IMAGE:figures/full_fig_p022_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18: Classical simulation of the quantum algorithm applied to the logistic equation with [PITH_FULL_IMAGE:figures/full_fig_p025_18.png] view at source ↗
read the original abstract

Nonlinear differential equations play a crucial role in modeling a wide range of phenomena, yet their solutions remain notoriously difficult to obtain. With the rapid development of quantum computing, quantum algorithms for efficiently solving such equations are actively being explored. One promising approach is based on Carleman linearization, which transforms nonlinear differential equations into linear systems. However, this method suffers from exponential divergence beyond a certain time scale. By reformulating the solutions in terms of eigenvalues and eigenvectors, we identify that this divergence originates from the Laurent expansion outside its neighborhood of convergence. To address this issue, we insert a regularized function to the divergent solution hinted by analytical continuation. We validate this divergence-correction method on both the logistic equation and some other partial differential equations like KPP-Fisher equations and Phase-Field models under periodic conditions. We implement our method for the logistic equation using the Linear Combination of Unitaries (LCU) quantum algorithm, providing a detailed complexity and error analysis.

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.

Referee Report

3 major / 9 minor

Summary. This manuscript addresses the long-time divergence problem in Carleman linearization, a technique that lifts nonlinear ODEs into infinite-dimensional linear systems. The authors identify the divergence as arising from evaluating a Laurent expansion outside its radius of convergence. For the logistic equation, they derive a convergence condition c < (1−x₀)/x₀ using a Möbius conformal map ζ = cω/(1−ω) and construct a regularized function f_{M,c}(k,t) = I_{1−ω}(k, M−k+1) (a regularized incomplete beta function) that replaces the divergent spectral factor. The method is extended to KPP–Fisher equations (quadratic nonlinearity, L=3 and L=5) and phase-field models (cubic nonlinearity, requiring a modified map η = e^{2t}). A quantum implementation using LCU with block encoding is presented for the logistic equation, including a complexity and error analysis.

Significance. The paper tackles a genuine and well-known obstacle in Carleman-based quantum algorithms for nonlinear ODEs. The logistic-equation derivation (Sec. II.B–III) is clean and self-contained: the convergence condition (Eq. 18) follows analytically from the Möbius map properties and the geometric series, and the regularized function emerges from a binomial expansion identity rather than being postulated. The explicit error bound (Eq. 60) for the logistic case is a concrete, falsifiable result. The quantum resource estimate (Sec. VI), including the block-encoding construction in Appendix A, is a useful contribution. However, the extension to PDEs (Sec. IV–V) relies on numerical validation without a priori convergence guarantees, which limits the significance of those results.

major comments (3)
  1. §IV–V, Eqs. (29), (40): The regularized function f_{M,c} is rigorously justified only for the logistic equation, where the spectral coefficients a_k are known in closed form (Eq. 12) and decay geometrically. For the KPP–Fisher and phase-field PDEs, the spectral coefficients a_k are not known in closed form, and their growth rate as k → ∞ is never characterized. Convergence of the regularized series requires not just |ω| < 1 but also convergence of the outer spectral sum Σ_k a_k (cω/(1−ω))^k. If |a_k| grows faster than (1/c)^k, the regularized series diverges regardless of M. No convergence condition analogous to Eq. (18) is derived for the PDE cases. The paper should either (a) derive such a condition or bound on a_k growth for the PDE systems, or (b) explicitly state as a limitation that the PDE extensions are heuristic, supported only by numerical evidence on small systems (L=3,5; K≤14
  2. §IV.A, Eq. (31): The degeneracy-breaking perturbation ε_i = s·(K+1)(i−L+1) with s ≤ 10⁻⁴ is introduced ad hoc to stabilize eigenvector computation. The condition number κ(P_K) appears in the error bound (Eq. 64) but is never quantified for any system. For the non-normal Carleman matrices in the PDE cases, κ(P_K) could be large, and the perturbation itself introduces an unquantified error into the solution. The paper should discuss the magnitude of κ(P_K) for the systems studied and bound the perturbation error, or at minimum state clearly that this is an open issue affecting the reliability of the PDE results.
  3. §VI, Eq. (67): The total error bound includes κ(P_K)·ε_poly, but for the logistic equation with K=M and n=K−1, the authors claim ε_poly = 0 in exact arithmetic. However, the LCU normalization factor Λ (Eq. 54) depends on the coefficients β_i = a_i(t)·α_BE^i, and the authors note that Λ may increase with K. The gate complexity (Eq. 71) scales as O(ΛK²logK/ε_meas), but no bound on Λ as a function of K is provided. Without this, the quantum advantage claim is incomplete: if Λ grows exponentially in K, the method may not be efficient. The paper should provide at least a scaling estimate for Λ in the logistic case.
minor comments (9)
  1. Abstract: 'analytical continuation' should be 'analytic continuation' (also appears in the title and throughout the body).
  2. §II.B, Eq. (13): The text states 'which coincides with the exact solution (7) whenever the geometric series converges' — the convergence condition |x₀ζ/(x₀−1)| < 1 should be stated explicitly here for clarity, even though it appears later in Eq. (18).
  3. §III.A, Eq. (22): The notation '∈ [0,1]' after the definition of f_{M,c} is ambiguous — it is unclear whether this is a claim that f_{M,c} always lies in [0,1], or a domain specification. If it is a claim, it should be justified.
  4. §IV.A, Fig. 9 caption: 'Comparision' should be 'Comparison'. The same typo appears in Figs. 11, 14, and 16 captions.
  5. §V, Eq. (40): The expression uses both ζ and ω' notation, but the relationship between the modified map variable η and the original ζ is not restated here, making the equation hard to follow without referring back to Eq. (38).
  6. §VI, Eq. (50): The block-encoding definition uses 's max_{i,j} |A_{ij}|' on the left side, but the right side has (⟨0···0|⊗I)U(A)(I⊗|0···0⟩). The normalization factor α_BE = s·max|A_{ij}| is defined in Eq. (51), but the notation in Eq. (50) is unusual — typically one writes α_BE · A = ⟨0|U(A)|0⟩. The current form may confuse readers familiar with standard block-encoding conventions.
  7. §VII: The conclusion states 'non-dissipative types of diffusion-reaction differential equations' in the abstract, but KPP–Fisher and phase-field models are dissipative. This appears to be an error in the abstract.
  8. Appendix A, Eq. (A12): The 4×4 matrix shown after the Carleman matrix A includes numerical artifacts (e.g., 1.8369702e-16+0.j, −5.5109106e-16+0.j) that should be cleaned up or noted as floating-point noise.
  9. References [25] and [33] appear to be duplicate citations of the same work by Itani and Succi.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive report. The referee correctly identifies the core contribution (logistic-equation derivation, explicit error bound, quantum resource estimate) and the main gap (PDE extensions lack a priori convergence guarantees). We agree with all three major comments and will revise accordingly: (1) we will explicitly state the PDE extensions as heuristic/numerically validated, with the spectral-coefficient growth issue flagged as an open problem; (2) we will add discussion of κ(P_K) and the perturbation error as open issues; (3) we will add a scaling estimate for Λ in the logistic case. No standing objections remain.

read point-by-point responses

  1. Referee: §IV–V, Eqs. (29), (40): The regularized function f_{M,c} is rigorously justified only for the logistic equation, where the spectral coefficients a_k are known in closed form (Eq. 12) and decay geometrically. For the KPP–Fisher and phase-field PDEs, the spectral coefficients a_k are not known in closed form, and their growth rate as k → ∞ is never characterized. Convergence of the regularized series requires not just |ω| < 1 but also convergence of the outer spectral sum Σ_k a_k (cω/(1−ω))^k. If |a_k| grows faster than (1/c)^k, the regularized series diverges regardless of M. No convergence condition analogous to Eq. (18) is derived for the PDE cases. The paper should either (a) derive such a condition or bound on a_k growth for the PDE systems, or (b) explicitly state as a limitation that the PDE extensions are heuristic, supported only by numerical evidence on small systems (L=3,5; K≤14

    Authors: The referee is correct on all counts. For the logistic equation, the closed-form coefficients a_k = (x_0/(1−x_0))^k (Eq. 12) decay geometrically, and the convergence condition (18) follows directly. For the KPP–Fisher and phase-field PDEs, we do not have closed-form expressions for the spectral coefficients, nor have we characterized their growth rate as k → ∞. The referee's observation that convergence of the regularized series requires |a_k| to not grow faster than (1/c)^k is mathematically precise and we agree that without bounding a_k growth, no convergence guarantee analogous to Eq. (18) can be claimed for the PDE cases. We will adopt option (b): the revised manuscript will explicitly state in Sections IV–V that the PDE extensions are heuristic, supported by numerical evidence on small systems (L=3,5; K≤14), and that deriving a priori bounds on the spectral coefficient growth for these systems is an open problem. We will also add the convergence requirement on Σ_k a_k (cω/(1−ω))^k as a stated condition that must be verified, making clear that our numerical results do not constitute a proof of convergence for general PDE systems. revision: yes

  2. Referee: §IV.A, Eq. (31): The degeneracy-breaking perturbation ε_i = s·(K+1)(i−L+1) with s ≤ 10⁻⁴ is introduced ad hoc to stabilize eigenvector computation. The condition number κ(P_K) appears in the error bound (Eq. 64) but is never quantified for any system. For the non-normal Carleman matrices in the PDE cases, κ(P_K) could be large, and the perturbation itself introduces an unquantified error into the solution. The paper should discuss the magnitude of κ(P_K) for the systems studied and bound the perturbation error, or at minimum state clearly that this is an open issue affecting the reliability of the PDE results.

    Authors: The referee raises a valid concern. The degeneracy-breaking perturbation (Eq. 31) was introduced for numerical stability of eigenvector computation, but we did not quantify κ(P_K) for any system, nor did we bound the error introduced by the perturbation itself. For the non-normal Carleman matrices arising in the PDE cases, κ(P_K) could indeed be large, and this directly affects the reliability of the error bound (Eq. 64). We will revise the manuscript to: (1) report the numerically computed values of κ(P_K) for the specific systems studied (L=3 KPP–Fisher, L=5 KPP–Fisher, L=3 phase-field) at the truncation orders used; (2) discuss the scaling of κ(P_K) with K based on these numerical observations; (3) state explicitly that bounding the perturbation error from the degeneracy-breaking term is an open issue. We agree that without this analysis, the reliability of the PDE results is not fully characterized. revision: yes

  3. Referee: §VI, Eq. (67): The total error bound includes κ(P_K)·ε_poly, but for the logistic equation with K=M and n=K−1, the authors claim ε_poly = 0 in exact arithmetic. However, the LCU normalization factor Λ (Eq. 54) depends on the coefficients β_i = a_i(t)·α_BE^i, and the authors note that Λ may increase with K. The gate complexity (Eq. 71) scales as O(ΛK²logK/ε_meas), but no bound on Λ as a function of K is provided. Without this, the quantum advantage claim is incomplete: if Λ grows exponentially in K, the method may not be efficient. The paper should provide at least a scaling estimate for Λ in the logistic case.

    Authors: The referee correctly identifies a gap in the quantum complexity analysis. While ε_poly = 0 in exact arithmetic for the logistic case with n = K−1, the LCU normalization factor Λ = Σ_{i=0}^{n} |β_i| with β_i = a_i(t)·α_BE^i is not bounded, and without this the efficiency of the quantum implementation cannot be fully assessed. We will add a scaling estimate for Λ in the logistic case. Specifically, for the logistic Carleman matrix with sparsity s=2 and α_BE = K (since the maximum matrix entry is K), the interpolation coefficients a_i(t) are determined by the polynomial p_n matching e^{kt} f_{M,c}(k,t) at k=1,...,K. We will provide a numerical study of Λ as a function of K for the parameter regime used in our implementation (c=4, x_0=0.1, t∈[0,10]), and discuss whether the growth is polynomial or exponential. We agree that if Λ grows exponentially, the method would not be efficient, and we will state this explicitly. We will also note that the interplay between the exponential decay of the regularization error (1−r)^K and the potential growth of Λ creates a trade-off that must be optimized, and we will characterize this trade-off at least numerically. revision: yes

Circularity Check

0 steps flagged

No circularity found; the derivation chain is self-contained with externally cited standard techniques

full rationale

The paper's central derivation chain is mathematically self-contained and does not reduce to its inputs by construction. (1) The divergence origin (Sec. II.B, Eq. 13) is derived from the eigenbasis expansion (Eq. 11) and the closed-form coefficients (Eq. 12), yielding a geometric series whose finite convergence radius is a standard analytic fact. (2) The Möbius conformal map (Eq. 14) is a standard technique cited to external authors Takahasi and Mori [34–36], not self-citation. (3) The convergence condition c < (1−x₀)/x₀ (Eq. 18) follows from requiring |r| = |cx₀/(1−x₀)| < 1 in the geometric series — it is derived analytically, not fitted. (4) The regularized function f_{M,c}(k,t) = I_{1−ω}(k, M−k+1) (Eq. 23) emerges from truncating the binomial expansion of (cω/(1−ω))^k at order M (Eqs. 20–22), which is a mathematical identity relating the truncated binomial sum to the regularized incomplete beta function. (5) The numerical phase diagrams (Figs. 6, 7, 17) validate the analytic bound rather than defining it — the bound exists independently. (6) The quantum implementation (Sec. VI) uses standard block encoding [31] and LCU techniques with an explicit error decomposition (Eq. 67) that is self-contained for the logistic equation. The paper is transparent about the gap in extending to PDEs: 'Deriving sharp a priori bounds for these general systems is left as future work.' This is a completeness limitation, not circularity. No self-citations are load-bearing for the central argument, no fitted parameters are renamed as predictions, and no definitions are circular.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 0 invented entities

The paper introduces no new physical entities, particles, or forces. The regularized function f_{M,c} is derived from standard special functions (incomplete beta), not invented. The free parameters (c, M, s) are standard numerical truncation/regularization parameters. The main burden is the domain assumption that the spectral decomposition is available and well-conditioned, which is explicitly violated in the degenerate KPP-Fisher case requiring ad hoc perturbation.

free parameters (3)
  • c (conformal map scale) = varies by problem (e.g., c=4 for logistic x₀=0.1, c=1 for KPP-Fisher L=3)
    Free parameter of the Möbius map, constrained by analytic bound c < (1−x₀)/x₀ for logistic. Chosen by the user; affects convergence rate and stability.
  • M (mapped series truncation) = typically M=K
    Truncation order of the conformally mapped series. The paper recommends M=K but does not derive an optimal ratio. Too large M relative to K causes instability.
  • s (degeneracy-breaking perturbation) = ≤10⁻⁴
    Random perturbation parameter introduced in Eq. (31) to break eigenvalue degeneracy in KPP-Fisher L=3. Ad hoc; no systematic justification for the magnitude or form.
axioms (3)
  • domain assumption The Carleman matrix A admits a spectral decomposition with eigenvalues and eigenvectors that can be computed or approximated.
    The entire regularization framework (Sec. III onward) depends on the eigenvalue decomposition of A. For the logistic equation this is explicit (Eq. 10); for PDEs it is assumed numerically.
  • domain assumption The conformal map that fixes divergence depends on the singularity structure of the exact solution, which is assumed known or inferable.
    Sec. V shows the cubic case requires a different map (η=e^{2t}) because branch points are on the imaginary axis. The paper acknowledges 'the optimal conformal map should depend on the singularity structure' (Sec. VII) but does not provide a systematic selection procedure.
  • domain assumption The regularized incomplete beta function I_{1−ω}(k, M−k+1) correctly generalizes to non-integer eigenvalues without modification.
    Used for L=5 KPP-Fisher (Sec. IV.B) where eigenvalues are non-integer. The gamma-function definition (Eq. 22) supports this, but no rigorous proof of convergence properties for non-integer spectra is given.

pith-pipeline@v1.1.0-glm · 22349 in / 2990 out tokens · 514157 ms · 2026-07-08T21:45:28.378970+00:00 · methodology

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

36 extracted references · 36 canonical work pages · 15 internal anchors

  1. [1]

    (39) and the divergence can be fixed

    In this case, In this case, the regularized function is modified into fM,c(k,t)=  I1−ω( k 2 ,M− k 2 +1),k>0, 1,k≤0. (39) and the divergence can be fixed. With this modification, the numerical approximation takes the form y(t)=lim K→∞  KX k=−2K akekt  =lim K→∞ 2KX k=0 a−k ζ−k + KX k=1 ak ζkI1−ω′ k 2,M− k 2 +1 !! . (40) where we fol...

    work page

  2. [2]

    Zeigler, H

    B. Zeigler, H. Pr ¨ahofer, and T. G. Kim,Theory of Modeling and Simulation, 2nd ed.Academic Press, 2000

    work page 2000

  3. [3]

    Study of the diffusion equation with growth of the quantity of matter and its application to a biological problem,

    A. N. Kolmogorov, I. G. Petrovsky, and N. S. Piskunov, “Study of the diffusion equation with growth of the quantity of matter and its application to a biological problem,”Bull. Moscow Univ. Math. Ser . A1(1937) 1–25

    work page 1937

  4. [4]

    The wave of advance of advantageous genes,

    R. A. Fisher, “The wave of advance of advantageous genes,”Ann. Eugen.7(1937) 355–369

    work page 1937

  5. [5]

    On the theory of phase transitions,

    L. D. Landau, “On the theory of phase transitions,”Zh. Eksp. Teor . Fiz.7(1937) 19–32

    work page 1937

  6. [6]

    On the theory of superconductivity,

    V . L. Ginzburg and L. D. Landau, “On the theory of superconductivity,”Zh. Eksp. Teor . Fiz.20 (1950) 1064–1082

    work page 1950

  7. [7]

    Free energy of a nonuniform system. i. interfacial free energy,

    J. W. Cahn and J. E. Hilliard, “Free energy of a nonuniform system. i. interfacial free energy,”J. Chem. Phys.28(1958) 258–267

    work page 1958

  8. [8]

    Ground state structures in ordered binary alloys with second neighbor interactions,

    S. M. Allen and J. W. Cahn, “Ground state structures in ordered binary alloys with second neighbor interactions,”Acta Metall.27(1979) 1085–1095

    work page 1979

  9. [9]

    Phase-field method for computationally efficient modeling of solidification with arbitrary interface kinetics,

    A. Karma and W.-J. Rappel, “Phase-field method for computationally efficient modeling of solidification with arbitrary interface kinetics,”Phys. Rev. E53(1996) R3017–R3020

    work page 1996

  10. [10]

    Bellman,Dynamic Programming

    R. Bellman,Dynamic Programming. Princeton University Press, Princeton, 1957

    work page 1957

  11. [11]

    Exponential quantum speedup in simulating coupled classical oscillators,

    R. Babbush, D. W. Berry, R. Kothari, R. D. Somma, and N. Wiebe, “Exponential quantum speedup in simulating coupled classical oscillators,” inProc. IEEE FOCS. 2023

    work page 2023

  12. [12]

    A Theory of Trotter Error

    A. M. Childs, Y . Su, M. C. Tran, N. Wiebe, and S. Zhu, “Theory of trotter error with commutator scaling,”Phys. Rev. X11no. 1, (2021) 011020,arXiv:1912.08854 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  13. [13]

    Tailoring Term Truncations for Electronic Structure Calculations Using a Linear Combination of Unitaries

    R. Meister, S. C. Benjamin, and E. T. Campbell, “Tailoring term truncations for electronic structure calculations using a linear combination of unitaries,”Quantum6(2022) 637,arXiv:2007.11624 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  14. [14]

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

    A. Gily ´en, Y . Su, G. H. Low, and N. Wiebe, “Quantum singular value transformation and beyond: Exponential improvements for quantum matrix arithmetics,” inProc. 51st ACM SIGACT Symp. 33 Theory Comput.2018.arXiv:1806.01838 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  15. [15]

    A Grand Unification of Quantum Algorithms

    J. M. Martyn, Z. M. Rossi, A. K. Tan, and I. L. Chuang, “Grand unification of quantum algorithms,” PRX Quantum2no. 4, (2021) 040203,arXiv:2105.02859 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  16. [16]

    Linear combination of Hamiltonian simulation for nonunitary dynamics with optimal state preparation cost

    D. An, J.-P. Liu, and L. Lin, “Linear combination of hamiltonian simulation for nonunitary dynamics with optimal state preparation cost,”Phys. Rev. Lett.131no. 15, (2023) 150603, arXiv:2303.01029 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  17. [17]

    Quantum algorithm for partial differential equations of nonconservative systems with spatially varying parameters,

    Y . Sato, H. Tezuka, R. Kondo, and N. Yamamoto, “Quantum algorithm for partial differential equations of nonconservative systems with spatially varying parameters,”Phys. Rev. Appl.23no. 1, (2025) 014063

    work page 2025

  18. [18]

    Efficient Hamiltonian Simulation for Solving Option Price Dynamics

    J. Gonzalez-Conde, ´A. Rodr´ıguez-Rozas, E. Solano, and M. Sanz, “Efficient hamiltonian simulation for solving option price dynamics,”Phys. Rev. Res.5no. 4, (2023) 043220,arXiv:2101.04023 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  19. [19]

    Quantum simulation of partial differential equations via Schrodingerisation: technical details

    S. Jin, N. Liu, and Y . Yu, “Quantum simulation of partial differential equations: Applications and detailed analysis,”Phys. Rev. A108no. 3, (2023) 032603,arXiv:2212.14703 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  20. [20]

    Quantum algorithm for nonlinear differential equations

    S. Lloyd, G. D. Palma, C. Gokler, B. Kiani, Z.-W. Liu, M. Marvian, F. Tennie, and T. Palmer, “Quantum algorithm for nonlinear differential equations,”arXiv:2011.06571 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  21. [21]

    Efficient quantum algorithm for dissipative nonlinear differential equations

    J.-P. Liu, H. Ø. Kolden, H. K. Krovi, N. F. Loureiro, K. Trivisa, and A. M. Childs, “Efficient quantum algorithm for dissipative nonlinear differential equations,”Proc. Natl. Acad. Sci. U.S.A.118no. 35, (2021) e2026805118,arXiv:2011.03185 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  22. [23]

    (Cited on page 3)

    H. Krovi, “Improved quantum algorithms for linear and nonlinear differential equations,”Quantum7 (Feb., 2023) 913.https://doi.org/10.22331/q-2023-02-02-913

    work page doi:10.22331/q-2023-02-02-913 2023

  23. [24]

    Further improving quantum algorithms for nonlinear differential equations via higher-order methods and rescaling,

    P. C. S. Costa, P. Schleich, M. E. S. Morales, and D. W. Berry, “Further improving quantum algorithms for nonlinear differential equations via higher-order methods and rescaling,”npj Quantum Information11no. 1, (Aug., 2025) .http://dx.doi.org/10.1038/s41534-025-01084-z

    work page doi:10.1038/s41534-025-01084-z 2025

  24. [25]

    Quantum algorithms for general nonlinear dynamics based on the Carleman embedding

    D. Jennings, K. Korzekwa, M. Lostaglio, A. T. Sornborger, Y . Subasi, and G. Wang, “Quantum algorithms for general nonlinear dynamics based on the carleman embedding.” 2025. https://arxiv.org/abs/2509.07155

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  25. [26]

    Analysis of carleman linearization of lattice boltzmann,

    W. Itani and S. Succi, “Analysis of carleman linearization of lattice boltzmann,”Fluids7no. 1, 34 (2022) 24

    work page 2022

  26. [27]

    Quantum Algorithm for Lattice Boltzmann (QALB) Simulation of Incompressible Fluids with a Nonlinear Collision Term

    W. Itani, K. R. Sreenivasan, and S. Succi, “Quantum algorithm for lattice boltzmann simulation of incompressible fluids with a nonlinear collision term,”Phys. Fluids36no. 1, (2024) 017112, arXiv:2304.05915 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [28]

    Three Carleman routes to the quantum simulation of classical fluids

    C. Sanavio, R. Scatamacchia, C. D. Falco, and S. Succi, “Three carleman routes to the quantum simulation of classical fluids,”Phys. Fluids36no. 5, (2024) 057143,arXiv:2402.16686

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  28. [29]

    Efficient quantum algorithm for nonlinear reaction-diffusion equations and energy estimation

    J.-P. Liu, D. An, D. Fang, J. Wang, G. H. Low, and S. Jordan, “Efficient Quantum Algorithm for Nonlinear Reaction–Diffusion Equations and Energy Estimation,”Commun. Math. Phys.404no. 2, (2023) 963–1020,arXiv:2205.01141 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  29. [30]

    Divergence-free algorithms for solving nonlinear differential equations on quantum computers,

    K. Endo and K. Z. Takahashi, “Divergence-free algorithms for solving nonlinear differential equations on quantum computers,”arXiv:2411.16233 [quant-ph]

    work page arXiv

  30. [31]

    Globalizing the carleman linear embedding method for nonlinear dynamics,

    I. Novikau and I. Joseph, “Globalizing the carleman linear embedding method for nonlinear dynamics,”arXiv:2510.15715 [quant-ph]

    work page arXiv

  31. [32]

    Explicit Quantum Circuits for Block Encodings of Certain Sparse Matrices

    D. Camps, L. Lin, R. Van Beeumen, and C. Yang, “Explicit Quantum Circuits for Block Encodings of Certain Sparse Matrices,”SIAM J. Matrix Anal. Appl.45no. 1, (2024) 801–827, arXiv:2203.10236 [quant-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  32. [33]

    Application de la th ´eorie des ´equations int´egrales lin´eaires aux syst`emes d’´equations diff´erentielles non lin´eaires,

    T. Carleman, “Application de la th ´eorie des ´equations int´egrales lin´eaires aux syst`emes d’´equations diff´erentielles non lin´eaires,”Acta Mathematica59(Dec., 1932) 63–87

    work page 1932

  33. [34]

    Analysis of Carlemann Linearization of Lattice Boltzmann

    W. Itani and S. Succi, “Analysis of carlemann linearization of lattice boltzmann,”Fluids7no. 1, (2022) 24,arXiv:2111.11327 [physics.flu-dyn]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  34. [35]

    Convergence acceleration of taylor series by change of variables (a study on numerical algorithms),

    H. Takahasi and M. Mori, “Convergence acceleration of taylor series by change of variables (a study on numerical algorithms),”RIMS Kokyuroku172(1973) 78–87

    work page 1973

  35. [36]

    Quadrature formulas obtained by variable transformation,

    M. Mori, “Quadrature formulas obtained by variable transformation,”Numer . Math.35no. 3, (1980) 163–174

    work page 1980

  36. [37]

    Analytic continuation of some special functions by variable transformation,

    H. Takahasi and M. Mori, “Analytic continuation of some special functions by variable transformation,”Japan Journal of Applied Mathematics1no. 3, (1984) 337–346. 35

    work page 1984