Paving the way to a $\operatorname{T}$-coercive method for the wave equation

Carolina Urz\'ua-Torres; Daniel Hoonhout; Richard L\"oscher

REVIEW 1 major objections 4 minor 10 references

Paving the way to a $\operatorname{T}$-coercive method for the wave equation

T0 review · 1 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A frequency-dependent transformation makes the wave equation's weak form coercive, so discretizations stay stable and accurate as the frequency parameter grows.

desk verdict The T_mu construction is genuinely new and the coercivity identity is right, but the advertised mu-independent error estimates are wrong; Section 5's algebraic bound needs a fix. read the letter →

arxiv 2509.02288 v1 pith:XE2MB6QH submitted 2025-09-02 math.NA cs.NA

Daniel Hoonhout , Richard Löscher , Carolina Urzúa-Torres This is my paper

classification math.NAcs.NA MSC 65M6065N3035L05

keywords T-coercivitywaveequationspace-timefiniteelementmethodtransformationoperatorGalerkin-Bubnovformulationmu-independentstabilitySobolevspacesordinarydifferential

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

The paper takes the first step toward a space-time variational method for the wave equation in which the bilinear form is coercive with a constant independent of the frequency parameter. Using the model ODE u''+mu u=f, it constructs a transformation operator T_mu that modifies only the test space, leaving the trial space a standard Sobolev space. The key result is that applying T_mu turns the non-coercive bilinear form into the H^1_mu inner product, giving a unique solution and best-approximation error estimates whose constants do not depend on mu. A reader should care because this removes the usual degradation as mu grows and points toward unconditionally stable, adaptive space-time finite element methods for the wave equation.

What carries the argument

The load-bearing object is the real-valued transformation operator T_mu. It is the real part of the composition (D_+)^(-1)D_-, where D_±=±i d/dt+sqrt(mu), written explicitly as an integral operator with cosine and sine kernels. Its role is to convert the weakly coercive bilinear form b_mu(u,v)=<-u',v'>+mu<u,v> into the full H^1_mu inner product on the second slot: b_mu(u,T_mu w)=<u',w'>+mu<u,w>. Because the coercivity constant is 1, all stability and error bounds become independent of mu.

What would settle it

Compute T_mu u explicitly for u(t)=t on (0,1) and symbolically check whether (T_mu u)''+mu T_mu u = -u''+mu u holds as a distribution on test functions vanishing at 0. More directly, assemble the discrete matrix for b_mu(phi_i,T_mu phi_j) with piecewise-linear hat functions for large mu and check that the smallest generalized eigenvalue relative to the H^1_mu mass matrix stays at 1; if it decays with mu, the claimed mu-independent coercivity is false.

Watch

Extended reading notes

Core claim

For the ODE u''+mu u=f with zero initial conditions and mu>0, the paper defines T_mu w(t)=integral from t to T of [cos(sqrt(mu)(t-s)) w'(s) - sqrt(mu) sin(sqrt(mu)(t-s)) w(s)] ds, the real part of (D_+)^(-1)D_- with D_±=±i d/dt+sqrt(mu). This operator maps H^1_{0,}(0,T) into H^1_{,0}(0,T) and satisfies the distributional identity (T_mu u)''+mu T_mu u = -u''+mu u in the dual space. From this, the paper proves b_mu(u,T_mu w)=<u',w'>+mu<u,w>, so b_mu(u,T_mu u)=||u||^2_{H^1_mu}. The result is a Galerkin-Bubnov formulation that is T_mu-coercive with coercivity constant 1, hence unconditionally stable, with Cea-type best-approximation error estimates independent of mu. At mu=0 the operator reduces

Load-bearing premise

The whole argument hinges on a delicate integration-by-parts identity: after applying the transformation, the second derivative of the transformed function plus mu times itself must equal -u''+mu u in the weak sense. If that identity is wrong, the coercivity equality that powers the method collapses.

Editorial extensions

If this is right

The discrete Galerkin-Bubnov method with piecewise-linear finite elements is unconditionally stable and satisfies ||u-u_h||_{H^1_mu} <= inf_{w_h} ||u-w_h||_{H^1_mu} <= c(h^{s-1}+mu h^s)|u|_{H^s} for s in [1,2], with no mu-dependent constant.
Numerical examples with mu=1, 1000, and 100000 show matching convergence rates, confirming that the parameter dependence does not degrade stability.
Because T_mu depends on mu explicitly, its extension to the wave equation via Fourier expansion in space acts in both time and space, giving a route to a coercive space-time formulation on bounded Lipschitz domains.
Only the test space is altered, so the trial space remains a standard Sobolev space and the error analysis is the usual Cea argument.
For f in L2, the right-hand side functional <f,T_mu w> is bounded by (T/sqrt(2))||f||_{L2}||w||_{H^1_mu}, placing L2 inside the class F for which the mu-independent solution bound holds.

Reading between the lines

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

The identity b_mu(u,T_mu w)=inner product means the method effectively symmetrizes the nonsymmetric space-time form; the same strategy could apply to first-order hyperbolic systems or variable-coefficient problems by choosing D_+ differently.
The real-part projection discards the imaginary component of the complex transformation; using the full complex-valued test space might yield an even simpler identity and could improve conditioning, at the cost of complex arithmetic.
For the wave equation, the natural extension replaces mu with the spatial eigenvalue and applies T_mu in time; an open question the paper does not address is how spatial boundary terms interact with the transformation on Lipschitz domains.
The mu-independent coercivity suggests the formulation is a good starting point for goal-oriented adaptivity in space and time, since standard residual-based error estimators would not need frequency-weighted constants.

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

The T_mu operator is the real thing. Explicit mu dependence is new relative to Steinbach–Zank's H_T, and the core identity b_mu(u,T_mu w) = + mu is proved cleanly. I checked the delicate distributional step in Lemma 3.1: (T_mu u)'' + mu T_mu u = -u'' + mu u follows from the representation using D_+^-1 D_- and the terminal condition; the computation holds. So the T_mu-coercivity and the unconditionally stable Bubnov formulation are solid. Good credit: the paper keeps the trial space standard and only modifies the test space, which is a nice feature.

The soft spot is the error estimate claim. The abstract and Section 5 say 'error estimates independent of mu,' but that is not true in the H^1_mu norm. The norm contains mu |u|^2_L2, so inf_{w_h} ||u-w_h||_{H^1_mu} grows at least like sqrt(mu) times the L2 distance for fixed h. The displayed bound c(h^{s-1}+mu h^s)|u|_{H^s} is also algebraically wrong: standard interpolation gives sqrt(mu) h^s, not mu h^s. Even with that correction, the bound is mu-dependent. The numerics are fine—they show stability in practice—but the theoretical claim overreaches. The authors need to either prove a mu-independent bound in a norm without the mu-weighted L2 term, or explicitly state the mu-dependence.

The paper is careful to say the wave equation extension is future work, so the significance is limited to a model ODE. That's fine as a first step, but it means the main claim to watch is the error estimate, and right now it's overstated. The citation pattern is reasonable: [8] is the correct comparator, and the relationship to the Steinbach–Zank H_T is spelled out in Remark 3.1.

Verdict: the core contribution is worth a serious referee, but the error-estimate section needs correction before this is publishable as is. I'd engage with it after that fix.

Referee Report

1 major / 4 minor

Summary. The paper studies the model ODE u'' + μu = f on (0,T) with homogeneous initial conditions. It constructs a transformation operator Tμ = Re((D_+)^{-1}D_-), where D_± = ±i∂_t + √μ, mapping H^1_{0,}(0,T) into H^1_{,0}(0,T), and proves the identity bμ(u,Tμw) = ⟨u',w'⟩ + μ⟨u,w⟩. This yields Tμ-coercivity of the bilinear form with coercivity constant 1 in the μ-dependent norm ∥·∥_{H^1_μ}, a well-posed modified variational formulation, a stability bound for a class F of right-hand sides containing L², and a conforming piecewise-linear discretization. Numerical experiments for μ ∈ {1,1000,10^5} are presented.

Significance. The explicit construction of Tμ is elegant, the proof of the coercivity identity is essentially correct and self-contained, and the treatment of the right-hand side (Remark 4.1) gives a clean μ-independent stability bound for L² data. As a stepping stone toward space-time T-coercive methods for the wave equation, the paper is potentially useful. However, the advertised 'error estimates independent of μ' is not currently supported: the H^1_μ norm itself depends on μ, and the displayed interpolation estimate in Section 5 is both algebraically incorrect and μ-dependent. This overclaim affects the central message and must be fixed before the paper can be accepted.

major comments (1)

[Section 5 (displayed error estimate; also Abstract and Conclusions)] The paper's main advertised claim, 'error estimates independent of μ', is contradicted by the displayed bound ∥u-u_h∥_{H^1_μ} ≤ c(h^{s-1}+μh^s)|u|_{H^s}. The norm ∥·∥_{H^1_μ} defined in Section 2 contains μ∥·∥²_{L²}. Hence for any w_h∈X_h, ∥u-w_h∥_{H^1_μ} ≥ √μ ∥u-w_h∥_{L²}. For fixed h and u∉X_h, the L² distance to X_h has a positive infimum, so the H^1_μ best-approximation error necessarily grows at least like √μ as μ→∞. Moreover, the interpolation estimate is algebraically wrong: from ∥u-I_hu∥_{L²} ≤ ch^s|u|_{H^s} and ∥(u-I_hu)'∥_{L²} ≤ ch^{s-1}|u|_{H^s}, the correct bound is c(h^{s-1}+√μ h^s)|u|_{H^s}, not c(h^{s-1}+μh^s)|u|_{H^s}. Even with this correction the bound is μ-dependent. The Abstract and Conclusions must be revised to state the actual μ-dependence, or the authors must prove a genuinely μ-independent estimate in a norm such as the H¹ seminorm. Table 1 reports only H¹-semino

minor comments (4)

[Lemma 3.1, proof] The distributional identity (Tμu)''+μTμu = -u''+μu in [H^1_{0,}(0,T)]' is the most delicate step. Since Tμu satisfies only the terminal condition, the pairing with w∈H^1_{0,} has a boundary contribution at t=T; this is handled by the identity (Tμu)'(T)=-u'(T), but the proof should spell out this cancellation explicitly for readability.
[Section 5, Table 1] The text says the method 'gives optimal orders of convergence in H¹ right from the start'. For the singular example u(t)=t²(T-t)^{3/4}∈H^s, s<5/4, the expected H¹ order is s-1≈0.25, so the asymptotic rates around 0.26-0.31 are consistent; this qualification should be added. Also, only H¹-seminorm errors are listed for μ>1; the H^1_μ norm used in the theoretical estimate should be reported to support the stability claims.
[Section 5, Eq. (10)] The method is called 'Galerkin-Bubnov', but in the original variables the test space is Tμ(X_h), not X_h; after Lemma 4.1 it is equivalent to a standard Galerkin method for the symmetric form with a modified right-hand side. This terminology should be clarified.
[Theorem 4.2] The definition of F uses 'with some C>0 independent of μ'. It would be clearer to state that the constant C may depend on f but not on μ, as is then used in (9).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Tμ is explicitly constructed from D±, and the coercivity identity follows by direct computation, not from a fitted parameter or self-citation.

full rationale

The derivation chain is self-contained. Tμ is defined in Eq. (4) as Re(D_+^{-1}D_-) with D_±=±i∂t+√μ, and its closed form is computed in Eq. (5). Lemma 3.1 proves (Tμu)''+μTμu = -u''+μu by differentiating the explicit integral formula, using (Tμu)(T)=0 and (Tμu)'(T)=-u'(T). Lemma 4.1 then obtains bμ(u,Tμw)=<u',w'>+μ<u,w> and bμ(u,Tμu)=||u||²_H1μ by substitution and the representation u(t)=∫_0^t q(s) ds with the integration-by-parts identity (7). This is a proof of a proved identity, not a renaming: the relation bμ(u,Tμu)=||u||²_H1μ is an algebraic consequence of D_+D_-=∂²+μ and the chosen boundary conditions, and no parameter is fitted to any data. The only self-citation, [5] (Hoonhout-Löscher-Steinbach-Urzúa-Torres), appears in a survey list of other space-time formulations and is not load-bearing for the coercivity proof or the discrete error analysis. Theorem 2.1 is external (Zank's thesis). The Section 5 claim that inf_{w_h∈X_h}||u-w_h||_{H1μ} can be bounded independent of μ is a genuine correctness concern—the H1μ norm contains μ||·||²_L2, so the best-approximation error generally grows with μ—but that is an error in the claimed estimate, not a circular step. Therefore no circularity is present.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters; mu is given. The paper relies on standard Sobolev space theory and one external theorem from Zank's thesis. No invented entities.

assumptions (3)

domain assumption Theorem 2.1 (boundedness, bounded invertibility, surjectivity of b_mu on H^1_{0,} x H^1_{,0}) is taken from Zank's PhD thesis [10].
The paper relies on this external theorem for the well-posedness of the standard weak form, which motivates the need for the T_mu modification.
standard math Integration by parts formula (7): integral_0^T p(t) integral_0^t q(s) ds dt = integral_0^T integral_t^T p(s) ds q(t) dt.
Used in Lemma 3.1 and Lemma 4.1 to rewrite boundary integrals.
standard math Poincare inequality on H^1_{0,}(0,T) ensures the norm ||.||_{H^1_mu} is a norm.
Standard functional analysis background for the weighted norm.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Paving the way to a $\operatorname{T}$-coercive method for the wave equation." pith.science (2026). https://pith.science/paper/XE2MB6QH

@misc{pith2026250902288,
  author       = {Pith},
  title        = {Pith review of: Paving the way to a $\operatornameT$-coercive method for the wave equation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XE2MB6QH}},
  note         = {Machine review of arXiv:2509.02288}
}

abstract

In this paper, we take a first step toward introducing a space-time transformation operator $\operatorname{T}$ that establishes $\operatorname{T}$-coercivity for the weak variational formulation of the wave equation in space and time on bounded Lipschitz domains. As a model problem, we study the ordinary differential equation (ODE) $u'' + \mu u = f$ for $\mu>0$, which is linked to the wave equation via a Fourier expansion in space. For its weak formulation, we introduce a transformation operator $\operatorname{T}_\mu$ that establishes $\operatorname{T}_\mu$-coercivity of the bilinear form yielding an unconditionally stable Galerkin-Bubnov formulation with error estimates independent of $\mu$. The novelty of the current approach is the explicit dependence of the transformation on $\mu$ which, when extended to the framework of partial differential equations, yields an operator acting in both time and space. We pay particular attention to keeping the trial space as a standard Sobolev space, simplifying the error analysis, while only the test space is modified. The theoretical results are complemented by numerical examples.

Figures

Figures reproduced from arXiv: 2509.02288 by the authors.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

10 extracted references · 9 canonical work pages

[8]

Steinbach, M

O. Steinbach, M. Zank: Coercive space-time finite element methods for initial boundary value problems , ETNA 52 (2020) 154–194

work page 2020
[1]

Bignardi, A

P. Bignardi, A. Moiola: A space-time continuous and coercive formulation for the wave equation . arXiv: 2312.07268 (2025)

work page arXiv 2025
[2]

Ferrari, I

M. Ferrari, I. Perugia, E. Zampa: Inf-sup stable space-time discretization of the wave equation based on a first-order-in-time variational formulation. arXiv:2506.05886 (2025)

arXiv 2025
[3]

Fraschini, G

S. Fraschini, G. Loli, A. Moiola, G. Sangalli: An unconditionally stable space–time isogeometric method for the acoustic wave equation, Comput. Math. Appl. vol. 169, pp. 205–222 (2024)

work page 2024
[4]

F¨ uhrer, R

T. F¨ uhrer, R. Gonz´ alez, M. Karkulik: Well-posedness of first-order acoustic wave equations and space-time finite element approximation. IMA J. Numer. Anal., drae104 (2025)

work page 2025
[5]

Stable least-squares space-time boundary element methods for the wave equation

D. Hoonhout, R. L¨ oscher, O. Steinbach, C. Urz´ ua-Torres:Stable least squares space time boundary element methods for the wave equation. arXiv:2312.12547 (2023)

work page Pith review arXiv 2023
[6]

Adaptive least-squares space-time finite element methods

C. K¨ othe, R. L¨ oscher, O. Steinbach: Adaptive least-squares space-time finite element methods. arXiv:2309.14300 (2023)

work page Pith review arXiv 2023
[7]

In: Advanced finite element methods with applications

O.Steinbach, M- Zank: A stabilized space-time finite element method for the wave equation. In: Advanced finite element methods with applications. Lect. Notes Comput. Sci. Eng., vol. 128, pp. 341–370 (2019) 8

work page 2019

Show all 10 references

[9]

Steinbach, M

O. Steinbach, M. Zank: A generalized inf-sup stable variational formulation for the wave equation . J. Math. Anal. Appl. 505 (2022) 125457

2022
[10]

Zank: Inf-sup Stable Space-Time Methods for Time-Dependent Partial Dif- ferential Equations

M. Zank: Inf-sup Stable Space-Time Methods for Time-Dependent Partial Dif- ferential Equations. PhD Thesis, TU Graz (2020). 9

2020

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [8]

Steinbach, M

O. Steinbach, M. Zank: Coercive space-time finite element methods for initial boundary value problems , ETNA 52 (2020) 154–194

work page 2020

[2] [1]

Bignardi, A

P. Bignardi, A. Moiola: A space-time continuous and coercive formulation for the wave equation . arXiv: 2312.07268 (2025)

work page arXiv 2025

[3] [2]

Ferrari, I

M. Ferrari, I. Perugia, E. Zampa: Inf-sup stable space-time discretization of the wave equation based on a first-order-in-time variational formulation. arXiv:2506.05886 (2025)

arXiv 2025

[4] [3]

Fraschini, G

S. Fraschini, G. Loli, A. Moiola, G. Sangalli: An unconditionally stable space–time isogeometric method for the acoustic wave equation, Comput. Math. Appl. vol. 169, pp. 205–222 (2024)

work page 2024

[5] [4]

F¨ uhrer, R

T. F¨ uhrer, R. Gonz´ alez, M. Karkulik: Well-posedness of first-order acoustic wave equations and space-time finite element approximation. IMA J. Numer. Anal., drae104 (2025)

work page 2025

[6] [5]

Stable least-squares space-time boundary element methods for the wave equation

D. Hoonhout, R. L¨ oscher, O. Steinbach, C. Urz´ ua-Torres:Stable least squares space time boundary element methods for the wave equation. arXiv:2312.12547 (2023)

work page Pith review arXiv 2023

[7] [6]

Adaptive least-squares space-time finite element methods

C. K¨ othe, R. L¨ oscher, O. Steinbach: Adaptive least-squares space-time finite element methods. arXiv:2309.14300 (2023)

work page Pith review arXiv 2023

[8] [7]

In: Advanced finite element methods with applications

O.Steinbach, M- Zank: A stabilized space-time finite element method for the wave equation. In: Advanced finite element methods with applications. Lect. Notes Comput. Sci. Eng., vol. 128, pp. 341–370 (2019) 8

work page 2019