{"id":"708d60d0-b411-41e8-b8f5-ae98de35bd9a","arxiv_id":"2506.19978","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every GBD function carries a symmetric-matrix-valued measure whose decomposition separates absolutely continuous, Cantor, and jump contributions, and GSBD is exactly the class where the Cantor part vanishes.","lead":"This paper defines a matrix-valued measure for every function of generalized bounded deformation, extending the symmetric gradient beyond the classical BD space. It splits each such measure into absolutely continuous, Cantor, and jump parts, and characterizes the special subspace GSBD by vanishing of the Cantor part.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the Section 6 grid estimate is the load-bearing technical step, but it is internally coherent and the diagonal extraction preserves the required uniformity.","rationale":"The reader's ACCEPT with moderate confidence is appropriate. I focused on Section 6 because it is the asserted load-bearing technical premise. The apparent danger is that the bad grid contribution (6.5)-(6.7) might not vanish; this is handled by a two-scale argument: large m kills the unbounded-multiplicity set via Lemma 6.4, then k to infinity kills the bounded-multiplicity contribution via Lemma 6.5. The diagonal extraction is valid because each application of Lemma 6.8 can be taken relative to the previous K_n, and the bad-index sums are bounded by epsilon/2^n for all sufficiently large k, so letting n go to infinity gives the limit. I also checked the supporting points: integrability of f in Lemma 6.4 follows from Lemma 2.4 and H1(J)<infinity; the non-atomicity needed in Lemma 6.7 is justified by the reduction Ju = J1_u; the identity |sigma^xi_u| = lambda^xi_u restricted to Omega minus J1_u supports equality (8.3). No internal inconsistency or unproved circular step surfaced. The paper remains unverified by formal methods, but that is a verification-status caveat, not a mathematical objection. Hence no change to the reader's verdict.","tokens_in":864,"tokens_out":3066,"duration_ms":551923,"concrete_test":"Re-derive Lemma 6.4 in the case zeta = xi + eta, writing out the constants in (6.38)-(6.42) and verifying that the Riemann-sum set K^zeta_epsilon and the Egorov set I^zeta_epsilon from Lemma 3.1 are independent of the shift omega and that (6.43) holds uniformly in z2 in I^zeta_epsilon. Then check that the diagonal construction in Theorem 6.1, with U_k = intersection_{n<=n_k} U^k_n, still satisfies (6.70)-(6.71) for every zeta simultaneously; in particular, confirm L2(U_k) >= L2(U)-2epsilon and that K cap [k_n,+infinity) subset K_n yields both (6.79) and (6.80).","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the chain that actually carries the main theorem, I find no concrete flaw. The only part that could break Theorem 8.1 is the quadraticity proof in dimension 2, specifically the treatment of bad grid indices in Section 6. The proof rests on Lemma 6.4's bound H^1(pi_zeta(E_check^{k,zeta}_m)) <= C/m and on Lemma 6.5's non-atomic absolute continuity. I verified the logic: Lemma 6.4 reduces to the uniform Riemann-sum estimate (6.43), where f(s)=H^0(J^{zetabar}_{s pi_{zetabar}(zeta)}) is integrable by Lemma 2.4; the strip argument for h=3,...,6 is geometrically consistent. Lemma 6.5 follows from Lemma 6.7 because, after the reduction Ju = J1_u, |Du^zeta_y| has no atoms outside J^zeta_y. Lemma 6.8 and the diagonal extraction in Theorem 6.1 correctly transfer (6.70)-(6.71) to the bad-index sums. The proof is not machine-checked, so independent verification is advisable, but I do not see a surviving objection to the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper associates to every function u in GBD(Ω) a symmetric-matrix-valued bounded Radon measure μ_u, and more generally a family μ_{u,r}, defined through the limits of the directional derivatives D_ξ(τ_R(u·ξ)) on sets avoiding the large-jump set J_u^r. The main result, Theorem 8.1 and Corollary 8.6, proves the existence of such measures and their three-part decomposition μ_{u,r} = μ_u^a + μ_u^c + μ_{u,r}^j, with explicit formulas for the absolutely continuous part ∫ Eu dx and the jump part ∫_{J_u\\J_u^r} [u]⊙ν_u dH^{d-1}. In addition, Theorem 8.9 characterizes GSBD(Ω) as the set of u in GBD(Ω) for which μ_u^c = 0, giving an intrinsic, non-slicing characterization of GSBD. The proof is built on a lengthy quadraticity argument for the map ξ ↦ σ_u^ξ(B), proved first in dimension two by a discretization with careful control of bad grid indices, and then extended to higher dimensions by a Fubini-type slicing argument.","tokens_in":57091,"tokens_out":5273,"duration_ms":55759,"significance":"If correct, the paper closes a genuine gap in the theory of generalized bounded deformation: it provides a measure-theoretic analogue of the symmetric gradient Eu for functions that are not in BD(Ω), together with a Cantor part and a jump part. This is directly relevant to variational models of cohesive fracture, where GBD is the natural space once anti-plane symmetry is dropped. The characterization of GSBD by the vanishing of μ_u^c is a clean structural result that complements the slicing definition and the recent characterizations of Chambolle and Crismale. The proof is unusually detailed, with all technical lemmas and an appendix on measurability; the load-bearing estimate is Lemma 6.4, and I have checked its reduction to the uniform Riemann-sum bound (6.43) and the strip argument for h = 3,...,6. I find the argument coherent and do not see a surviving objection to the central claim. The paper is not machine-checked, so independent verification of the many algebraic identities in Sections 5–6 is advisable, but this is a matter of prudent verification, not a detected flaw.","major_comments":[],"minor_comments":[{"comment":"In the paragraph beginning 'The analysis of the fine properties...', the word 'reveales' should be 'reveals'.","section":"Section 1"},{"comment":"In the proof of Lemma 6.5, the derivative is written as Du^ξ_y in two places ('Du^ξ_y({t}) = 0' and 'the measure µ := |Du^ξ_y| (U^ζ_y\\J^ζ_y)'); for consistency with the statement of the lemma it should be Du^ζ_y.","section":"Lemma 6.5, proof"},{"comment":"The phrase 'ω = z1¯ζ + z2ζ with z1,∈ R' contains a stray comma and should read 'with z1 ∈ R'.","section":"Section 6, after (6.48)"},{"comment":"The transliterations 'ˇCebyˇ s¨ ev' and 'Cebyˇ s¨ ev' should be normalized to a single spelling, for example 'Čebyšev' or 'Chebyshev'.","section":"Section 6, Lemmas 6.4 and 6.8"},{"comment":"The chain of equalities in (8.1) is correct because σ_u^ξ(J_u^1)=0 by definition (4.1), but the notation σ_u^ξ(B) = lim_{R→∞} D_ξ(τ_R(u·ξ))(B\\J_u^1) is slightly compressed; a parenthetical reminder that the two occurrences of σ_u^ξ differ by a null set with respect to J_u^1 would help the reader.","section":"Theorem 8.1, Eq. (8.1)"}],"recommendation":"accept","confidential_remarks":"I recommend acceptance. The manuscript is very long and technical, and the referee process cannot guarantee line-by-line verification of every identity in Sections 5–6; the main risk is concentrated in the bad-grid-index estimates of Section 6, but I found no concrete flaw there. The citation to the 2025 paper by Chambolle and Crismale should be updated to its final published form if one becomes available before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is the real thing. Dal Maso and Donati associate to every GBD function a symmetric-matrix-valued Radon measure, prove the BD-style decomposition into absolutely continuous, Cantor, and jump parts, and show that u belongs to GSBD if and only if the Cantor part vanishes. That characterization is genuinely new and gives GSBD a slicing-free identity. The paper extends the BD structure theorem to a setting where the distributional symmetric gradient need not exist, rather than restating known results in heavier language.\n\nWhat I value most is that the proof is honest about where the difficulty lies. The whole argument reduces to proving that the function ξ↦σ^ξ_u(B) is quadratic. They do this in dimension two by a discretization: approximate σ^ζ on a parallelogram by Riemann sums on a shifted grid, check the discrete parallelogram identity on good cells, and show the bad cells contribute nothing in the limit. The bad-cell estimates in Section 6 are load-bearing. I walked through the chain: Lemma 6.4 reduces to integrability of the counting function via Lemma 2.4, and the strip argument for h=3,...,6 is geometrically sound. Lemma 6.5 uses non-atomicity of the sliced measure off the jump; Lemma 6.8 and the diagonal extraction preserve uniformity. I could not find a surviving objection. The passage to higher dimension by slicing and Fubini is standard in spirit and carefully executed. The formulas for the three parts in Proposition 8.5 are correct, and Corollary 8.6 handles the truncation level r properly.\n\nSoft spots, in proportion. The proof is very long and extremely technical, and it is not machine-checked. If there is a hidden bug, the most plausible location is the Section 6 estimates, where the constants and index sets are intricate; independent verification is appropriate. Minor: Theorem 8.10 depends on the Chambolle–Crismale characterization from an external preprint, so that part inherits an external preprint's status. Also, this is not a light read; you need GBD/GSBD background to appreciate where the difficulty sits. None of this shakes the central argument as far as I can see.\n\nThis paper is for specialists in calculus of variations and geometric measure theory, especially those working on fracture mechanics and cohesive models. It deserves a serious referee; if the referee confirms Section 6, accept. I would cite it if I were working on GSBD compactness or structure.","headline":"Serious structural paper: the new matrix-valued measure and the GSBD characterization look right, with the real risk concentrated in the Section 6 grid estimates.","tokens_in":57606,"tokens_out":2082,"would_cite":true,"duration_ms":25824,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q20","74A45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every function of generalised bounded deformation admits a symmetric matrix-valued measure that replaces the distributional symmetric gradient, and the vanishing of its Cantor part exactly characterizes the space GSBD.","keywords":["free discontinuity problems","functions of generalised bounded deformation","GSBD","symmetric gradient","matrix-valued measure","Cantor part","slicing","fine properties of functions"],"falsifier":"Evaluate formula (1.5) for a GBD function whose singular part is supported on a purely unrectifiable set that is not aligned with the coordinate axes, using $\\xi$, $\\eta$, $\\xi+\\eta$, and $\\xi-\\eta$ in a Borel set $B$ meeting that set; if the four numbers violate the parallelogram identity, the claimed quadraticity---and hence the existence of $\\mu_u$---fails.","tokens_in":56662,"feed_emoji":"📐","tokens_out":10396,"duration_ms":98384,"temperature":0.7,"pith_summary":"The paper proves that every function of generalised bounded deformation (GBD) carries a matrix-valued measure that does the work of a symmetric gradient even when no distributional gradient exists. For each threshold $r>0$ this measure $\\mu_{u,r}$ is obtained as the common limit of the directional derivatives of suitably truncated projections $\\tau_R(u\\cdot\\xi)$, and it splits into an absolutely continuous part with density the approximate symmetric gradient $Eu$, a jump part supported on the small jumps, and a Cantor part that is singular to Lebesgue measure and vanishes on sets that are $\\sigma$-finite for $H^{d-1}$. The same construction characterizes the smaller space GSBD: a GBD function belongs to GSBD exactly when its Cantor part is zero. This matters for fracture mechanics, where GBD is the natural frame for cohesive models and no distributional gradient is available.","feed_headline":"A symmetric-gradient measure exists for all GBD functions","feed_subtitle":"Its Cantor part vanishes precisely on GSBD, tying the new measure to fracture mechanics.","key_machinery":"The load-bearing object is the family of slice measures\n$$\n\\$\\sigma$^\\xi_u(B):=|\\xi|\\int_{\\Pi_\\xi} Du^\\xi_y\\big((B\\setminus $J^{1}$_u)^\\xi_y\\big)\\,$dH^{{d-1}}$(y),\n$$\nwhich integrate the one-dimensional BV derivative of the slice $u(y+t\\xi)\\cdot\\xi$, excluding points where the jump size is at least 1. The proof's central step is showing that for every Borel set $B$ the map $\\xi\\mapsto\\sigma^\\xi_u(B)$ is a quadratic form---2-homogeneous, lower bounded, and satisfying the parallelogram identity---so that by a standard characterization there is a symmetric matrix $\\mu_u(B)$ with $\\sigma^\\xi_u(B)=\\mu_u(B)\\xi\\cdot\\xi$. In dimension 2 the parallelogram identity is proved by discretising the slice integrals on the grid $\\{\\omega+(i/k)\\xi+(j/k)\\eta\\}$ and using a carefully chosen translation $\\omega$ so that grid segments meeting the jump set contribute a vanishing error; dimensions $d>2$ follow by a Fubini-type reduction to two-dimensional slices.","core_discovery":"The central result is that the slice data of $u\\in GBD(\\Omega)$ can be assembled into one symmetric-matrix-valued bounded Radon measure. Theorem 8.1 and Corollary 8.6 give, for every $r>0$, a measure $\\mu_{u,r}\\in M_b(\\Omega;\\mathbb{R}^{d\\times d}_{\\mathrm{sym}})$ such that for every unit vector $\\xi$\n$$\n\\mu_{u,r}(B)\\xi\\cdot\\xi=\\lim_{R\\to+\\infty} D_\\xi(\\tau_R(u\\cdot\\xi))(B\\setminus J^r_u)\n$$\nfor every Borel set $B$, with $\\tau_R$ the piecewise-linear truncation at $\\pm R/2$. The decomposition\n$$\n\\mu_{u,r}=\\mu^a_u+\\mu^c_u+\\mu^j_{u,r}\n$$\nholds with $\\mu^a_u(B)=\\int_B Eu\\,dx$, $\\mu^j_{u,r}(B)=\\int_{(J_u\\setminus J^r_u)\\cap B}[u]\\odot\\nu_u\\,dH^{d-1}$, and $\\mu^c_u$ singular with respect to $L^d$ and zero on $H^{d-1}$-$\\sigma$-finite Borel sets. Theorem 8.9 then characterises $GSBD(\\Omega)$ as the subspace of $GBD(\\Omega)$ consisting of functions with $\\mu^c_u=0$, giving an intrinsic definition of GSBD that no longer refers to slicing.","pith_inferences":["Editorial inference: if the decomposition is stable under the natural convergence for GBD, energies in cohesive fracture models could be written directly as sums over $\\mu^a_u$, $\\mu^c_u$, and $\\mu^j_u$, making the Cantor part the term that governs diffuse damage.","Editorial inference: the same slice-integral-to-quadratic-form mechanism may give a general criterion for when other 2-homogeneous, slice-defined set functions are represented by matrix-valued measures, which would apply to relaxations of free-discontinuity energies.","Editorial inference: the finite-direction criterion (Theorem 8.10) suggests a concrete algorithmic test for GSBD membership: verify SBV slicing behavior in $d(d+1)/2$ directions and compute $\\Lambda$; a positive test returns the quantitative bound $\\lambda_u(\\Omega)\\leq C_d\\Lambda$."],"forward_implications":["For every $u\\in GBD(\\Omega)$ and $r>0$, the directional measures $\\sigma^\\xi_u(B)$ are the quadratic form of a single symmetric-matrix-valued measure $\\mu_{u,r}$, so the limit in (1.3) exists simultaneously for all directions and all Borel sets avoiding $J^r_u$.","The three-part decomposition with explicit densities for the absolutely continuous and jump parts and a singular Cantor part vanishing on $H^{d-1}$-$\\sigma$-finite sets holds for every $u\\in GBD(\\Omega)$, not just for $BD$.","$GSBD(\\Omega)$ is exactly $\\{u\\in GBD(\\Omega): \\mu^c_u=0\\}$, an intrinsic characterization independent of slicing.","Using a recent finite-direction characterization of GBD, an $L^d$-measurable function belongs to $GSBD(\\Omega)$ as soon as the slice conditions hold for an orthonormal basis and their sums, with the quantitative bound $\\lambda_u(\\Omega)\\leq C_d\\Lambda$.","For $u\\in BD(\\Omega)$, $\\mu_{u,r}$ equals the distributional symmetric gradient $Eu$ restricted to $\\Omega\\setminus J^r_u$, so the new measure extends the classical BD structure theorem."],"supporting_citations":[{"why":"Introduces the spaces GBD and GSBD and supplies the slicing definition, the approximate symmetric gradient, and the jump-set rectifiability that the paper relies on.","marker":"[20]"},{"why":"Provides the structure theorem for BD whose three-part decomposition is the model being generalized to GBD.","marker":"[6]"},{"why":"Gives the recent characterization of GBD used in Theorem 8.10 to turn the intrinsic GSBD criterion into a finite-direction test.","marker":"[13]"},{"why":"Provides the estimate $H^{d-1}(J^1_u)<+\\infty$ that makes the small-jump integrand $[u]\\odot\\nu_u$ integrable.","marker":"[4]"},{"why":"Supplies the standard BV slicing theory used to relate one-dimensional slice derivatives $Du^\\xi_y$ to the measures $\\sigma^\\xi_u$ and $\\mu_u$.","marker":"[7]"},{"why":"Contains the characterization of quadratic functions used to convert the slice form $\\xi\\mapsto\\sigma^\\xi_u(B)$ into a symmetric matrix.","marker":"[19]"},{"why":"Supplies the Riemann-sum lemma underlying the discretization in Lemma 3.1 and the grid-shift estimates.","marker":"[21]"},{"why":"Provides geometric measure facts on coarea, projection, and slicing of Hausdorff measures used in the higher-dimensional Fubini argument.","marker":"[28]"}],"fun_headline_variants":["A matrix measure for every GBD function","GSBD means zero Cantor part in the new measure","New measure splits GBD into AC, Cantor, and jump parts","Beyond BD: symmetric gradient measure for all GBD"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument hinges on a grid-shift estimate in the plane: for every scale one can shift the grid so that the total variation of the slices that meet the jump set is negligible; if this estimate fails for some admissible function, the parallelogram identity---and with it the existence of $\\mu_u$---would collapse.","fun_headline_variants_meta":{"raw":{"variants":["A matrix measure for every GBD function","GSBD means zero Cantor part in the new measure","New measure splits GBD into AC, Cantor, and jump parts","Beyond BD: symmetric gradient measure for all GBD"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000867,"raw_usage":{"total_tokens":3775,"prompt_tokens":978,"completion_tokens":2797,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":594,"completion_tokens_details":{"reasoning_tokens":2731}},"tokens_in":594,"tokens_out":2797,"duration_ms":19147,"temperature":1.0,"reasoning_tokens":2731,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:21:22.931229+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate formula (1.5) for a GBD function whose singular part is supported on a purely unrectifiable set that is not aligned with the coordinate axes, using $\\xi$, $\\eta$, $\\xi+\\eta$, and $\\xi-\\eta$ in a Borel set $B$ meeting that set; if the four numbers violate the parallelogram identity, the claimed quadraticity---and hence the existence of $\\mu_u$---fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the spaces GBD and GSBD and supplies the slicing definition, the approximate symmetric gradient, and the jump-set rectifiability that the paper relies on."},{"cited_title":"Ambrosio, A","cited_arxiv_id":null,"evidence_quote":"Provides the structure theorem for BD whose three-part decomposition is the model being generalized to GBD."},{"cited_title":"A characterization of Generalized functions of Bounded Deformation","cited_arxiv_id":"2502.10861","evidence_quote":"Gives the recent characterization of GBD used in Theorem 8.10 to turn the intrinsic GSBD criterion into a finite-direction test."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the estimate $H^{d-1}(J^1_u)<+\\infty$ that makes the small-jump integrand $[u]\\odot\\nu_u$ integrable."},{"cited_title":"Dal Maso, An introduction to Γ-convergence, vol","cited_arxiv_id":null,"evidence_quote":"Contains the characterization of quadratic functions used to convert the slice form $\\xi\\mapsto\\sigma^\\xi_u(B)$ into a symmetric matrix."},{"cited_title":"Quasi-static evolution in brittle fracture: the case of bounded solutions","cited_arxiv_id":"math/0401198","evidence_quote":"Supplies the Riemann-sum lemma underlying the discretization in Lemma 3.1 and the grid-shift estimates."}],"review_version":2}