Pith. sign in

REVIEW 2 major objections 4 minor 19 references

Analysis of a time-stepping discontinuous Galerkin method for fractional diffusion-wave equation with nonsmooth data

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A low-order time-stepping discontinuous Galerkin method is shown to converge at near-optimal rates for fractional diffusion-wave equations even when the data are nonsmooth.

desk verdict First nonsmooth-data convergence rates for the low-order time-stepping DG method for fractional diffusion-wave equations; the analysis is careful, but the key rates rest on a regularity theorem imported from a companion paper, so the result is conditional. read the letter →

arxiv 1908.09189 v1 pith:KKCM4J5X submitted 2019-08-24 math.NA cs.NA

Binjie Li , Tao Wang , Xiaoping Xie This is my paper · ORCID
classification math.NAcs.NA MSC 65M6065M1265M1535R11
keywords fractionaldiffusion-waveequationdiscontinuousGalerkinmethodtimesteppingnonsmoothdatadiscreteLaplacetransformconvergenceanalysiserrorestimatesordinary
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 analyzes the low-order time-stepping discontinuous Galerkin method for the fractional diffusion-wave equation $u' - \Delta D^{-\alpha}_{0+} u = f$, using piecewise constant functions in time and continuous piecewise linear functions in space. It establishes convergence with nearly optimal rates under nonsmooth data: for $f=0$ and $u_0 \in L^2(\Omega)$ the error at time $t_j$ is $O(h^2 t_j^{-\alpha-1} + \tau t_j^{-1})$, for $u_0=0$ and $f \in L^2(0,T;\dot H^{\alpha/(\alpha+1)}(\Omega))$ it is $O(h + \sqrt{\ln(1/h)}\,\tau^{1/2})$, and for $u_0=0$ and $f \in {}_0H^{\alpha+1/2}(0,T;L^2(\Omega))$ it is $O(\ln(T/\tau)(\sqrt{\ln(1/h)}\,h^2 + \tau))$. These are the first nonsmooth-data convergence results for this scheme, closing a gap left by earlier smooth-data analyses and showing that uniform time grids suffice for first-order temporal accuracy up to a logarithmic factor.

What carries the argument

The proof is carried by the discrete Laplace transform of the numerical scheme together with an auxiliary function $$\psi(z) = \frac{e^z - 1}{\Gamma(2+\$\alpha$)} \sum_{k=1}^\infty $k^{{1+\alpha}}$ $e^{{-kz}}$,$$ which encodes the effect of the piecewise-constant time discretization on the fractional integral. The argument shows that $1 + \mu\psi(z)$ with $\mu = \lambda\tau^{1+\alpha}$ stays away from zero in a sector of the complex plane (Lemma 3.1) and satisfies pointwise lower bounds (Lemma 3.2), yielding $O(k^{-1})$ estimates for the scalar fractional ordinary equations to which the full problem is reduced (Theorems 3.1–3.3). Spatial and temporal interpolation estimates then transfer these scalar bounds to the full discrete solution.

What would settle it

Take $u_0=0$, $f(t,x) = t^{\alpha - 1/2}\sin(\pi x)$ on $\Omega=(0,1)$, $T=1$, which lies in ${}_0H^{\alpha+1/2}(0,T;L^2(\Omega))$. Compute the scheme with $h$ small and $\tau = 2^{-n}$ for $n=8,\dots,14$; if $\|u-U\|_{L^\infty(0,T;L^2)}/(\tau \ln(1/\tau))$ grows without bound as $n$ increases while the source norm stays finite, the claimed $O(\ln(T/\tau)\tau)$ rate fails.

Watch

Extended reading notes

Core claim

The central claim is that the low-order time-stepping DG scheme achieves nearly optimal convergence uniformly in time under nonsmooth data. The main estimate, stated as Theorem 4.4, is $$\|u - U\|_{L^\infty(0,T;$L^{2}$(\$\Omega$))} \lesssim \ln(T/\tau)\bigl(\sqrt{\ln(1/h)}\,$h^{2}$ + \tau\bigr)\, \|f\|_{{}$_0H^{{\alpha+1/2}}$(0,T;$L^{2}$(\$\Omega$))}$$ for $u_0=0$, and Theorem 4.1 gives the companion bound $\|u(t_j) - U_j\|_{L^2(\Omega)} \lesssim (h^2 t_j^{-\alpha-1} + \tau t_j^{-1})\|u_0\|_{L^2(\Omega)}$. Because the solution of this fractional problem generally develops a singularity at $t=0$, the achievement is that no graded temporal meshes and no extra regularity of the initial value are required for nearly optimal rates.

Load-bearing premise

The proof's rates rest on the weak-solution regularity bounds of Theorem 2.2, which the paper borrows from a companion analysis with only a 'trivial modification of the proof'; if those bounds fail, the error estimates have no foundation.

Editorial extensions

If this is right

  • Uniform time grids suffice: the temporal singularity at $t=0$ caused by nonsmooth data does not force graded meshes for the $O(\tau)$ rate up to a logarithmic factor.
  • For a constant-in-time source $f(t)=v \in L^2(\Omega)$, the error is $O(t^{-\alpha} h^2 + \tau)$, so the spatial error has a $t^{-\alpha}$ singularity while the temporal rate remains first order.
  • When $f \in H^{\alpha+1/2}$ with $f(0) \ne 0$, combining Theorems 4.2 and 4.4 gives a bound containing both a $t^{-\alpha}$ spatial term and a logarithmic temporal term (Remark 4.3).
  • The analyzed scheme coincides with the first discretization in the earlier work of McLean, Thomée, and Wahlbin [10], so the results close the nonsmooth-data gap for that method.

Reading between the lines

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

  • The same discrete-Laplace-transform machinery likely extends to semilinear versions of the fractional diffusion-wave equation, provided a matching regularity theorem is available.
  • The logarithmic factor $\ln(T/\tau)$ in Theorem 4.4 may be an artifact of the duality argument and the summation of jumps; numerical experiments suggest that on data with mild smoothness the observed rate can be closer to $O(\tau)$ than $O(\tau \ln(1/\tau))$.
  • The sharp dependence of the auxiliary bound on $\mu = \lambda\tau^{1+\alpha}$ indicates that the stability constant may degrade as $\alpha$ approaches 0 or 1; practitioners using extreme $\alpha$ values might need to check whether the predicted rates are still visible at practical resolutions.
  • Because the load-bearing regularity estimates are imported from a companion paper with a 'trivial modification', a careful independent check of Theorem 2.2 for the full parameter range would be prudent before relying on the advertised rates.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper analyzes the low-order time-stepping discontinuous Galerkin method for the fractional diffusion-wave equation u' - Δ D^{-α}_{0+}u = f, 0<α<1, with piecewise constant time and continuous piecewise linear space. The main results are four error estimates: Theorem 4.1 for nonsmooth initial data u0∈L²(Ω) and f=0; Theorem 4.2 for u0=0 and time-independent f∈L²(Ω); Theorem 4.3 for u0=0 and f∈L²(0,T; ˙H^{α/(α+1)}(Ω)), giving L∞(0,T;L²(Ω)) error O(h+√ln(1/h) τ^{1/2}); and Theorem 4.4 for u0=0 and f∈0H^{α+1/2}(0,T;L²(Ω)), giving the nearly optimal rate O(ln(T/τ)(√ln(1/h) h²+τ)). The proofs combine discrete Laplace transform arguments for two scalar fractional ODEs (Section 3), energy estimates and interpolation inequalities (Section 4), with the regularity of the weak solution imported from the companion paper [6]. Four numerical experiments are reported that are consistent with the claimed rates.

Significance. If the results are correct, this is a valuable contribution: it extends the convergence theory of the low-order DG method for fractional diffusion-wave problems to nonsmooth data, a regime that the earlier analyses in [10,14] did not cover. The Laplace-transform estimates in Section 3 are detailed and appear internally consistent, and the error estimates are stated with explicit structural constants rather than fitted parameters. The claimed rates are falsifiable and are tested numerically. The main weakness is that the central regularity Theorem 2.2 is not proved in the manuscript but imported from an unpublished companion paper by the same group, so the advertised error bounds are conditional on the validity of that transfer.

major comments (2)
  1. [§2.4, Theorem 2.2] Theorem 2.2 is the load-bearing regularity input for Theorems 4.3 and 4.4, yet it is introduced with the sentence 'by a trivial modification of the proof of [6, Theorems 4.2]' and no proof is given. Specifically, estimate (9) with γ=0 provides the 0H¹(0,T; ˙H^{α/(α+1)}) and 0H^{-α}(0,T; ˙H^{2+α/(α+1)}) controls used in the proof of Theorem 4.3; estimate (9) with γ=α+1/2 provides the 0H^{α+3/2}(0,T;L²) and 0H^{1/2}(0,T; ˙H²) controls used in (69); estimate (11) with γ=α+1/2 and the 1/√ε constant is used in (68) to obtain the √ln(1/h) h² factor in Theorem 4.4; and estimate (10) provides the C([0,T]; ˙H¹) embedding used in Theorem 4.3. If [6] is not yet published, the manuscript is not self-contained on this point. The authors must either prove Theorem 2.2 in this paper or provide a complete, checkable derivation from [6, Theorems 4.2], including the endpoint γ=α+1/2 with the 1/√ε bound, the range γ≥(α-3)/4, and the range 0≤β<∞. Until then, the right-hand sides of (45) and (46) are unsupported.
  2. [§2.3, Theorem 2.1(6)] The stability bound (6) in Theorem 2.1 is asserted with the sentence 'the proof of (6) is trivial and hence omitted.' Although this bound is apparently not used in the proofs of Theorems 4.1-4.4, it is still a stated theorem in a numerical analysis paper. The authors should either supply the proof, since it is claimed to be trivial, or downgrade the statement to a remark so that no unproved assertion remains in the main text. As it stands, this is a completeness gap, albeit not a load-bearing one for the central convergence claims.
minor comments (4)
  1. [§2.2] Lemmas 2.1-2.4 are supported by the sentence 'For the proofs of the above lemmas, we refer the reader to [6, Section 3].' Since these lemmas are used throughout the paper and [6] is a companion preprint, it would improve self-containedness to include the proofs or to cite a published source for these standard fractional-calculus facts.
  2. [§5] The numerical experiments compare discrete solutions only against the finest discrete solution U^{11,16}, not against the exact solution. The word 'verify' in the section introduction is therefore too strong; the experiments are consistent with the predicted rates but do not test the constants or the precise dependence on the data norms. A sentence acknowledging this limitation would be appropriate.
  3. [Throughout] There are several typos and infelicities: 'thoerem' in the transposition-technique paragraph of §2.4, 'the this paper' in §2.2, 'combing' in the proofs of Theorems 3.2 and 3.3, and 'inerit' in the proof of Theorem 3.3. The reference list also has inconsistent formatting, e.g., page ranges for [7] and [8] use a hyphen in one place and an en-dash in another. These should be corrected.
  4. [§4.1] In the proof of Theorem 4.3, the step 'by Theorem 2.2 and Lemma 4.3' is quite compressed. It would help the reader if the authors explicitly identified which choices of the interpolation parameters (β,γ,r,s,θ) in Lemma 4.3 yield the required norm ‖u‖_{0H^{(1+ε+εα)/2}(0,T; ˙H^{1-ε})} and how Theorem 2.2 supplies the endpoint norms. This is not a correctness issue, but it would make the argument much easier to verify.

Circularity Check

1 steps flagged · score 4.0 of 10

Convergence rates are conditional on a same-authors companion regularity theorem; no construction-level circularity in the DG error analysis.

  1. self citation load bearing [Section 2.4, Theorem 2.2, and its applications in the proofs of Theorems 4.3 and 4.4 (Section 4).]
    "Furthermore, by a trivial modification of the proof of [ 6, Theorems 4.2], we readily obtain the following regularity results."

    Theorem 2.2 is the central a priori regularity input for the final error estimates. It is not proved in this paper; the only justification is a trivial modification of the proof of [6], which is the same-authors submitted companion paper (Luo, Li, and Xie, arXiv:1901.02799). The later proofs rely on (9)-(11) directly: (68) uses (11) for the h^{2(1-ε)} spatial bound, (69) uses Theorem 2.2 and Lemma 4.3 for the temporal error, and Remark 4.1 explicitly derives the 'nearly optimal' regularity claim from Theorem 2.2. Thus the advertised rates in (45) and (46) stand or fall with an unstated companion proof.

full rationale

The paper's own derivation in Section 3, including the Laplace-transform analysis of the two scalar fractional ODEs (Theorems 3.1-3.3), is carried out in detail and is independent of the final convergence result. No parameter is fitted, and no error bound is assumed as an input; the discrete solution is defined by the DG scheme (4), and the proofs of Theorems 4.1-4.4 use standard finite-element estimates, interpolation inequalities, and the scalar ODE results. The numerical experiments compare discrete solutions with each other rather than with the exact solution, which is a verification design limitation rather than a circular derivation. The one load-bearing external input is Theorem 2.2, the space-time weak-solution regularity estimates, imported from same-authors companion paper [6] with only a 'trivial modification' of [6, Theorems 4.2]. Since [6] is a submitted arXiv paper by overlapping authors rather than an independent machine-checked or externally verified result, the advertised rates are conditional on that companion proof. The paper also leaves the proof of (6) in Theorem 2.1 to 'techniques used in the proof of Theorem 4.3' without giving the argument, but (6) is not used in the later convergence proofs, so it is an omitted proof rather than a circular step. Overall, the central convergence claim has independent content and does not reduce by construction to its inputs; the main circularity concern is the self-citation of the regularity theorem that supports the final rates.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

Two classes of inputs are imported rather than derived here: the fractional Sobolev norm lemmas and the weak-solution regularity theorem come from the authors' companion manuscript [6] (arXiv:1901.02799), and the semidiscrete spatial error bounds come from Lubich, Sloan and Thomée [5]. There are no fitted constants, no invented physical entities, and no ad-hoc model assumptions. The balancing parameters ε used in the proofs are standard analytic devices with no scientific content.

assumptions (4)
  • standard math Lemmas 2.1 to 2.4: norm equivalences and positivity bounds for Riemann-Liouville fractional integrals on 0H^β spaces hold as stated.
    Cited to [6, Section 3] in Remark 2.2; used to control the memory term in the proofs of Theorems 4.3, 4.4 and Lemma 4.5.
  • domain assumption Theorem 2.2: the weak solution u satisfies the regularity bound (9) for γ in [(α-3)/4,∞) and the embeddings (10)-(11).
    Inherited from [6, Theorem 4.2] via 'trivial modification' (Section 2.4). Theorems 4.3 and 4.4 use this regularity to convert source-term norms into the norms appearing in the error estimates.
  • standard math Semidiscrete finite element error estimates of Lubich, Sloan and Thomée [5, Theorems 2.1 and 2.2]: for f=0, ‖u(t)-u_h(t)‖ ≲ h² t^{-α-1} ‖u0‖; for constant-in-time source v, ‖u(t)-u_h(t)‖ ≲ t^{-α} h² ‖v‖.
    Used directly in the proofs of Theorems 4.1 and 4.2 to split spatial and temporal errors.
  • standard math Interpolation and sampling inequalities from Tartar [17] (Lemmas 4.1-4.3) and the projection estimates (48)-(53) hold.
    Basis for the sharp h-dependence and τ^{1/2} factors in Theorems 4.3 and 4.4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Analysis of a time-stepping discontinuous Galerkin method for fractional diffusion-wave equation with nonsmooth data." pith.science (2026). https://pith.science/paper/KKCM4J5X

@misc{pith2026190809189,
  author       = {Pith},
  title        = {Pith review of: Analysis of a time-stepping discontinuous Galerkin method for fractional diffusion-wave equation with nonsmooth data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KKCM4J5X}},
  note         = {Machine review of arXiv:1908.09189}
}
abstract

This paper analyzes a time-stepping discontinuous Galerkin method for fractional diffusion-wave problems. This method uses piecewise constant functions in the temporal discretization and continuous piecewise linear functions in the spatial discretization. Nearly optimal convergence rate with respect to the regularity of the solution is established when the source term is nonsmooth, and nearly optimal convergence rate $ \ln(1/\tau)(\sqrt{\ln(1/h)} h^2+\tau)$ is derived under appropriate regularity assumption on the source term. Convergence is also established without smoothness assumption on the initial value. Finally, numerical experiments are performed to verify the theoretical results.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [6]

    H. Luo, B. Li, and X. Xie. Convergence analysis of a Petrov-Galer kin method for fractional wave problems with nonsmooth data. submitted, arXiv:1901.02799, 2018

    work page Pith review arXiv 1901

  2. [1]

    Cuesta, C

    E. Cuesta, C. Lubich, and C. Palencia. Convolution quadrature t ime dis- cretization of fractional diffusion-wave equations. Mathematics of Compu- tation, 65(213):1–17, 1996. 29

    work page 1996

  3. [2]

    B. Jin, R. Lazarov, and Z. Zhou. An analysis of the l1 scheme for t he subdiffusion equation with nonsmooth data. IMA Journal of Numerical Analysis, 36(1):197–221, 2016

    work page 2016

  4. [3]

    Lions and E

    J. Lions and E. Magenes. Non-Homogeneous Boundary Value Pro blems and Applications. Springer, Berlin, 1972

    work page 1972

  5. [4]

    C. Lubich. Discretized fractional calculus. SIAM Journal on Mathematical Analysis, 17(3):704–719, 1986

    work page 1986

  6. [5]

    Lubich, I

    C. Lubich, I. Sloan, and V. Thom´ ee. Nonsmooth data error est imates for approximations of an evolution equation with a positive-type memory term. Mathematics of Computation , 65(213):1–17, 1996

    work page 1996

  7. [7]

    McLean and V

    W. McLean and V. Thom´ ee. Maximum-norm error analysis of a num erical solution via laplace transformation and quadrature of a fractional- order evolution equation. IMA Journal of Numerical Analysis , 30(1):208-230, 2010

    work page 2010

  8. [8]

    McLean and V

    W. McLean and V. Thom´ ee. Numerical solution via laplace transfo rms of a fractional order evolution equation. Journal of Integral Equations and Applications, 22(1):57-94, 2010

    work page 2010

Show all 19 references
  1. [9]

    McLean, V

    W. McLean, V. Thom´ ee, and L. B. Wahlbin. Numerical solution of a n evolution equation with a positive type memory term. Journal of the Aus- tralian Mathematical Society Series B. Applied Mathematic s, 35(1):23-70, 1993

    1993

  2. [10]

    McLean, V

    W. McLean, V. Thome, and L.B. Wahlbin. Discretization with variab le time steps of an evolution equation with a positive-type memory term . Journal of Computational and Applied Mathematics , 69(1):49 – 69, 1996

    1996

  3. [11]

    McLean and K

    W. McLean and K. Mustapha. A second-order accurate numer ical method for a fractional wave equation. Numerische Mathematik , 105(3):481–510, Jan 2007

    2007

  4. [12]

    McLean and K

    W. McLean and K. Mustapha. Time-stepping error bounds for f ractional diffusion problems with non-smooth initial data. Journal of Computational Physics, 293(C):201–217, 2015

    2015

  5. [13]

    Mustapha and D

    K. Mustapha and D. Sch¨ otzau. Well-posedness of hp-version discontinuous galerkin methods for fractional diffusion wave equations. IMA Journal of Numerical Analysis, 34(4):1426–1446, 2014

    2014

  6. [14]

    Mustapha and W

    K. Mustapha and W. McLean. Discontinuous galerkin method for an evo- lution equation with a memory term of positive type. Mathematics of Computation, 78(268):1975–1995, 2009

    1975

  7. [15]

    Podlubny

    I. Podlubny. Fractional Differential Equations. Academic Press, 1998. 30

    1998

  8. [16]

    Samko, A

    S. Samko, A. Kilbas, and O. Marichev. Fractional Integrals and Deriva- tives: Theory and Applications. USA: Gordon and Breach Science Pub- lishers, 1993

    1993

  9. [17]

    L. Tartar. An Introduction to Sobolev Spaces and Interpolat ion Spaces. Springer, Berlin, 2007

    2007

  10. [18]

    Thom´ ee

    V. Thom´ ee. Galerkin Finite Element Methods for Parabolic Proble ms. Springer, Berlin, 2006

    2006

  11. [19]

    D. Wood. The computation of polylogarithms, technical report 15-29. 1992. 31

    1992

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.