Pith. sign in

REVIEW 3 major objections 5 minor 25 references

General-domain FC-based shock-dynamics solver II: Non-smooth domains, accuracy and parallel performance

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A spectral shock solver now treats domains with corners and scales almost perfectly across 1,620 cores.

desk verdict Genuinely new corner-patch construction with strong scaling and mostly clean validations; the no-slip Euler wall treatment needs a diagnostic, not a rewrite. read the letter →

arxiv 2506.19370 v1 pith:HDT5642R submitted 2025-06-24 math.NA cs.NAphysics.flu-dyn

Daniel V. Leibovici , Oscar P. Bruno This is my paper · ORCID
classification math.NAcs.NAphysics.flu-dyn MSC 65M7065M5565Y0576L0576N15
keywords Fouriercontinuationshockcapturingartificialviscosityneuralnetwork2DEulerequationsnon-smoothdomainsoverlappingpatchesparallelscaling
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 extends a Fourier-continuation spectral solver for shock dynamics to domains whose boundaries contain corners, and adds a distributed-memory parallel implementation. The central claim is that two new patch types—C1-type patches for corners with interior angle greater than 180 degrees and C2-type patches for corners with interior angle less than 180 degrees—make the overlapping-patch method stable where a naive smooth extension past a corner fails, while the neural-network-selected artificial viscosity keeps shocks sharp and regular regions smooth. If correct, the method would be the first spectral scheme able to handle shocks in general non-smooth domains, and its near-perfect weak scaling would make large multi-obstacle simulations practical. The paper supports the claim by comparing wedge shock angles with inviscid theory, reproducing shock-prism interactions, checking energy conservation over 120 wall reflections, and measuring scaling on up to 1,620 cores.

What carries the argument

The load-bearing construction is the corner patch. For a C2-type corner, the patch is the image of the unit square $Q=[0,1]^2$ under a parametrization of the form $M_{C2}(q_1,q_2)=\ell_A(q_1)+\ell_B(q_2)-C$; for a C1-type corner, the same sum is used but the parameter domain is the L-shaped set $L=([0,1]\times[1/2,1])\cup([1/2,1]\times[0,1/2])$. Each patch is divided into overlapping subpatches, and stability requires the minimum-overlap condition: every grid point near a patch side that is not on the physical boundary must lie at least $2n_v+1$ points deep inside another patch, with the additional requirement that S-type patches extend up to and including the corner point. The artificial viscosity is built per subpatch from a neural-network smoothness classification of the Mach number proxy variable, then blended across the patch union by smooth windowing functions, so dissipation is localized near discontinuities while Fourier continuation and SSPRK-4 time stepping provide spectral accuracy and stability.

What would settle it

Refine the Mach 3.5 and Mach 10 wedge-flow tests beyond the 31.5-million-point meshes reported here, for example with adaptive refinement near the tip, and fit the oblique-shock angle far from the tip: if the fitted angle converges to a value outside the Rankine-Hugoniot predictions ($34.6^\circ$ and $25.8^\circ$, respectively) instead of approaching them, the artificial-viscosity boundary layer is not reproducing the intended wall physics.

Watch

Extended reading notes

Core claim

The paper claims that the FC-SDNN spectral solver, previously demonstrated for smooth or rectangular domains, can be extended to non-smooth domains with corners by adding C1- and C2-type corner patches and by imposing a minimum-overlap condition that requires smooth-boundary patches to reach all the way to the corner. With this extension and the existing neural-network-selected artificial viscosity, the solver resolves shocks and contact discontinuities sharply, produces smooth level-set contours without Gibbs oscillations, and does not require positivity-preserving limiters. The parallel implementation shows essentially perfect weak scaling, reported as 99 to 104 percent efficiency, and strong scaling efficiencies in the 81 to 97 percent range for the tested configurations. The paper further claims that, to the authors' knowledge, this is the first spectral scheme to handle shocks in general domains.

Load-bearing premise

The solver imposes sticky, insulating wall conditions on obstacles even though it is solving the inviscid Euler equations, trusting that the artificial viscosity creates a physically correct thin boundary layer; if that layer does not match the real viscous flow, the comparisons to inviscid shock theory and experiments are comparisons to the wrong model.

Editorial extensions

If this is right

  • The method can be applied to supersonic and hypersonic Euler flows around obstacles with corners, such as wedges, prisms, and arrays of cylinders, without positivity-preserving limiters.
  • Shock and contact discontinuity features remain sharp while regular flow regions show smooth contours, so features such as slip lines and vortices remain visible in Schlieren-type visualizations.
  • Near-perfect weak scaling means problem size can be increased by adding cores with essentially fixed time per step, making larger obstacle arrays and finer meshes practical on distributed-memory machines.
  • Energy remains close to conserved, with the reported value $"bar{E}"\approx 1.37\pm 0.01$ against the exact value 1.375 over 120 wall reflections, indicating that artificial viscosity and filtering do not dissipate the solution excessively.
  • Computed oblique-shock angles for wedge flows move toward the inviscid Rankine-Hugoniot values as the mesh is refined, supporting the physical fidelity of the boundary treatment.

Reading between the lines

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

  • The solver imposes sticky, insulating wall conditions on obstacles even though it solves the inviscid Euler equations, so it is effectively solving a viscous regularized problem near walls; if that interpretation is right, the same machinery could be pointed at genuine Navier-Stokes problems at finite Reynolds numbers.
  • The C1 corner construction, based on an L-shaped parameter domain, likely generalizes to arbitrary corner angles and to three-dimensional edges, although the paper demonstrates only two-dimensional examples.
  • Because the paper reuses a single pretrained neural-network classifier across problems, an untested implication is that the same classifier may transfer to other conservation-law systems or to finer meshes without retraining.
  • The scaling tests assign one core per subpatch, and the reported weak-scaling efficiencies suggest communication per subpatch does not grow with problem size; the decomposition should therefore remain efficient at much larger core counts if the network hardware can sustain the communication pattern.
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

3 major / 5 minor

Summary. This Part II paper extends the FC-SDNN spectral shock solver of Part I to non-smooth domains by introducing corner patches of types C1 and C2, adds an MPI parallel implementation, and validates the method on several test problems for the 2D Euler equations: oblique shocks past a wedge compared with Rankine-Hugoniot theory, the Riemann4 benchmark, an energy-conservation study, shock/obstacle interactions, and weak/strong scaling studies. The central claims are that the extended method enables stable and accurate treatment of domains with corners, resolves shocks and contacts sharply while remaining smooth in regular regions, and achieves near-perfect weak scalability.

Significance. If established, the paper would be a significant contribution: it appears to be the first spectral scheme demonstrated for shock dynamics in general domains with corners, and the parallel scaling results are strong. The C1/C2 corner-patch construction is a nontrivial algorithmic advance, and the numerical experiments show several positive properties: the wedge shock angle converges toward the Rankine-Hugoniot value with refinement (35.2 deg versus 34.6 deg at M=3.5), total energy stays within about 1 percent over 120 wall reflections, and the weak-scaling efficiencies in Tables 1 and 2 are essentially perfect. The ANN-based localized viscosity also produces visibly smooth contours away from discontinuities. The main weakness is the modeling choice of imposing no-slip boundary conditions while solving the inviscid Euler equations, which directly affects the wedge validation that is the only quantitative accuracy test for the new corner-patch construction.

major comments (3)
  1. [7.1.1, 2, 7.2.4, Table 4] The wedge validation is built on a questionable modeling choice: the Euler equations are solved with full no-slip and adiabatic wall conditions (Section 7.1.1; Section 2), justified by claiming that the artificial viscosity creates a viscous boundary layer. However, the artificial viscosity I-(38) is a mesh-dependent numerical regularization, and the paper provides no measurement of the resulting boundary-layer thickness, displacement effect, or effective Reynolds number. The residual shock-angle errors in Table 4 (0.6 deg at M=3.5 and 1.5 deg at M=10 at the finest mesh) could therefore be the displacement effect of the numerical boundary layer rather than ordinary discretization error, and the comparison with the inviscid Rankine-Hugoniot relation (36) may be a comparison with the wrong model. The authors should either impose slip conditions for the inviscid wedge problem, or quantify the boundary-layer displacement and demonstrate that the inviscid limit is approached as h goes to zero, or re-frame the test as a Navier-Stokes validation with a physical viscosity.
  2. [7.2.7, abstract] The abstract states that results are 'compared to previous theoretical and experimental results', but the experimental validation in Section 7.2.7 is only qualitative: the Schardin problem is referenced and simulations are shown, but no experimental measurement (shock position, pressure trace, or density field) is plotted or compared. The paper should either add a quantitative experimental comparison or temper the experimental-validation claim to what the data actually support.
  3. [7.2, Conclusions] The conclusion that 'the approach does not require use of problem-dependent algorithmic parameters' is contradicted by the CFL choice reported in Section 7.2: CFL=0.25 is used for all simulations involving C1-type corners and patches, whereas CFL=0.5 is used in all other cases. This indicates that at least one algorithmic parameter was adjusted according to the problem geometry. The claim should be reworded to reflect the fact that while most parameters are fixed, the CFL number is problem-dependent in the present implementation.
minor comments (5)
  1. [3.2.2, Eq. (23)] Equation (23) defines L^{C1}_{p,2} twice and never defines L^{C1}_{p,3}; the third line should read L^{C1}_{p,3}.
  2. [Throughout] There are several typos and grammatical slips, e.g., 'results produces', 'the the', 'evidences', and 'suffices' in Section 7.1.1; these should be corrected in a revision.
  3. [3.2.2] The term 'equiaxed' is non-standard; consider using 'isotropic' or defining the meaning explicitly in Remark 4.
  4. [7.2.1] The multi-patch versus single-patch comparison in Section 7.2.1 is presented visually only; a quantitative error measure (e.g., an L2 difference between the two solutions) would strengthen the validation.
  5. [7.1.3-7.1.5, 7.2.2] The scaling tables and the energy-conservation study report single runs without error bars or repeated executions; a brief statement on run-to-run variability would improve confidence in the claimed efficiencies and in the 1 percent energy bound.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's validation claims are tested against independent external benchmarks, and the cited Part I methods are provenance rather than derived targets.

full rationale

The paper's central results—shock-angle comparisons against the Rankine-Hugoniot relation, the Riemann4 comparison with the single-patch solver, the energy-conservation test against the exact value 1.375, the Schardin-problem qualitative benchmark, and the measured parallel scaling efficiencies—are all evaluated against independent data or closed-form values, not against quantities fitted from the same simulations. The ANN-based viscosity and time-stepping algorithms are taken from the authors' Part I and from [6,24], but the present paper does not derive its validation from those citations: the numerical solutions are produced and then compared with external theory and experiment. The omission of ANN training details and exact weights is a transparency and reproducibility limitation, not a circular step, because no target output of the paper is used to define the classifier or the viscosity operator. The use of no-slip and adiabatic boundary conditions in an Euler solver with artificial viscosity is a modeling concern that could affect the physical fidelity of the wedge validation, but it does not make the derivation circular: the computed shock angle is not imposed as an input, and the table shows systematic convergence toward the inviscid theoretical value rather than an identity. No equation in the paper was found to be equivalent by construction to an input, and no load-bearing uniqueness claim is imported solely from the authors' prior work. Accordingly, under the hard rule that circularity must be exhibited by a specific reduction, no significant circularity is present.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the stability of the new corner-patch overlap rules, which are justified by numerical experimentation rather than proof; on the generalization of a pretrained neural-network classifier to new geometries; on the unusual no-slip wall model for Euler flows; and on the availability of the unpublished Part I for the underlying algorithm. Several hand-chosen constants (CFL, nv, nf, n0, n1) are fixed for all tests. No new physical entities are postulated.

free parameters (5)
  • CFL number = 0.5 (general), 0.25 (C1-corner problems)
    Chosen per problem class for stability with stretched meshes; contradicts the paper's claim of no problem-dependent parameters.
  • ANN weights and biases = not provided (trained in Part I/[6])
    Smoothness classifier tau is load-bearing for viscosity placement; trained offline, not shipped.
  • Artificial viscosity coefficient(s) from I-(38) = not restated in Part II
    Viscosity magnitude is set by a formula involving mesh size and maximum wave speed; any scaling coefficient is inherited from Part I and effectively tuned in earlier work.
  • Spectral filter parameters (I-(32)) = not restated in Part II
    Exponential filter applied every time step to control high-frequency modes; shape parameters inherited from Part I.
  • Subpatch discretization constants = nv=9, nf=5, n0=83, n1=43
    Hand-chosen constants used in all test cases; they affect overlap width and subpatch resolution.
assumptions (5)
  • domain assumption The Euler equations are solved with full no-slip and adiabatic boundary conditions on obstacle surfaces, justified by the artificial viscosity acting as a physical viscous boundary layer.
    Section 7.1.1 and Section 2; if the artificial viscosity does not reproduce the relevant viscous boundary-layer physics, all obstacle comparisons (wedge angles, shock-prism interaction) are against the wrong physical model.
  • domain assumption The minimum-overlap condition and the special S-C1 overlap rule (S patches must reach the corner point) guarantee stability.
    Section 3.2.3 states this is based on numerical experimentation, not proof; the stability of the whole time-stepping method rests on it.
  • ad hoc to paper The ANN smoothness classifier trained in Part I/[6] generalizes to the new corner-patch geometries and the Mach 1.5-10 flow regimes used here.
    Algorithm 2 line 8 loads pretrained weights; no retraining or separate validation of classifier accuracy on corner patches is reported.
  • ad hoc to paper C1 corner patch arms are approximately equal in width and length, so a single grid size h can be used.
    Section 3.2.2 Remark 4; restricts the class of corners the method can handle cleanly.
  • standard math Rankine-Hugoniot oblique shock theory for an infinite wedge is the correct benchmark for the finite-wedge simulations.
    Used in Section 7.2.4 as the reference for shock deflection angles; finite wedge and no-slip walls introduce deviations that are assumed to vanish with refinement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of General-domain FC-based shock-dynamics solver II: Non-smooth domains, accuracy and parallel performance." pith.science (2026). https://pith.science/paper/HDT5642R

@misc{pith2026250619370,
  author       = {Pith},
  title        = {Pith review of: General-domain FC-based shock-dynamics solver II: Non-smooth domains, accuracy and parallel performance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HDT5642R}},
  note         = {Machine review of arXiv:2506.19370}
}
read the original abstract

This contribution Part II of a two-part series, extends the general-domain FC-SDNN (Fourier Continuation Shock-Detecting Neural Network) introduces in Part I to enable treatment of non-smooth domains, it introduces a parallel implementation of the scheme with high-quality weak and strong scalability properties, and it illustrates the overall methodology for a variety of tests for the 2D Euler equations--including supersonic and hypersonic flows and shocks past obstacles with corners. The results produces by the new methods are compared to previous theoretical and experimental results, and the high parallel scalability of the algorithm is demonstrated in both weak and strong scaling cases. Thanks to its use of a localized yet smooth artificial viscosity term--whose support is confined to regions near flow discontinuities identified by an artificial neural network--the algorithm maintains minimal numerical dissipation away from discontinuities. Overall, the method delivers accurate, sharp resolution of shocks and contact discontinuities, while producing smooth numerical solutions in regular flow regions--as evidences by the near-complete absence of spurious oscillations in level-set contours, even under strong shocks and high-speed flow conditions.

Figures

Figures reproduced from arXiv: 2506.19370 by the authors.

Figure 1
Figure 1. Patch types used in the overlapping-patch decomposition of a general domain Ω, including [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Domain mappings for a C1-type patch (left image pair), a C2-type patch (middle image pair), and an S-type patch (right image pair). To conclude this section it is relevant to note that a straightforward patching strategy around a C1 corner point obtained as a union of two patches extruded separately from the curves ⌢ AD and ⌢ BE results in a pair of patches that cannot satisfy the inter-patch overlap condition requi… view at source ↗
Figure 3
Figure 3. Subpatch decomposition of the L-parameter-space used for patches ΩR p with R = C1; for this illustration the geometric parameter values r C1 p = 4, n1 = 9, and nv = 3 were used. According to the text prescriptions the patch is partitioned into a total of 3 4 [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Illustration of the types of geometries used for the test problems in Section 7.1.1. Left [PITH_FULL_IMAGE:figures/full_fig_p021_4.png]
Figure 5
Figure 5. Figure 5: FC-SDNN numerical solutions for the Riemann4 test case at [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]
Figure 6
Figure 6. Figure 6: Numerical values of the total energy E¯ (left panel) and its defect ∆E¯ from the exact value E¯ exact = 1.375 (right panel) as functions of time in the time interval 0 ≤ t ≤ 100—within which the shock, contact discontinuity and rarefaction wave were reflected 60 times …
Figure 7
Figure 7. Figure 7: Supersonic flows past a triangular wedge at time [PITH_FULL_IMAGE:figures/full_fig_p028_7.png]
Figure 8
Figure 8. Figure 8: Supersonic flow past a triangular prism at time [PITH_FULL_IMAGE:figures/full_fig_p030_8.png]
Figure 9
Figure 9. Figure 9: Density Schlieren visualization of a Mach 10 shock-wave mitigation by a 3x3 matrix of [PITH_FULL_IMAGE:figures/full_fig_p031_9.png]
Figure 10
Figure 10. Figure 10: Shock–prism interaction at two different shock speeds and times: Mach 1.5 at [PITH_FULL_IMAGE:figures/full_fig_p032_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 24 canonical work pages

  1. [1]

    Albin and O

    N. Albin and O. P. Bruno. A spectral FC solver for the compressible Navier–Stokes equations in general domains I: Explicit time-stepping. Journal of Computational Physics , 230:6248–6270, 2011

    work page 2011

  2. [2]

    Amlani and O

    F. Amlani and O. P. Bruno. An FC-based spectral solver for elastodynamic problems in general three-dimensional domains. Journal of Computational Physics , 307:333–354, 2016

    work page 2016

  3. [3]

    Baggag, H

    A. Baggag, H. Atkins, and D. Keyes. Parallel implementation of the discontinuous galerkin method. Technical report, 1999

    work page 1999

  4. [4]

    J. W. Banks, W. Henshaw, D. Schwendeman, and A. Kapila. A study of detonation propagation and diffraction with compliant confinement. Combustion Theory and Modelling , 12(4):769–808, 2008

    work page 2008

  5. [5]

    O. P. Bruno, M. Cubillos, and E. Jimenez. Higher-order implicit-explicit multi-domain compressible Navier-Stokes solvers. Journal of Computational Physics , 391:322–346, 2019

    work page 2019

  6. [6]

    O. P. Bruno, J. S. Hesthaven, and D. V. Leibovici. FC-based shock-dynamics solver with neural- network localized artificial-viscosity assignment. Journal of Computational Physics: X , 15:100110, 2022

    work page 2022

  7. [7]

    O. P. Bruno and D. V. Leibovici. General-domain FC-based shock-dynamics solver i: Basic elements. Part I of the present two-part paper sequence. , 2025

    work page 2025

  8. [8]

    Chang and K.-S

    S.-M. Chang and K.-S. Chang. On the shock–vortex interaction in Schardin’s problem. Shock Waves, 10(5):333–343, 2000. 33

    work page 2000

Show all 25 references
  1. [9]

    J. Chao, A. Haselbacher, and S. Balachandar. A massively parallel multi-block hybrid compact–weno scheme for compressible flows. Journal of Computational Physics , 228(19):7473–7491, 2009

    2009

  2. [10]

    Chaudhuri, A

    A. Chaudhuri, A. Hadjadj, and A. Chinnayya. On the use of immersed boundary methods for shock/obstacle interactions. Journal of Computational Physics , 230(5):1731–1748, 2011

    2011

  3. [11]

    Chaudhuri, A

    A. Chaudhuri, A. Hadjadj, O. Sadot, and G. Ben-Dor. Numerical study of shock-wave mitigation through matrices of solid obstacles. Shock Waves, 23:91–101, 2013

    2013

  4. [12]

    Chesshire and W

    G. Chesshire and W. D. Henshaw. Composite overlapping meshes for the solution of partial differential equations. Journal of Computational Physics , 90(1):1–64, 1990

    1990

  5. [13]

    S. Das, H. T. Toh, A. B. H. Abu, S. Das, K. Mondal, and L. K. Quen. A preliminary numerical study of high speed compressible flow around a solid triangular prism

  6. [14]

    Dresdner, D

    G. Dresdner, D. Kochkov, P. Norgaard, L. Zepeda-Nunez, J. A. Smith, M. P. Brenner, and S. Hoyer. Learning to correct spectral methods for simulating turbulent flows. arXiv preprint arXiv:2207.00556, 2022

    2022 arXiv

  7. [15]

    Gottlieb

    S. Gottlieb. On high order strong stability preserving Runge-Kutta and multi step time discretizations. Journal of Scientific Computing , 25(1):105–128, 2005

    2005

  8. [16]

    Guermond, M

    J.-L. Guermond, M. Kronbichler, M. Maier, B. Popov, and I. Tomas. On the implementation of a robust and efficient finite element-based parallel solver for the compressible navier-stokes equations. Computer Methods in Applied Mechanics and Engineering , 389:114250, 2022

    2022

  9. [17]

    Guermond, R

    J.-L. Guermond, R. Pasquetti, and B. Popov. Entropy viscosity method for nonlinear conservation laws. Journal of Computational Physics , 230(11):4248–4267, 2011

    2011

  10. [18]

    Kornelus and D

    A. Kornelus and D. Appel¨ o. Flux-conservative Hermite methods for simulation of nonlinear conser- vation laws. Journal of Scientific Computing , 76(1):24–47, 2018

    2018

  11. [19]

    P. D. Lax and X.-D. Liu. Solution of two-dimensional Riemann problems of gas dynamics by positive schemes. SIAM Journal on Scientific Computing , 19(2):319–340, 1998

    1998

  12. [20]

    Maier and M

    M. Maier and M. Kronbichler. Efficient parallel 3d computation of the compressible euler equations with an invariant-domain preserving second-order finite-element scheme. ACM Transactions on Par- allel Computing , 8(3):16:1–30, 2021

    2021

  13. [21]

    Mehta, T

    Y. Mehta, T. Jackson, J. Zhang, and S. Balachandar. Numerical investigation of shock interaction with one-dimensional transverse array of particles in air. Journal of Applied Physics , 119(10), 2016

    2016

  14. [22]

    Nazarov and A

    M. Nazarov and A. Larcher. Numerical investigation of a viscous regularization of the euler equations by entropy viscosity. Computer Methods in Applied Mechanics and Engineering , 317:128–152, 2017

    2017

  15. [23]

    Schardin

    H. Schardin. High frequency cinematography in the shock tube. The Journal of Photographic Science, 5(2):17–19, 1957

    1957

  16. [24]

    Schwander, D

    L. Schwander, D. Ray, and J. S. Hesthaven. Controlling oscillations in spectral methods by local artificial viscosity governed by neural networks. Journal of Computational Physics , 431:110144, 2021

    2021

  17. [25]

    Sonntag and C.-D

    M. Sonntag and C.-D. Munz. Efficient parallelization of a shock capturing for discontinuous galerkin methods using finite volume sub-cells. Journal of Scientific Computing , 70:1262–1289, 2017. 34

    2017

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.