A 100x Faster Beam Propagation Method for Nonlinear Optical Wave Propagation Based on Discrete Exterior Calculus

Amgad Abdrabou; R. El-Ganainy

REVIEW 3 major objections 6 minor 26 references

A 100x Faster Beam Propagation Method for Nonlinear Optical Wave Propagation Based on Discrete Exterior Calculus

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims a discrete-exterior-calculus beam propagation method solves the nonlinear Schrödinger equation on unstructured meshes with spectral-level accuracy and over 100x speedup versus FFT-based spectral methods in waveguides with

desk verdict A plausible DEC-based NLSE solver, but the paper overstates speed and accuracy; the validation needs an independent reference. read the letter →

arxiv 2608.03587 v1 pith:SUSXT4XM submitted 2026-08-04 physics.optics physics.comp-ph

Amgad Abdrabou , R. El-Ganainy This is my paper

classification physics.opticsphysics.comp-ph MSC 35Q5565M6078A60

keywords discreteexteriorcalculusbeampropagationmethodnonlinearSchrödingerequationunstructuredmeshesHodgestaroperatorstep-indexopticalfiberGibbsphenomenonspectral

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 aims to establish that a beam propagation method built on discrete exterior calculus can solve the nonlinear Schrödinger equation in structured optical waveguides much faster than FFT-based spectral solvers—more than 100x in the demonstrated cases—while keeping errors at or below the spectral method's. The method discretizes fields as discrete differential forms on an unstructured mesh that conforms to the waveguide cross-section, so the step-index interface is represented by the mesh instead of by a staircased Cartesian grid. Benchmarks on a fundamental soliton, an off-axis elliptic Gaussian, and an optical vortex show second-order convergence and speedups of about 73x and 150x at fixed error thresholds. If the claim holds, multiscale nonlinear photonic simulations that currently take days would take hours, and optimization of photonic crystal fibers, multimode waveguides, and similar devices becomes practical.

What carries the argument

The machinery is the discrete exterior calculus pair of operators on a primal simplicial complex and its dual: the combinatorial exterior derivative D0 (an incidence matrix encoding which vertices bound which edges) and the metric-dependent Hodge stars H0 and H1. H0 is diagonal, built from barycentric dual-cell areas, which makes its inverse trivial; H1 is built from Whitney 1-forms via a Galerkin construction. The discrete transverse Laplacian is -H0^{-1} D0^T H1 D0, and the potential and Kerr nonlinearity enter through per-triangle constant values and the precomputed sparse matrix Q^T D Q, where Q averages vertex values onto triangles. These operators make both the linear and nonlinear rig

What would settle it

Run the elliptic-Gaussian and vortex benchmarks with an independent reference that resolves the step-index interface exactly (for example, a conforming high-order finite-element solution or an exact modal expansion on the circular fiber), recompute Erel for both SIP and DEC, and check whether the time to reach Erel around 5e-3 and 8e-3 still differs by roughly two orders of magnitude. If the spectral method's Gibbs-contaminated reference was the only thing making DEC look 100x faster, the gap will shrink or vanish.

Watch

Extended reading notes

Core claim

The central claim is that replacing the uniform Cartesian grid of a spectral beam propagation method with a geometry-conforming simplicial mesh, and discretizing the NLSE through discrete exterior calculus, removes the main bottleneck for waveguides with sharp refractive-index steps: Gibbs oscillations at material interfaces. The resulting DEC-BPM is second-order convergent O(h^2), reaches agreement with the spectral reference using roughly an order of magnitude fewer unknowns, and at the benchmark error thresholds runs about 73x faster for the elliptic Gaussian and 150x faster for the vortex under the same adaptive RK45 time integrator. The paper also claims the formulation preserves d^2=0

Load-bearing premise

The speedup numbers rest on comparing against reference solutions generated by the spectral method itself; if that reference has ringing errors at the sharp fiber interface, the reported error and speedup are not measured against the true solution.

Editorial extensions

If this is right

For step-index fibers and similar discontinuous-index waveguides, DEC-BPM reaches a given accuracy with about 6x fewer degrees of freedom than the spectral method, and the wall-clock advantage grows as accuracy demands increase.
The method's second-order convergence is stable across symmetric solitons, asymmetric off-axis beams, and vortex beams, while the spectral method's convergence stagnates at first order because of Gibbs ringing at the same interface.
Because the discrete operators are sparse and H0 is diagonal, adaptive explicit time stepping is cheap; the per-step cost scales with mesh nonzeros rather than with FFT's n log n.
Higher-order diffraction terms in generalized NLSEs can be added by repeated products of the discrete Laplacian, giving a systematic route beyond the standard NLSE.
The same DEC operators support absorbing boundary conditions and perfectly matched layers by complex-stretching Hodge stars, which is not possible in FFT methods without ad hoc absorbing potentials.

Reading between the lines

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

The reported speedups are relative to the spectral method's own dense-grid solution; an independent, interface-exact reference could change the magnitude of the speedup, and for waveguides without sharp index discontinuities FFT-based solvers may remain the faster choice.
Because the normalized NLSE is identical to the Gross-Pitaevskii equation, the framework should transfer to Bose-Einstein condensate simulations; the practical blocker is the same boundary-handling question, not the discretization.
The algorithmic construction of higher-order operators suggests a route to vectorial and polarization-dependent propagation, but only if the discrete Hodge stars and exterior derivatives are built for vector bundle-valued forms, which the paper does not demonstrate.
A testable extension would be to add a PML to the DEC code and let the vortex propagate well beyond z=0.1; if the speedup persists at longer distances, the current boundary-limited benchmark is the only obstacle.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The method in this paper is genuinely new: applying discrete exterior calculus to scalar NLSE beam propagation, with a diagonal H0 Hodge star that enables explicit Runge-Kutta integration and a Galerkin H1 built from Whitney forms. The algorithmic construction of higher-order operators from lower-order ones is a useful byproduct. The soliton benchmark is the strongest part—DEC holds the profile to ~1e-6 over z=25, while SIP wanders in the 1e-3 to 1e-2 range. That result is clean and credible.

The soft spots, in order of size. First, the "spectral-level accuracy" claim is wrong on its face: every measured convergence curve for DEC is second-order O(h^2). That is polynomial, not spectral. The abstract should say "second-order" or "high-order" and stop there.

Second, "over 100x" is not uniformly supported. The Gaussian benchmark gives 73x, and only the vortex gives 150x. The blanket claim in the abstract and conclusion is not backed by the paper's own numbers.

Third—and this is the load-bearing problem—the validation protocol has a circular component. For the elliptic Gaussian and vortex cases, the reference solution is generated by SIP on a dense grid, the same spectral method the paper claims to outperform. Since the paper itself documents SIP's staircase Gibbs errors at the step-index interface, that reference is not an independent ground truth. The error E_rel measures distance to that SIP solution, not to the true continuum solution. The vortex case is especially exposed, because the paper attributes the gap to the interface, exactly where the SIP reference is least reliable. Without an independent, demonstrably converged reference (a conforming FEM solve, an analytic test, or a grid-convergence check against a different method), the reported speedups at fixed error thresholds are not robust evidence of true accuracy.

Minor points: the abstract promises absorbing boundary conditions as a feature, but they are only planned; the non-soliton benchmarks stop at z=0.1 because of Neumann boundaries. The common-integrator protocol is fair, but it makes speedups sensitive to chosen tolerances and thresholds.

Who is this for? Photonics researchers developing numerical methods for the NLSE, or anyone doing waveguide optimization who wants a geometry-conforming alternative to spectral solvers. The paper deserves a serious referee, but it needs major revision before publication: temper the accuracy language, correct the speedup claims, and add an independent reference test for the benchmark cases.

Referee Report

3 major / 6 minor

Summary. The paper introduces DEC-BPM, a beam propagation method for the nonlinear Schrödinger equation based on discrete exterior calculus. The method discretizes the NLSE directly on unstructured simplicial meshes using discrete exterior derivatives and Hodge stars, avoiding weak formulations and FFT-based periodic grids. Propagation uses an explicit adaptive RK45 integrator with sparse matrix-vector products. The authors benchmark DEC-BPM against the split-step/spectral SIP method on three test cases: a fundamental soliton in a step-index fiber, an off-axis elliptic Gaussian beam, and an elliptic optical vortex. The reported results show second-order convergence for DEC-BPM, roughly 73x and 150x speedups over SIP at given error thresholds in the two non-soliton cases, and long-term shape preservation for the soliton. The abstract and conclusion claim 'spectral-level accuracy' and 'speedups exceeding 100x.'

Significance. If the claimed performance is reproducible, DEC-BPM would be a valuable addition to the toolbox for nonlinear waveguide simulation, especially for geometries with sharp material interfaces where FFT-based spectral methods suffer from Gibbs oscillations. The method has several genuinely attractive features: the Hodge star H0 is diagonal, enabling explicit time integration without mass-matrix inversion; the discrete operators preserve topological identities such as d^2=0; higher-order diffraction operators are constructed by matrix products of the discrete Laplacian; and the unstructured mesh conforms to curved interfaces. These are concrete algorithmic contributions. However, the central quantitative claims—'spectral-level accuracy' and 'over 100x' speedup—are not supported by the evidence as presented, and the validation contains a self-referential element in the benchmark reference generation. The paper therefore requires substantive revision before the headline claims can be accepted.

major comments (3)

[Abstract; Section II, 'Numerical Benchmarking and Performance'; Section III] The claim of 'spectral-level accuracy' is directly contradicted by the paper's own measured convergence. In Figs. 7(a) and 10(a), the DEC method exhibits second-order convergence O(h^2), i.e., O(N^{-1}) in the number of degrees of freedom, and the text repeatedly states 'second-order DEC formulation' and 'clear second-order convergence.' This is algebraic, not spectral, accuracy. The abstract and conclusion must be revised to describe the accuracy honestly as high-order algebraic (second-order) convergence, not spectral-level accuracy.
[Section II, 'Numerical Benchmarking and Performance' (reference solutions; Eq. 15)] For the elliptic Gaussian and vortex benchmarks, the reference solution is obtained by SIP on a dense 512x512 grid—the same spectral method being outperformed. The paper itself documents that SIP suffers from staircase-induced Gibbs oscillations and non-monotone convergence at the step-index interface (Figs. 7 and 10). Thus E_rel (Eq. 15) measures error relative to a reference that may carry uncontrolled discretization error. The speedups of 73x and 150x are therefore not demonstrated against the true continuum solution. The authors should validate at least one non-soliton benchmark against an independent, converged reference—for example, a much finer DEC solution, a mesh-refinement study with Richardson extrapolation, or a manufactured solution—and report the error of both methods against that reference. Without this, the central speedup claim is unverified.
[Section II, 'Fundamental soliton'] The soliton benchmark is self-referential: the launch field is computed self-consistently on the same mesh used for propagation, so it is a discrete eigenmode of the discrete evolution operator by construction. The near-machine-zero propagation error therefore does not test the method's ability to propagate generic, non-eigenmode fields. Moreover, no speedup is reported for this case. Consequently, the only quantitative support for the 'over 100x' claim comes from one benchmark (the vortex at E_rel=8e-3, ~150x), while the other non-soliton benchmark gives only 73x at E_rel=5e-3. The headline claim should be qualified accordingly, and the soliton test should be complemented by a non-eigenmode initial condition that is not a discrete eigenmode of the same operator.

minor comments (6)

[Section II, 'DEC beam propagation'] Typo: 'we now write the DEC form of the NLES' should be 'NLSE.'
[Section II, Eq. (13)] The projection operator Q = (1/6)|D1||D0| is introduced without derivation or a reference. A brief explanation of the 1/6 factor would improve reproducibility.
[Section II, 'Fundamental soliton'] The sentence 'The two solvers are mutually consistent at the 10^-3 level' conflates agreement with the reference with accuracy. Since the reference is itself a numerical SIP result, 'mutually consistent' is the correct wording; consider using it throughout.
[Section II, 'Numerical Benchmarking and Performance'] The statement that RK45 tolerances are 'fixed relative tolerance rtol=10^-9 and absolute tolerance atol=10^-11' for both methods is useful, but it would be helpful to report the actual number of accepted steps or the ranges of step sizes, since stiffness differences affect wall-clock time.
[References] Reference [15] is listed as (2026), which is plausible for the arXiv version but should be updated with a journal reference if available. Also, the arXiv number [13] for Desbrun et al. is math/0508341; verify it is the published version cited.
[Section II, 'Elliptic optical vortex'] The discussion of the vortex benchmark's larger speedup is thoughtful and correctly attributes it to interface sampling rather than field smoothness. However, the three-point explanation would be more convincing if a quantitative mode-overlap analysis were provided.

Circularity Check

1 steps flagged · score 4.0 of 10

Soliton shape-preservation benchmark is circular: the launch field is constructed as a discrete eigenmode of the propagation operator, so the reported near-zero error is by construction; the main >100x speedup claim rests on the separate, non-tautological SIP-relative benchmarks.

fitted input called prediction [Section II, 'Fundamental soliton' benchmark (paragraph introducing Figs. 4 and 5)]
"Since the soliton is a stationary solution of the NLSE, any distortion of its shape during propagation provides a direct measure of numerical error. This example therefore serves as a stringent test of the DEC method. ... Each soliton is computed self-consistently on the same grid or mesh subsequently used for its propagation, so that the launch field is a discrete eigenmode of the operator that evolves it."

The benchmark input is not an independent initial condition; it is obtained by solving the discrete eigenproblem for the exact operator L+N used in the propagation equation i∂Ψ/∂z = LΨ + N(|ψ|²)Ψ (Eq. 10). By construction, a discrete eigenmode evolves with constant intensity up to the RK45 time-integration error, so the reported 'shape-preservation' error near 10⁻⁶ is the residual of the eigen-solver plus integrator error, not a test of whether DEC accurately reproduces the continuum soliton. The paper itself states that the launch field is a discrete eigenmode of the evolving operator, making the 'stringent test' claim tautological. The SIP comparison is likewise confounded because the SIP soliton is also self-consistent on its own grid.

full rationale

The derivation of the DEC operators is a standard, parameter-free construction: the discrete exterior derivative is combinatorial, the Hodge stars are geometric (diagonal H0, Whitney-form H1), and the NLSE discretization L = -H0^{-1}(D0^T H1 D0 + H0(V)), N = -H0^{-1}H0(χ|ψ|²) follows directly without fitting constants to a target result. No load-bearing self-citation or imported uniqueness theorem was found; the citation to the authors' prior DEC-SIE work (ref. 15) supports a standard primal-dual mesh concept and is not load-bearing. The main 'over 100x' speedup claim is supported by the elliptic-Gaussian (73x) and vortex (150x) benchmarks, which compare DEC against a common SIP reference and are not derivation-tautological, although the SIP reference's own Gibbs error at the step-index interface is a correctness risk rather than a circularity. However, the soliton benchmark, which is presented as a 'stringent test' of accuracy and long-term stability, is circular: the launch field is computed self-consistently as a discrete eigenmode of the very operator used for propagation, so its intensity invariance is guaranteed by construction. This reduces one of the three claimed confirmations of 'spectral-level accuracy' to a fitted initial condition, meriting a score of 4: partial circularity in a supporting benchmark while the central speedup claim retains independent content.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method has no fitted constants and introduces no new physical entities. The free parameters are benchmark artifacts (accuracy thresholds) plus a normalization choice. The main axioms are the physical model, the convergence of the discrete operators, and the specific validation choices made in the test cases.

free parameters (2)

Normalization constants p1, qL, qNL = set to 1
Chosen for convenience in Eq. (1); a scaling choice, not fitted to data.
Benchmark accuracy thresholds = E_rel = 5e-3 (Gaussian), 8e-3 (vortex)
The reported speedups (73x, 150x) are quoted at hand-picked error thresholds; different thresholds would give different speedups, affecting the headline 'over 100x' claim.

assumptions (4)

domain assumption The NLSE with a step-index potential and Kerr nonlinearity accurately models light propagation in the tested fiber
The benchmarks and performance claims depend on this physical model being a valid description of the experimental system (Section II, test cases).
standard math DEC with Whitney-form Galerkin Hodge star yields a convergent discrete Laplacian on unstructured meshes
Relies on established results cited as [11,12,19]; the paper applies them without re-derivation (Section II, Discrete Hodge star).
ad hoc to paper The soliton launch field is a discrete eigenmode of the same discrete evolution operator
Each soliton is 'computed self-consistently on the same grid or mesh subsequently used for its propagation,' so the long-distance test measures internal consistency rather than absolute error against the continuum problem (Section II, Fundamental soliton).
domain assumption Neumann boundaries do not perturb the short-distance benchmarks
For z=0.1 with a radius-4 domain, radiated light has not yet reached the boundary; the paper acknowledges this limitation and extends the Gaussian case with a larger window in Supplementary Note 6 (Section II, final remark).

how reviews work

0 comments

Cite this review

Pith. "Pith review of A 100x Faster Beam Propagation Method for Nonlinear Optical Wave Propagation Based on Discrete Exterior Calculus." pith.science (2026). https://pith.science/paper/SUSXT4XM

@misc{pith2026260803587,
  author       = {Pith},
  title        = {Pith review of: A 100x Faster Beam Propagation Method for Nonlinear Optical Wave Propagation Based on Discrete Exterior Calculus},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SUSXT4XM}},
  note         = {Machine review of arXiv:2608.03587}
}

read the original abstract

Efficient simulation of nonlinear light propagation in complex photonic structures remains a major challenge because these systems combine intricate transverse geometries with propagation over distances spanning many diffraction lengths. Existing numerical methods often require computationally intensive uniform discretizations or struggle to accurately represent complex material boundaries, limiting the practical simulation of multiscale nonlinear photonic devices. Here we introduce a computational framework for solving the nonlinear Schr\"odinger equation that accelerates simulations by more than two orders of magnitude (over 100x) compared with conventional approaches while maintaining spectral-level accuracy, thereby reducing the computational time required to solve large-scale and multiscale problems from days to hours. The method is based on discrete exterior calculus, enabling geometry-conforming discretization directly on unstructured meshes without the weak formulations required by conventional finite-element methods. In contrast to Fourier-based spectral solvers, it avoids global oversampling, eliminates Gibbs-type oscillations at material interfaces, naturally incorporates absorbing boundary conditions, and preserves the topological structure of the underlying differential operators. A key feature of the framework is that higher-order propagation operators are generated algorithmically from lower-order discrete operators, providing a systematic route to extending simulations beyond the standard nonlinear Schr\"odinger equation. Benchmarks on fundamental solitons, asymmetric beams, and optical vortices confirm spectral-level accuracy while demonstrating computational speedups exceeding 100x.

Figures

Figures reproduced from arXiv: 2608.03587 by the authors.

**Figure 2.** FIG. 2. (a) A scalar function [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** FIG. 3. (a) Geometric representation of a two-dimensional primal simplicial complex and its barycentric dual. The top panel shows a local [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: FIG. 4. Initial intensity profile of soliton solutions obtained via the [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 6.** Figure 6: By intentionally breaking the cylindrical symmetry of [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

**Figure 5.** Figure 5: FIG. 5. Fundamental soliton propagation in a step-index fiber. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

**Figure 7.** Figure 7: FIG. 7. Performance benchmark for the off-axis elliptic Gaussian [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

**Figure 8.** Figure 8: FIG. 8. Visual comparison of the computed transverse intensity profiles ( [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

**Figure 9.** Figure 9: FIG. 9. Initial transverse intensity profiles ( [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

**Figure 11.** Figure 11: FIG. 11. Visual comparison of the computed transverse intensity profiles ( [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

26 extracted references · 25 canonical work pages

[1]

Weideman, J. A. C. & Herbst, B. M. Split-step methods for the solution of the nonlinear Schr¨odinger equation.SIAM J. Numer. Anal.23, 485 (1986)

work page 1986
[2]

Taha, T. R. & Ablowitz, M. I. Analytical and numerical aspects of certain nonlinear evolution equations. II. Numerical, nonlin- ear Schr¨odinger equation.J. Comput. Phys.55, 203 (1984)

work page 1984
[3]

N.Spectral Methods in MATLAB(SIAM, 2000)

Trefethen, L. N.Spectral Methods in MATLAB(SIAM, 2000)

work page 2000
[4]

Zouraris, G. E. On the convergence of a linear two-step finite element method for the nonlinear Schr¨odinger equation.ESAIM Math. Model. Numer. Anal.35, 389 (2001)

work page 2001
[5]

& Zhang, Z

Li, P. & Zhang, Z. Efficient finite element methods for semiclas- sical nonlinear Schr ¨odinger equations with random potentials. ESAIM Math. Model. Numer. Anal.59, 3249 (2025)

work page 2025
[6]

& Zhang, Z

Chen, J., Li, S. & Zhang, Z. Efficient multiscale methods for the semiclassical Schr¨odinger equation with time-dependent poten- tials.Comput. Methods Appl. Mech. Eng.369, 113232 (2020)

work page 2020
[7]

Deschamps, G. A. Electromagnetics and differential forms. Proc. IEEE69, 676 (1981)

work page 1981
[8]

Katz, V . J. Differential forms—Cartan to de Rham.Arch. Hist. Exact Sci.33, 321 (1985)

work page 1985

Show all 26 references

[9]

Flanders, H.Differential Forms with Applications to the Physi- cal Sciences2nd edn (Dover Publications, 1989)

1989
[10]

Bossavit, A.Computational Electromagnetism: Variational Formulations, Complementarity, Edge Elements(Academic Press, 1998)

1998
[11]

& Bossavit, A

Tarhasaari, T., Kettunen, L. & Bossavit, A. Some realizations of a discrete Hodge operator: a reinterpretation of finite element techniques.IEEE Trans. Magn.35, 1494 (1999)

1999
[12]

N.Discrete Exterior Calculus

Hirani, A. N.Discrete Exterior Calculus. PhD thesis, California Institute of Technology (2003)

2003
[13]

N., Leok, M

Desbrun, M., Hirani, A. N., Leok, M. & Mars- den, J. E. Discrete exterior calculus. Preprint at https://arxiv.org/abs/math/0508341 (2005)

2005 arXiv
[14]

Grady, L. J. & Polimeni, J. R. Introduction to discrete calculus. InDiscrete Calculus: Applied Analysis on Graphs for Compu- tational Science13–89 (Springer, 2010)

2010
[15]

& Gomez, L

Abdrabou, A. & Gomez, L. J. A hybrid DEC-SIE framework for potential-based electromagnetic analysis of heterogeneous media.J. Comput. Phys.553, 114726 (2026)

2026
[16]

N., Kalyanaraman, K

Hirani, A. N., Kalyanaraman, K. & VanderZee, E. B. Delaunay Hodge star.Comput.-Aided Des.45, 540 (2013)

2013
[17]

& M ¨oller, T

Dyer, R., Zhang, H. & M ¨oller, T. V oronoi–Delaunay duality and Delaunay meshes. InProc. ACM Symposium on Solid and Physical Modeling415–420 (ACM, 2007)

2007
[18]

Press, 1957)

Whitney, H.Geometric Integration Theory(Princeton Univ. Press, 1957)

1957
[19]

S., Hirani, A

Mohamed, M. S., Hirani, A. N. & Samtaney, R. Comparison of discrete Hodge star operators for surfaces.Comput.-Aided Des. 78, 118 (2016)

2016
[20]

& Kettunen, L

Lohi, J. & Kettunen, L. Whitney forms and their extensions.J. Comput. Appl. Math.393, 113520 (2021)

2021
[21]

The definition of molecular resonance curves by the method of exterior complex scaling.Phys

Simon, B. The definition of molecular resonance curves by the method of exterior complex scaling.Phys. Lett. A71, 211 (1979)

1979
[22]

& Kosloff, D

Kosloff, R. & Kosloff, D. Absorbing boundaries for wave prop- agation problems.J. Comput. Phys.63, 363 (1986)

1986
[23]

Riss, U. V . & Meyer, H.-D. Investigation on the reflection and transmission properties of complex absorbing potentials.J. Chem. Phys.105, 1409 (1996)

1996
[24]

Manolopoulos, D. E. Derivation and reflection properties of a transmission-free absorbing potential.J. Chem. Phys.117, 9552 (2002)

2002
[25]

G., Palao, J

Muga, J. G., Palao, J. P., Navarro, B. & Egusquiza, I. L. Com- plex absorbing potentials.Phys. Rep.395, 357 (2004)

2004
[26]

Dormand, J. R. & Prince, P. J. A family of embedded Runge– Kutta formulae.J. Comput. Appl. Math.6, 19 (1980)

1980

Pith tools

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

[1] [1]

Weideman, J. A. C. & Herbst, B. M. Split-step methods for the solution of the nonlinear Schr¨odinger equation.SIAM J. Numer. Anal.23, 485 (1986)

work page 1986

[2] [2]

Taha, T. R. & Ablowitz, M. I. Analytical and numerical aspects of certain nonlinear evolution equations. II. Numerical, nonlin- ear Schr¨odinger equation.J. Comput. Phys.55, 203 (1984)

work page 1984

[3] [3]

N.Spectral Methods in MATLAB(SIAM, 2000)

Trefethen, L. N.Spectral Methods in MATLAB(SIAM, 2000)

work page 2000

[4] [4]

Zouraris, G. E. On the convergence of a linear two-step finite element method for the nonlinear Schr¨odinger equation.ESAIM Math. Model. Numer. Anal.35, 389 (2001)

work page 2001

[5] [5]

& Zhang, Z

Li, P. & Zhang, Z. Efficient finite element methods for semiclas- sical nonlinear Schr ¨odinger equations with random potentials. ESAIM Math. Model. Numer. Anal.59, 3249 (2025)

work page 2025

[6] [6]

& Zhang, Z

Chen, J., Li, S. & Zhang, Z. Efficient multiscale methods for the semiclassical Schr¨odinger equation with time-dependent poten- tials.Comput. Methods Appl. Mech. Eng.369, 113232 (2020)

work page 2020

[7] [7]

Deschamps, G. A. Electromagnetics and differential forms. Proc. IEEE69, 676 (1981)

work page 1981

[8] [8]

Katz, V . J. Differential forms—Cartan to de Rham.Arch. Hist. Exact Sci.33, 321 (1985)

work page 1985