{"id":"b9028d76-f208-4895-8c1c-0f6d97b2bea8","arxiv_id":"2506.09012","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Every convolution-invertible complex measure on R^n is a convolution of a translation, finitely many orthogonal embeddings of a fixed one-dimensional measure σ, and an exponential exp(ν).","lead":"The authors prove a complete characterization of complex measures on R^n that are invertible under convolution, extending a 1971 theorem of Taylor from one dimension to all dimensions. The result describes every such measure as a shift, finitely many orthogonal copies of a fixed one-dimensional factor, and an exponential of a complex measure, with direct applications to quasi-infinitely divisible distributions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim is conditional on Taylor's Proposition 4.3: if the spectrum of the subalgebras {αδ0 + fλ^q⊗ζ^{n−q}_C} is not the same as in M(R^n) for some q, the GFS inversion step in Theorem 8.1 collapses.","rationale":"I read the paper as a complete structural classification. The reduction to Taylor's Theorem 1.2, the application of Hewitt's topology theorem, and the three case analyses (q=n, q=0, 1≤q≤n−1) are coherent. The Banach algebra-valued GFS machinery (Sections 7–8) is intricate but internally consistent: Lemma 7.8 and Theorem 7.9 correctly transfer the scalar GFS characterization to semisimple commutative unital Banach algebras with connected Gelfand space, and A = (Cδ^q_0 + L(R^q))^∧ qualifies. The distinguished-logarithm argument in Theorem 5.3 for n≥2 is standard and correct. The only place where the argument depends on an unproved, explicitly nontrivial external assertion is Proposition 4.3; the reader's weakest assumption pinpoints the same. Since that theorem is published and the authors state it as a direct consequence of Taylor's results, I do not regard it as an internal flaw, but it is the most load-bearing assumption. A careful check of the hypotheses of [28, Thm. 3.3/Prop. 4.1] for all q would settle whether the concern lands. No change to the reader's verdict.","tokens_in":37911,"tokens_out":42397,"duration_ms":358544,"concrete_test":"Independently verify Taylor [28, Thm. 3.3 and Prop. 4.1] for the subalgebra B_q := {αδ0 + fλ^q⊗ζ^{n−q}_C : C ⊂ R^{n−q} countable, f ∈ L^1(λ^q⊗ζ^C)} of M(R^n), by checking that B_q is the measure algebra corresponding to the topology R^q × R^{n−q}_d in the sense of [28, Prop. 4.1], and that the spectrum of any element of B_q computed in B_q equals its spectrum in M(R^n). If this equality fails for some q (especially q=1), the proof of Theorem 8.1(i)→(ii) collapses; if it holds, the dependency is sound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 2.1 reduces an arbitrary invertible measure to factors U(αδ0 + fλ^q⊗ζ^{n−q}_C) (Proposition 4.2), and then must show that invertibility of such a factor is preserved with the inverse in the same subalgebra. Proposition 4.3 supplies exactly this via Taylor's spectral-stability results, but the paper does not prove it and explicitly calls it non-trivial. In Section 8, this is load-bearing because invertibility of bµ(·, z2) in GFS(R^{n−q}, A) is asserted from the fact that the inverse µ' has the same form; without Proposition 4.3, bµ'(·, z2) would not be known to lie in A, and the application of Theorem 7.9 would not go through. This is a genuine dependency on an external deep theorem rather than an internal flaw; no counterexample or inconsistency is present. Still, the main theorem's correctness is conditional on Taylor's Proposition 4.3 holding for every q ∈ {0,...,n}.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper characterizes convolution-invertible complex measures on R^n. The main theorem (Theorem 2.1) states that µ ∈ M(R^n) is invertible if and only if it admits a factorization µ = δ_γ^n * U_1(σ^{*m_1} ⊗ δ_0^{n-1}) * ... * U_p(σ^{*m_p} ⊗ δ_0^{n-1}) * exp(ν), where σ is the one-dimensional signed measure with characteristic function (1+iz)/(1-iz), U_i are orthogonal matrices, m_i are integers, γ ∈ R^n, ν ∈ M(R^n), and p ∈ N_0. This directly extends Taylor's 1971 one-dimensional characterization. The authors also give an equivalent characteristic-function formulation (Theorem 2.4) with a complex Lévy-type measure, and a version for finite signed measures (Corollary 2.2). The proof uses Taylor's general decomposition theorem for invertible measures on locally compact abelian groups (Theorem 1.2), Hewitt's classification of group topologies on R^n (Lemma 4.1), and a substantial new development of Banach-algebra-valued generalized Fourier series, including a distinguished logarithm theorem for semisimple commutative unital Banach algebras with connected Gelfand space and an invertibility criterion for GFS(R^n, A) (Theorems 7.4, 7.7, 7.9). These tools are applied in Section 8 to the intermediate-dimensional case 1 ≤ q ≤ n-1, which is the core technical novelty.","tokens_in":38030,"tokens_out":11938,"duration_ms":117328,"significance":"If the main theorem is correct, it provides a complete and surprisingly simple structural description of all convolution-invertible complex measures on R^n, showing that only one-dimensional σ-factors, shifts, and exponentials are needed. The paper is well structured and carefully distinguishes its own contributions from deep external results (Taylor's measure-algebra spectral theory, Hewitt's classification), with Proposition 4.3 explicitly cited as nontrivial and load-bearing. The new Banach-algebra machinery for generalized Fourier series is interesting in its own right and likely to be useful beyond this paper. The applications to quasi-infinitely divisible distributions and to convolution roots of all orders (Theorem 10.1) are natural and credibly derived. The proofs are detailed and internally consistent; the main theorem is conditional on Taylor's spectral-stability results, but those are standard published theorems and the dependence is clearly flagged rather than hidden.","major_comments":[],"minor_comments":[{"comment":"In condition (v'), the notation λ^1 should be λ^n: the measure ν is on R^n, so the density g should be in L^1(R^n) and the expression should read ν(dx) = g(x) λ^n(dx). The same typo appears in the proof sentence 'µ = exp(gλ1 + βδ_0^n)'.","section":"Theorem 5.3"},{"comment":"The indexing in the statement is incorrect: the functions f_i are listed for i = 1, ..., q, but there are p factors, so it should read i = 1, ..., p.","section":"Proposition 4.2"},{"comment":"The statement contains a duplicated 'Let n ∈ N': it appears once before the definition of σ and again immediately after.","section":"Theorem 2.1"},{"comment":"In the proof of (i) implies (iv), the assertion that the non-integer coefficients m_j/k cannot be absorbed into a different representation is described as 'easily seen'; a short justification via mutual singularity of the measures τ_{u_j} (supported on distinct lines) and the finiteness of ν_0 would make the argument more transparent.","section":"Theorem 10.1"}],"recommendation":"accept","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a major result. Berger and Lindner give the full analogue of Taylor's 1971 theorem for R^n, and the answer is strikingly clean: every invertible complex measure is a shift, a convolution of finitely many line measures U_i(σ^{*m_i} ⊗ δ_0^{n-1}), and exp(ν). They also get the Lévy–Khintchine type representation and several Cramér–Wold devices. The n-dimensional theorem is new, and the intermediate cases in Taylor's factorization are genuinely hard. The Banach algebra machinery they build—GFS with values in a semisimple commutative unital algebra with connected Gelfand space, the distinguished logarithm theorem, and the invertibility characterization in Theorem 7.9—is a real contribution. The paper is careful about the topology classification and Haar measures along the way.\n\nThe main theorem is downstream of Taylor's Theorem 1.2 and Proposition 4.3. Section 8 leans on Proposition 4.3 to know that the inverse of a measure in Cδ_0 + L(R^q) ⊗ discrete stays in the same subalgebra, which lets the inverse's characteristic function land in GFS(R^{n-q}, A). If that statement failed for some q, the proof of Theorem 8.1 would collapse. But it is a published theorem of Taylor, and the paper flags it as nontrivial. I do not count this as a flaw; it is a normal external dependency. The stress-test note correctly identifies the load-bearing point, but it is not an internal gap. A referee should verify that Proposition 4.3 covers every q as stated, since the paper's interpretation of spectral stability is exactly what makes the GFS inversion step go through.\n\nThis paper is for specialists in measure algebras, convolution operators, and quasi-infinitely divisible distributions. It deserves a serious referee. The approach is sound, the new GFS results look correct, and the main theorem is important. I would send it to peer review without hesitation. If I were refereeing, I would ask for a short remark making the dependence on Proposition 4.3 impossible to miss.","headline":"A complete, carefully proved characterization of invertible complex measures on R^n; the central argument is conditional on Taylor's spectral-stability results, but the dependency is explicit and legitimate.","tokens_in":38664,"tokens_out":3590,"would_cite":true,"duration_ms":41628,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["43A05","43A20","60E10","60E07"],"pacs":[],"model":"deepseek-v4-flash","headline":"A complex measure on R^n is convolution-invertible exactly when it factors as a shift, finitely many copies of a fixed one-dimensional signed measure on lines through the origin, and the exponential of a complex measure.","keywords":["invertible complex measures","convolution","characteristic functions","generalized Fourier series","quasi-infinitely divisible distributions","Lévy–Khintchine representation","Banach algebra valued functions","measure algebra"],"falsifier":"The claim would be falsified by an invertible complex measure on $R^{2}$ whose characteristic function, under the Lévy–Khintchine representation of Theorem 2.4, requires the singular measure Λ on the unit sphere to be supported on infinitely many points; the theorem asserts Λ is always finitely supported.","tokens_in":37598,"feed_emoji":"📐","tokens_out":9462,"duration_ms":98163,"temperature":0.7,"pith_summary":"This paper proves a complete structural characterization of invertible complex measures on Euclidean space. It shows that a complex measure on R^n is invertible under convolution exactly when it can be built from a shift, finitely many copies of a single fixed one-dimensional signed measure placed on lines through the origin, and the exponential of a complex measure. The same description yields a Lévy–Khintchine type formula for the characteristic function of every such measure, with the non-finite part always supported on finitely many lines with a fixed radial density. This settles the multidimensional analogue of a problem solved in one dimension in 1971. In fact, the multidimensional structure is no more complicated than the one-dimensional one.","feed_headline":"Invertible complex measures on R^n factor along lines","feed_subtitle":"A 1971 one-dimensional theorem now covers every dimension via shifts, exponentials, and one fixed signed measure.","key_machinery":"The proof is carried by three components. First, a structural decomposition theorem for invertible measures on locally compact abelian groups reduces any invertible µ to a convolution of factors each consisting of a point mass plus a density with respect to Haar measure on R^q × $R_d^{{n-q}}$. Second, the fixed signed measure σ, with characteristic function (1+iz)/(1-iz), is the single building block that accounts for the non-exponential part; its characteristic function satisfies exp(∫($e^{{ixz}}$-1) m $e^{{-|x|}}$/x dx) = ($σ^{{∧}}$(z))^m, which is used to rewrite the σ-factors as Lévy measures on lines. Third, for the intermediate dimensions q ∈ {1,...,n-1}, the paper develops a theory of Banach algebra-valued generalized Fourier series and a distinguished logarithm for functions taking values in a semisimple commutative unital Banach algebra with connected Gelfand space, which lets it characterize when the characteristic function of such a factor is invertible in the algebra GFS($R^{{n-q}}$, A).","core_discovery":"On the paper's own terms, the central claim is Theorem 2.1: a complex measure µ on R^n is convolution-invertible if and only if it can be written as δ_γ^n * U_1($σ^{{*m_1}}$ ⊗ $δ_0^{{n-1}}$) * ... * U_p($σ^{{*m_p}}$ ⊗ $δ_0^{{n-1}}$) * exp(ν), where σ is the fixed signed measure whose characteristic function is (1+iz)/(1-iz), the U_j are orthogonal matrices, m_j are integers, γ is a shift vector, and ν is a complex measure. Equivalently, at the level of characteristic functions, µ is invertible exactly when its Fourier transform has the form c exp(iγ·z + ∫($e^{{iz·x}}$-1)(ν_0+ν_1)(dx)) with ν_0 finite and ν_1 given by $r^{{-1}}$$e^{{-r}}$ dr times a finite signed measure on the unit sphere, so the non-finite part lives only on finitely many lines through the origin. When n=1 the theorem reduces to the 1971 one-dimensional characterization.","pith_inferences":["Because the non-finite part of the Lévy measure is always supported on finitely many lines, invertibility in M(R^n) is far more rigid than the mere non-vanishing of the characteristic function; this suggests that the failure of the naive condition inf |Xhat µ| > 0 to be sufficient is already captured by line-like obstructions.","The Banach algebra valued generalized Fourier series machinery could transfer to other settings where one studies invertibility of functions with values in a commutative Banach algebra, such as convolution algebras on products of Euclidean and discrete abelian groups.","The paper's description of invertible measures gives an explicit way to construct probability measures that are quasi-infinitely divisible but not infinitely divisible: take a probability measure equal to δ_γ * σ^{*u} * exp(ν) with u ≠ 0 and normalize; the signed density e^{-r}/r along a line yields the quasi-Lévy measure.","One could test whether the finiteness of the support of Λ on the sphere is genuinely necessary by trying to construct an invertible measure whose characteristic function requires a continuum of lines; Theorem 2.4 rules that out, so a positive example would refute the theorem as stated."],"forward_implications":["Every invertible complex measure on R^n has a Lévy–Khintchine representation with a complex Lévy type measure that integrates min(1,|x|).","For absolutely continuous-plus-point-mass measures and for discrete measures, invertibility is equivalent to invertibility of all one-dimensional projections.","An invertible finite signed measure can be factored with a signed measure ν and the normalization factor µ(R^n), and its inverse is again a signed measure.","An invertible complex measure has convolution roots of all orders if and only if it is a shift times an exponential, i.e. the σ-factors are absent."],"supporting_citations":[{"why":"Gives the one-dimensional characterization of invertible complex measures that this paper extends, and the general decomposition theorem for invertible measures on locally compact abelian groups used as the starting point.","marker":"[25]"},{"why":"Supplies the spectral-stability result (Proposition 4.3) that the inverse of a factor of the form αδ_0 + f λ^q ⊗ ζ_C preserves the form and that invertibility in the subalgebra is the same as in M(R^n).","marker":"[28]"},{"why":"Provides background results on measure algebras, convolution operators, and the invertibility criterion for discrete complex measures used in Section 6.","marker":"[27]"},{"why":"Gives the characterization of invertible discrete complex measures via characteristic functions bounded away from zero, used for the q=0 case and for GFS(R^n,C).","marker":"[1]"},{"why":"Classifies the locally compact group topologies on R^n finer than the Euclidean topology as R^q × R_d^{n-q}, which determines the Haar measures and the form of the factors in Proposition 4.2.","marker":"[11]"},{"why":"Supplies the Wiener–Lévy theorem used in the absolutely continuous case (q=n) to show that the distinguished logarithm of the characteristic function is the Fourier transform of an L^1 function.","marker":"[22]"},{"why":"Provides the one-dimensional characterization for measures αδ_0 + fλ^1 (the q=1 case) that is invoked in the proof of Proposition 5.2.","marker":"[4]"}],"fun_headline_variants":["Invertible measures on R^n factor along one-dimensional lines","Convolution-invertible measures in R^n are line-built","Taylor's 1971 measure result extended to all dimensions","n-D invertible measures: only finitely many lines matter","Complex measures invertible on R^n: line factorization theorem"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on the assumption that the inverse of each factor that is a point mass plus a density on R^q × $R_d^{{n-q}}$ stays in the same subalgebra and that invertibility in that subalgebra is equivalent to invertibility in the full measure algebra M(R^n); if that spectral-stability input failed for some q, the chain of reasoning would break.","fun_headline_variants_meta":{"raw":{"variants":["Invertible measures on R^n factor along one-dimensional lines","Convolution-invertible measures in R^n are line-built","Taylor's 1971 measure result extended to all dimensions","n-D invertible measures: only finitely many lines matter","Complex measures invertible on R^n: line factorization theorem"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00025,"raw_usage":{"total_tokens":1639,"prompt_tokens":1118,"completion_tokens":521,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":734,"completion_tokens_details":{"reasoning_tokens":437}},"tokens_in":734,"tokens_out":521,"duration_ms":5843,"temperature":1.0,"reasoning_tokens":437,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:58:01.666307+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The claim would be falsified by an invertible complex measure on $R^{2}$ whose characteristic function, under the Lévy–Khintchine representation of Theorem 2.4, requires the singular measure Λ on the unit sphere to be supported on infinitely many points; the theorem asserts Λ is always finitely supported.","supporting_citations":[{"cited_title":"Taylor, The cohomology of the spectrum of a measure algebra","cited_arxiv_id":null,"evidence_quote":"Gives the one-dimensional characterization of invertible complex measures that this paper extends, and the general decomposition theorem for invertible measures on locally compact abelian groups used as the starting point."},{"cited_title":"Taylor, On the spectrum of a measure","cited_arxiv_id":null,"evidence_quote":"Supplies the spectral-stability result (Proposition 4.3) that the inverse of a factor of the form αδ_0 + f λ^q ⊗ ζ_C preserves the form and that invertibility in the subalgebra is the same as in M(R^n)."},{"cited_title":"Taylor, Measure algebras","cited_arxiv_id":null,"evidence_quote":"Provides background results on measure algebras, convolution operators, and the invertibility criterion for discrete complex measures used in Section 6."},{"cited_title":"Alexeev and A","cited_arxiv_id":null,"evidence_quote":"Gives the characterization of invertible discrete complex measures via characteristic functions bounded away from zero, used for the q=0 case and for GFS(R^n,C)."},{"cited_title":"Hewitt, A remark on characters of locally compact groups","cited_arxiv_id":null,"evidence_quote":"Classifies the locally compact group topologies on R^n finer than the Euclidean topology as R^q × R_d^{n-q}, which determines the Haar measures and the form of the factors in Proposition 4.2."},{"cited_title":"Rudin, Fourier analysis on groups","cited_arxiv_id":null,"evidence_quote":"Supplies the Wiener–Lévy theorem used in the absolutely continuous case (q=n) to show that the distinguished logarithm of the characteristic function is the Fourier transform of an L^1 function."},{"cited_title":"Berger, On quasi-infinitely divisible distributions with a point mass","cited_arxiv_id":null,"evidence_quote":"Provides the one-dimensional characterization for measures αδ_0 + fλ^1 (the q=1 case) that is invoked in the proof of Proposition 5.2."}],"review_version":1}