{"id":"efba4e19-f75c-41a6-9d08-009c5b57ed95","arxiv_id":"2412.02309","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Computing the inverse Jacobian exactly with automatic differentiation, instead of by Taylor series, lets the Q1STc+ element pass the membrane patch test that Q1STc fails and improves mesh-distortion behavior.","lead":"This paper improves a low-order 3D finite element method by using automatic differentiation to compute the inverse Jacobian exactly instead of approximating it with a Taylor series. The resulting element, Q1STc+, passes a standard membrane patch test that its predecessor failed and stays accurate on distorted meshes, with code released publicly.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Q1STc+ still uses det(J0) as volume measure and T0 at the element center; the acknowledged solids patch-test failure (14.5%, Table 3) may stem from these center approximations, not from the bilinear stress truncation.","rationale":"The paper is honest and the central comparison is meaningful: Q1STc+ passes the membrane patch test exactly while Q1STc does not, and the force response is insensitive to the tested mesh distortion while Q1STc drifts. These are genuinely falsifiable demonstrations, and the public code deposit is a positive. My concern does not overturn those findings. It targets the scope and the causal diagnosis of the remaining failure. The reader's conditional verdict is appropriate; I would not change it. The most load-bearing residual issue is that the formulation still uses det(J0) as the volume measure and T0 as the enhanced-strain transformation. These are center-point geometry approximations, independent of the inverse Jacobian, and they are not mentioned in the abstract's 'approximation-free' claim. The paper's own solids patch test (Table 3) shows errors up to 14.5%, which it attributes to the bilinear Taylor truncation of stress/strain quantities. That attribution is plausible but not established; the same symptom can arise from using det(J0) and T0, because the orthogonality condition (12b) is then not integrated with the true volume and transformation fields. If the volume measure is the dominant cause, the paper's proposed trilinear expansion will not fix the solids patch test. A single computational variant with exact det(J(ξ)) in the integrals would separate the two causes. Whether the result supports the Taylor-truncation diagnosis or the volume-measure diagnosis, the conditional verdict stands; only the revision guidance changes. No parameter was fitted to the benchmarks, and the failure is openly reported, so there is no reason to escalate to reject.","tokens_in":28737,"tokens_out":6963,"duration_ms":80848,"concrete_test":"Implement a variant of Q1STc+ in which all hourglass and enhanced-strain integrals use the exact volume element det(J(ξ)) (e.g., 2x2x2 Gauss quadrature) while keeping the AD inverse Jacobian and the center-valued T0; then rerun the solids patch test of Section 8.1 (Table 3). If the Sxy, Sxz, and Syz errors drop below about 1%, the solids failure is caused by the det(J0) volume approximation; if the errors persist, it is caused by the bilinear Taylor truncation of Eq. (19) or by the center-valued T0. This single change isolates the volume-measure approximation from the inverse-Jacobian computation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central improvement claim rests on replacing the Taylor expansion of J^{-1} (Eqs. 21-24) with an exact AD evaluation. However, the element formulation still contains two center-point geometry approximations independent of J^{-1}: (i) the volume measure in all element integrals is dV0_e = det(J0) dξ dη dζ (Section 7, below Eq. 26), not the exact det(J(ξ)) measure; (ii) the enhanced strain operator uses T0 evaluated at the element center (Eq. 18), not the full field T(ξ). Consequently, 'approximation-free' applies only to the inverse Jacobian, not to the geometry treatment. The solids patch test (Table 3) fails with errors up to 14.5%, and Section 9 attributes this to 'non-sufficient approximation of the strain and constitutive quantities by a Taylor series expansion,' proposing trilinear terms. But the same failure can be explained by the det(J0)/T0 center approximations: for an exact constant-stress field, the orthogonality condition (12b) is enforced with det(J0) and T0, so the incompatible modes W can be spuriously activated when det(J) and T vary over the element. If that is the actual cause, the proposed trilinear fix will not restore the solids patch test. This does not disprove the demonstrated membrane patch and distortion improvements, but it means the paper's diagnosis of its own limitation is untested, and the scope of the 'enhanced' claim is narrower than 'approximation-free geometry.'","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents Q1STc+, a modification of the single-Gauss-point hourglass-stabilized hexahedral element Q1STc, in which the Taylor-series approximation of the inverse Jacobian used by Q1STc is replaced by an exact automatic differentiation (AD) evaluation. The authors compare Q1STc and Q1STc+ on the membrane patch test and the solid patch test, on a cube-under-compression locking benchmark, and on an elasto-plastic asymmetrically notched specimen with distorted meshes, using the mixed U-P-SBFEM formulation as a reference. The main reported outcomes are that Q1STc+ reproduces the analytical membrane patch stresses to four decimals while Q1STc does not, that Q1STc+ is insensitive to the tested mesh distortions in the notched specimen while Q1STc is not, and that Q1STc+ converges in the cube compression where Q1 locks. The paper also reports that both Q1STc and Q1STc+ fail the solid patch test, with Q1STc+ errors up to 14.5%, and attributes this to the bilinear Taylor truncation of strain and constitutive quantities.","tokens_in":28934,"tokens_out":6218,"duration_ms":68536,"significance":"If the results hold, the contribution provides a simple implementation-level improvement to a widely used reduced-integration element family: replacing a hand-derived Taylor approximation of the inverse Jacobian by an AD evaluation improves membrane patch behavior and distortion robustness without changing the element topology or the stabilization framework. The public availability of the element and material routines (Zenodo DOI) is a concrete strength, as is the use of an external analytical benchmark (Macneal and Harder membrane patch test) and an independent locking-free reference (U-P-SBFEM) for key comparisons. The significance is, however, moderated by the paper's own solid patch test failure, which leaves the scope of the improvement narrower than the phrase 'exact geometry' might suggest; the central claim is better stated as 'exact inverse Jacobian with remaining center-point approximations elsewhere in the formulation.'","major_comments":[{"comment":"The paper's conclusion states that 'the geometry of the system was calculated exactly using AD,' but the element formulation still uses the center-point volume measure dV0_e = det(J0) dξ dη dζ in all element integrals (Section 7, below Eq. (26)) and the center-point transformation T0 in the enhanced B-operator (Eq. (18)). Thus 'approximation-free' is demonstrated only for the inverse Jacobian, not for the geometry treatment as a whole. This distinction is load-bearing because the solid patch test failure documented in Table 3 could originate from these remaining center-point approximations rather than from the bilinear truncation in Eq. (19). The authors should either temper the conclusion or provide a diagnostic experiment: evaluate the solid patch test with an exact det(J(ξ)) and full T(ξ) field (e.g., via numerical quadrature) to identify which approximation actually produces the observed errors.","section":"§7, Eq. (26); §4, Eq. (18); §9"},{"comment":"The paper attributes the solid patch test failure to the 'non-sufficient approximation of the strain and constitutive quantities by a Taylor series expansion' and proposes extending the expansion to trilinear terms. This diagnosis is not tested by any numerical experiment. The same failure is equally consistent with the remaining det(J0) and T0 center-point approximations identified in Eq. (18) and Eq. (26); since a trilinear stress/strain expansion would not remove those center-point approximations, the proposed fix may not restore the solid patch test. The authors should either implement and test the trilinear extension or present a numerical comparison that isolates the two error sources before offering this explanation as the likely cause.","section":"§8.1, Table 3; §9"},{"comment":"The convergence study for the cube-under-compression example plots only Q1 and Q1STc+, with no reference solution. The text uses this figure to support the claim that Q1STc+ avoids volumetric locking, but a plateau in the displacement-versus-mesh curve only demonstrates mesh insensitivity; it does not by itself show that the converged value is the correct one. Adding the U-P-SBFEM result already used elsewhere in the paper, or another well-established reference solution, to Figure 7 would directly support the locking-free claim and remove the ambiguity.","section":"§8.2, Figure 7"}],"minor_comments":[{"comment":"The element name is written inconsistently as 'Q1Stc+' in Table 1 and in several places in Section 8.1; it should be 'Q1STc+' everywhere.","section":"§8.1, Table 1 and text"},{"comment":"The material model name is written both as 'St.Venant Kirchhoff' and 'St. Venant-Kirchhoff'; please unify the spelling.","section":"§8.1"},{"comment":"In Figure 7, adding a horizontal reference line or a U-P-SBFEM curve, as suggested in the major comments, would make the convergence behavior immediately interpretable.","section":"§8.2, Figure 7"},{"comment":"The unit of the kinematic hardening parameter b is listed as '-' in Table 2, but the term b/(2a) tr(dev(Θ)^2) in Eq. (7) suggests that b has units of stress; please clarify the units or the normalization.","section":"§2, Eq. (7); Table 2"}],"recommendation":"major_revision","confidential_remarks":"The core numerical contribution is plausible and the membrane patch and distortion results are well supported. The main reservation is that the paper's own diagnosis of the solid patch test failure is untested and the 'exact geometry' wording overstates the actual scope of the approximation-free part. If the authors can perform the suggested diagnostic test (exact det(J) and T versus trilinear stress expansion) or clearly limit their claims, the paper would be suitable for publication. The open-source code and the use of an independent reference element are positive features that strengthen the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I'll cut to the chase: this paper does something small and real. Q1STc+ computes the inverse Jacobian exactly with AD instead of the Taylor series used in Q1STc, and the result is that the membrane patch test is now passed to four decimals (Table 1), the force response stops drifting under the tested mesh distortions (Fig. 11b), and the element converges on the cube compression where Q1 locks (Fig. 7). Those are concrete, reproducible improvements, and the paper earns credit for showing them and for openly reporting that the solids patch test still fails with errors up to 14.5% (Table 3). No parameter was fit to any benchmark; the hourglass tangent and effective shear modulus come from the element state via the prior Q1STc construction.\n\nWhere I part company with the authors is the 'approximation-free' framing. It is true that J^{-1} is no longer approximated, but the element still uses det(J0) as the volume measure in every integral and T0 at the element center for the enhanced strain operator (Eq. 18, Section 7). So the geometry handling is only partially exact. The stress-test note makes a sharper point: the solids patch test failure could be caused by those center-point approximations, not by the bilinear truncation of stress that the authors blame. They propose trilinear terms as the fix, but that diagnosis is untested. If the real culprit is det(J0)/T0, the trilinear expansion won't restore the patch test. I'd want the authors to actually run that experiment or at least discuss the alternative.\n\nMinor: the validation element U-P-SBFEM shares co-authors and the material formulation, so it is not an independent benchmark. The Zenodo deposit is a plus, but reproduction requires the proprietary FEAP/AceGen/Mathematica stack and the deposit lacks a commit hash and run instructions.\n\nVerdict: this is a competent, honest, incremental contribution. It deserves a serious referee. I'd send it out, but condition the acceptance on narrowing the abstract-level claims and on either fixing or cleanly scoping the solids patch test. Whoever referees it should ask the authors to test whether the remaining patch failure is a det(J0)/T0 artifact before they spend a paper on trilinear terms.","headline":"A modest, honest improvement to a one-point hex element: exact AD inverse Jacobian fixes the membrane patch test and distortion sensitivity, but the solids patch test still fails and the 'approximation-free' label overreaches.","tokens_in":29650,"tokens_out":2527,"would_cite":true,"duration_ms":25493,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that replacing the Taylor-series approximation of the inverse Jacobian with an exact automatic-differentiation computation turns the one-point hourglass-stabilized brick element Q1STc into a version, Q1STc+, that passes…","keywords":["hourglass stabilization","reduced integration","automatic differentiation","inverse Jacobian","enhanced assumed strain","volumetric locking","elasto-plasticity","patch test"],"falsifier":"Rerun the seven-element solid patch test with Q1STc+ after adding a trilinear $\\xi\\eta\\zeta$ term to the Taylor expansion behind Equation (19). If the stress errors do not fall from their reported 14.5% peak toward the analytical values $2001.5$ and $400.5\\ \\mathrm{N/mm^2}$, then the residual failure is not caused by the bilinear truncation and the paper's explanation of the missing patch-test pass is wrong.","tokens_in":28404,"feed_emoji":"📐","tokens_out":10934,"duration_ms":101718,"temperature":0.7,"pith_summary":"The paper is trying to establish that a reduced-integration three-dimensional brick element can be improved by replacing the Taylor-series approximation of its inverse Jacobian with an exact evaluation computed by automatic differentiation. The upgraded element, named Q1STc+, is an eight-node brick with a single integration point and hourglass stabilization, and the claimed payoff is concrete: it passes the membrane patch test—a benchmark where a distorted patch of elements must reproduce a uniform stress—where its predecessor does not, and its force response no longer drifts when meshes are distorted. The authors do not claim the element is finished: both the old and new versions still fail the three-dimensional solid patch test, with stress errors up to 14.5%, and the paper points to missing trilinear terms in the stress expansion as the likely cause. If the claim is right, exact Jacobian handling is a practical, low-cost upgrade for a widely used class of one-point elements, and the residual patch-test failure marks an explicit next step.","feed_headline":"Exact Jacobian lets a one-point brick pass the membrane test","feed_subtitle":"Automatic differentiation replaces the Taylor-series inverse Jacobian and removes mesh-distortion drift in Q1STc+.","key_machinery":"The load-bearing mechanism is the pointwise inverse Jacobian and its evaluation. In Q1STc, $J^{-1}$ is expanded as $J^{-1} \\approx (J^0)^{-1} - \\sum_i (J^0)^{-1} J^{\\xi_i}(J^0)^{-1}\\xi_i$, so all geometry enters the analytically integrated residual through a linear approximation taken at the element center. Q1STc+ replaces that formula with the exact matrix inverse $j = J(\\xi)^{-1}$ computed by automatic differentiation, so the cartesian strain components inherit the true geometry pointwise. The rest of the hourglass stabilization still works through a Taylor expansion, up to bilinear terms, of the second Piola-Kirchhoff stress and the compatible and enhanced B-operators, with the volume element still approximated as $dV_e \\approx \\det(J^0)\\,d\\xi\\,d\\eta\\,d\\zeta$; only the inverse Jacobian is approximation-free.","core_discovery":"On its own terms, the paper's central claim is that the Taylor-series expansion of the inverse Jacobian, used in the Q1STc formulation to keep the one-point element analytically integrable, is both unnecessary and harmful. With automatic differentiation, the inverse Jacobian $J^{-1}$ can be evaluated exactly at any point of the element, so the enhanced formulation Q1STc+ computes the geometry exactly while keeping all other ingredients of Q1STc unchanged. The payoff is that Q1STc+ passes the membrane patch test for linear elasticity, with stresses $1334.20$ and $400.50\\ \\mathrm{N/mm^2}$ reproduced to four decimals and no spurious out-of-plane stress; in the elasto-plastic version it agrees with a locking-free reference element where Q1STc overestimates the reaction force once plasticity starts; and on distorted meshes the force-displacement response no longer depends on distortion amplitude. The paper also states the limitation plainly: the solid patch test is still not passed, and the cause is attributed to the bilinear truncation of the Taylor expansion for the strain and constitutive quantities, not to the Jacobian approximation.","pith_inferences":["Because the volume measure still uses $\\det(J^0)$ and the enhanced strain operator still uses $T^0$ at the element center, 'approximation-free' applies only to the inverse Jacobian; testing exact evaluation of those two quantities as well would show whether the remaining 14.5% solid-patch error is purely a truncation effect.","The membrane patch test passes on a thin structure where out-of-plane stress vanishes, so the gain from exact Jacobian geometry is most visible where the structure is dominated by in-plane response; the same improvement may not carry over to a full three-dimensional stress state, as the solid patch test shows.","The proposed trilinear extension is directly testable in the same automatic-differentiation framework: add the $\\xi\\eta\\zeta$ term to the expansion behind Equation (19) and rerun the seven-element patch, and if exactness is restored, the truncation hypothesis is confirmed.","Since the material derivations assume volume-preserving plasticity with $J_p=1$, the element's behavior under non-volume-preserving models such as pressure-dependent plasticity remains an open empirical question."],"forward_implications":["A one-point hourglass-stabilized brick can pass the membrane patch test for a St. Venant-Kirchhoff material without any extra integration points, because exact Jacobian geometry removes the stress scatter that the Taylor-based element produced.","In the elasto-plastic membrane test, Q1STc+ matches the locking-free reference curve, so the geometry upgrade also fixes the over-stiff plastic response of Q1STc.","Mesh distortion in the plane of the asymmetrically notched specimen no longer shifts the force-displacement curve for Q1STc+, whereas Q1STc drifts visibly for distortions of $\\pm 0.5$ mm.","In the cube-compression benchmark, Q1STc+ converges already at $10\\times10\\times10$ elements while the full-integration Q1 element still locks at $18\\times18\\times18$ elements.","The solid patch test remains an open failure for both formulations, and the paper's suggested remedy is to extend the Taylor expansion to the trilinear terms."],"supporting_citations":[{"why":"Supplies the base Q1STc formulation whose Taylor-approximated inverse Jacobian is replaced by the exact computation.","marker":"[7]"},{"why":"Establishes the linear Taylor approximation of the inverse Jacobian and transformation matrix that Q1STc+ removes.","marker":"[65, 66]"},{"why":"Provides the locking-free mixed scaled boundary element used as the reference solution for validation.","marker":"[61]"},{"why":"Defines the membrane and solid patch test problems and the analytical stresses the elements must reproduce.","marker":"[51]"},{"why":"Introduces the enhanced assumed strain two-field variational foundation of the element.","marker":"[68]"},{"why":"Extends the enhanced strain concept to three-dimensional trilinear elements, the lineage of Q1STc.","marker":"[67]"}],"fun_headline_variants":["AD exact Jacobian: one-point brick clears membrane test","No Taylor series: AD exact Jacobian passes membrane test","Exact Jacobian via AD removes mesh distortion drift","One-point brick's exact Jacobian passes membrane test"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the bilinear Taylor truncation of strain and stress quantities suffices for hourglass stabilization and for capturing the element response, so that only the Jacobian needs to be exact; the paper's own seven-element solid patch test contradicts that assumption, with errors up to 14.5%.","fun_headline_variants_meta":{"raw":{"variants":["AD exact Jacobian: one-point brick clears membrane test","No Taylor series: AD exact Jacobian passes membrane test","Exact Jacobian via AD removes mesh distortion drift","One-point brick's exact Jacobian passes membrane test"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001148,"raw_usage":{"total_tokens":4761,"prompt_tokens":947,"completion_tokens":3814,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":3749}},"tokens_in":563,"tokens_out":3814,"duration_ms":27541,"temperature":1.0,"reasoning_tokens":3749,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:37:45.295422+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rerun the seven-element solid patch test with Q1STc+ after adding a trilinear $\\xi\\eta\\zeta$ term to the Taylor expansion behind Equation (19). If the stress errors do not fall from their reported 14.5% peak toward the analytical values $2001.5$ and $400.5\\ \\mathrm{N/mm^2}$, then the residual failure is not caused by the bilinear truncation and the paper's explanation of the missing patch-test pass is wrong.","supporting_citations":[{"cited_title":"Sauren and S","cited_arxiv_id":null,"evidence_quote":"Provides the locking-free mixed scaled boundary element used as the reference solution for validation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the membrane and solid patch test problems and the analytical stresses the elements must reproduce."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the enhanced assumed strain two-field variational foundation of the element."}],"review_version":1}