Skip to content

Reproductions

Self-contained, validated reproductions of results from the literature. Each folder under reproductions/ contains a parameters.json, a reproduce.py that asserts its result against an analytic/closed-form reference, and a README.md with the ground truth and outcome. Run them all with:

python -m pytest tests/test_reproductions.py
Reproduction Reference Stack Headline outcome
eftekhar_2019_parametric_cascades Eftekhar et al., Nat. Commun. 10, 1638 (2019) multimode_gnlse (3 GRIN modes) + chunked taper ⚠️ MFD-oscillation periods match the window-avg L_si to 0.1 % / 2.4 %; acceleration ratio x3.33 vs window model x3.25….
stolen_lin_1978_spm Stolen & Lin, Phys. Rev. A 17, 1448 (1978) GNLSE (Kerr) + pulse ✅ spectrum matches closed form to ~1e-13; peak-count rule exact.
macleod_quarter_wave_dbr Macleod, Thin-Film Optical Filters; Born & Wolf (textbook) dbr TMM ✅ peak reflectance matches exact closed form to <1e-6; stopband width within 7%.
shg_textbook Boyd, Nonlinear Optics (3rd ed.), Ch. 2; Fejer et al., IEEE JQE 28, 2631 (1992) chi2 (RK4IP) ✅ η matches tanh²(κL) to ≤8e-15; first-order QPM recovers tanh²((2/π)κL) to 0.12%.
gordon_1986_ssfs Gordon, Opt. Lett. 11, 662 (1986) GNLSE + Raman + soliton ✅ measured +1.08 nm / 20 m vs Gordon +0.91 nm (ratio 1.19).
dudley_2006_cherenkov_dw Akhmediev & Karlsson, Phys. Rev. A 51, 2602 (1995) phase_matching + GNLSE ✅ dispersive_wave_roots = analytic 699.3 nm.
dudley_2006_scg Dudley, Genty & Coen, Rev. Mod. Phys. 78, 1135 (2006) GNLSE + Raman + phase_matching + soliton ✅ 9 figure scripts (Figs. 3, 4, 5, 6, 7, 8, 9, 10, 23).
kuznetsov_ma_2012_breather Kibler et al., Sci. Rep. 2, 463 (2012) GNLSE (NLSE) + phase_matching ✅ exact Kuznetsov–Ma soliton reproduced over one period: peak 7.6129 W vs 7.6130 W, intensity L2 9e-4, T₀/z_p exact.
narhi_2016_mi_breathers Närhi et al., Nat. Commun. 7, 13675 (2016) GNLSE (NLSE) + phase_matching + SplitStepEngine ✅ MI sideband 46.41 GHz = paper 46.4; exact Peregrine ratio 8.9996 (theory 9).
tomlinson_1985_wave_breaking Tomlinson, Stolen & Johnson, Opt. Lett. 10, 457 (1985) GNLSE (NLSE) ✅ steep edges → flat top → oscillations; steepening onset at 0.57 z_WB; scaling z ∝ P₀^(−0.54) (theory −0.5) with….
marcuse_menyuk_wai_1997_manakov_pmd Marcuse, Menyuk & Wai, JLT 15, 1735 (1997) vector_gnlse (manakov coupling + RandomBirefringenceEngine) ✅ Eq. (30) 8/9 law exact (ratio 9/8, shapes invariant to 7×10⁻⁴ over 30 z₀).
renninger_wise_2013_grin_solitons Renninger & Wise, Nat. Commun. 4, 1719 (2013) multimode_gnlse (3 GRIN modes, isotropic tensors) ✅ temporal locking 0.06 of linear walk-off.
menyuk_1987_birefringent_pulses Menyuk, IEEE JQE QE-23, 174 (1987) vector_gnlse (incoherent 2/3 XPM + coherent FWM branch) ⚠️ Eq. (9)/(10) soliton filaments at machine precision (shape L2 7×10⁻⁶); coherent-FWM filament deviation O(1/Rδ) as….
krupa_2019_multimode Krupa et al., APL Photonics 4, 110901 (2019) (GPI: Krupa, PRL 116, 183901 (2016)) multimode_gnlse (modal GNLSE + phase_offsets grating) ✅ ξ = 0.6157 mm / f_m = 124.98 THz (PRL 0.615 / 125.0); √h·f_m ladder to 0.5 % for h ≤ 3 (h = 1..5 measured); f₁ ×4-power….
mumtaz_2013_multimode_jlt Mumtaz, Essiambre & Agrawal, JLT 31, 398 (2013) (spec source of v2 multimode Eq. 12/29) multimode_gnlse (Manakov xpm_weights) + Eq. 6 stochastic harness ✅ Table-II walk-offs exact (≤3e-6).
wai_menyuk_1991_random_birefringence_solitons Wai, Menyuk & Chen, Opt. Lett. 16, 1231 (1991) vector_gnlse (RandomBirefringenceEngine, Wai-1991 Eq.-(2) rotation law) ⚠️ Fig. 1 shadow peak ratio 1.10 vs Eq. (5).
wright_2015_self_organized_instability Wright et al., Nat. Photon. 10, 471 (2016 online 2015) multimode_gnlse (FWM Jacobian + fwm_weights; the oam_l gate is inert here — all labels 0) + phase_offsets ✅ check A: analytic STMI ladder vs digitized Fig. 3d circles to −5.2/−0.9/−2.8 % (tol 15 %); check-B' probe: engine….
guasoni_2015_generalized_mi_multimode Guasoni, Phys. Rev. A 92, 033849 (2015) multimode_gnlse (XPM-coupled deck) + Eq.-(8)/(9) eigen solver ✅ Fig. 3 anchor to 0.7 % (g₁ = 0.9071 / g₂ = 0.7070 vs paper 0.90 / 0.71); band structure asserted — the local unsaturated gain gives band/edge 4.13 at z = 0.10 m (the flat end-to-end readout was saturation).
raissi_2019_pinn_nlse Raissi, Perdikaris & Karniadakis, J. Comput. Phys. 378, 686 (2019) reproduction-local PINN (torch float64, Adam + L-BFGS) + SplitStepEngine data ✅ §I Schrödinger example: PINN rel-L2 3.62e-3 on |h| (the magnitude the paper's Fig. 2 plots) / 1.52e-3 on |h|² vs paper 1.97e-3; the complex-field 6.1e-3 is kept as the conservative number — the "3× gap" was a metric mismatch, not a defect.
dw_timing_gas_hollowcore Brahms & Travers, arXiv:2101.04014 (2021) TaperedGNLSESolver gas β(ω,z) (Marcatili–Schmeltzer + Boerzsoenyi He) + RDW arrival-time statistics ⚠️ REPRODUCED 2026-09-30 (plasma-free subset): transmission 86.8 % vs paper ~87 %.
kibler_2010_peregrine Kibler et al., Nat. Phys. 6, 790 (2010) GNLSE (β₂-only) + breathers (general_sfb AB, peregrine_soliton) ✅ REPRODUCED 2026-09-30: engine tracks the analytic AB pointwise over the growth leg (peak ≤ 2 %); a = 0.42….
hult_2007_rk4ip Hult, JLT 25, 3770 (2007) RK4IPIntegrator + GNLSEOperator (from heidt_adaptive.py) ⚠️ REPRODUCED 2026-10-01, both decks. Deck B keeps the house two-exponential silica Raman model rather than the paper's Hollenbeck–Cantrell modal sum (recorded deviation, shown non-limiting).
shg_lnoi_shg Wang et al., Opt. Express 25(6), 6963 (2017) — citation unverified chi2 mode-overlap (shg_coupling_overlap, pgln_overlap) + materials.db LiNbO₃ extraordinary Sellmeier ⚠️ Partial: overlap machinery on the article geometry with scalar toy modes — g = 203.4 vs 77.4 and g' = 9.93 vs 34.5 are convention gaps; Boyd-normalised σ = 1331.3 vs 1330.8 (0.03 %).
heidt_2009_adaptive_step Heidt, JLT 27, 3984 (2009) reproduction-local adaptive-step layer (heidt_adaptive.py: SSFIntegrator/RK4IPIntegrator, LocalErrorStepper/CQEStepper, counted-FFT cost) + GNLSEOperator ⚠️ REPRODUCED 2026-10-01, all 16 checks green in 290 s.
huang_202x_pcgnlse_attractors Huang et al., arXiv:2607.05244 (2026) SplitStepEngine(conserving_shock=True) + signed γ + folder-local dimensionless pcGNLSE/moment-ODE layer ⚠️ REPRODUCED partly vs recorded caveats (2026-10-02): ENGINE SIGN BENCH green — pcGNLSE γ>0 redshift path identical to….
poletti_2008_multimode Poletti & Horak, J. Opt. Soc. Am. B 25, 1645 (2008) multimode_gnlse (Eq. 7 fwm_weights / xpm_weights) + folder-local LP mode solver / overlap-integral layer ✅ Sec. 4.A symmetry analysis, all rules emerging from the Eq. (7) quadrature: Eq. (18) 1360 survivors/type (forbidden entries 7.8e-17), Eq. (19) 1360 → 340, the eleven Eq. (16) identities to 2.2e-16, Sec. 5 saving of 31–47× on the O(M⁴N) term, and the Eq. (14) polarisation closure on the engine (opposite-σ power exactly 0 vs 1.09 for an isotropic control). ⚠️ The paper's Fig. 1 counts need a vectorial mode solver — folder README.

✅ = the paper's claim is reproduced within its stated tolerance. ⚠️ = reproduced, with a deviation or an out-of-scope arm recorded in the folder README.md and in the repository ISSUES.md register.

Reproducing the papers also surfaced real bugs (inverted GNLSE dispersion sign, Raman response direction); see the repository REPORT.md for the full list.