{"id":"fa3002f5-585e-4afe-bb51-2b97643a60ca","arxiv_id":"1908.04533","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A boundary integral equation method computes conformal capacity, hyperbolic capacity, and elliptic capacity for a wide class of planar doubly connected domains, with relative errors near 1e-14 on smooth test cases.","lead":"This paper presents a boundary integral method for computing conformal capacity and related invariants of planar doubly connected domains, using conformal maps onto annuli. It validates the method against exact formulas for many geometries and provides MATLAB code.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 8.2's antipodal set E* = {-1/a} omits the complex conjugate; Section 8.5's elliptic capacity for the non-symmetric amoeba is computed for the wrong complementary domain and the cape < caph claim is unverified.","rationale":"The reader's CONDITIONAL verdict is well supported: the BIE method is standard, the code is public, and the many exact-solution validations in Sections 3-7 give strong evidence for the core capacity algorithm on smooth and piecewise smooth boundaries. The iterative [33] preimage is indeed used as a black box without convergence analysis, but Section 4.1 and Section 6 provide relative errors near 1e-14 for slit and half-plane slit configurations against exact formulas, so the auxiliary mapping is at least empirically controlled where testable. The more dangerous point is the Section 8.2 definition: 'antipodal' and the cited elliptic-capacity theory require z -> -1/\\bar z, while the paper writes z -> -1/z. The disk and square examples hide this typo because those sets are conjugation-symmetric; the amoeba example is non-symmetric and is exactly where the discrepancy would appear. Recomputing Table 12 with the conjugate is a quick decisive test, and a second test with two disjoint non-symmetric disks would settle the definition unambiguously. If the concern lands, the elliptic-capacity results for non-symmetric sets need revision, but the core method and the rest of the paper remain valid; hence I keep the reader's CONDITIONAL verdict rather than escalating to REJECT.","tokens_in":23635,"tokens_out":24612,"duration_ms":237733,"concrete_test":"Confirm from [13] the intended definition of E*. Then recompute the last column of Table 12 for the amoeba of Section 8.5 using the corrected boundary eta*(t) = -1/conj(eta(t)) and z2 = -1/conj(z1) in annq with type 'u'. If any table entry changes by more than 1e-10, the published cape values are for the wrong domain. As a secondary, exact check, take the non-symmetric disk E = {z : |z - 0.5i| <= 0.1}; compute the modulus of the annulus between E and -1/E versus E and -1/conj(E) using the known formula for two disjoint disks, and compare with the two corresponding sqrt(q) outputs from annq.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is in Section 8.2: 'we define the antipodal set E* = {-1/a : a in E}.' In the Duren-Kuhnau theory cited as [13], the elliptic capacity is tied to the fixed-point-free anti-Mobius involution z -> -1/\\bar z, which is the true antipodal map on the Riemann sphere; the displayed map z -> -1/z is a different holomorphic involution with fixed points +/- i. The two definitions agree only for sets invariant under conjugation, so for the genuinely non-symmetric amoeba-shaped set of Section 8.5 the doubly connected domain G between E and E* is not the annulus whose modulus is cape(E). The equality cape(E) = sqrt(q) is then applied to the wrong complementary domain, and the reported Table 12 values plus the numerical confirmation of cape(E) < caph(E) do not test the claimed elliptic-capacity theorem. This is a formula-level error independent of the BIE solver, and it undermines the elliptic-capacity branch of the paper's advertised scope rather than merely the auxiliary mapping step.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a boundary integral equation method, based on the generalized Neumann kernel, for computing the conformal capacity of doubly connected planar domains by numerically constructing the conformal map onto a circular annulus. The authors supply a MATLAB implementation, annq, and validate it against exact formulas for several model geometries: two confocal ellipses, a square-in-square domain, two collinear slits, a segment and a circle, and a half-plane with a vertical slit. They report relative errors around 1e-14 for smooth boundaries and roughly twelve correct digits for corner domains. The method is then applied to slit and polygon geometries and to the numerical computation of hyperbolic and elliptic capacities of compact sets in the unit disk.","tokens_in":23832,"tokens_out":9658,"duration_ms":106590,"significance":"If the advertised claims are fully supported, the paper provides a useful, versatile, and easily modifiable computational tool for conformal invariants. Its strengths are the independent exact-formula benchmarks, the explicit MATLAB code, and the fact that the reported errors are not fitted to the benchmarks. The ring-capacity and hyperbolic-capacity results are convincing. However, the elliptic-capacity branch, which is part of the advertised scope, is currently undermined by an incorrect definition of the antipodal set in Section 8.2 and by a consequently invalid numerical test in Section 8.5. The central computational method itself appears sound, and the defect is localized and fixable.","major_comments":[{"comment":"The definition of the antipodal set E* = {-1/a : a in E} is not the involution used in the Duren-Kuhnau elliptic capacity theory cited as [13]. The elliptic capacity of a set in the disk is associated with the fixed-point-free anti-Mobius antipodal map a -> -1/\\bar a; the map a -> -1/a printed in the paper is a different holomorphic involution with fixed points ±i, and the two definitions agree only for sets invariant under complex conjugation. As a consequence, for the non-conjugation-symmetric amoeba of Section 8.5 the doubly connected domain between E and the incorrectly defined E* is not the annulus whose modulus yields cape(E), and the values labeled cape(E) in Table 12 together with the reported numerical confirmation of cape(E) < caph(E) do not test the stated inequality. The definition also breaks down when 0 belongs to E. Please replace E* by {-1/\\bar a : a in E}, interpreted on the Riemann sphere when needed, recompute Section 8.5, and ensure that the code underlying annq uses the corrected definition.","section":"Section 8.2, Eqs. (47)-(48), and Section 8.5, Table 12"},{"comment":"The general two-slit and half-plane-with-slit computations rely on the iterative method of [33] to construct a smooth doubly connected model domain, but the paper does not analyze the convergence or the accuracy of this auxiliary step. The benchmarks in Sections 4.1 and 6 with exact formulas show that the composed method is accurate in those cases, including the iterative step; however, for the genuinely non-symmetric configurations in Section 4.3 and Figure 16 no error estimate or refinement study for the auxiliary mapping is supplied. Because the advertised scope includes slit geometries, the authors should either quantify the error introduced by this step, for example by varying the stopping criterion of the iteration or comparing successive iterates, or explicitly state that these specific results are benchmark-dependent.","section":"Sections 4 and 6"}],"minor_comments":[{"comment":"There are several typographical errors: \"using using MATLAB R2017a\" in Section 2.3, \"In view of of (7)\" near Eq. (8), and \"the the auxiliary map\" in Section 7.3.","section":"Sections 2.3 and 7.3"},{"comment":"The caption of Figure 21 mentions a quantity c(E) that is not defined in the text; it should be defined or removed.","section":"Section 8.3, Figure 21"},{"comment":"The \"exact\" capacity at the bottom of Table 9 is itself obtained by numerically solving equations (43) with Mathematica; the reported agreement of 1.5e-9 is several orders of magnitude larger than the other errors in the paper, so the precision of the Mathematica solution and the source of this discrepancy should be stated.","section":"Section 7.1, Table 9"},{"comment":"The header \"z2\\z1\" is confusing because the row entries appear to be values of z2 while the columns are values of z1; the layout should be clarified.","section":"Table 5"}],"recommendation":"major_revision","confidential_remarks":"This is a useful numerical study and the main computational method is well validated for ring domains and hyperbolic capacity. The requested revision is driven by the localized but load-bearing error in the elliptic-capacity section: the antipodal map is defined incorrectly, so the reported elliptic-capacity example does not test the theorem it claims to confirm. Once that definition is corrected and the affected numerical results are recomputed, I would expect the paper to be suitable for publication. I have no concerns about novelty disclosure; the methodology is largely prior work of the first author, but the systematic validation and the range of applications are appropriate for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: a solid, useful computational paper, and I would send it out. The engine is the first author's generalized Neumann-kernel BIE, but the application to a wide family of condenser geometries—slits, slit-plus-curve, half-plane slits, thin rectangles—and to hyperbolic and elliptic capacity is the contribution. The validation is the paper's strength: 1e-14 relative errors against independent exact formulas on smooth cases, 12 digits on corner cases, and the code is on GitHub with enough MATLAB snippets to reproduce the experiments. Section 5.2's exact interpolation between a segment and a disk is a nice independent result.\n\nThe reader's conditional verdict and the stress-test focus on Section 8.2. I don't think the conjugate objection lands. The kernel in (47) is |z-w|/|1+zw|, and the natural adjoint set is E*={-1/a:a in E}. The code confirms it: for the amoeba example z1=0.25+0.5i lies in E and z2=-1/z1 is chosen as the auxiliary point in the complementary domain, which only makes sense for that definition. The spherical antipodal map z -> -1/\\bar z is a different involution and is not the one for this capacity. I would still ask for one addition in the elliptic-capacity section: the amoeba example in 8.5 is the only non-symmetric test and it is not independently verified. Table 12 demonstrates convergence, but the cap_er_2>1$, for which the exact capacity is $2\\pi/\\log(r_1/r_2)$; with $r_2=2$, $r_1=4$, and $n=2^{12}$ discretization points per boundary, the method should give a relative error near $10^{-14}$. If it does not, the claimed smooth-boundary accuracy fails.","tokens_in":23417,"feed_emoji":"📐","tokens_out":12748,"duration_ms":123145,"temperature":0.7,"pith_summary":"This paper aims to make conformal invariants numerically routine. Its central claim is that one boundary integral equation, based on the generalized Neumann kernel, can map a ring-shaped domain conformally onto an annulus $\\{q<|z|<1\\}$ and thereby compute its capacity as $2\\pi/\\log(1/q)$. The same pipeline, with small auxiliary maps attached, handles smooth curves, polygons, slits, segment-circle rings, and half-plane slit domains. In tests with known exact values, the reported relative errors are near $10^{-14}$ for smooth boundaries and several correct digits for cornered boundaries. The same computed $q$ also gives hyperbolic capacity ($q$) and elliptic capacity ($\\sqrt{q}$) for compact connected sets in the unit disk.","feed_headline":"Computes smooth conformal capacities to 14 digits with one solver","feed_subtitle":"The same boundary-integral routine also handles slits, polygons, and half-plane geometries.","key_machinery":"The central object is the generalized Neumann kernel $N(s,t)$, a continuous kernel built from the boundary parametrization and an auxiliary function $A$; with its singular companion $M(s,t)$ it forms the integral equation $(I-N)\\rho=-M\\gamma$, which is uniquely solvable. The load-bearing identity is $q=e^{h_2-h_1}$, where $h_1,h_2$ are the two values of the piecewise constant function $h=[M\\rho-(I-N)\\gamma]/2$ on the two boundary components. This identity turns the numerical solution of a linear integral equation into the modulus of the annulus conformally equivalent to the given domain, and therefore into the capacity $2\\pi/\\log(1/q)$. The paper packages that whole computation in one function, `annq`, whose output feeds every example in the text.","core_discovery":"The paper establishes that the conformal capacity, hyperbolic capacity, and elliptic capacity of planar domains can all be extracted from a single computation: the conformal modulus $q$ of an associated doubly connected domain. Solving the generalized Neumann kernel integral equation on the two boundary curves yields a piecewise constant function $h$, and the identity $q=e^{h_2-h_1}$ gives the modulus of the annulus to which the domain maps; conformal invariance then forces $\\mathrm{cap}(G)=2\\pi/\\log(1/q)$. For a compact connected set $E\\subset\\mathbb{D}$, the same $q$ with $G=\\mathbb{D}\\setminus E$ is the hyperbolic capacity, and with $G$ the region between $E$ and its antipodal set $E^*$, $\\sqrt{q}$ is the elliptic capacity. Slit and half-plane boundaries are reduced to smooth rings by elementary maps or an iterative pre-mapping, and the same solver applies. Exact-formula comparisons show relative errors near $10^{-14}$ for smooth boundaries, and for cornered boundaries the values match known benchmarks to several digits.","pith_inferences":["Since several other conformal invariants, such as extremal length and moduli of curve families in doubly connected domains, are also functions of the annulus modulus $q$, the same computed $q$ could supply those values with no additional integral solving.","A graded-mesh version of the trapezoidal rule, which the paper notes is needed at corners, should close the accuracy gap seen for square and polygon boundaries while keeping the same pipeline.","The unquantified pre-mapping error for slit domains could be measured empirically by comparing the iterative pre-mapping results with the exact formulas for two-real-slits or half-plane slits.","Because geometry enters only through the boundary parametrization, coupling the solver with automatic differentiation could produce geometric gradients of capacity, enabling shape optimization."],"forward_implications":["A new doubly connected geometry needs only a boundary parametrization; the same solver returns its capacity, hyperbolic capacity, or elliptic capacity without a new derivation.","For cornered geometries such as square-in-square, polygon-in-polygon, and rectangle-slits, the method produces stable values that in the tested tables match exact or benchmark values to several digits, so it can be used as a cross-check for other numerical schemes.","The $O(n \\log n)$ cost and reported runtimes below roughly ten seconds make parameter studies feasible, such as tracing capacity as a slit moves or as a boundary component changes shape.","The inequality $\\mathrm{cap_e}(E)\\le \\mathrm{cap_h}(E)$ for compact connected subsets of the unit disk becomes numerically checkable for arbitrary shapes, as demonstrated for the non-symmetric amoeba region."],"supporting_citations":[{"why":"Supplies the fast boundary integral equation solver used for every capacity computation in the paper.","marker":"[30]"},{"why":"Provides the theorems that the annulus modulus is $q=e^{h_2-h_1}$ and gives the explicit conformal mapping formulas.","marker":"[28]"},{"why":"Establishes the solvability of the generalized Neumann kernel integral equation and the properties of its kernels.","marker":"[45]"},{"why":"Supplies the iterative pre-mapping used to reduce slit and half-plane slit domains to smooth rings.","marker":"[33]"},{"why":"Gives the exact capacity formulas used as benchmarks for two slits, segment-circle, and half-plane slit domains.","marker":"[44]"},{"why":"Supplies the elliptic-integral and $\\mu$-function identities behind the exact square-in-square capacity formula.","marker":"[2]"},{"why":"Provides earlier modulus and capacity benchmark values for square-in-square rings that the paper compares against.","marker":"[17]"},{"why":"Defines elliptic capacity and states the inequality $\\mathrm{cap_e}(E)\\le \\mathrm{cap_h}(E)$ tested numerically in Section 8.","marker":"[13]"},{"why":"Defines hyperbolic capacity through the limit formula used to connect it to the annulus modulus $q$.","marker":"[42]"}],"fun_headline_variants":["One solver computes three capacities to 14 digits","Same integral equation yields hyperbolic and elliptic capacity","Conformal capacity: smooth or slit, one code, 14 digits","14-digit precision for conformal, hyperbolic, elliptic capacity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"For slit and half-plane geometries, the capacity result relies on an auxiliary numerical pre-mapping that converts the slit into a smooth curve, and the paper does not analyze that pre-mapping's error.","fun_headline_variants_meta":{"raw":{"variants":["One solver computes three capacities to 14 digits","Same integral equation yields hyperbolic and elliptic capacity","Conformal capacity: smooth or slit, one code, 14 digits","14-digit precision for conformal, hyperbolic, elliptic capacity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000205,"raw_usage":{"total_tokens":1355,"prompt_tokens":871,"completion_tokens":484,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":487,"completion_tokens_details":{"reasoning_tokens":419}},"tokens_in":487,"tokens_out":484,"duration_ms":6248,"temperature":1.0,"reasoning_tokens":419,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:40:02.713850+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Use the confocal-ellipse ring with $r_1>r_2>1$, for which the exact capacity is $2\\pi/\\log(r_1/r_2)$; with $r_2=2$, $r_1=4$, and $n=2^{12}$ discretization points per boundary, the method should give a relative error near $10^{-14}$. If it does not, the claimed smooth-boundary accuracy fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the fast boundary integral equation solver used for every capacity computation in the paper."},{"cited_title":"Nasser , Numerical conformal mapping via a boundary integral equati on with the gener- alized Neumann kernel , SIAM J","cited_arxiv_id":null,"evidence_quote":"Provides the theorems that the annulus modulus is $q=e^{h_2-h_1}$ and gives the explicit conformal mapping formulas."},{"cited_title":"Wegmann and M","cited_arxiv_id":null,"evidence_quote":"Establishes the solvability of the generalized Neumann kernel integral equation and the properties of its kernels."},{"cited_title":"Nasser and C","cited_arxiv_id":null,"evidence_quote":"Supplies the iterative pre-mapping used to reduce slit and half-plane slit domains to smooth rings."},{"cited_title":"Lecture Not es in Mathematics , Springer-Verlag, Berlin, 1988","cited_arxiv_id":null,"evidence_quote":"Gives the exact capacity formulas used as benchmarks for two slits, segment-circle, and half-plane slit domains."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the elliptic-integral and $\\mu$-function identities behind the exact square-in-square capacity formula."},{"cited_title":"Hakula, A","cited_arxiv_id":null,"evidence_quote":"Provides earlier modulus and capacity benchmark values for square-in-square rings that the paper compares against."},{"cited_title":"Duren and R","cited_arxiv_id":null,"evidence_quote":"Defines elliptic capacity and states the inequality $\\mathrm{cap_e}(E)\\le \\mathrm{cap_h}(E)$ tested numerically in Section 8."},{"cited_title":"V asil’ev, Moduli of Families of Curves for Conformal and Quasiconform al Mappings , Springer-Verlag, Berlin, 2002","cited_arxiv_id":null,"evidence_quote":"Defines hyperbolic capacity through the limit formula used to connect it to the annulus modulus $q$."}],"review_version":1}