{"id":"2951874a-2179-4e0f-bb37-b0b559327f4a","arxiv_id":"2504.19960","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Closed-form and manufactured solutions for the EMI cardiac cell model are derived and numerically verified through mortar FEM with operator splitting.","lead":"The paper derives exact and manufactured analytical solutions for the EMI model of cardiac electrophysiology, for one cell, two coupled cells, and N cells. It verifies them with mortar finite element simulations, offering benchmarks for testing EMI solver codes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Experiment 2's printed two-cell solution violates v=u_i-u_e and the interface flux condition, and the §3.2 existence condition for the gap-junction IVP is self-contradictory; Eq. (13) is not verified as written, so the central benchmark claim needs correction before it can be trusted.","rationale":"We agree with the CONDITIONAL verdict, but for a slightly different primary reason. The zero-flux manufactured benchmark is a genuine limitation: because ∇X·n=0 on all interfaces, the Section 5.2.3 test does not exercise the normal-flux or gap-junction coupling that defines the EMI model. That concern is valid and should be stated as a caveat. However, the more load-bearing problem is internal correctness: the printed Experiment 2 solution in Section 5.2.2 does not satisfy the EMI interface equations, and the existence condition after eq. (13l) is logically inconsistent. Since a benchmark is only useful if the reference solution actually solves the model, these inconsistencies directly undermine the strongest claim that eq. (13) provides a valid analytical solution. The single-cell solution and the N-cell manufactured solution appear coherent, and the two-cell derivation may be salvageable as a one-parameter family with corrected example formulas, so rejection would be too harsh. A conditional acceptance requiring correction of the two-cell formulas, clarification of the gap-junction IVP condition, and an explicit statement of the zero-flux limitation of the manufactured benchmark is the appropriate outcome, which matches the reader's overall verdict.","tokens_in":16074,"tokens_out":17418,"duration_ms":172954,"concrete_test":"Use a symbolic algebra system to substitute the Section 5.2.2 formulas for ue, u_i^(k), and v^(k) into the residuals of (12d), (12e), and (12g) at ρ=ρ1=5 for k=1,2, evaluating R_e=u_i^(k)-ue-v^(k) and R_f=(σi∂ρu_i^(k)-σe∂ρu_e) on the membrane. If either residual is nonzero, the printed two-cell solution is invalid. If the GitHub repository code uses different formulas, this confirms a typographical error and the corrected formulas should be inserted; if the code matches the printed formulas, the Experiment 2 convergence tables should be rerun with a genuine solution of (12).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that eqs. (11), (13), and (14) are exact references for EMI solver verification. The least secure link is the two-cell family. In §3.2, after eq. (13l), the authors state that w=(v0^(1)-v0^(2))e^{-t/(C_m R_m)} satisfies the gap-junction IVP either when v0^(1)=v0^(2) and w_rest=w0=0, or when v0^(1)≠v0^(2), w0=0, and C_m^(1,2)R_m^(1,2)=C_m R_m. The second case is impossible: at t=0 the expression equals v0^(1)-v0^(2)≠0, not w0=0. In the printed Experiment 2 example (§5.2.2), substituting k=2, ρ1=5, C=R=1, v_rest=5, A=-50e^{-t/10}cos(t) into (12e) gives u_i^(k)-u_e at ρ=5 that does not equal the printed v^(k); for instance at t=0 the difference is about 2 while v^(2) is about 10.8, and the stated initial value v0^(2)=30 is not reproduced. The flux condition (12d) also fails by a factor of 5: both σi∂ρu_i^(2) and σe∂ρu_e are radial, but at ρ=5 the printed u_i^(2) gives ∂ρu_i^(2)=-0.4e^{-t/10}cos(t), whereas ∂ρu_e=-2e^{-t/10}cos(t), so the normal fluxes do not match. The manufactured solution's zero normal flux condition in §3.3 is a real scope limitation, but it does not make eq. (14) invalid; the two-cell inconsistencies make eq. (13) unverified as printed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents three families of reference solutions for the EMI model of cardiac electrophysiology: a single two-dimensional cell solution in polar coordinates, a two-cell solution in spherical coordinates, and an N-cell manufactured solution in Cartesian coordinates. For each it gives a derivation, parameter restrictions, and a numerical experiment using a mortar finite-element/operator-splitting implementation. The stated goal is to provide exact benchmarks that EMI solver codes should reproduce.","tokens_in":16481,"tokens_out":15903,"duration_ms":145260,"significance":"The single-cell family is clean and self-consistent: with the stated parameter choices, Eq. (11) satisfies the Laplace equations, the membrane relation v = u_i - u_e, and the initial condition, and it exercises a nonzero membrane current. The manufactured solution is a standard MMS construction with explicit forcing, and the accompanying code and data are publicly available. However, the two-cell family in Eq. (13) is not correct as printed, the Experiment 2 formulas do not satisfy the EMI interface conditions, and none of the benchmarks exercises the gap-junction current coupling. The paper therefore has a useful core but cannot currently be used as a reliable reference for EMI verification.","major_comments":[{"comment":"The statement that w=(v0^(1)-v0^(2))e^{-t/(C_m R_m)} satisfies the gap-junction IVP (12j)-(12k) in the case v0^(1)≠v0^(2), w0=0 is inconsistent: at t=0 the expression equals v0^(1)-v0^(2)≠0, not w0=0. The correct compatibility condition is w0=v0^(1)-v0^(2), w_rest=0, and C_m^(1,2)R_m^(1,2)=C_mR_m. As printed, the derivation of Eq. (13) does not establish a valid family for unequal initial transmembrane potentials.","section":"Section 3.2, after Eq. (13l)"},{"comment":"The printed Experiment 2 formulas do not solve the EMI model. At ρ=5 and t=0, u_i^(2)-u_e = 2, whereas the printed v^(2) equals 1954/181≈10.8, and the stated initial value v0^(2)=30 is not reproduced. The normal derivatives also fail (12d): ∂ρ u_e = -2e^{-t/10}cos(t) while ∂ρ u_i^(2) = -0.4e^{-t/10}cos(t), so the membrane flux balance is violated by a factor of 5. Consequently, Eq. (13) is not verified by Experiment 2, and the error norms in Table 2 are not errors against a solution of the stated problem.","section":"Section 5.2.2, Eqs. (12d), (12e), (12g)"},{"comment":"The manufactured solution is constructed with ∇X·n=0 on all membranes and gap junctions, so I_m^(k)=I_m^(k,ℓ)=0 on every interface. This means Experiment 3 verifies only the volume equations and the membrane/gap-junction ODEs with zero normal-flux coupling; it does not exercise the transmission conditions (4b) and (4d) that distinguish the EMI model from decoupled Poisson problems. The abstract's claim that the solutions verify 'the accuracy of numerical simulations of the EMI model' should be tempered to state this scope limitation explicitly.","section":"Section 3.3, after Eq. (14e)"}],"minor_comments":[{"comment":"The superscript w^(2,1) is used here while the system is written for w^(1,2); standardize the notation.","section":"Section 3.2, Eq. (13l)"},{"comment":"The caption repeats u_i^(2) twice; it should refer to u_i^(1) and u_i^(2).","section":"Figure 6 caption"},{"comment":"The table contains spacing artifacts such as '5 .929' and '2 .736'; please reformat the error entries.","section":"Table 1"},{"comment":"The gap-junction resistance R_m^(k,l) is not specified, although the choice g^(k,l)=0 with w=(k-l)e^{-t}X implicitly requires R_m^(k,l)=1; state this parameter.","section":"Section 5.2.3"},{"comment":"The statement that the error domain for u_i^(1), u_i^(2), and w is 4≤ρ≤5 is ambiguous because w is an interface variable; clarify how surface errors are computed and whether the reported norms depend on the artificial boundary at ρ=3.","section":"Table 2 and Section 5.2.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's core idea is reasonable and the single-cell and manufactured parts are usable, but the two-cell solution and its numerical experiment are not. I would be willing to review a revision in which the two-cell family is corrected, the Experiment 2 formulas and tables are regenerated, and the no-flux limitation of the manufactured solution is stated prominently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The single-cell and N-cell manufactured benchmarks are usable, but the two-cell exact solution as printed does not satisfy the EMI equations. The derivation in §3.2 is coherent, but the printed Experiment 2 formulas in §5.2.2 don't reproduce the stated initial condition, violate v = u_i − u_e at ρ=5, and fail the membrane flux condition by a factor of five. As printed, eq. (13) is not verified.\n\nWhat's actually new: a family of analytical solutions for two coupled semi-spherical cells (new for the EMI literature) and a manufactured solution for N cells. The single-cell derivation is clean, and the N-cell manufactured solution is self-consistent by construction. The authors provide code and data, and the numerical convergence tables show the expected behavior. This is a legitimate extension of prior single-cell manufactured solutions.\n\nSoft spots, in proportion. First, the two-cell printed example is wrong in a way that is probably typographical but nevertheless load-bearing for the paper's central claim. The second is a logical contradiction in §3.2: the statement that (13l) satisfies the gap-junction IVP when v0^(1) ≠ v0^(2) and w0 = 0 is impossible, because at t=0 the expression equals the difference, not zero. The only consistent case is equal initial v's and zero resting potential. Third, both the two-cell family and the manufactured solution impose zero normal flux at the gap junctions and membranes, so neither benchmark exercises the normal-flux coupling that is the defining feature of the EMI model. That is a scope limitation, not an invalidity, but it means passing these tests doesn't certify a solver's interface handling. The truncated error norms in Experiment 2 are minor and disclosed.\n\nIf the two-cell formulas are corrected, this becomes a solid suite of benchmarks. As it stands, I'd trust the single-cell and manufactured parts, and treat the two-cell example as under repair. The paper deserves a serious referee: the errors are checkable and likely fixable, and the benchmark suite fills a real gap for EMI solver verification. I'd accept it for review with the expectation that the two-cell example is corrected or removed.","headline":"The single-cell and N-cell manufactured benchmarks are usable, but the two-cell exact solution as printed does not satisfy the EMI equations, so the paper's strongest new claim needs correction.","tokens_in":17012,"tokens_out":6246,"would_cite":true,"duration_ms":59958,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M60","65M12","65M15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives exact reference solutions for the Extracellular-Membrane-Intracellular cardiac model and verifies that a mortar finite-element solver reproduces them under refinement.","keywords":["Extracellular-Membrane-Intracellular model","Manufactured solution","Cardiac electrophysiology","Operator-splitting method","Mortar finite element method","Analytical solution","Convergence verification","Benchmark"],"falsifier":"Substitute equations (11), (13), and (14) into the original interface conditions (4b) and (4d): in every case the normal-flux term vanishes because $\\nabla X \\cdot \\hat{n} = 0$, so the solutions are exact for zero-flux data. To test whether they serve as a general benchmark, construct an EMI problem with a nonzero constant normal flux on a membrane and check whether any member of these families can satisfy it; the continuity condition shows no member can, so the family's validity is restricted to the zero-flux case.","tokens_in":15828,"feed_emoji":"❤️","tokens_out":6823,"duration_ms":70284,"temperature":0.7,"pith_summary":"This paper tries to give numerical analysts working on cardiac tissue models a set of exact reference solutions for the Extracellular-Membrane-Intracellular (EMI) model: a family for one circular cell in two dimensions, a family for two coupled semi-spherical cells in three dimensions, and a manufactured solution for N rectangular cells in three dimensions. These are the kind of test problems a simulation code must reproduce, so they let developers measure discretization error and convergence order. The paper verifies them by running a mortar finite element solver with operator splitting and showing that the L2 errors decrease as the mesh and time step shrink. If the solutions are correct, they give the EMI community concrete benchmarks where previously mainly single-cell manufactured solutions existed.","feed_headline":"Exact solutions give EMI cardiac simulations a benchmark to hit","feed_subtitle":"One-, two-, and many-cell reference solutions let solver developers measure how accurate their code really is.","key_machinery":"The machinery is separation of variables under radial symmetry, plus the method of manufactured solutions for the Cartesian case. In the radial problems, assuming the potentials depend only on the radius reduces the Laplace equations to one-dimensional integrals, and the equal-conductivity assumption makes the interface flux condition reduce to equality of the leading integration constants; the remaining constants are fixed by the ordinary differential equations for the passive membrane and gap-junction potentials. In the manufactured solution, the spatial factor $X = \\cos(2\\pi x/\\alpha_g)\\cos(2\\pi y/\\beta_g)\\cos(2\\pi z/\\gamma_g)$ is chosen so that $\\nabla X \\cdot \\hat{n} = 0$ on all interior interfaces, which trivializes the normal-flux coupling, while the time factors are exponentials and forcing terms are added to make the potentials satisfy Poisson equations and the passive cell models.","core_discovery":"The central claim is that equations (11), (13), and (14) are valid analytical and manufactured solutions of the EMI model with passive cell models, and that they are usable as benchmark references. The first family solves a single radially symmetric cell in polar coordinates; the second solves two coupled radially symmetric semi-spherical cells in spherical coordinates; both require equal intracellular and extracellular conductivities. The third is a true manufactured solution for N identical cuboidal cells in Cartesian coordinates, built by separation of variables with a cosine profile whose normal derivative vanishes on every membrane and gap junction, together with explicit forcing terms. Numerical experiments with the mortar finite element method and operator splitting reproduce all three sets of solutions, with L2 errors that decrease under mesh and time-step refinement.","pith_inferences":["Because every constructed solution has zero normal flux across membranes and gap junctions, a solver that mishandles the flux-coupling interface conditions could still pass all three benchmarks; a test that enforces a nonzero transmembrane current is still needed.","Extending the derivation to unequal intracellular and extracellular conductivities, or to anisotropic conductivities, would produce benchmarks that actually exercise the transmembrane current, which none of the current solutions do.","One could adapt the manufactured-solution construction to a spatial factor with nonzero normal derivative and solve the resulting consistency condition on the time factors; the paper's zero-flux choice is sufficient for verification but not necessary."],"forward_implications":["EMI solver developers can use these solutions as convergence benchmarks for spatial and temporal discretization in both two and three dimensions.","The single-cell and two-cell families provide time-dependent reference data, so they test time integrators and operator-splitting error in addition to spatial accuracy.","The manufactured solution works for any number N of cells in a sheet, allowing scaling tests and parallel implementations to be checked against a known answer.","Because the benchmark family covers coupled cells and gap junctions, it goes beyond the previous single-cell manufactured solutions used in the EMI literature."],"supporting_citations":[{"why":"Supplies the EMI model equations and the mortar finite element framework that the paper's numerical verification builds on.","marker":"[1]"},{"why":"Earlier manufactured-solution validation of an EMI solver for a single cell; the paper extends this approach to multiple cells.","marker":"[3]"},{"why":"Earlier splitting-based solver for cell-by-cell models with a manufactured-solution benchmark; provides comparison for the splitting approach.","marker":"[4]"},{"why":"Finite element numerical analysis of the EMI model; supplies the spatial discretization context for the convergence tests.","marker":"[5]"}],"fun_headline_variants":["Analytical EMI solutions now available for solver verification","Benchmark EMI cardiac models with exact analytical solutions","EMI model gets exact solutions for testing numerical codes","New analytical solutions verify EMI cardiac simulations","Reference solutions for EMI cardiac electrophysiology released"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a solution with zero normal flux across all membranes and gap junctions is enough to verify an EMI solver, because the proposed benchmarks never exercise the coupling currents that distinguish the EMI model from a set of decoupled diffusion problems.","fun_headline_variants_meta":{"raw":{"variants":["Analytical EMI solutions now available for solver verification","Benchmark EMI cardiac models with exact analytical solutions","EMI model gets exact solutions for testing numerical codes","New analytical solutions verify EMI cardiac simulations","Reference solutions for EMI cardiac electrophysiology released"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000556,"raw_usage":{"total_tokens":2588,"prompt_tokens":829,"completion_tokens":1759,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":445,"completion_tokens_details":{"reasoning_tokens":1689}},"tokens_in":445,"tokens_out":1759,"duration_ms":11439,"temperature":1.0,"reasoning_tokens":1689,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:39:26.991119+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Substitute equations (11), (13), and (14) into the original interface conditions (4b) and (4d): in every case the normal-flux term vanishes because $\\nabla X \\cdot \\hat{n} = 0$, so the solutions are exact for zero-flux data. To test whether they serve as a general benchmark, construct an EMI problem with a nonzero constant normal flux on a membrane and check whether any member of these families can satisfy it; the continuity condition shows no member can, so the family's validity is restricted to the zero-flux case.","supporting_citations":[{"cited_title":"Berre, M","cited_arxiv_id":null,"evidence_quote":"Earlier manufactured-solution validation of an EMI solver for a single cell; the paper extends this approach to multiple cells."},{"cited_title":"A splitting, discontinuous Galerkin solver for the cell-by-cell electroneutral Nernst-Planck framework","cited_arxiv_id":"2404.08320","evidence_quote":"Earlier splitting-based solver for cell-by-cell models with a manufactured-solution benchmark; provides comparison for the splitting approach."},{"cited_title":"Fokoué, Y","cited_arxiv_id":null,"evidence_quote":"Finite element numerical analysis of the EMI model; supplies the spatial discretization context for the convergence tests."}],"review_version":1}