Pith. sign in

REVIEW 4 major objections 5 minor 8 references

Numerical Study of Dissipative Weak Solutions for the Euler Equations of Gas Dynamics

T0 review · 4 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read Different numerical schemes converge to different dissipative weak solutions of the compressible Euler equations.

desk verdict A careful numerical comparison whose central interpretive claim—order-averages at one grid being K-convergence—doesn't follow from the cited theorems. read the letter →

arxiv 2601.17452 v3 pith:2VDX6A2O submitted 2026-01-24 math.NA cs.NA

Shaoshuai Chu , Michael Herty , Alexander Kurganov , Maria Lukacova-Medvidova , Changsheng Yu This is my paper · ORCID
classification math.NAcs.NA MSC 65M0865M1276M1276M2076N1035L65
keywords dissipativeweaksolutionsEulerequationsgasdynamicsK-convergenceYoungmeasuresmeasure-valuedKelvin-Helmholtzinstabilityenergydefect
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 studies first- through ninth-order versions of three finite-volume and finite-difference schemes applied to two-dimensional Riemann problems and a Kelvin-Helmholtz instability. It argues that when the flow develops fine-scale oscillations, the numerical approximations do not converge strongly to a single weak solution; instead, their weak limits are generalized dissipative weak solutions that depend on the numerical scheme. The paper shows that averages of solutions across orders of accuracy converge strongly, and time-averaged flows look similar across methods, even when pointwise fields differ. It proposes entropy production and energy defect as criteria to select among these non-unique limits. A sympathetic reader would care because this pins down what simulations of turbulent compressible flow actually compute.

What carries the argument

The load-bearing object is the dissipative weak solution: a measure-valued solution in which the momentum equation carries an additional Reynolds stress defect and the total energy carries an energy defect, both encoded in a Young measure over density, momentum, and entropy. The mechanism used to extract a strong limit is K-convergence: averaging a weakly convergent sequence of approximations (here, averages over the six formal orders at fixed mesh resolution) can produce strong convergence to a DW solution. The paper also uses the time-average of the averaged density and histogram approximations of the Young measure's probability density to assess convergence and scheme dependence.

What would settle it

Run one of the schemes on the Kelvin-Helmholtz test at meshes 256, 512, 1024, and 2048 and form order-averaged densities at each resolution. If these averages do not converge, or converge to different limits on different refinement paths, the claim that order-averaging yields strong convergence to a single DW solution fails. If a very-high-resolution reference simulation shows all schemes' averages approaching one common field, then the scheme-dependence of the DW limit would be contradicted.

Watch

Extended reading notes

Core claim

The paper's central claim is that in regimes where the Euler equations develop fine-scale oscillations, the limit of a consistent numerical approximation is not a unique weak solution but a generalized dissipative weak solution that depends on the chosen numerical method. Using first- through ninth-order versions of three finite-volume/finite-difference methods on the same problems, the authors observe that pointwise density fields from different schemes diverge, while order-averaged densities converge within each scheme yet differ across schemes. They interpret the scheme-dependent limits as different DW solutions, characterized by non-Dirac Young measures, and use entropy-production and en

Load-bearing premise

The load-bearing premise is that averaging six numerical solutions of different formal orders on one fixed 1024 by 1024 mesh behaves like the K-convergence averaging from the theory, which applies to subsequences of consistent approximations on refined meshes converging to zero mesh size. The paper itself states that rigorous consistency is not available for the high-order schemes, so this identification is an assumption rather than a theorem.

Editorial extensions

If this is right

  • In oscillatory flow regimes, no single consistent numerical method is guaranteed to approximate the same weak solution; the scheme selects a particular dissipative weak limit.
  • Averaging approximations of increasing formal order at a fixed resolution yields strongly convergent-looking sequences, and these order-averaged fields can distinguish genuine DW solutions from plain weak solutions.
  • Time-averaged density fields of order-averaged solutions look nearly identical across different schemes, suggesting that statistical measures are more transferable than pointwise fields.
  • The limiting Young measures are not Dirac deltas in regions with oscillations, and their approximated probability densities depend on the numerical scheme.
  • Entropy production and energy defects provide computable selection criteria, but the rankings of schemes differ depending on which criterion is used; the paper leaves the choice open.

Reading between the lines

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

  • The identification of order-averaging on one fixed grid with the K-convergence averaging of the theory is an extrapolation: the theorem requires a sequence of consistent approximations on refined meshes, while the experiments average six orders on one mesh and the paper concedes that rigorous consistency for the high-order schemes is not available.
  • If scheme-dependent DW limits occur generically, then a practical consequence is that uncertainty in turbulent compressible-flow simulations should be quantified not just by mesh-refinement error but by comparing different numerical fluxes; convergence of a single scheme may be convergence to a scheme's own DW limit.
  • The robustness of time-averaged order-averaged fields suggests a testable extension: statistical quantities such as mean density, momentum, or energy across an ensemble of initial perturbations should be reproducible across qualitatively different numerical methods even when snapshot fields are not.
  • A natural next step the paper does not take is to prove uniqueness of the DW solution that maximizes entropy production (or extremizes energy defect) within a suitable class; if such uniqueness held, the selection criteria would promote an observation about schemes into a definition of the physically relevant limit.
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.

Referee Report

4 major / 5 minor

Summary. The paper reports a numerical study of the compressible Euler equations using three families of high-order schemes (LCDCU, LDCU, VFV) at formal orders 1, 2, 3, 5, 7, and 9, applied to two-dimensional Riemann problems and a Kelvin-Helmholtz instability. The central interpretive claim is that, in oscillatory regimes, different numerical methods converge only weakly and that averages over the different formal orders at a fixed grid converge strongly, in the sense of K-convergence, to scheme-dependent dissipative weak (DW) solutions. The paper also computes approximate Young-measure PDFs and compares entropy production and energy defects as potential selection criteria.

Significance. If the central claim were justified, the paper would provide valuable numerical evidence for the non-uniqueness of DW solutions and for scheme-dependence of limits of oscillatory numerical approximations, with practical criteria for selecting physically relevant solutions. The manuscript also makes a useful contribution by extending a family of A-WENO schemes up to ninth order and by comparing three different numerical fluxes on challenging benchmark problems. The authors are transparent about the lack of a rigorous consistency analysis for high-order methods and make the code available on request. However, the main theoretical interpretation is not supported by the experiments as presented.

major comments (4)
  1. [Section 4 (paragraph after notation) and Theorem 2.5] The averaging procedure used in the numerical study is not an instance of the K-convergence theorem. Theorem 2.5 applies to a consistent approximation indexed by mesh refinement h_n → 0 and requires Cesàro averages along a subsequence. The paper instead averages six solutions of different formal orders at the single mesh Δx = Δy = 1/1024. Moreover, Section 1 states that 'a rigorous consistency analysis is not available for high-order methods', so the sequences are not known to satisfy Definition 2.1, and the consistency errors e[φ] are never measured. The abstract and Section 5 claim that order-averaged solutions converge to DW solutions in the K-convergence sense; this is unsupported.
  2. [Section 4.3, Figure 4.13, Table 4.2] The evidence for 'K-convergence of the sequence of averages' consists of the L1 errors ||ρ̄_n - ρ̄_6|| for n = 1,...,5 relative to the sixth-order reference. This measures only closeness to a particular finite-order average; it does not establish existence of a limit as n → ∞. With n ≤ 6 and no mesh refinement, the fitted power laws in Table 4.2 are descriptive curve fits, not convergence tests. The claim that the average sequence 'is convergent' is therefore not demonstrated.
  3. [Sections 4.1–4.2, Figures 4.7 and 4.12] The approximate PDFs σ_ℓ and σ̄_n are spatial histograms of density values over a subdomain Ω̃, not the parametrized Young measure V_{x,y,t} appearing in Definition 2.3. The latter is a pointwise measure in phase space at each (x,y,t). A spatial histogram combines oscillations in space and time and cannot, by itself, establish that the limiting DW solutions differ. The connection between the computed histograms and the measure-valued limit is not made rigorous.
  4. [Section 4.4, Tables 4.3 and 4.4] The selection criteria are evaluated on S_n(t) and D_E^n(t) computed from order-averaged fields. Since these fields are not shown to be K-convergent limits, the comparisons may reflect properties of the averaging procedure rather than of genuine DW solutions. In addition, S_n(t) is the total entropy, not an entropy production rate; the paper should define the 'entropy production' that criteria (i) refers to, and clarify how it is measured.
minor comments (5)
  1. [Definition 2.1] The continuity equation consistency error e_2[φ] is introduced for φ ∈ C^1_c, but the convergence condition is stated for φ ∈ C^2_c. Please clarify the required test-function regularity.
  2. [Section 4 notation] The averaged quantities ρ̄_n, m̄_n, S̄_n are introduced informally. Please give explicit definitions, e.g., ρ̄_n = (1/n) Σ_{ℓ=1}^n ρ_ℓ, and state that these are functions on Ω×(0,T).
  3. [Table 4.2] The least-squares fits use only five data points (n = 1,...,5). Report the number of degrees of freedom or confidence intervals so the reader can judge the fit quality.
  4. [Section 3.2] The Heaviside functions in the A-WENO flux formulas make the order selection implicit. A short table or sentence summarizing which correction terms are active for r = 3,5,7,9 would improve readability.
  5. [Reproducibility] The data availability statement says FORTRAN codes are available upon request. A permanent repository or DOI, and a description of the pseudo-random number generation for the KH initial data, would strengthen reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the scheme-dependent DW limits are exhibited numerically, not derived from the cited framework; the main weakness is an unproved identification of order-averages with K-convergence, which is a validity issue rather than circularity.

full rationale

The paper's central claim—different numerical methods may converge to different DW solutions—rests on direct numerical measurements (density plots, order-averaged profiles, Young-measure PDFs, and entropy/energy-defect tables) and not on a parameter fitted to the same target. The cited results [16,18,19] supply the theoretical notions (consistent approximation, DW solution, K-convergence) and are from overlapping authors, but they are used as an interpretive framework, not as the evidence for scheme dependence. No equation in the paper is defined in terms of the target claim, and no fitted quantity is renamed as a prediction. The most serious methodological weakness is the identification in Section 4 of Cesàro averages over formal orders of accuracy at a fixed mesh Δx=Δy=1/1024 with the K-convergence averages of Theorem 2.5, which is defined for consistent approximations on refined meshes h_n→0. The paper itself concedes in Section 1 that 'a rigorous consistency analysis is not available for high-order methods,' so the high-order sequences are not known to satisfy Definition 2.1. That makes the DW-limit conclusion under-supported, but it is not circular: the numerical outputs are not constructed from the theorem, and the theorem does not force the observed scheme dependence. The self-citations to [16,18,19] and to submitted preprints [17,20] do not weaken the argument circularly because the central observations would stand or fall on the numerical experiments themselves. Thus no specific circular step can be quoted with an equation-level reduction; score 0.

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

The central claim is an empirical observation, so the ledger does not contain fitted constants that engineer the result. The load-bearing axioms are the mapping of order-averaging onto K-convergence and the unproved consistency/boundedness of high-order schemes; both are assumptions the reader must accept to move from the figures to the DW-solution interpretation.

free parameters (4)
  • Generalized minmod limiter parameter θ = 1.3
    Set by hand in Eq. (3.3); controls numerical diffusion in the second-order finite-volume reconstructions and may affect the fine-scale structure of the computed solutions.
  • CFL numbers = 0.45 for LCDCU/LDCU, 0.1 for VFV
    Chosen in Section 4 for stability; all schemes use SSP-RK3 time stepping, so the time-step size differs by method and contributes to method-dependent dissipation.
  • Young-measure histogram bin count = 30 bins
    Introduced in Section 4 to partition the density range in each subdomain; the choice is arbitrary and affects the resolution and visual appearance of the PDF approximations.
  • Least-squares power-law coefficients for order-averaging errors = C n^{-α} values in Table 4.2, e.g., 0.1335 n^{-0.98} for KH/LCDCU
    Fitted to L1 errors for n=1,...,5 in Section 4.3; these fitted rates are used to claim K-convergence behavior, but they are descriptive fits, not derived quantities.
assumptions (4)
  • ad hoc to paper Theorem 2.5 (K-convergence) applies to averages over increasing formal order at fixed mesh resolution
    The theorem is stated for subsequences of consistent approximations on refined meshes h_n→0; the paper applies it to order-indexed averages at Δx=Δy=1/1024 (Section 4, Section 4.3) with no proof that this is a valid instance of the theorem.
  • domain assumption High-order LCDCU, LDCU, and VFV schemes generate consistent approximations in the sense of Definition 2.1
    Consistency of the approximation sequence is required for Theorems 2.4 and 2.5; the paper admits in Section 1 that 'a rigorous consistency analysis is not available for high-order methods.'
  • domain assumption The computed numerical solutions are uniformly bounded in L∞, so Theorems 2.4 and 2.5 apply
    Theorem 2.4 assumes uniform L∞ boundedness of the approximation sequence; the paper does not verify this for the oscillatory KH or Configuration 3 solutions.
  • ad hoc to paper Entropy-production and energy-defect selection criteria (i)-(iii) are physically meaningful
    The criteria are justified in Sections 4.4 and 5 by reference to submitted preprints [17,20] from the same research group; no independent published validation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Numerical Study of Dissipative Weak Solutions for the Euler Equations of Gas Dynamics." pith.science (2026). https://pith.science/paper/2VDX6A2O

@misc{pith2026260117452,
  author       = {Pith},
  title        = {Pith review of: Numerical Study of Dissipative Weak Solutions for the Euler Equations of Gas Dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2VDX6A2O}},
  note         = {Machine review of arXiv:2601.17452}
}
abstract

We study dissipative weak (DW) solutions of the Euler equations of gas dynamics using the first-, second-, third-, fifth-, seventh-, and ninth-order local characteristic decomposition-based central-upwind (LCDCU), low-dissipation central-upwind (LDCU), and viscous finite volume (VFV) methods, whose higher-order extensions are obtained via the framework of the alternative weighted essentially non-oscillatory (A-WENO) schemes. These methods are applied to several benchmark problems, including several two-dimensional Riemann problems and a Kelvin-Helmholtz instability test. The numerical results demonstrate that for methods converging only weakly in space and time, the limiting solutions are generalized DW solutions, approximated in the sense of ${\cal K}$-convergence and dependent on the numerical scheme. For all of the studied methods, we compute the associated Young measures and compare the DW solutions using entropy production and energy defect criteria.

Figures

Figures reproduced from arXiv: 2601.17452 by the authors.

Figure 4.1
Figure 4.1. Configuration 2: Density computed by the first- (top row), third- (middle row), and ninth-order (bottom row) LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. Configuration 4. In this example, the initial data are (ρ, u, v, p)(x, y, 0) =    (1.1, 0, 0, 1.1), x > 0.5, y > 0.5, (0.5065, 0.8939, 0, 0.35), x < 0.5, y > 0.5, (1.1, 0.8939, 0.8939, 1.1), x < 0.5, y < 0.5, (0.5065, 0, 0.8… view at source ↗
Figure 4.2
Figure 4.2. Configuration 4: Density computed by the first- (top row), third- (middle row), and ninth-order (bottom row) LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. Configuration 3. In this example, the initial data are (ρ, u, v, p)(x, y, 0) =    (1.5, 0, 0, 1.5), x > 0.8, y > 0.8, (0.5323, 1.206, 0, 0.3), x < 0.8, y > 0.8, (0.138, 1.206, 1.206, 0.029), x < 0.8, y < 0.8, (0.5323, 0, 1.2… view at source ↗
Figure 4.3
Figure 4.3. Configuration 3: Density computed by the first- (top row), third- (middle row), and seventh￾order (bottom row) LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. computed densities, which are denoted by ρ T ℓ := 1 T Z T 0 ρℓ(·, t) dt, where T is the final time [PITH_FULL_IMAGE:figures/full_fig_p013_4_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4.4
Figure 4.4. Figure 4.4: Configuration 3: ρe3 (top row), ρe4 (middle row), and ρe5 (bottom row) computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. 4.2 Kelvin-Helmholtz (KH) Instability In the last example, we consider the KH instability problem taken f…
Figure 4.5
Figure 4.5. Figure 4.5: Configuration 3: ρ T 1 (top row), ρ T 3 (middle row), and ρ T 5 (bottom row) computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes [PITH_FULL_IMAGE:figures/full_fig_p015_4_5.png]
Figure 4.6
Figure 4.6. Figure 4.6: Configuration 3: ρe T 5 computed by the LCDCU (left), LDCU (middle), and VFV (right) schemes. density profiles show similarities for different n, suggesting that the sequence of averages is convergent. Additionally, we observe that for different methods, the limits o…
Figure 4.7
Figure 4.7. Figure 4.7: Configuration 3: σe3, σe4, and σe5 computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes in the subdomains Ωe1 (top row) and Ωe2 (bottom row). a n 1 a n 2 b n 1 b n 2 1 6.848086824246653e-08 9.373025805955863e-03 -0.973625473853271…
Figure 4.8
Figure 4.8. Figure 4.8: KH Instability: Density computed by the first- (top row), third- (middle), and ninth-order (bottom row) LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. where ρe6 serves as a reference solution. In [PITH_FULL_IMAGE:figures/full_fig_p017_4_8.png]
Figure 4.9
Figure 4.9. Figure 4.9: KH Instability: ρe3 (top row), ρe4 (second row), ρe5 (third row), and ρe6 (bottom row) computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes. 4.4 Selection Criteria As seen in the previous results (Configuration 3 for the 2-D Riema…
Figure 4.10
Figure 4.10. Figure 4.10: KH Instability: ρ T 1 (top row), ρ T 3 (middle row), and ρ T 6 (bottom row) computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes [PITH_FULL_IMAGE:figures/full_fig_p019_4_10.png]
Figure 4.11
Figure 4.11. Figure 4.11: KH Instability: ρe T 6 computed by the LCDCU (left), LDCU (middle), and VFV (right) schemes. The turbulent energy and total energy are defined by E n 1 (t) := Z Ω Een(x, y, t) dxdy = Z Ω 1 n Xn ℓ=1  |mℓ | 2 2ρℓ + ρℓe(ρℓ , Sℓ)  dxdy, E n 2 (t) := Z Ω E [PITH_FULL_…
Figure 4.12
Figure 4.12. Figure 4.12: KH Instability: σe3, σe4, σe5, and σe6 computed by the LCDCU (left column), LDCU (middle column), and VFV (right column) schemes in the subdomains Ωe1 (top row) and Ωe2 (bottom row) [PITH_FULL_IMAGE:figures/full_fig_p020_4_12.png]
Figure 4.13
Figure 4.13. Figure 4.13: The L1-errors of the average density computed by the studied schemes for Configuration 2 (left), Configuration 4 (middle), and KH instability (right) problems. Possible selection criteria for DW solutions are (i) the maximization of entropy production, (ii) the mini…
Figure 4.14
Figure 4.14. Figure 4.14: Configuration 3: Time evolution of the average entropy S 5 (left) and energy defects D5 E (right) computed by the LCDCU, LDCU, and VFV schemes. LCDCU LDCU VFV 1 T R T 0 S 5 (t) dt -0.3778 -0.3778 -0.3741 1 T R T 0 E 5 1 (t) dt 1.3690 1.3690 1.3541 1 T R T 0 E 5 2 (t…
Figure 4.15
Figure 4.15. Figure 4.15: KH Instability: Time evolution of the average entropy S 6 (left) and energy defects D6 E (right) computed by the LCDCU, LDCU, and VFV schemes. LCDCU LDCU VFV 1 T R T 0 S 6 (t) dt 1.049343 1.05036 1.05482 1 T R T 0 E 6 1 (t) dt 6.4375 6.4375 6.4375 1 T R T 0 E 6 2 (t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

8 extracted references · 2 linked inside Pith

  1. [1]

    Borges, M

    [1]R. Borges, M. Carmona, B. Costa, and W. S. Don,An improved weighted essentially non- oscillatory scheme for hyperbolic conservation laws, J. Comput. Phys., 227 (2008), pp. 3191–3211. Numerical Study of Dissipative Weak Solutions23 [2]T. Buckmaster, C. de Lellis, L. Sz ´ekelyhidi, Jr., and V. Vicol,Onsager’s conjecture for admissible weak solutions, Com...

    2008

  2. [33]

    [11]S. Chu, A. Kurganov, and R. Xin,New more efficient A-WENO schemes, J. Sci. Comput., 104 (2025). Paper No

    2025

  3. [53]

    De Lellis and L

    [12]C. De Lellis and L. Sz ´ekelyhidi, Jr.,On admissibility criteria for weak solutions of the Euler equations, Arch. Ration. Mech. Anal., 195 (2010), pp. 225–260. [13]R. J. DiPerna,Measure-valued solutions to conservation laws, Arch. Ration. Mech. Anal., 88 (1985), pp. 223–270. [14]W. S. Don and R. Borges,Accuracy of the weighted essentially non-oscillat...

    2010

  4. [56]

    Lie and S

    [28]K.-A. Lie and S. Noelle,On the artificial compression method for second-order nonoscillatory central difference schemes for systems of conservation laws, SIAM J. Sci. Comput., 24 (2003), pp. 1157–1174. [29]H. Liu,A numerical study of the performance of alternative weighted ENO methods based on various numerical fluxes for conservation law, Appl. Math....

    2003

  5. [152]

    Feireisl, M

    [16]E. Feireisl, M. Luk ´aˇcov´a-Medviˇdov´a, B. She, and Y. W ang,Computing oscillatory solutions of the Euler system viaK-convergence, Math. Models Methods Appl. Sci., 31 (2021), pp. 537–576. [17]E. Feireisl and M. Luk ´aˇcov´a-Medviˇdov´a,Well-posedness of the Euler system of gas dy- namics. Submitted. Preprint available athttps://arxiv.org/abs/2512.18...

    arXiv 2021

  6. [1792]

    Chertock, S

    [4]A. Chertock, S. Chu, M. Herty, A. Kurganov, and M. Luk ´aˇcov´a-Medviˇdov´a,Local characteristic decomposition based central-upwind scheme, J. Comput. Phys., 473 (2023). Paper No. 111718. [5]A. Chertock, S. Chu, and A. Kurganov,Adaptive high-order A-WENO schemes based on a new local smoothness indicator, E. Asian. J. Appl. Math., 13 (2023), pp. 576–609...

    arXiv 2023

  7. [2011]

    Gottlieb, C.-W

    [24]S. Gottlieb, C.-W. Shu, and E. Tadmor,Strong stability-preserving high-order time dis- cretization methods, SIAM Rev., 43 (2001), pp. 89–112. [25]Y. Jiang, C.-W. Shu, and M. P. Zhang,An alternative formulation of finite difference weighted ENO schemes with Lax-Wendroff time discretization for conservation laws, SIAM J. Sci. Comput., 35 (2013), pp. A11...

    2001

  8. [2021]

    24S. Chu, M. Herty, A. Kurganov, M. Luk ´aˇcov´a-Medviˇdov´a & C. Yu [20]E. Feireisl, M. Luk ´aˇcov´a-Medviˇdov´a, and C. Yu,Oscillatory approximations and maxi- mum entropy principle for the Euler system of gas dynamics. Submitted. Preprint available at https://arxiv.org/abs/2505.02070. [21]U. S. Fjordholm, S. Mishra, and E. Tadmor,On the computation of ...

    arXiv 2016

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.