{"id":"59cc5260-9a98-411b-88a2-32f5273f825b","arxiv_id":"1908.06508","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For simple 2D geometries, the scattering transport inverse source problem is injective for isotropic and vector sources, and the gauge for higher-degree sources is fully described.","lead":"This paper proves that inverse source problems in radiative transport with scattering and variable refractive index are injective for certain anisotropic sources, and exactly characterizes the non-injectivity for general sources. It does so by reducing the transport problem to recent attenuated tensor tomography results, giving reconstruction algorithms with the same strength as the underlying ray transform.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Most load-bearing issue is the unsupported strength of the 'constructive' claim: Theorem 2 inherits a non-explicit invariant-distribution step from [11], and a displayed formula in §4.3.1 is algebraically wrong.","rationale":"The reader's weakest-assumption diagnosis is the right one: the central claims in Theorems 2 and 3 inherit their main substance from the external attenuated tensor tomography result, Theorem 11 of [11]. The paper does not prove that theorem, and Remark 12 shows that its recovery step is not fully constructive for general simple surfaces. That makes the unqualified 'constructively invertible' assertion in the abstract and Theorem 2 stronger than what is established. I also verified the internal reduction: the triangular system for p, the H1/H0 projection equations, and the use of subcriticality are coherent given Theorem 11, so I do not see a separate fatal gap in the main argument. The algebraic error in §4.3.1 is real but localized; the same case is treated again in the later general-scattering subsection, so it does not undermine the theorem's overall correctness. The correct verdict remains CONDITIONAL: the displayed formula should be corrected and the constructivity claims qualified consistently across the abstract, Theorem 2, and Remark 12.","tokens_in":14937,"tokens_out":30545,"duration_ms":334346,"concrete_test":"Inspect the proof of the recovery step in [11, Thms 1-2] and determine whether h is produced by an explicit finite algorithm for every simple surface or only shown to exist via invariant distributions from [22]; if the latter, the abstract and Theorem 2 must replace 'constructively invertible' with 'invertible modulo a non-explicit existence step.' Independently recompute the §4.3.1 Case (2) formula by substituting u0 = B + tilde f0 into A = k0 u0 - a tilde f0 to confirm the printed expression is incorrect.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The advertised constructive inversion is only as strong as the recovery step in Theorem 11 of [11]. If that theorem is taken as a black box, the reduction from Ma,k to attenuated tensor tomography is internally coherent; but the paper itself concedes in Remark 12 that for general simple surfaces the reconstruction of the analytic part relies on invariant distributions whose existence is microlocal and for which 'a fully constructive approach remains to be found.' Hence the unqualified wording 'constructively invertible' in the abstract and Theorem 2 is not supported at that level of generality; what is proved is invertibility modulo a non-constructive existence result. This is a real mismatch with the central claim as stated, though it does not by itself invalidate the injectivity or gauge theorems. A secondary internal issue: in §4.3.1, Case (2), the displayed formula for tilde f0 is algebraically inconsistent with the equations preceding it. From A = k0 u0 - a tilde f0 and B = u0 - tilde f0, substitution gives A = k0 B + (k0 - a) tilde f0, so tilde f0 = (A - k0 B)/(k0 - a), not A/(k0 - a) - k0 B as printed. The later general-scattering argument covers this case, so the theorem is repairable, but the proof as written contains an error.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the inverse source problem for the stationary linear transport equation with attenuation and scattering on a simply connected two-dimensional domain with variable index of refraction modeled by a simple Riemannian metric. The scattering kernel is assumed to have finite harmonic content in the angular deviation. The main results are: (Theorem 1) well-posedness and continuity of the source-to-boundary map; (Theorem 2) injectivity and constructive inversion for sources of the form f = f0 + X⊥ f⊥ and for vector-field sources; and (Theorem 3) a full characterization of the kernel of the source-to-boundary map for arbitrary finite-degree sources, namely Ma,k(f)=0 iff f = Xp + ap - Sp for some boundary-vanishing p of one degree lower. The proofs reduce the transport problem to the attenuated tensor tomography problem, using the solenoidal decomposition of Theorem 11 from a prior paper by one of the authors and solving the remaining reconstruction steps by triangular elliptic systems.","tokens_in":15211,"tokens_out":6132,"duration_ms":63596,"significance":"The paper represents a substantial contribution to inverse transport theory: it removes smallness assumptions on the scattering kernel and replaces them with a finite harmonic-content condition, and it gives an exact gauge description for finite-degree sources on simple surfaces. The forward theory in Section 3 is careful and self-contained, including an L2 treatment of trace issues and a maximal-accretivity proof of well-posedness. The reduction to attenuated tensor tomography is clearly explained, and the reliance on [11] is explicit and legitimate; this is not a circular argument. The main advertised feature, 'constructive invertibility,' is, however, stronger than what the proof supports for general simple surfaces, because the underlying recovery step from [11] relies on non-explicit invariant distributions (Remark 12). With this qualification, the paper's injectivity and gauge results are significant and likely correct.","major_comments":[{"comment":"The displayed reconstruction formula for tilde-f0 is algebraically incorrect as written. With A := k0 u0 - a tilde-f0 and B := u0 - tilde-f0, substitution gives A = k0 B + (k0 - a) tilde-f0, so tilde-f0 = (A - k0 B)/(k0 - a). The printed formula, tilde-f0 = A/(k0 - a) - k0 B, omits the division by (k0 - a) on the second term. Since the general scattering argument in Section 4.3.2 recovers this case by another route, the error is local and does not invalidate the theorem, but the displayed formula must be corrected.","section":"§4.3.1, Case (2)"},{"comment":"The unqualified claim that the map is 'constructively invertible' in the abstract and in Theorem 2 is not supported at the stated level of generality. The proof invokes Theorem 11 of [11], whose reconstruction step for the analytic part relies on invariant distributions whose existence is obtained by microlocal arguments and for which, as the manuscript itself states in Remark 12, 'a fully constructive approach remains to be found.' Thus, for a general simple surface, the inversion is only as constructive as the non-explicit existence step in [11]. The claim should be qualified, e.g., as 'invertible, with reconstruction reduced to existing attenuated tensor tomography methods,' or the constructive claim should be restricted to cases where those invariant distributions are explicit, such as the Euclidean disk.","section":"Abstract and Theorem 2; Remark 12"}],"minor_comments":[{"comment":"The proof determines f⊥ only up to an additive constant. This does not affect the reconstructed source because X⊥ applied to a constant is zero, but the text should state this explicitly to explain why the constant ambiguity is harmless.","section":"§4.3.1, Case (1)"},{"comment":"Lemma 10 is stated as a 'fact' without proof or reference. Since it is load-bearing for the reconstruction steps, please add a proof or a precise citation to the literature.","section":"Lemma 10"},{"comment":"Several displayed equations in Section 4.3.1 are unnumbered, which makes it hard to refer to the erroneous formula precisely. In the revised version, please number the key displayed equations and add the corrected formula for tilde-f0.","section":"General presentation"}],"recommendation":"major_revision","confidential_remarks":"The reviewer's stress-test correctly identifies the algebraic slip in §4.3.1 and the overstatement of constructivity. Both are fixable within the scope of the paper, so I do not see grounds for rejection. The reliance on Theorem 11 of [11] is a legitimate external reference rather than a circular dependency; however, the authors should make the limits of the constructive claim explicit in the abstract and Theorem 2, since the current wording overstates what is proved for general simple surfaces."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on Bal-Monard 1908.06508. The paper does something genuinely new: it reduces the inverse source problem for the transport equation with finite-degree scattering to attenuated tensor tomography on simple surfaces, giving injectivity for isotropic plus solenoidal sources and for vector fields (Theorem 2), and a full gauge characterization for sources of higher degree (Theorem 3). Unlike earlier work by Fujiwara–Sadiq–Tamasan, which is restricted to constant index of refraction, this handles variable speed and anisotropic sources. The forward theory in Section 3 is careful: the L2 trace estimates and semigroup argument for Theorem 1 are clean and complete. The reliance on Theorem 11 from Krishnan–Mishra–Monard is legitimate—that's an external theorem, not a re-derivation of the target, so the circularity burden is low.\n\nThe problems are both in the reconstruction part. First, the displayed formula for tilde f0 in Section 4.3.1, Case (2), is wrong as written. Given A = k0 u0 - a tilde f0 and B = u0 - tilde f0, the correct inversion is tilde f0 = (A - k0 B)/(k0 - a). The expression printed, 1/(k0-a)(k0u0 - a tilde f0) - k0(u0 - tilde f0), does not simplify to tilde f0. Since the later general-scattering argument in §4.3.2 covers this case, the theorem is repairable, but the proof as printed contains an algebraic error that a referee should flag.\n\nSecond, the 'constructive' wording is oversold. The abstract and Theorem 2 say constructively invertible, but the reconstruction inherits a non-explicit step from [11]: the analytic part is recovered via invariant distributions whose existence is microlocal, and Remark 12 concedes a fully constructive approach remains to be found. So what's proved is invertibility modulo a non-constructive existence result for general simple surfaces. That's still a solid theorem, but the claim should be qualified.\n\nThe gauge theorem (Theorem 3) looks right to me, and the backward direction is straightforward. The degree-descent argument is coherent.\n\nWho's this for? People working on inverse transport, SPECT, or tensor tomography. It deserves a serious referee—send it out. With the formula corrected and the constructivity language toned down, I'd be happy to see it accepted.","headline":"Worth reading and worth refereeing; the new reduction to attenuated tensor tomography is solid, but the reconstruction formula in §4.3.1 is algebraically wrong and the 'constructive' claim is stronger than the proof supports.","tokens_in":15709,"tokens_out":2975,"would_cite":true,"duration_ms":25280,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R30","44A12","53C65"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper establishes that in a two-dimensional simple geometry with finite-degree scattering, the source-to-boundary map is injective and constructively invertible for isotropic and vector-field sources, and fully characterizes the…","keywords":["inverse source problem","radiative transfer","attenuated X-ray transform","tensor tomography","simple Riemannian surface","scattering kernel","injectivity modulo gauge","variable index of refraction"],"falsifier":"On the Euclidean unit disk, where explicit inversion formulas are known, implement the claimed reconstruction for a smooth isotropic source under a strongly scattering finite-harmonic kernel: the algorithm must recover the source exactly, so any numerical case where the reconstructed source differs, or where two distinct isotropic sources produce the same boundary data, would refute Theorem 2.","tokens_in":14733,"feed_emoji":"📡","tokens_out":4943,"duration_ms":51168,"temperature":0.7,"pith_summary":"This paper asks when an internal light source inside a scattering, absorbing medium with a smoothly varying index of refraction can be recovered from the radiation leaving the boundary. In a two-dimensional simple geometry, it proves that the source-to-boundary measurement map is injective and constructively invertible for the two source classes that matter in applications: isotropic sources (possibly with a tangential-gradient component) and vector-field sources. For sources with arbitrary finite angular dependence, the map is not injective, and the paper gives the complete description of its kernel: the only invisible sources are those generated by a boundary-vanishing function through the transport operator f = Xp + ap - Sp. The proofs work by reducing the scattering problem to the attenuated X-ray transform on tensor fields and using a recently established solenoidal decomposition of that transform.","feed_headline":"Inverse source map is injective for isotropic and vector-field sources","feed_subtitle":"Boundary data determine these sources even with strong scattering; all invisible sources are exactly one explicit gauge term.","key_machinery":"The load-bearing object is the attenuated X-ray transform Ia on functions of finite degree, viewed as a transform on tensor fields, together with its solenoidal decomposition: every degree-m integrand splits uniquely as f = (X + a)p + h with p vanishing at the boundary and h solenoidal, and h is recoverable from Ia f. The argument rewrites the transport equation as Xu + au = f + Su, observes that the measured data equal Ia[f + Su], and uses the reconstructible solenoidal representative to eliminate the unknown scattering term by solving a descending triangular system of elliptic boundary-value problems. The finite harmonic content of k keeps all integrands finite-degree, reducing an infinite-dimensional transport problem to finitely many Poisson equations.","core_discovery":"The central claim is that the inverse source problem for the linear transport equation Xu + au = Su + f on the unit tangent bundle of a simple Riemannian surface is governed exactly by the kernel of the attenuated tensor tomography transform. When the scattering kernel k has finite harmonic degree and absorption dominates scattering, the source-to-boundary operator is injective on sources of the form f = f0 + X⊥ f⊥ and on vector-field sources, and the inversion is constructive: from the boundary data one reconstructs the full right-hand side modulo the known gauge, then solves a free transport equation and a triangular system of elliptic equations to recover f. For sources of finite degree m, the operator has zero data exactly when f = Xp + ap - Sp for some boundary-vanishing p of degree m - 1, so the non-injectivity is entirely accounted for by this gauge.","pith_inferences":["The explicit gauge characterization suggests a stability program: two sources with close boundary data should differ by a gauge term plus a small solenoidal part, and the quantitative version of that statement is a natural next step not proved here.","The finite-harmonic condition on k is likely not essential; a scattering kernel with rapidly decaying Fourier coefficients should produce approximate recovery, with error controlled by the tail of the coefficients, though the paper does not formulate such an estimate.","The entire approach is two-dimensional because it relies on complex-analytic integrating factors; a genuine three-dimensional analogue would need a different mechanism, and the paper itself flags three dimensions as open.","For time-dependent sources, applying a Fourier transform in time would reduce each frequency to a stationary problem of the type treated here, so the injectivity and gauge results may carry over to time-harmonic emission settings."],"forward_implications":["Isotropic sources, the case relevant to optical molecular imaging, are uniquely recoverable from boundary measurements in simple refractive geometries with arbitrary-strength finite-harmonic scattering, with no smallness assumption on the scattering kernel.","Vector-field sources, relevant to Doppler tomography, are also uniquely and constructively recoverable.","For higher-degree anisotropic sources, the measurement map has a kernel, but the kernel is fully characterized, so the data determine the source up to the explicit gauge f = Xp + ap - Sp.","The reconstruction procedure never differentiates and never solves a transport equation with scattering; it requires only free-transport integration and Poisson solves, which makes it computationally efficient.","In the Euclidean disk, the reconstruction becomes fully explicit through existing fan-beam inversion formulas for the attenuated ray transform."],"supporting_citations":[{"why":"Supplies the solenoidal decomposition theorem for the attenuated ray transform that is the basis for reconstructing a representative from the data.","marker":"[11]"},{"why":"Provides the holomorphic integrating factors whose existence and regularity underpin the decomposition theorem for smooth attenuation.","marker":"[22]"},{"why":"Gives explicit inversion formulas for the attenuated tensor transform in the Euclidean disk, making the reconstruction fully explicit there.","marker":"[19]"},{"why":"Supplies inversion formulas and range characterizations for the attenuated geodesic ray transform used to recover solenoidal components.","marker":"[1]"},{"why":"Establishes the tensor tomography results on surfaces that underlie the gauge-kernel theorem used in the non-injective case.","marker":"[20]"},{"why":"Supplies the maximal-accretive operator theory used to prove well-posedness of the forward transport problem and continuity of the measurement map.","marker":"[8]"}],"fun_headline_variants":["Inverse transport source problem injective modulo explicit gauge","Boundary data fix transport sources up to one gauge term","All non-injective transport sources equal one gauge term","Attenuated tensor tomography gives constructive inversion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reconstruction inherits the assumption, from the cited solenoidal decomposition theorem for the attenuated ray transform, that every degree-m integrand on a simple surface uniquely splits as (X + a)p + h with boundary-vanishing p and recoverable solenoidal h, and that the holomorphic integrating factors behind that decomposition exist at the required regularity.","fun_headline_variants_meta":{"raw":{"variants":["Inverse transport source problem injective modulo explicit gauge","Boundary data fix transport sources up to one gauge term","All non-injective transport sources equal one gauge term","Attenuated tensor tomography gives constructive inversion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00151,"raw_usage":{"total_tokens":5969,"prompt_tokens":776,"completion_tokens":5193,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":392,"completion_tokens_details":{"reasoning_tokens":5130}},"tokens_in":392,"tokens_out":5193,"duration_ms":37858,"temperature":1.0,"reasoning_tokens":5130,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:42:51.004306+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On the Euclidean unit disk, where explicit inversion formulas are known, implement the claimed reconstruction for a smooth isotropic source under a strongly scattering finite-harmonic kernel: the algorithm must recover the source exactly, so any numerical case where the reconstructed source differs, or where two distinct isotropic sources produce the same boundary data, would refute Theorem 2.","supporting_citations":[{"cited_title":"On solenoidal-injective and injective ray transforms of tensor fields on surfaces","cited_arxiv_id":"1807.10730","evidence_quote":"Supplies the solenoidal decomposition theorem for the attenuated ray transform that is the basis for reconstructing a representative from the data."},{"cited_title":"Salo and G","cited_arxiv_id":null,"evidence_quote":"Provides the holomorphic integrating factors whose existence and regularity underpin the decomposition theorem for smooth attenuation."},{"cited_title":"II: atten uated transforms , In- verse Problems and Imaging, 12 (2018), pp","cited_arxiv_id":null,"evidence_quote":"Gives explicit inversion formulas for the attenuated tensor transform in the Euclidean disk, making the reconstruction fully explicit there."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies inversion formulas and range characterizations for the attenuated geodesic ray transform used to recover solenoidal components."},{"cited_title":"Paternain, M","cited_arxiv_id":null,"evidence_quote":"Establishes the tensor tomography results on surfaces that underlie the gauge-kernel theorem used in the non-injective case."},{"cited_title":"Dautray and J.-L","cited_arxiv_id":null,"evidence_quote":"Supplies the maximal-accretive operator theory used to prove well-posedness of the forward transport problem and continuity of the measurement map."}],"review_version":1}