{"id":"f12dd5a8-e87d-4a9e-9008-7dc146cd4ea2","arxiv_id":"2607.06707","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":3.5,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In 2D FDTD, choosing Courant number Sc equal to the relative refractive index nr eliminates numerical dispersion and enables stable simulation of media with nr < 1 and left-handed media.","lead":"Setting the Courant number equal to the medium refractive index removes numerical dispersion in 2D FDTD for homogeneous nondispersive media, including sub-unity and negative-index cases. Practitioners modeling exotic or low-density media get a simple stability and accuracy rule that standard vacuum-tuned FDTD lacks.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Proposition 3 cancels monochromatic phase error, but the paper never shows that broadband packets remain free of numerical dispersion under the same Sc=nr.","rationale":"The Reader correctly flags the missing rigorous stability proof for Proposition 4 and the unexplained backward pulse; both are genuine limitations. The more load-bearing soft spot for the central claim, however, is the unjustified leap from monochromatic phase-velocity cancellation (Proposition 3) to the assertion that numerical dispersion vanishes for arbitrary broadband signals. That leap is what the abstract and Corollary 4 advertise as the main result. The monochromatic algebra is textbook and solid; the broadband extension is only visually supported. A single two-tone propagation test would settle whether residual dispersion remains. Because the monochromatic core still holds and the numerical illustrations are consistent with it, the Reader’s CONDITIONAL verdict is unchanged; the stability and artifact issues remain secondary conditions that should also be addressed.","tokens_in":13920,"tokens_out":561,"duration_ms":7279,"concrete_test":"Propagate a linear superposition of two monochromatic plane waves whose frequencies differ by a factor of two (both well-resolved, Nλ≥20) under Sc=nr for a fixed nr\neq1; measure the relative phase and envelope distortion after a propagation distance of at least 50 wavelengths. If either quantity grows systematically with distance, the broadband claim fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim rests on Proposition 3: substituting Sc=nr into the monochromatic ratio (27) yields ~cp/cp=1 for every Nλ. That algebra is correct for a pure plane wave of fixed ω. The paper then asserts (Corollary 4 and the abstract) that the same choice “eliminates the numerical dispersion” for signals of any shape and spectral composition, and illustrates this only with two broadband packets (Gaussian and Ricker) in Figs. 5 and 7. Because the Yee dispersion relation (17) is nonlinear in frequency, exact cancellation of phase velocity at every frequency does not automatically guarantee that a finite-bandwidth packet retains its shape; residual group-velocity or amplitude errors could still accumulate. The paper supplies no analytic argument that the full discrete transfer function becomes nondispersive for a continuum of frequencies when Sc=nr, nor any quantitative measure (e.g., L2 shape error versus propagation distance) that would confirm the claim beyond visual inspection of two particular waveforms. If that gap is real, the extension from monochromatic cancellation to “any spectral composition” is unsupported, and the optimality statement is overstated.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"The manuscript studies numerical dispersion of the 1D Yee FDTD scheme for nondispersive homogeneous media with refractive index nr = √(εr μr) ≠ 1. Using shift operators, it derives the discrete dispersion relation (17) and the monochromatic phase-velocity ratio (27). Proposition 3 shows that the algebraic choice Sc = nr makes ˜cp/cp = 1 for every wavelength discretization Nλ, eliminating monochromatic phase error. Numerical examples with Gaussian and Ricker pulses (Figs. 5, 7) and a left-handed medium (Fig. 8) illustrate that the same choice also works for broadband packets and for media with nr < 1 or εr = μr = −1. Proposition 4 asserts that the scheme (2)–(5) diverges for Sc > nr, supported by a numerical illustration (Fig. 6). The authors conclude that Sc = nr is the unique optimal Courant number for the class of media considered.","tokens_in":14138,"tokens_out":1172,"duration_ms":12351,"significance":"If the claims hold, the paper supplies a simple, parameter-free rule that removes the usual numerical-dispersion limitation of Yee FDTD in homogeneous media and simultaneously extends a standard update scheme to optically less dense and left-handed media. The monochromatic derivation (Lemma 1, Propositions 1–3) is clean and the algebraic cancellation Sc = nr is exact, not fitted. The numerical illustrations for nr < 1 and for negative-index media are of practical interest and are rarely shown with ordinary Yee updates. These strengths make the work potentially useful for FDTD practitioners, provided the broadband and stability gaps identified below are closed.","major_comments":[{"comment":"Proposition 3 and the subsequent claim in Corollary 4 / the abstract that Sc = nr “eliminates the numerical dispersion … for signals of any shape and spectral composition” rest only on monochromatic cancellation of (27). Because the discrete dispersion relation (17) is nonlinear in frequency, exact phase-velocity matching at every ω does not automatically guarantee that a finite-bandwidth packet retains its shape (group-velocity or amplitude errors may remain). The paper supplies no analytic argument that the full discrete transfer function becomes nondispersive for a continuum of frequencies, nor any quantitative shape-error metric (e.g., L2 residual versus propagation distance). The visual agreement of two particular waveforms (Gaussian, Ricker) in Figs. 5 and 7 is insufficient to support the general claim.","section":"Proposition 3, Corollary 4, Abstract"},{"comment":"Proposition 4 asserts that the update scheme (2)–(5) diverges whenever Sc > nr and is therefore the unique stability boundary that makes Sc = nr optimal. The only support offered is a single numerical illustration (Fig. 6) and an argument “at the physical level of rigor.” No von Neumann, matrix, or energy stability analysis is given. Without such a proof the forbidden region B in Fig. 4 and the uniqueness statement of Corollary 4 remain incompletely justified; a mesh-dependent or source-dependent stability boundary would weaken the optimality claim.","section":"Proposition 4, Fig. 6, Corollary 4"}],"minor_comments":[{"comment":"The paper is written for a 1-D spatial grid (one spatial + one temporal dimension) yet repeatedly refers to the “2D case.” Standard FDTD terminology reserves “2D” for two spatial dimensions; the wording should be corrected to avoid confusion.","section":"Abstract, Introduction, Eq. (1)"},{"comment":"Remark 11 and Fig. 7 document a spurious backward pulse Pb whose amplitude grows as nr \to 0. The phenomenon is left unexplained and unquantified; a short discussion of its origin (TF/SF source, floating-point, or analytic inconsistency) and a bound on its relative amplitude would strengthen the practical utility of the method.","section":"Remark 11, Fig. 7"},{"comment":"Several figures (especially Figs. 1, 2, 5) lack axis labels or legends that make the Courant number and refractive index immediately readable; adding them would improve reproducibility.","section":"Figs. 1, 2, 5"},{"comment":"The open questions listed at the end are valuable, but the first two (accuracy for nr < 1 and origin of Pb) are already partially answerable from the material in the paper and could be addressed briefly rather than left entirely open.","section":"Some open questions"}],"recommendation":"major_revision","confidential_remarks":"The monochromatic algebra is correct and the numerical examples for nr < 1 and left-handed media are interesting; the manuscript is therefore salvageable. The two load-bearing gaps (broadband justification and stability proof) are standard FDTD issues that the authors should be able to close with a short analytic argument or a systematic numerical study. I would not reject on novelty grounds alone, but the paper currently overstates the generality of its main claim."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful takeaway is simple: set the Courant number equal to the medium refractive index and the classical Yee phase-velocity error vanishes for every Nλ. That is an immediate algebraic consequence of the standard dispersion relation, but the paper does the service of writing it cleanly, showing the figures for nr > 1, nr < 1, and simultaneous negative ε, μ, and stating the practical limits.\n\nWhat is actually new is modest but real. The monochromatic cancellation (Prop. 3) is textbook once you adopt the medium-based Courant definition some sources already use. The explicit demonstrations for optically thinner media and a left-handed snapshot (Fig. 8), plus the clear warning that Sc > nr produces rapid self-excitation (Fig. 6), are the parts that go beyond the usual vacuum “magic step” discussion. The shift-operator derivation of the dispersion relation is clean and the numerical illustrations for Gaussians and Ricker wavelets look correct inside the stated domain.\n\nSoft spots are real but proportional. Prop. 4 (divergence for Sc > nr) rests only on a numerical example and a “physical-level” argument; a von Neumann or energy proof is missing, so the uniqueness claim and the forbidden region B are not fully closed. The stress-test concern about broadband packets is partly right: exact monochromatic phase matching does not automatically guarantee shape preservation for a continuum of frequencies, and the paper offers only visual inspection of two waveforms rather than an L2 error or group-velocity analysis. Still, when Sc = nr the discrete relation collapses to the continuous one for every frequency, so residual numerical dispersion should be second-order truncation, not the usual Yee phase lag; the claim is therefore overstated in wording but not empty. The unexplained backward pulse for small nr is a genuine open artifact the authors themselves flag.\n\nThis is for people who run FDTD on homogeneous dielectrics, low-index media, or simple left-handed slabs and want a better default time step. It is not a general theory for inhomogeneous or dispersive media. I would send it to peer review: the math is transparent, the numerics are reproducible, and the practical recommendation is useful once the stability gap and the broadband wording are tightened. Worth a look if you care about FDTD defaults; not a must-cite for everyone.","headline":"Solid applied note that Sc = nr cancels monochromatic phase error and works for nr < 1 and left-handed media; core algebra is textbook, stability and broadband claims are only partially supported.","tokens_in":14815,"tokens_out":567,"would_cite":false,"duration_ms":7280,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Setting the Courant number equal to the refractive index removes numerical dispersion from 2D FDTD wave simulations and works for rarefied and left-handed media.","keywords":["FDTD","Courant number","numerical dispersion","Yee grid","nondispersive media","negative refraction","left-handed media","phase velocity"],"falsifier":"Run the identical Yee update (2)–(5) with Sc slightly larger than nr on a sequence of successively refined grids; if the solution remains bounded for some meshes while nr is held fixed, the asserted unique optimality of Sc = nr fails.","tokens_in":14778,"feed_emoji":"∿️","tokens_out":623,"duration_ms":7921,"temperature":0.7,"pith_summary":"Standard Yee FDTD calculations of electromagnetic pulses in homogeneous nondispersive media produce artificial numerical dispersion whenever the Courant number is left at the vacuum value Sc = 1. This paper derives the exact numerical dispersion relation on the staggered grid and shows that the phase velocity on the grid becomes identical to the physical phase velocity for every wavelength discretization once Sc is set equal to the relative refractive index nr = √(εr μr). That single choice therefore eliminates numerical dispersion, restores correct pulse shapes, and simultaneously stabilizes the update scheme for media optically thinner than vacuum as well as for left-handed media with simultaneously negative permittivity and permeability. The result is presented as four statements that fix both the optimal Courant number and the hard stability boundary Sc ≤ nr.","feed_headline":"Magic Courant number cancels FDTD numerical dispersion","feed_subtitle":"Set Sc equal to the refractive index and pulses stay clean even in rarefied or left-handed media","key_machinery":"The closed-form numerical dispersion relation (Proposition 1) obtained by substituting plane-wave modes into the staggered Yee updates, which yields the phase-velocity ratio (27) whose only free parameter that can be tuned to unity is the Courant number Sc.","core_discovery":"For any real relative permittivity and permeability belonging to the paper's domain of definition, the choice Sc = √(εr μr) makes the Yee-grid phase velocity coincide exactly with the continuous-space phase velocity independently of the wavelength sampling Nλ, thereby removing numerical dispersion; the same choice is the unique optimal value that keeps the discrete algorithm stable and extends it to nr < 1 and to left-handed media.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Optimal Courant number eliminates FDTD numerical dispersion","Set Sc to refractive index for exact FDTD phase velocity","Sc equal to n removes numerical dispersion for any media","Courant Sc=√(εrμr) cancels Yee-grid numerical dispersion","Sc=n extends FDTD cleanly to nr<1 and left-handed media"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The claim that the scheme always diverges when Sc exceeds nr rests only on numerical illustrations and a physical-level argument, not a complete discrete stability proof.","fun_headline_variants_meta":{"raw":{"variants":["Optimal Courant number eliminates FDTD numerical dispersion","Set Sc to refractive index for exact FDTD phase velocity","Sc equal to n removes numerical dispersion for any media","Courant Sc=√(εrμr) cancels Yee-grid numerical dispersion","Sc=n extends FDTD cleanly to nr<1 and left-handed media"]},"model":"grok-4.5","effort":"low","cost_usd":0.005908,"raw_usage":{"total_tokens":1502,"prompt_tokens":680,"num_sources_used":0,"completion_tokens":94,"cost_in_usd_ticks":59080000,"prompt_tokens_details":{"text_tokens":680,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":728,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":680,"tokens_out":94,"duration_ms":17365,"temperature":1.0,"reasoning_tokens":728,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-10T22:57:54.702365+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Run the identical Yee update (2)–(5) with Sc slightly larger than nr on a sequence of successively refined grids; if the solution remains bounded for some meshes while nr is held fixed, the asserted unique optimality of Sc = nr fails.","supporting_citations":[],"review_version":1}