pith. sign in

arxiv: 2601.02079 · v2 · pith:ZBZOORNLnew · submitted 2026-01-05 · 🧮 math.NA · cs.NA

Asymptotic condition numbers for linear ordinary differential equations: the generic real case

Pith reviewed 2026-05-21 16:30 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords asymptotic condition numberslinear ordinary differential equationsreal casegeneric casepointwise condition numberinitial-value sensitivitynumerical stability
0
0 comments X

The pith

For generic real linear ODEs the long-time limits of directional and pointwise condition numbers are determined explicitly.

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

The paper determines the long-time asymptotic behavior of two condition numbers that quantify how a small relative perturbation in the initial value of a real linear ODE y' = A y propagates to the solution at later times. One number tracks a perturbation in a fixed direction; the other takes the worst-case direction. The analysis is carried out for the generic case of the real matrix A, extending an earlier treatment that was restricted to complex coefficients. A sympathetic reader cares because these limits tell how sensitive a real dynamical system is to initial data over very long intervals, which directly affects the reliability of long-time numerical simulations.

Core claim

For a real linear ordinary differential equation y'(t) = A y(t) whose coefficient matrix A is in the generic case, the asymptotic (long-time) behaviors of the directional pointwise condition number and of the pointwise condition number are determined.

What carries the argument

The directional pointwise condition number (effect of a perturbation in one chosen direction) and the pointwise condition number (worst-case direction over all possible perturbations), both measured with relative errors.

If this is right

  • The condition numbers approach explicit limits or grow/decay at rates governed by the real parts of the eigenvalues of A.
  • Long-time sensitivity of the solution can be predicted from the spectrum and eigenvectors of A without computing the full fundamental matrix for every perturbation.
  • The same limits supply a practical a-priori bound on the amplification of initial rounding errors in long-time integration of generic real linear systems.

Where Pith is reading between the lines

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

  • The derived limits could be used to choose step-size or integrator order automatically so that accumulated error remains below a tolerance over a prescribed time horizon.
  • Similar asymptotic analysis might extend to linear systems with periodic coefficients or to small nonlinear perturbations around a linear flow.
  • Numerical experiments on random real matrices with distinct eigenvalues would provide an immediate independent check of the formulas.

Load-bearing premise

The real matrix A satisfies generic properties such as distinct eigenvalues or a Jordan structure that avoids special cancellations allowing the asymptotic limits to be derived in closed form.

What would settle it

For a concrete generic real matrix A with known eigenvalues, compute the two condition numbers numerically at successively larger times t and check whether they approach the explicit limits stated in the paper.

read the original abstract

The paper \cite{M0} studied, for a \emph{complex} linear ordinary differential equation $y^\prime(t)=Ay(t)$, the long-time propagation to the solution $y(t)$ of a perturbation of the initial value. By measuring the perturbations with relative errors, this paper introduced a directional pointwise condition number, defined for a specific initial value and for a specific direction of perturbation of this initial value, and a pointwise condition number, defined for a specific initial value and the worst-case scenario for the direction of perturbation. The asymptotic (long-time) behaviors of these two condition numbers were determined. The present paper analyzes such asymptotic behaviors in depth, for a \emph{real} linear ordinary differential equation in a generic case.

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

2 major / 2 minor

Summary. The manuscript extends prior work on complex linear ODEs to the real case, determining the long-time asymptotic behaviors of the directional pointwise condition number (for a fixed initial value and perturbation direction) and the pointwise condition number (worst-case direction over perturbations) for the system y'(t) = A y(t), where A is real and satisfies generic assumptions (distinct eigenvalues or conjugate pairs without special Jordan structure). The analysis measures perturbations via relative errors and identifies the dominant eigenvalue or conjugate pair as controlling the growth.

Significance. If the central derivations hold, the results complete the asymptotic picture for real-valued linear systems, which are the standard setting in applications. The generic-case assumption is standard and avoids polynomial prefactors, allowing clean exponential asymptotics tied to the spectral radius. Explicit formulas for the limits would be a useful addition to the literature on condition numbers for ODE initial-value problems.

major comments (2)
  1. §4, Theorem 4.2: the derivation of the asymptotic limit for the directional condition number appears to rely on the absence of resonance between the dominant eigenvalue and other modes; a brief expansion showing why genericity precludes the logarithmic or polynomial corrections seen in non-generic Jordan cases would strengthen the claim.
  2. §5.1, Eq. (5.3): the expression for the pointwise condition number limit is stated to be independent of the initial vector norm, but the preceding reduction step invokes a normalization that is not explicitly verified to be uniform over all generic real matrices; a short remark on the uniformity would remove any ambiguity.
minor comments (2)
  1. The notation for the directional perturbation vector v(t) is introduced in §2 but reused with a different scaling in §4; a single consistent definition or a clarifying remark would improve readability.
  2. Several references to the complex-case predecessor [M0] are given without page or theorem numbers; adding specific cross-references would help readers trace the extensions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive evaluation, and constructive suggestions. We respond to each major comment below and will incorporate clarifications in the revised manuscript.

read point-by-point responses
  1. Referee: §4, Theorem 4.2: the derivation of the asymptotic limit for the directional condition number appears to rely on the absence of resonance between the dominant eigenvalue and other modes; a brief expansion showing why genericity precludes the logarithmic or polynomial corrections seen in non-generic Jordan cases would strengthen the claim.

    Authors: We agree that an explicit clarification strengthens the presentation. Under the stated generic assumptions (distinct eigenvalues or simple conjugate pairs with no nontrivial Jordan structure), the spectral gap to the dominant mode is strictly positive and there are no resonances with subdominant modes. Consequently, the solution and its perturbations are asymptotically dominated by the leading term without polynomial prefactors or logarithmic corrections. We will add a short remark immediately after Theorem 4.2 making this reasoning explicit. revision: yes

  2. Referee: §5.1, Eq. (5.3): the expression for the pointwise condition number limit is stated to be independent of the initial vector norm, but the preceding reduction step invokes a normalization that is not explicitly verified to be uniform over all generic real matrices; a short remark on the uniformity would remove any ambiguity.

    Authors: We thank the referee for noting this point. The normalization step is homogeneous of degree zero and the dominant eigenspace (or invariant subspace for a conjugate pair) depends continuously on the matrix entries within the open set of generic real matrices. This ensures the limiting expression remains uniform. We will insert a brief clarifying sentence in §5.1 confirming the uniformity over the generic class. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation extends prior complex-case results to distinct real generic setting

full rationale

The paper determines asymptotic long-time behaviors of directional and pointwise condition numbers for real linear ODE y'=Ay in the generic case, building on the complex-case analysis in cited prior work [M0]. Genericity assumptions (distinct eigenvalues or specific Jordan structure) are standard technical conditions to avoid cancellations and are not self-referential or fitted to the target quantities. The central claims are derived from the spectral properties of the real matrix A and the solution structure of the linear system, without reducing to self-definitions, renamed empirical patterns, or load-bearing self-citations that lack independent verification. The extension to the real case introduces new analysis rather than recycling inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard properties of linear ODEs and the generic-case assumption for real matrices; no free parameters, invented entities, or non-standard axioms are visible in the abstract.

axioms (1)
  • domain assumption The real matrix A satisfies generic properties allowing derivation of asymptotic limits for condition numbers without degenerate cases.
    Invoked to restrict the analysis to the generic real case as stated in the abstract.

pith-pipeline@v0.9.0 · 5647 in / 1148 out tokens · 36595 ms · 2026-05-21T16:30:08.361471+00:00 · methodology

discussion (0)

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