A variational approach to the analysis of the continuous space-time FEM for the wave equation

Sergio G\'omez

REVIEW 2 major objections 3 minor 1 cited by

A variational approach to the analysis of the continuous space-time FEM for the wave equation

T0 review · 2 major / 3 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper proves that the continuous space-time finite element method for the Hamiltonian wave equation is unconditionally stable and, for smooth data, converges at quasi-optimal rates without CFL restrictions.

desk verdict The unconditional stability result is real and valuable; the nonhomogeneous Dirichlet error analysis has an unjustified simplification that undermines Theorem 4.9 as written. read the letter →

arxiv 2501.11494 v2 pith:HOIEMEJW submitted 2025-01-20 math.NA cs.NA

Sergio Gómez This is my paper · ORCID

classification math.NAcs.NA MSC 65M6065M1235L04

keywords space-timefiniteelementmethodwaveequationHamiltonianformulationunconditionalstabilityapriorierrorestimatesposterioricontinuousGalerkinintimenonhomogeneousDirichletboundaryconditions

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

This paper proves that the continuous space-time finite element method for the wave equation, a Galerkin scheme approximating position and velocity together in space and time, is unconditionally stable: the discrete energy stays bounded in terms of the data with no restriction linking the mesh size to the time step. The same argument yields quasi-optimal error estimates: for smooth solutions the velocity error is order $h^{p+1}+\tau^{q+1}$, the gradient of the position error is order $h^p+\tau^{q+1}$, and the position error is order $h^{p+1}+\tau^{q+1}$, with constants independent of the final time $T$. The analysis also handles nonhomogeneous Dirichlet boundary data through specially chosen discrete liftings, and supplies a constant-free, reliable a posteriori estimator for the semidiscrete-in-time version. A sympathetic reader would care because Galerkin time discretizations of wave problems have often required CFL conditions or suffered exponentially growing constants from Gronwall arguments, and the paper removes both defects for this scheme.

What carries the argument

The load-bearing mechanism is a nonstandard family of discrete test functions. On each time slab $I_n$, the test function is $\Pi^t_{q-1}(\phi_n \Pi^t_{q-1} v_{h,\tau})$, where $\phi_n(t)=1-\lambda_n(t-t_{n-1})$ with $\lambda_n=1/(2\tau_n)$; inserting this into the velocity equation and using the perturbed first equation converts slab integrals into boundary energies at $t_n$ and $t_{n-1}$ together with two nonnegative Legendre-polynomial terms. Combined with a weak bound at discrete times, an inverse estimate turns the slab-wise $L^2$ control into $C^0([0,T];L^2)$ control of both $v_{h,\tau}$ and $c\nabla u_{h,\tau}$, avoiding Gronwall arguments and any CFL condition. A second ingredient is the projection $P_\tau$ used to define discrete liftings of nonhomogeneous Dirichlet data; its stability in $C^0[0,T]$ is what keeps the boundary treatment from spoiling the convergence rates.

What would settle it

Reproduce the numerical experiment of Section 6.1 with the manufactured solution $u=\cos(\sqrt{2}\pi t)\cos(\pi x)\sin(\pi y)$ and degrees $p=8$, $q=1,\dots,4$; if the temporal-error-dominated rates for $\|u-u_{h,\tau}\|_{C^0L^2}$ or $\|v-v_{h,\tau}\|_{C^0L^2}$ fall below $q+1$ when the boundary data are handled with the paper's specially constructed liftings, the quasi-optimality claim of Theorem 4.9 would be contradicted.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is Theorem 3.5: for the fully discrete space-time continuous FEM (2.2), the discrete solution obeys an energy bound of the form $$\frac{1}{2}\left(\|v_{h,\tau}\|^2_{$C^{0}$([0,T];$L^{2}$)} + \|c\,\nabla u_{h,\tau}\|^2_{$C^{0}$([0,T];$L^{2}$)}\right) \lesssim \frac{1}{2}\left(\|v_0\|^2_{$L^{2}$}+\|c\,\nabla u_0\|^2_{$L^{2}$}\right)+\|f\|^2_{$L^{1}$(0,T;$L^{2}$)}+\|c\,\nabla \Upsilon\|^2_{$L^{1}$(0,T;$L^{2}$)},$$ with no mesh-size or time-step restriction and with a hidden constant depending only on the time approximation degree $q$. From this, Theorem 4.9 derives the quasi-optimal rates $\|v-v_{h,\tau}\|_{C^0L^2}\lesssim h^{p+1}+\tau^{q+1}$, $\|c\nabla(u-u_{h,\tau})\|_{C^0L^2}\lesssim h^p+\tau^{q+1}$, and $\|u-u_{h,\tau}\|_{C^0L^2}\lesssim h^{p+1}+\tau^{q+1}$ for sufficiently smooth solutions, with constants independent of $T$. The author frames the contribution as replacing the earlier CFL condition of the original method and the exponential-in-$T$ constants of previous analyses, while requiring only $f\in L^1(0,T;L^2)$ for stability.

Load-bearing premise

The quasi-optimal convergence rates require the exact solution and the data to be smooth enough, roughly two time derivatives of the solution, enough spatial derivatives for the finite-element approximations, and boundary data agreeing with the initial data at the initial time, so that the interpolation and projection errors used in the proof are controlled; if the solution is less regular, only stability, not a rate, is guaranteed.

Editorial extensions

If this is right

Users of this space-time FEM can choose time steps independently of the spatial mesh; the scheme remains stable for arbitrarily small or varying $\tau_n$, with a stability constant depending only on the time polynomial degree $q$.
For smooth data, the scheme achieves order $h^{p+1}+\tau^{q+1}$ for velocity and position in $C^0L^2$ and order $h^p+\tau^{q+1}$ for the energy gradient, with error constants independent of $T$, so long-time simulations do not inherit the exponential drift typical of Gronwall-based estimates.
Nonhomogeneous Dirichlet data require the discrete liftings built with the projection $P_\tau$; the paper's experiments show that naive interpolation of the boundary datum degrades the observed time convergence.
The postprocessed approximation $u^\star_{h,\tau}(t)=u_{h,\tau}(0)+\int_0^t v_{h,\tau}\,ds$ superconverges in time as $O(\tau^{q+2})$ when $q>1$.
The semidiscrete-in-time scheme admits a reliable a posteriori error bound in $C^0([0,T];L^2)$ with no unknown constant, providing a computable target for adaptive time-stepping.

Reading between the lines

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

One consequence the paper leaves implicit is that the stability argument is largely agnostic to the spatial operator, so the same weighted test functions should yield CFL-free $C^0$ energy bounds for any self-adjoint second-order elliptic operator in place of $-\nabla\cdot(c^2\nabla)$; replacing the operator and re-running the proof would be a direct test of this extension.
The paper's own experiments with singular solutions of the form $u(t)=t^\alpha$ show convergence at $O(\tau^\alpha)$ instead of quasi-optimal order; this suggests that an adaptive time mesh driven by the a posteriori estimator could restore near-optimal rates for low-regularity data, an extension the paper does not attempt.
Because the energy bound is independent of $T$ and free of CFL conditions, the method could be combined with aggressive time-step adaptivity over very long horizons without the usual exponential blow-up of stability constants; verifying this would only require the paper's estimator and a long-time numerical test.

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

Short version: Theorem 3.5 is a genuine advance and the proof is worth taking seriously. The nonhomogeneous Dirichlet error analysis, however, has a gap in Lemma 4.6 that a referee should catch: the initial error for the discrete velocity is simplified by dropping a term that is not generally small.

What is actually new: the weighted-test-function argument gives unconditional stability for the French–Peterson cG method in a C0([0,T];X)-type energy norm, with constants independent of T. That removes the CFL restriction from [26] and avoids the exponential T-constants in [8,9,33,46]. The proof is detailed, and the Legendre identities for the nonnegative boundary terms check out. The homogeneous-Dirichlet convergence analysis follows cleanly from Theorem 3.5. The constant-free a posteriori estimator for the semidiscrete-in-time scheme is a useful extra.

The soft spot: the treatment of nonhomogeneous Dirichlet boundary data. In Lemma 4.6, the discrete initial velocity error is reduced to Π_{hτ}e_v(0) = -◦Π_h(Id - ◦R_h)w, with w = v0 - ∂_t u_{gD}|_{Σ0}. To get this, the term R_h(∂_t u_{gD}|_{Σ0} - I^∂_h v0|∂Ω) is silently dropped. The paper provides no proof that this term vanishes, and in general it does not. For u(x,t) = x + t on (0,1)×(0,T), v0 ≡ 1, and the natural lifting u_{gD} = x + t, we have w = 0 but v0,h = I^∂_h 1, which is not close to 1 in L2 for standard choices of the boundary interpolant. The error at t = 0 alone is O(1). The numerical experiments in Section 6.1 avoid the problem because they take v0 = 0 and a lifting whose time derivative at t = 0 is zero, so the missing term vanishes. The claim that Theorem 4.9 gives quasi-optimal rates for nonhomogeneous Dirichlet data is therefore not established as written. This is fixable—choose v0,h differently, impose a stronger property on I^∂_h, or estimate the dropped term under extra regularity—but it needs to be fixed.

Who this is for: people working on space-time continuous FEM for wave equations, particularly the French–Peterson method and long-time integration. The stability result alone is worth publishing, and the paper deserves a serious referee; the nonhomogeneous part should be revised after the gap is addressed.

Referee Report

2 major / 3 minor

Summary. The paper analyzes the continuous space-time finite element method of French and Peterson for the Hamiltonian formulation of the wave equation. The main results are: (i) an unconditional stability estimate in C0([0,T];X)-type energy norms for the fully discrete scheme, with constants independent of T; (ii) quasi-optimal a priori error estimates in C0([0,T];L2(Ω)) and C0([0,T];H1(Ω))-type norms, including a treatment of nonhomogeneous Dirichlet boundary conditions; (iii) a constant-free reliable a posteriori error estimate for the semidiscrete-in-time formulation; and (iv) numerical experiments in 2+1 dimensions illustrating h-, τ-, and (p,q)-convergence, plus tests of the a posteriori estimator for smooth and singular solutions. The stability analysis uses nonstandard test functions involving Legendre polynomials, and the error analysis uses the projections Pτ, Rh, and discrete liftings of the boundary data.

Significance. If the results hold, this is a substantial contribution to the analysis of space-time continuous Galerkin methods for the wave equation. The unconditional stability estimate in Theorem 3.5, with no CFL condition and no Grönwall-type growth in T, significantly improves earlier analyses, and the explicit treatment of nonhomogeneous Dirichlet data addresses a known source of suboptimal convergence. The a posteriori estimator with explicit constants is also valuable. The proofs are detailed and largely self-contained, and the numerical experiments are carefully matched to the theoretical rates. However, the convergence claims for the nonhomogeneous case with nonzero initial velocity rest on an unproven—and generally false—simplification of the initial error, so the main convergence theorem is not correct as stated for that case.

major comments (2)

[§4.2 and Lemma 4.6] The simplification of the initial error in Lemma 4.6 is not justified and is false for nonzero v0 with nonhomogeneous Dirichlet data. With v0,h defined in (4.9) and ṽ0 = v0 − ∂_t u_{gD}|_{Σ0}, the proof asserts Π_{hτ}e_v(0) = R_h v0 − v0,h = ◦R_h ṽ0 − ◦Π_h ṽ0. This requires R_h(∂_t u_{gD}|_{Σ0}) = I^∂_h v0|_{∂Ω}, i.e., the Ritz projection of the time derivative of the lifting equals the boundary interpolant; no such property is stated or proved, and it fails generically. For example, on Ω=(0,1), take u(x,t)=x+t+t sin(πx), so v0(x)=1+sin(πx), gD(0,t)=t, gD(1,t)=1+t, and choose the natural lifting u_{gD}=u. Then ṽ0=0, and (4.9) gives v0,h = I^∂_h v0|_{∂Ω}, which with the natural boundary extension is the constant 1; hence ‖v0−v0,h‖_{L2(0,1)} = ‖sin(πx)‖_{L2} is O(1). Lemma 4.6's formula predicts Π_{hτ}e_v(0)=0, and Theorem 4.9 predicts O(h^{ℓ+1}+τ^{m+1}) for ‖v−v_{h,τ}‖_{C0L2}, which is contradicted at t=0. The numerical experiments in §6.1 use v0=0 and therefore do not exercise this case. The flaw is fixable, for instance by setting v0,h = R_h v0 (which is compatible with vD_{h,τ}(0)=I^∂_h v0|_{∂Ω}) or by carrying the missing term ◦R_h(∂_t u_{gD}|_{Σ0} − I^∂_h v0|_{∂Ω}) through the error analysis, but as written Theorem 4.9's nonhomogeneous-Dirichlet convergence claim is not correct.
[§4.2, Eq. (4.8)-(4.9)] The discrete lifting operator I^∂_h is not uniquely defined by the phrase 'interpolant of the restriction to ∂Ω', and the validity of the method depends on the unspecified interior extension. In (4.8), uD_{h,τ}=Pτ I^∂_h gD and vD_{h,τ}=Pτ I^∂_h ∂_t gD are used as discrete liftings, and in (4.9) the same operator is added to ◦Π_h(v0−∂_t u_{gD}|_{Σ0}). But a map C0(∂Ω)→V^p_h is not determined by boundary values alone; different extensions produce different discrete solutions and different initial errors, and the error identity in Lemma 4.6 silently assumes a particular extension that equals the Ritz projection of the continuous lifting's time derivative. This definitional gap should be closed by specifying the construction (e.g., a standard extension of the boundary interpolant, or a Ritz-type lifting) and the subsequent analysis should be checked against that choice.

minor comments (3)

[§5, Eq. (5.13)] The final equality in (5.13) is missing a square on the norm; it should read ‖u−u⋆_τ‖²_{C0([0,T];L2(Ω))} to match the preceding line and the use in (5.15).
[Theorem 5.1 statement] The statement of Theorem 5.1 contains garbled summation notation ('m=1∑ n=1' and 'C•(q) m=1∑ n=1'), which should be corrected to explicit sums such as ∑_{n=1}^{m-1} and a properly typeset definition of C•(q).
[Lemma 4.6] The notation 'v0 := v0 − ∂_t u_{gD}|_{Σ0}' in Lemma 4.6 reuses the symbol v0 for two different objects; the reader must track that the new v0 has a tilde or another name in the rest of the proof.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the stability and error estimates are derived from variational identities and standard external projection estimates; the only self-citation is non-load-bearing.

full rationale

Score 0. The central claims—Theorem 3.5 (unconditional C0 energy stability) and Theorem 4.9 (quasi-optimal a priori rates)—are derived, not assumed. The stability proof uses test functions constructed inside the proof from the discrete solution and standard L2/Ritz projections; no fitted parameter is calibrated to the target quantities. The a priori analysis decomposes the error into projection errors and discrete errors, bounding each with textbook results (Lemma 3.1 from [18], Lemma 3.2 from [13], Lemma 4.4 from [19], Lemmas 4.2–4.3 from [20] and [18]); none of these assume the paper's conclusions. The a posteriori bound (Theorem 5.1) follows from residual identities and known Poincaré/Nagy constants, again with no input–output identification. The only self-citation is [29] (the author's prior work with V. Nikolić), cited as a source of the weight-function technique; the technique is fully re-proved in Section 3, so the citation is not load-bearing. The skeptic's objection to Lemma 4.6 (the simplification of the initial discrete error for nonhomogeneous Dirichlet data) identifies a potential omitted proof of a Ritz-projection identity, which is a correctness risk rather than circularity: nothing in that step equates the predicted rates with their inputs by construction. Accordingly, no circular step is exhibited.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

No free parameters: the method contains no fitted constants; the hidden constants in the estimates depend only on polynomial degrees, shape regularity, and the ratio c_star/c^star. No new physical or mathematical entities are postulated. The axioms listed are the regularity and standard lemmas on which the proofs rest.

assumptions (7)

domain assumption Elliptic regularity of Omega: if phi in H^1_0(Omega) and Delta phi in L^2(Omega), then phi in H^2(Omega) (Assumption 1).
Used in Lemma 4.4 for L^2-error estimates of the Ritz projection, which underlies the h^{ell+1} rates in Theorem 4.9.
domain assumption Data and solution regularity Assumption 2: f in L^1(0,T;L^2), gD in H^2(0,T;H^s(boundary)), u0,v0 in H^r, c in C^0 cap W^{1,infinity} with positive bounds, u in H^2(0,T;H^r), and div(c^2 grad u) in H^1(0,T;L^2).
Needed for the a priori error estimates and for the discrete liftings via P_tau I^boundary_h; if these fail, the quoted rates are not guaranteed.
standard math Well-posedness of the weak solution: existence and uniqueness with u in C^0([0,T];H^1), partial_t u in C^0([0,T];L^2), and partial_tt u in L^2(0,T;H^{-1}).
Invoked at the end of Section 1, citing [34, Thm. 2.1 in Part I].
standard math Stability of the time L^2 projection Pi_t^r in L^p and polynomial inverse estimates (Lemmas 3.1 and 3.2).
Quoted from [18, Thm. 18.16] and [13, Thm. 4.5.11]; used throughout the stability proof; constants depend only on q.
standard math Approximation properties of the Ritz projection R_h (Lemma 4.4) and the time projection P_tau (Lemma 4.3).
Quoted from [19, Thms. 33.2-33.3] and [20, Rem. 70.10]; used to convert discrete-error bounds into h,tau rates.
domain assumption For the a posteriori section: gD=0, c constant positive, u0 in H^2 cap H^1_0, v0 in H^1_0, f in H^1(0,T;L^2), and Omega satisfies Assumption 1.
Stated in (5.1); the semidiscrete a posteriori bound of Theorem 5.1 is proven only under these assumptions.
domain assumption Initial-boundary compatibility: gD(.,0)=u0 on the boundary and partial_t gD(.,0)=v0 on the boundary.
Required by the conforming space V^{p,q}_{h,tau} and by the construction of discrete liftings in (4.8)-(4.9).

how reviews work

0 comments

Cite this review

Pith. "Pith review of A variational approach to the analysis of the continuous space-time FEM for the wave equation." pith.science (2026). https://pith.science/paper/HOIEMEJW

@misc{pith2026250111494,
  author       = {Pith},
  title        = {Pith review of: A variational approach to the analysis of the continuous space-time FEM for the wave equation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HOIEMEJW}},
  note         = {Machine review of arXiv:2501.11494}
}

abstract

We present a stability and convergence analysis of the space-time continuous finite element method for the Hamiltonian formulation of the wave equation. More precisely, we prove a continuous dependence of the discrete solution on the data in a $C^0([0, T]; X)$-type energy norm, which does not require any restriction on the meshsize or the time steps. Such stability estimates are then used to derive a priori error estimates with quasi-optimal convergence rates, where a suitable treatment of possible nonhomogeneous Dirichlet boundary conditions is pivotal to avoid loss of accuracy. Moreover, based on the properties of a postprocessed approximation, we derive a constant-free, reliable a posteriori error estimate in the $C^0([0, T]; L^2(\Omega))$ norm for the semidiscrete-in-time formulation. Several numerical experiments are presented to validate our theoretical findings.

Figures

Figures reproduced from arXiv: 2501.11494 by the authors.

**Figure 1.** h-convergence (in log-log scale) of the errors in (6.1) for Method I corresponding to the problem with exact solution (u, v) in (6.2). The numbers in the yellow boxes are the empirical convergence rates [PITH_FULL_IMAGE:figures/full_fig_p022_1.png] view at source ↗

**Figure 2.** Plot of the difference of the discrete solutions uh,τ (left panel) and vh,τ (right panel) at t1 = 1/8 corresponding to Method I and Method II with p = q = 3 for the problem with exact solution (u, v) in (6.2). We fix a space–time mesh with h ≈ 1.77 × 10−1 and τ = 0.25. In [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗

**Figure 3.** τ-convergence (in log-log scale) of the errors in (6.1) for Method I (left panel) and Method II (right panel) corresponding to the problem with exact solution (u, v) in (6.2). The results for the approximation of the Dirichlet boundary conditions according to Section 4.2 are shown in solid lines, whereas those corresponding to Lagrange interpolation are shown in dashed lines. so, we define the effectivity index as E… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: (p, q)-convergence (in semilogy scale) of the errors in (6.1) for Method I corresponding to the problem with exact solution (u, v) in (6.2). Let QT be given by (−1, 1)2 × (0, 1). We consider the (2 + 1)-dimensional problem (1.2) with initial conditions, Dirichlet bound…

**Figure 5.** Figure 5: Left panel: comparison of the error keukC0([0,T ];L2(Ω)) (solid lines) and the estimator η (dashed lines) for the problem with smooth exact solution (u, v) in (6.4) and ψ(t) = cos(4t). Right panel: corresponding effectivity indices in (6.3). Singular solution. Similarl…

**Figure 6.** Figure 6: Left panel: comparison of the error keukC0([0,T ];L2(Ω)) (solid lines) and the estimator η (dashed lines) for the problem with singular exact solution (u, v) in (6.4) and ψ(t) = t 2.25 (first row), ψ(t) = t 2.5 (second row). Right panel: corresponding effectivity indic…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Inf-sup stable space-time discretization of the wave equation based on a first-order-in-time variational formulation
math.NA 2025-06 conditional novelty 7.0 of 10

A conforming space-time discretization of the wave equation using exponential time weights is proven unconditionally stable and quasi-optimal for general tensor product spaces.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages · cited by 1 Pith paper

[26]

D. A. French and T. E. Peterson. A space-time ﬁnite element method for the second order wave equation. Centre for Mathematics and its Applications, Australian National U niversity, 1991

work page 1991
[1]

Acosta and R

G. Acosta and R. G. Dur´ an. An optimal Poincar´ e inequality inL1 for convex domains. Proc. Amer. Math. Soc. , 132(1):195–202, 2004. 25

work page 2004
[2]

S. Adjerid. A posteriori ﬁnite element error estimation for seco nd-order hyperbolic problems. Com- put. Methods Appl. Mech. Engrg. , 191(41-42):4699–4719, 2002

work page 2002
[3]

Anselmann and M

M. Anselmann and M. Bause. Numerical study of Galerkin-collocat ion approximation in time for the wave equation. In Mathematics of wave phenomena. Selected papers based on the presentations at the conference, Karlsruhe, Germany, July 23–27, 2018 , pages 15–36. Cham: Birkh¨ auser, 2020

work page 2018
[4]

Anselmann, M

M. Anselmann, M. Bause, S. Becher, and G. Matthies. Galerkin-c ollocation approximation in time for the wave equation and its post-processing. ESAIM Math. Model. Numer. Anal. , 54(6):2099–2123, 2020

work page 2020
[5]

P. F. Antonietti, I. Mazzieri, N. Dal Santo, and A. Quarteroni. A high-order discontinuous Galerkin approximation to ordinary diﬀerential equations with applications to elastodynamics. IMA J. Numer. Anal., 38(4):1709–1734, 2018

work page 2018
[6]

P. F. Antonietti, I. Mazzieri, and F. Migliorini. A space-time discont inuous Galerkin method for the elastic wave equation. J. Comput. Phys. , 419:109685, 26, 2020

work page 2020
[7]

A. K. Aziz and P. Monk. Continuous ﬁnite elements in space and time for the heat equation. Math. Comp., 52(186):255–274, 1989

work page 1989

Show all 46 references

[8]

Bales and I

L. Bales and I. Lasiecka. Continuous ﬁnite elements in space and t ime for the nonhomogeneous wave equation. Comput. Math. Appl. , 27(3):91–102, 1994

1994
[9]

Bales and I

L. Bales and I. Lasiecka. Negative norm estimates for fully discre te ﬁnite element approximations to the wave equation with nonhomogeneous L2 Dirichlet boundary data. Math. Comp., 64(209):89–115, 1995

1995
[10]

Bangerth, M

W. Bangerth, M. Geiger, and R. Rannacher. Adaptive Galerkin ﬁ nite element methods for the wave equation. Comput. Methods Appl. Math. , 10(1):3–48, 2010

2010
[11]

Bause, U

M. Bause, U. K¨ ocher, F. A. Radu, and F. Schieweck. Post-pr ocessed Galerkin approximation of improved order for wave equations. Math. Comp. , 89(322):595–627, 2020

2020
[12]

Bernardi and E

C. Bernardi and E. S¨ uli. Time and space adaptivity for the seco nd-order wave equation. Math. Models Methods Appl. Sci. , 15(2):199–225, 2005

2005
[13]

S. C. Brenner and L. R. Scott. The mathematical theory of ﬁnite element methods , volume 15 of Texts in Applied Mathematics . Springer, New York, third edition, 2008

2008
[14]

H. Brezis. Functional analysis, Sobolev spaces and partial diﬀerenti al equations . Universitext. Springer, New York, 2011

2011
[15]

Cangiani, Z

A. Cangiani, Z. Dong, and E. H. Georgoulis. hp-version space-time discontinuous Galerkin methods for parabolic problems on prismatic meshes. SIAM J. Sci. Comput. , 39(4):A1251–A1279, 2017

2017
[16]

Chaumont-Frelet and A

T. Chaumont-Frelet and A. Ern. Damped energy-norm a poste riori error estimates for fully discrete approximations of the wave equation using C2-reconstructions for the fully discrete wave equa- tion with the leapfrog scheme. Accepted for publication in ESAIM Mat h. Model. Nume...
[17]

Z. Dong, L. Mascotto, and Z. Wang. A priori and a posteriori e rror estimates of a DG-CG method for the wave equation in second order formulation. arXiv:2411.03264, 2024

2024 arXiv
[18]

Ern and J.-L

A. Ern and J.-L. Guermond. Finite elements I—Approximation and interpolation , volume 72 of Texts in Applied Mathematics . Springer, Cham, 2021

2021
[19]

Ern and J.-L

A. Ern and J.-L. Guermond. Finite elements II—Galerkin approximation, elliptic and m ixed PDEs, volume 73 of Texts in Applied Mathematics . Springer, Cham, 2021

2021
[20]

Ern and J.-L

A. Ern and J.-L. Guermond. Finite elements III—ﬁrst-order and time-dependent PDEs , volume 74 of Texts in Applied Mathematics . Springer, Cham, 2021

2021
[21]

L. C. Evans. Partial diﬀerential equations , volume 19 of Graduate Studies in Mathematics . American Mathematical Society, Providence, RI, second edition, 2010. 26

2010
[22]

Ferrari and S

M. Ferrari and S. Fraschini. Stability of conforming space-time isogeometric methods for the wave equation. Math. Comp. , 2025. doi:10.1090/mcom/4062

2025 doi
[23]

Ferrari, S

M. Ferrari, S. Fraschini, G. Loli, and I. Perugia. Unconditionally s table space-time isogeometric discretization for the wave equation in Hamiltonian formulation. Acce pted for publication in ESAIM Math. Model. Numer. Anal., 2025

2025
[24]

Fraschini, G

S. Fraschini, G. Loli, A. Moiola, and G. Sangalli. An unconditionally st able space-time isogeometric method for the acoustic wave equation. Comput. Math. Appl. , 169:205–222, 2024

2024
[25]

D. A. French. A space-time ﬁnite element method for the wave e quation. Comput. Methods Appl. Mech. Engrg., 107(1-2):145–157, 1993

1993
[27]

D. A. French and T. E. Peterson. A continuous space-time ﬁnit e element method for the wave equation. Math. Comp. , 65(214):491–506, 1996

1996
[28]

E. H. Georgoulis, O. Lakkis, C. G. Makridakis, and J. M. Virtanen . A posteriori error estimates for leap-frog and cosine methods for second order evolution prob lems. SIAM J. Numer. Anal. , 54(1):120–136, 2016

2016
[29]

G´ omez and V

S. G´ omez and V. Nikoli´ c. Space–time DG–CG ﬁnite element meth od for the Westervelt equation. Accepted for publication in IMA J. Numer. Anal., 2025

2025
[30]

M. J. Grote, O. Lakkis, and C. Santos. A posteriori error est imates for the wave equation with mesh change in the leapfrog method. arXiv:2411.16933, 2024

2024 arXiv
[31]

G. M. Hulbert and T. J. R. Hughes. Space-time ﬁnite element met hods for second-order hyperbolic equations. Comput. Methods Appl. Mech. Engrg. , 84(3):327–348, 1990

1990
[32]

C. Johnson. Discontinuous Galerkin ﬁnite element methods for s econd order hyperbolic problems. Comput. Methods Appl. Mech. Engrg. , 107(1-2):117–129, 1993

1993
[33]

Karakashian and C

O. Karakashian and C. Makridakis. Convergence of a continuou s Galerkin method with mesh mod- iﬁcation for nonlinear wave equations. Math. Comp. , 74(249):85–102, 2005

2005
[34]

Lasiecka and R

I. Lasiecka and R. Triggiani. Recent advances in regularity of se cond-order hyperbolic mixed prob- lems, and applications. In Dynamics reported. Expositions in dynamical systems. New s eries. Volume 3, pages 104–162. Berlin: Springer-Verlag, 1994

1994
[35]

Makridakis and R

C. Makridakis and R. H. Nochetto. A posteriori error analysis f or higher order dissipative methods for evolution problems. Numer. Math. , 104(4):489–514, 2006

2006
[36]

L. E. Payne and H. F. Weinberger. An optimal Poincar´ e inequalit y for convex domains. Arch. Rational Mech. Anal. , 5:286–292, 1960

1960
[37]

Sch¨ otzau and C

D. Sch¨ otzau and C. Schwab. Time discretization of parabolic pr oblems by the hp-version of the discontinuous Galerkin ﬁnite element method. SIAM J. Numer. Anal. , 38(3):837–875, 2000

2000
[38]

Ch. Schwab. p- and hp-ﬁnite element methods . Numerical Mathematics and Scientiﬁc Computation. The Clarendon Press, Oxford University Press, New York, 1998. T heory and applications in solid and ﬂuid mechanics

1998
[39]

L. R. Scott and S. Zhang. Finite element interpolation of nonsmo oth functions satisfying boundary conditions. Math. Comp. , 54(190):483–493, 1990

1990
[40]

Strauss and L

W. Strauss and L. Vazquez. Numerical solution of a nonlinear Kle in-Gordon equation. J. Comput. Phys., 28(2):271–278, 1978

1978
[41]

Tantardini and A

F. Tantardini and A. Veeser. The L2-projection and quasi-optimality of Galerkin methods for parabolic equations. SIAM J. Numer. Anal. , 54(1):317–340, 2016

2016
[42]

B. v. Sz. Nagy. ¨Uber Integralungleichungen zwischen einer Funktion und ihrer Ableit ung. Acta Univ. Szeged. Sect. Sci. Math. , 10:64–74, 1941. 27

1941
[43]

N. J. Walkington. Combined DG-CG time stepping for wave equatio ns. SIAM J. Numer. Anal. , 52(3):1398–1417, 2014

2014
[44]

M. Zank. Inf–sup stable space–time methods for time–depend ent partial diﬀerential equations. volume 36 of Monographic Series TU Graz: Computation in Engineering and Science. Verlag der TU Graz, Graz , 2019

2019
[45]

M. Zank. The Newmark method and a space-time FEM for the sec ond-order wave equation. In Numerical mathematics and advanced applications—ENUMATH 2019, volume 139 of Lect. Notes Comput. Sci. Eng. , pages 1225–1233. Springer, Cham, 2021

2019
[46]

Zhao and H

Z. Zhao and H. Li. Convergence of a space-time continuous Gale rkin method for the wave equation. J. Inequal. Appl. , pages Paper No. 280, 18, 2016. 28

2016

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [26]

D. A. French and T. E. Peterson. A space-time ﬁnite element method for the second order wave equation. Centre for Mathematics and its Applications, Australian National U niversity, 1991

work page 1991

[2] [1]

Acosta and R

G. Acosta and R. G. Dur´ an. An optimal Poincar´ e inequality inL1 for convex domains. Proc. Amer. Math. Soc. , 132(1):195–202, 2004. 25

work page 2004

[3] [2]

S. Adjerid. A posteriori ﬁnite element error estimation for seco nd-order hyperbolic problems. Com- put. Methods Appl. Mech. Engrg. , 191(41-42):4699–4719, 2002

work page 2002

[4] [3]

Anselmann and M

M. Anselmann and M. Bause. Numerical study of Galerkin-collocat ion approximation in time for the wave equation. In Mathematics of wave phenomena. Selected papers based on the presentations at the conference, Karlsruhe, Germany, July 23–27, 2018 , pages 15–36. Cham: Birkh¨ auser, 2020

work page 2018

[5] [4]

Anselmann, M

M. Anselmann, M. Bause, S. Becher, and G. Matthies. Galerkin-c ollocation approximation in time for the wave equation and its post-processing. ESAIM Math. Model. Numer. Anal. , 54(6):2099–2123, 2020

work page 2020

[6] [5]

P. F. Antonietti, I. Mazzieri, N. Dal Santo, and A. Quarteroni. A high-order discontinuous Galerkin approximation to ordinary diﬀerential equations with applications to elastodynamics. IMA J. Numer. Anal., 38(4):1709–1734, 2018

work page 2018

[7] [6]

P. F. Antonietti, I. Mazzieri, and F. Migliorini. A space-time discont inuous Galerkin method for the elastic wave equation. J. Comput. Phys. , 419:109685, 26, 2020

work page 2020

[8] [7]

A. K. Aziz and P. Monk. Continuous ﬁnite elements in space and time for the heat equation. Math. Comp., 52(186):255–274, 1989

work page 1989