{"id":"af1cf42f-0163-486e-a6bf-bd9699d0afd2","arxiv_id":"2505.04504","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"A new propagation formalism for atomic vapors in oblique magnetic fields properly handles non-orthogonal electric field modes and removes unphysical spectral features.","lead":"This paper corrects how light is modeled when it passes through rubidium vapor in a magnetic field that points at an angle to the laser, because the two allowed electric field modes are not perpendicular. The correction removes spurious features in simulations and matches measured transmission spectra, which matters for designing better atomic filters and magnetometers.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The coefficients in Eq. (11) are derived via the full 3D overlap matrix S even though the incident field is purely transverse; the paper bounds the resulting error by asserting m_i,z < 10^-3 without demonstrating this at the resonant detunings where the mode overlap is largest, so the quantitative…","rationale":"I agree with the reader that the treatment of the longitudinal mode components is the weakest point in the derivation. The paper is transparent: it states in Section 2.2 that Eq. (8) is valid only when m_i,z is negligible, and it reports a numerical bound of 10^-3 for the systems studied. However, the central claim is not merely qualitative; the paper advertises 55-57% transmission errors in the old model and the removal of 'unphysical features.' These large differences occur at detunings where the mode overlap is maximal, and it is exactly there that the coefficient calculation is most sensitive: the overlap matrix S approaches singularity, and any error imported through the longitudinal components is amplified by the condition number. The paper does not report the condition number or the longitudinal components as a function of detuning, so the magnitude of this amplification is unknown. That said, the effect is formally second order in m_i,z, and even with a condition number of ~10-100 the error would remain below the experimental RMSE. The independent experimental agreement at 0.5-0.7% RMSE across two regimes gives strong support to the central claim. I therefore regard this as a verification gap rather than a demonstrated flaw; the proposed check would close it. Because the concern is real but small and the paper's central claim is well supported by the experiment and the internal logic, I would not change the reader's ACCEPT verdict. We agree that this is the weakest assumption, but it does not, on the present evidence, overturn the conclusion.","tokens_in":21393,"tokens_out":30170,"duration_ms":288005,"concrete_test":"Using the publicly deposited ElecSus code, reproduce the Fig. 4 (Regime I) and Fig. 6 (Regime II) calculations and, at every detuning, output m_i,z, the overlap |<m1|m2>|, and the condition number of S. Then recompute the transmission with the exact transverse-matching coefficients obtained by solving the 2x2 system [m1_x m2_x; m1_y m2_y][c1;c2] = [E_in_x;E_in_y] instead of Eq. (11). If the resulting transmission differs from the paper's Eq. (11) result by less than 0.1% (well below the 0.5-0.7% fit RMSE and far below the 55% model discrepancy), the concern is settled and the central claim stands as stated. If the difference is comparable to the claimed corrections, the boundary treatment needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, namely that the old formalism mispredicts transmission by up to 55-57%, rests entirely on the mode coefficients c1, c2 computed from Eq. (11). Those coefficients are obtained by solving the normal equations with the 3D overlap matrix S = [<m_i|m_j>], while the incident field has no z-component. Because each mode has a small longitudinal component m_i,z, the vector decomposition in Eq. (8) is overdetermined; the paper's least-squares solution differs from the physically required matching of only the transverse E components by an amount of order |m_i,z|^2. The paper asserts m_i,z < 10^-3 of the mode norm (Section 2.2), which would make this error negligible. However, this bound is not demonstrated over the resonant detunings where |<m1|m2>| peaks and where the old/new models differ by ~50%. In that regime S becomes ill-conditioned (its off-diagonal approaches 1), which can amplify the m_i,z-induced error through the condition number of S. If m_i,z grows above 10^-3 near resonance, or if the condition number is large, the coefficients, and thus the headline correction, could be distorted by the approximate boundary treatment rather than being a faithful consequence of mode non-orthogonality. The paper flags the approximation but does not supply the numerical check that would make it safely load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses propagation of weak probe light through an alkali-metal vapor in an external magnetic field whose direction is oblique to the laser wavevector. The authors argue that the two eigenmodes of the wave equation are generally non-orthogonal and frequency dependent, and that existing propagation models (ElecSus 3) neglect this non-orthogonality, producing unphysical transmission features. They derive a revised formalism (ElecSus 4) in which the incident field is decomposed onto the two modes by inverting the overlap (Gram) matrix, each mode propagates with its own complex refractive index, and the transmitted intensity is computed from the reconstructed output field. The formalism is validated with weak-probe transmission spectroscopy of the Rb D2 line in two regimes: Regime I at B = 240 G and theta_B = 80 deg with varying input polarization, and Regime II at B = 2.5 kG with theta_B between 100 and 130 deg and two fixed polarizations. The reported fits give RMSE values of about 0.5-0.7%, and the old formalism is shown to disagree with the new one by up to 55-57% in transmission in regions of strongest mode overlap.","tokens_in":21743,"tokens_out":5726,"duration_ms":62690,"significance":"If the formalism is correct, it fills a genuine gap in the modeling of arbitrary-angle magneto-optics in thermal vapors: previous propagation models used orthogonal-mode assumptions that are exact only in the Faraday and Voigt geometries. The paper shows that the correction can be large, removes unphysical spectral features, and is compatible with careful weak-probe experiments. The correction introduces no new free parameters; it uses the same electric-susceptibility engine as ElecSus 3, and the experimental validation uses physically constrained simultaneous fits with global parameters. The open-source ElecSus implementation, the deposited dataset with a DOI, and the explicit comparison to ElecSus 3 fits under different constraints are additional strengths. The main caveat is that one boundary-condition approximation in Section 2.2 is asserted but not numerically quantified at the resonant detunings where the central quantitative claims are made.","major_comments":[{"comment":"The derivation of the mode coefficients relies on the assertion that the z-components of the modes satisfy m_i,z < 10^-3 of the mode norm, and that the incident field can therefore be decomposed in the transverse plane. This assumption is load-bearing: Eq. (10) solves the normal equations with the full 3D overlap matrix S, while the incident field has only x and y components, so the system is overdetermined unless the z-components are exactly zero. The Euclidean solution used in Eq. (11) is then only an approximation to the physically required matching of the transverse electric-field components, and the error in the coefficients can be amplified by the condition number of S when the modes become nearly parallel. The text states the bound but does not demonstrate it at the detunings where |<m1|m2>| is largest and where ElecSus 3 and ElecSus 4 differ by ~50%. Please add a numerical evaluation of max_i |m_i,z| over the full detuning range and over the theta_B and theta_E values used in Figs. 3-6, together with the condition number of S and a comparison of c_i and the resulting transmission computed with and without including the z-components. This check is needed to confirm that the headline 55-57% disagreement is a consequence of mode non-orthogonality rather than an artifact of the approximate boundary treatment.","section":"§2.2, Eqs. (8)-(11)"},{"comment":"The same z-component approximation is used at the exit interface, where E_out,z = 0 is set by assuming that the longitudinal component reflects off the boundary. This is self-consistent only if the z-component is indeed negligible throughout the cell, including at resonance. Since the modes are frequency dependent and the disagreement between old and new models is largest at resonance, the numerical check requested above should also cover the exit boundary treatment. Without that check, the quantitative claim that the old formalism mispredicts transmission by 55-57% is not fully supported.","section":"§2.2, Eq. (9) and §4"}],"minor_comments":[{"comment":"In the version of the manuscript supplied for review, the matrix in Eq. (4) is rendered with stray non-mathematical characters, making the equation difficult to verify. Please check the typeset file and ensure the matrix elements appear correctly.","section":"Eq. (4)"},{"comment":"The supplemental document contains at least two unresolved citation placeholders shown as \"[? ]\" (in the first paragraph and in the Fig. S3 caption text). These should be replaced with proper references.","section":"Supplemental document, Section 1 and Fig. S3 caption"},{"comment":"The text says that \"since the only variable changed during the experiment was theta_E, each spectrum should have the same atom parameters,\" but then notes that the vapor cell was not actively temperature stabilized and that temperature may vary. Please clarify whether T was fitted as a single global parameter or allowed to vary per spectrum; the description is slightly ambiguous.","section":"§4.1.2"},{"comment":"The caption states that residuals quantify the deviation of data from the ElecSus 4 fit, but the residual panels are not clearly visible in the figure as rendered. Please ensure the residual traces are visible and labeled in the final figure.","section":"Fig. 6 caption"}],"recommendation":"major_revision","confidential_remarks":"The core derivation and experimental validation are credible, and I expect the paper to be publishable after the authors supply the missing numerical check on the z-component approximation. The concern is specific and fixable: quantify max|m_i,z| and the condition number of S at the resonant detunings used in the paper, and show that the 55-57% disagreement is robust to the boundary treatment. No additional experiments are needed. The paper fits the journal's scope and the data availability statement is a strength."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know before reading. The paper catches a real bug in ElecSus, the widely used open-source model for atomic vapor magneto-optics. The old formalism decomposes the incident field onto the two eigenmodes as if they were orthogonal; in an oblique magnetic field they are not. The fix is the inverse overlap matrix, which is textbook but applied here for the first time, and it removes large unphysical transmission features (55–57% disagreement in the strongest cases). The second thing: the experimental validation is careful and convincing. Two regimes, B = 240 G and 2.5 kG, four polarization angles in the first regime, a theta_B sweep in the second, simultaneous global fits with physically reasonable parameters, and RMSE around 0.5–0.7%. That is real evidence.\n\nThe derivation is self-contained and the reasoning is clear. The old model's failure is exactly what you expect when you impose orthogonality on a birefringent, dichroic medium. The supplementary heatmaps over theta_B and theta_E are helpful, and the data are deposited. Citation pattern is fine — ElecSus is their own package but it was validated independently in earlier papers.\n\nSoft spot: the longitudinal field components. The paper states that m_i,z < 10^-3 of the mode norm and then sets E_out,z = 0, but it never demonstrates that bound numerically across the detunings where the overlap matrix peaks and becomes ill-conditioned. The stress-test raises a legitimate point: matching only the transverse E at the interface is not the same as the least-squares 3D projection of Eq. (11), and the difference is O(|m_i,z|^2) amplified by the condition number of S. If that amplification is large, the quantitative disagreement with the old model could be distorted. The good fit to experiment suggests the effect is small, but I would still ask the authors to plot max(|m_i,z|) and cond(S) over the absorption spectrum. The fit itself is not a fatal circularity — parameters are constrained and within measurement error — though it is fitting, not parameter-free prediction.\n\nSummary: solid paper for the atomic vapor optics community, particularly anyone designing filters or magnetometers in arbitrary field geometries. It deserves peer review and, after a modest revision that adds the z-component check, likely acceptance. I would cite it and I'd probably bring it to the group, because it's a clean example of a modeling correction with strong experimental backing.","headline":"A genuinely useful correction to a widely used magneto-optics model, validated by careful experiments; the only real caveat is an unverified smallness assumption about longitudinal mode components.","tokens_in":22240,"tokens_out":4458,"would_cite":true,"duration_ms":46548,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["32.60.+i","42.25.Bs"],"model":"deepseek-v4-flash","headline":"Magnetized atomic vapors need non-orthogonal light modes: a corrected propagation formalism removes unphysical spectral features and matches Rb D2 transmission data.","keywords":["non-orthogonal electric field modes","atomic vapor","magneto-optical transmission","weak-probe spectroscopy","rubidium D2 line","mode overlap matrix","arbitrary magnetic field geometry","atomic photonic devices"],"falsifier":"Repeat the Regime I transmission measurement (natural Rb D2, cell length 2 mm, temperature near 120 C, B = 240 G, theta_B = 84 degrees, with incident linear polarization scanned over theta_E approximately 0-172 degrees) and look at the spectrum near $\\Delta$ approximately 3 GHz: the old formalism predicts a sharp unphysical transmission feature with up to 55-57% disagreement, while the new formalism predicts a smooth, Voigt-like profile, and data matching the sharp feature would refute the central claim. A second check is to measure the longitudinal field component just after the cell exit, which the formalism assumes is negligible (m_i,z < $10^{-3}$).","tokens_in":2264,"feed_emoji":"🧲","tokens_out":5516,"duration_ms":167787,"temperature":0.7,"pith_summary":"When a laser beam passes through an atomic vapor in a magnetic field that is oblique to the beam, the two electric field modes of the medium are, in general, not perpendicular to each other. Earlier propagation models treat these modes as orthogonal, and this paper shows that the mistake produces sharp, unphysical features in predicted transmission spectra, with disagreements of up to 55-57% in transmitted intensity. The paper derives a corrected propagation formalism that keeps the modes' nonzero overlap explicitly, and verifies it with weak-probe transmission spectroscopy of the rubidium D2 line: simultaneous fits of temperature, field magnitude, and field angle agree with experiment to within 0.5-0.7% root-mean-square error. The result matters because atomic filters, magnetometers, and frequency references all rely on accurate transmission models in exactly these oblique field geometries.","feed_headline":"Non-orthogonal modes fix magnetized vapor spectra","feed_subtitle":"Correcting the mode-overlap error removes unphysical peaks and matches Rb D2 spectra to 0.7 percent.","key_machinery":"The load-bearing object is the mode-overlap matrix $S$ with entries $S_{ij} = \\langle \\hat{\\mathbf{m}}_i | \\hat{\\mathbf{m}}_j \\rangle = \\hat{\\mathbf{m}}_i^\\dagger \\cdot \\hat{\\mathbf{m}}_j$, built from the two eigenmodes solved out of the wave equation with the magnetized vapor's dielectric tensor; its diagonal entries are 1 and its off-diagonal entry is the complex mode overlap. The argument runs through the coefficient formula $\\mathbf{c} = S^{-1} M \\boldsymbol{\\epsilon}_{\\mathrm{inc}}$: when the modes are orthogonal, $S$ is the identity and the old formalism is recovered, and the explicit $S^{-1}$ factor is precisely what re-weights the modes where the overlap is large. Each mode then accrues its own phase $\\exp(\\mathrm{i} n_i k_0 L)$, and the exit field is reassembled as $E_0[c_1 t(n_1)\\hat{\\mathbf{m}}_1 + c_2 t(n_2)\\hat{\\mathbf{m}}_2]$. The modes are visualized as frequency-dependent polarization ellipses whose tilt and handedness track the regions where the old and new predictions disagree.","core_discovery":"The central claim is that the two eigenmodes of the wave equation for a magnetized atomic vapor -- the polarization states that propagate without changing shape -- are generally non-orthogonal whenever the magnetic field is neither parallel nor perpendicular to the beam, with a frequency-dependent complex overlap $\\langle \\hat{\\mathbf{m}}_1 | \\hat{\\mathbf{m}}_2 \\rangle$ that vanishes only in the Faraday and Voigt geometries (or if the susceptibilities were real). The paper's corrected formalism computes the incident-field decomposition coefficients as the solution of the overlap matrix equation $\\mathbf{c} = S^{-1} M \\boldsymbol{\\epsilon}_{\\mathrm{inc}}$, where $S_{ij} = \\langle \\hat{\\mathbf{m}}_i | \\hat{\\mathbf{m}}_j \\rangle$, and propagates each mode with its own complex phase $t(n_i) = \\exp(\\mathrm{i} n_i k_0 L)$, so that $\\mathbf{E}_{\\mathrm{out}} = E_0[c_1 t(n_1)\\hat{\\mathbf{m}}_1 + c_2 t(n_2)\\hat{\\mathbf{m}}_2]$. The previous formalism implicitly set $S$ to the identity matrix, mis-weighting the modes wherever their overlap is significant; the paper shows this produces unphysical sharp transmission features, while the corrected formalism removes them and fits measured natural-abundance Rb D2 spectra in two regimes -- 240 G near the Voigt geometry and 2.5 kG at oblique angles between 100 and 130 degrees -- with residuals of 0.5-0.7% root-mean-square error.","pith_inferences":["The same $S^{-1}$ correction applies to any propagation model that expands fields in eigenmodes of an anisotropic medium -- lossy crystals, magnetized plasmas, structured waveguides -- because non-orthogonal eigenvectors are generic for non-Hermitian propagation operators; the thermal vapor is a clean, continuously tunable test bed for that broader class of problems.","A direct test the paper does not perform: polarization tomography of the exit beam could independently reconstruct the complex mode overlap $\\langle \\hat{\\mathbf{m}}_1 | \\hat{\\mathbf{m}}_2 \\rangle$ and compare it with the model's prediction, upgrading the indirect transmission fit into a direct measurement of non-orthogonality.","The boundary treatment (negligible $m_{i,z}$ inside the cell and $E_{\\mathrm{out},z} = 0$ at the exit) sets a boundary on where the formalism applies; a variant with a full Fresnel boundary condition for the longitudinal component would make the prediction quantitative at larger fields or sharper angles where the z-component grows.","Because the old and new formalisms differ by up to ±57% transmission near $\\theta_E \\approx 82^\\circ$-$172^\\circ$, re-analyzing previously published oblique-geometry filter spectra might reveal systematic residuals that the corrected formalism would remove -- a cheap retrospective test of the claim."],"forward_implications":["Models that ignore mode non-orthogonality (the previous ElecSus formalism and its relatives) predict unphysical spectral features of up to 55-57% transmission in oblique-field geometries, and those predictions should not be used for design or parameter extraction there.","With the corrected formalism, weak-probe transmission spectra can be fitted with shared global parameters for temperature, field magnitude, and field angle, achieving 0.5-0.7% root-mean-square error across both small-field and intermediate-field regimes.","The frequency-dependent mode overlap, tunable through the incident polarization angle, becomes a control handle for atomic photonic devices; the paper names suppression of single-cell atomic-filter bandwidth as the concrete follow-on application.","At the Faraday and Voigt geometries the overlap vanishes and the new model coincides with the old one -- a built-in consistency check that holds across the parameter-space heatmaps."],"supporting_citations":[{"why":"Supplies the electric susceptibility calculation for the alkali ensemble that builds the dielectric tensor used in the wave equation.","marker":"[23]"},{"why":"Defines the ElecSus 3 propagation formalism that the paper corrects; it is the baseline that neglects mode non-orthogonality and produces the unphysical features.","marker":"[24]"},{"why":"Supplies the simultaneous fitting technique used to constrain global parameters, and the statement that modes are orthogonal in Faraday and Voigt geometries.","marker":"[33]"},{"why":"Provides the experimental calibration procedure (etalon-to-frequency conversion, reference-cell line centering) and the hyperfine Paschen-Back field estimate used to choose regimes.","marker":"[34]"},{"why":"Gives the wave-equation form and the approach for solving refractive indices and modes that the new formalism extends with the overlap matrix.","marker":"[53]"},{"why":"Supplies the oblique-geometry atomic filter context (theta_B about 80 degrees) that motivates the work, and the bandwidth-suppression application cited in the conclusions.","marker":"[56]"},{"why":"Prior statement that the modes are non-orthogonal and frequency dependent, and the pointer to exceptional-point coalescence as future work.","marker":"[59]"}],"fun_headline_variants":["Overlap correction removes phantom Rb peaks","Overlap-aware propagation matches Rb D2 to 0.7%","Non-orthogonal modes explain Rb D2 spectra","Correcting mode overlap yields accurate magnetized spectra"],"cache_read_input_tokens":24320,"weakest_assumption_plain":"The whole calculation rests on the modes' beam-direction components being negligible: the paper assumes $m_{i,z} < 10^{-3}$ of each mode's norm so that the incident field is spanned by the transverse parts alone, and it assumes the small longitudinal component of the output field reflects off the exit interface and can be set to zero; if those components were significant at the conditions tested, the mode coefficients and the predicted transmission would be inaccurate.","fun_headline_variants_meta":{"raw":{"variants":["Overlap correction removes phantom Rb peaks","Overlap-aware propagation matches Rb D2 to 0.7%","Non-orthogonal modes explain Rb D2 spectra","Correcting mode overlap yields accurate magnetized spectra"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000539,"raw_usage":{"total_tokens":2892,"prompt_tokens":1057,"completion_tokens":1835,"prompt_tokens_details":{"cached_tokens":896},"prompt_cache_hit_tokens":896,"prompt_cache_miss_tokens":161,"completion_tokens_details":{"reasoning_tokens":1772}},"tokens_in":161,"tokens_out":1835,"duration_ms":20512,"temperature":1.0,"reasoning_tokens":1772,"cache_read_input_tokens":896,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:27:10.042039+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the Regime I transmission measurement (natural Rb D2, cell length 2 mm, temperature near 120 C, B = 240 G, theta_B = 84 degrees, with incident linear polarization scanned over theta_E approximately 0-172 degrees) and look at the spectrum near $\\Delta$ approximately 3 GHz: the old formalism predicts a sharp unphysical transmission feature with up to 55-57% disagreement, while the new formalism predicts a smooth, Voigt-like profile, and data matching the sharp feature would refute the central claim. A second check is to measure the longitudinal field component just after the cell exit, which the formalism assumes is negligible (m_i,z < $10^{-3}$).","supporting_citations":[{"cited_title":"ElecSus: Extension to arbitrary geometry magneto-optics,","cited_arxiv_id":null,"evidence_quote":"Defines the ElecSus 3 propagation formalism that the paper corrects; it is the baseline that neglects mode non-orthogonality and produces the unphysical features."},{"cited_title":"Laser spectroscopy of hot atomic vapours: from’scope to theoretical fit,","cited_arxiv_id":null,"evidence_quote":"Provides the experimental calibration procedure (etalon-to-frequency conversion, reference-cell line centering) and the hyperfine Paschen-Back field estimate used to choose regimes."},{"cited_title":"Optimized ultra-narrow atomic bandpass filters via magneto-optic rotation in an unconstrained geometry,","cited_arxiv_id":null,"evidence_quote":"Supplies the oblique-geometry atomic filter context (theta_B about 80 degrees) that motivates the work, and the bandwidth-suppression application cited in the conclusions."},{"cited_title":"Improving Magneto-Optical Filter Performance: Cascading and Oblique B-fields","cited_arxiv_id":null,"evidence_quote":"Prior statement that the modes are non-orthogonal and frequency dependent, and the pointer to exceptional-point coalescence as future work."}],"review_version":1}