{"id":"bade1135-ee60-4ec9-a3da-f9cdc3adaee9","arxiv_id":"1908.02566","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A Bessel-function comparison bound for boundary-to-volume integrals yields new sharp eigenvalue estimates for Laplacians and Dirac operators on manifolds with boundary.","lead":"On manifolds with nonnegative Ricci curvature and boundary mean curvature bounded below, this paper derives a sharp Bessel-function bound for the ratio of boundary to bulk integrals of functions satisfying a Laplacian inequality. The bound yields new eigenvalue estimates for Dirichlet, Robin, and Dirac operators, plus a new Robin Laplacian on differential forms.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Even-dimensional case of Theorem 3.1 uses incorrect constants A and B that do not satisfy the stated initial conditions","rationale":"The reader's weakest_assumption identifies the Ricci/H0 curvature premise as the load-bearing geometric input. That is a stated scope restriction, not an internal inconsistency. My concern is different and more concrete: in the proof of the central Theorem 3.1, the even-dimensional linear system for the Bessel-coefficients is solved incorrectly. The published A and B do not reproduce y(0)=∫M f and y'(0)=-∫∂ f, so the comparison solution y does not have the initial data required for the argument. This is an internal error in the proof of the main claim, not merely a limitation of the hypothesis. Since the ratio A/B is unaffected by the erroneous common scale, the theorem's inequality (7) may still be derivable after correction, and I found no direct counterexample to the claim itself. I also verified the comparison principle F≥y from (8) is valid via the integrating-factor/Wronskian argument, so that step is not the weak point. The proof error affects even dimensions only, but Theorem 3.1 is used throughout Sections 3, 4, and 5, so it is load-bearing for the paper's applications. The reader's CONDITIONAL verdict remains appropriate: the paper needs a correction even if the final estimates are likely true.","tokens_in":33039,"tokens_out":47717,"duration_ms":432140,"concrete_test":"For n=2, H0=1, λ=1, choose arbitrary positive test data I=∫M f=1 and J=∫∂ f=1. Evaluate the paper's formulas for A and B in the even case of Theorem 3.1: A=-1/(2π)(Y_1(1)-Y_0(1)), B=-1/(2π)(-J_1(1)+J_0(1)). Then compute y(0)=A J_1(1)+B Y_1(1) and y'(0)=-(A J_0(1)+B Y_0(1)). The test fails if y(0)≠1 or y'(0)≠-1; numerically the paper's values give approximately 0.101 and -0.101. If the test reproduces the published constants, the initial conditions are violated, confirming the proof error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 3.1, the even-dimensional case solves a 2x2 linear system for the constants A and B multiplying J_{n/2} and Y_{n/2}. The paper states A = -H0/(2π)(Y_{n/2}(a)∫∂ f dμ - √λ Y_{n/2-1}(a)∫M f dμ) and B = -H0/(2π)(-J_{n/2}(a)∫∂ f dμ + √λ J_{n/2-1}(a)∫M f dμ), with a=√λ/H0. Using Lommel's formula (47), the determinant of the system is J_ν(a)Y_{ν-1}(a) - Y_ν(a)J_{ν-1}(a) = -2H0/(π√λ), where ν=n/2. Solving correctly gives A = π/(2H0)Y_ν(a)∫∂ f dμ - π√λ/(2H0)Y_{ν-1}(a)∫M f dμ and B = π/(2H0)(-J_ν(a)∫∂ f dμ + √λ J_{ν-1}(a)∫M f dμ). The paper's constants differ by a factor H0^2/π^2 and the opposite sign. For a concrete check in dimension n=2 with H0=1, λ=1, and data ∫M f = ∫∂ f = 1, the paper's A and B give y(0)=AJ_1(1)+BY_1(1) ≈ 0.101 and y'(0)=-√λ(AJ_0(1)+BY_0(1)) ≈ -0.101, whereas the correct values are 1 and -1. Thus the constructed solution y does not satisfy the initial conditions y(0)=∫M f and y'(0)=-∫∂ f, so the comparison F≥y is not applied to the stated boundary data. The ratio -A/B is scale-invariant, so the final inequality (7) may still be salvageable, but the proof as written is invalid in even dimensions.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a comparison estimate for the quotient of the boundary integral to the bulk integral of a function f satisfying Δf ≤ λf on a compact Riemannian manifold with boundary, under Ricci curvature nonnegative and boundary mean curvature bounded below by a positive constant. The central result, Theorem 3.1, bounds this quotient from below by a quotient of Bessel functions and characterizes equality by Euclidean geodesic balls. The authors then apply this estimate to recover Faber-Krahn inequalities for the Dirichlet and Robin Laplacians, to obtain new lower bounds for Dirac eigenvalues under CHI, gAPS, and mgAPS boundary conditions, and to study a Robin-type Hodge Laplacian on differential p-forms, including ellipticity, self-adjointness, variational characterization, and eigenvalue estimates. The appendix collects the Bessel-function identities used in the proofs.","tokens_in":33410,"tokens_out":14365,"duration_ms":148584,"significance":"If the main comparison is valid, the paper offers an elegant and fairly unified way to derive several sharp spectral estimates from one Bessel-function comparison, including known Faber-Krahn results and new Dirac and form-eigenvalue bounds. The equality statements, the explicit nature of the bounds, and the careful treatment of the p-form Robin extension are genuine strengths. The proof is mostly standard comparison geometry, but the Bessel solution is a nontrivial addition. However, the proof of the main theorem contains a localized algebraic error in the even-dimensional case, and the theorem as stated does not formally cover the nonnegative functions used in several applications. These issues are fixable but currently block acceptance.","major_comments":[{"comment":"The constants A and B displayed in the even-dimensional case do not solve the stated linear system. With ν=n/2, I_M=∫_M f dμ, I_∂=∫_∂M f dμ, and a=√λ/H0, Cramer's rule applied to J_ν(a)A+Y_ν(a)B=I_M and √λ J_{ν-1}(a)A+√λ Y_{ν-1}(a)B=I_∂ gives A_c=(π√λ/(2H0))Y_{ν-1}(a)I_M-(π/(2H0))Y_ν(a)I_∂ and B_c=(π/(2H0))J_ν(a)I_∂-(π√λ/(2H0))J_{ν-1}(a)I_M. The paper's constants are both equal to (H0²/π²) times these values, so the resulting function y(r) satisfies y(0)=(H0²/π²)I_M and y'(0)=-(H0²/π²)I_∂ rather than y(0)=I_M and y'(0)=-I_∂. For n=2, H0=1, λ=1, I_M=I_∂=1, this gives y(0)≈0.101 and y'(0)≈-0.101 instead of 1 and -1. Consequently the comparison F(r)≥y(r) is not established for the stated boundary data, and the derivation of (7) and its corollaries rests on an invalid step in even dimensions. Since A and B are multiplied by a common factor, the ratio -A/B and the first zero R0 are unchanged, so the final inequality may be salvageable; nevertheless the proof must be corrected and the equality case re-verified.","section":"Section 3, proof of Theorem 3.1 (even-dimensional case)"},{"comment":"Theorem 3.1 is stated for a positive smooth function f, but it is applied to f=|ψ|², f=|ω|², and to Dirichlet eigenfunctions, which are only nonnegative and may vanish on substantial sets. Every step of the proof, in particular equations (2)-(8) and the equality-case argument using y(0)=∫_M f>0, only requires f≥0 and ∫_M f>0. The theorem statement should therefore be relaxed to 'nonnegative, not identically zero', or an approximation argument should be supplied. Without this change, Theorems 4.1, 4.3, and 5.5 are not formally consequences of Theorem 3.1 as written.","section":"Theorem 3.1 and applications in Sections 3-5"}],"minor_comments":[{"comment":"The assertion that Courant's nodal domain theorem implies λ_{1,p}(τ) is simple and that every associated eigenfunction cannot change sign is not valid for p-forms when p≥1; the first eigenvalue of a Hodge Laplacian with Robin-type boundary conditions can be multiple. This simplicity claim is not used in the later eigenvalue estimates, so it should be removed or replaced by a correct statement such as a min-max description allowing repeated eigenvalues.","section":"Theorem 5.2, proof"},{"comment":"The symbol ν is used both for the inward unit normal field and for the order of Bessel functions. The context usually makes the meaning clear, but the dual use is a recurring source of possible confusion and should be disambiguated in the notation section.","section":"Throughout"},{"comment":"The distributional inequality (4) is quoted from the comparison geometry literature without explicitly stating that it holds pointwise away from the cut locus and in the distributional sense across it. A one-sentence reminder of this regularity would help readers who are not specialists in the mean value lemma.","section":"Section 2, equation (4)"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is promising and the error in the even-dimensional constants is localized, but because it occurs in the proof of the central result, the revised manuscript must contain a corrected derivation of the constants and a re-verification that inequality (7) and the equality case still follow. The actual correct constants are given in my major comment; note that the paper's constants differ from the correct ones by the common factor H0²/π², so the ratio -A/B is unaffected, which suggests the final inequality may survive. The positivity/nonnegativity gap should also be fixed since it formally excludes the functions used in the Dirac and form applications."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my take. The genuinely new result is Theorem 3.1, a sharp lower bound for the quotient of boundary to bulk integrals under Ric >= 0, H >= H0 > 0, and Delta f <= lambda f. This goes beyond Guerini-Savo and Raulot-Savo by solving the comparison ODE in Bessel functions rather than reducing to a constant-coefficient or subharmonic case. The applications to Dirichlet and Robin Laplacians, Dirac eigenvalues, and forms are credible. The bibliography looks appropriate; self-citations are for standard material. So the paper deserves serious referee time.\n\nBut there is a real problem in the proof of Theorem 3.1 in even dimensions. The stated constants A and B for the Y_{n/2} solution do not solve the linear system coming from y(0) = integral_M f and y'(0) = -integral_dM f. Correct Cramer's rule using Lommel's identity gives A = (pi/(2H0))(sqrt(lambda) Y_{n/2-1}(a) integral_M f - Y_{n/2}(a) integral_dM f) and B = (pi/(2H0))(J_{n/2}(a) integral_dM f - sqrt(lambda) J_{n/2-1}(a) integral_M f). The paper has A = H0/(2pi)(...) and B = H0/(2pi)(...), off by the positive factor H0^2/pi^2. The ratio -A/B is invariant under this rescaling, so the final inequality (7) may still be salvageable, but as written y does not satisfy the stated initial conditions and the comparison F >= y is not justified. A referee should ask for a corrected computation.\n\nOther issues are minor. Theorem 3.1 assumes f > 0, but applications use nonnegative f = |psi|^2 or |omega|^2; this needs a short justification. In Section 5, the principal symbol for top-degree forms looks inconsistent with the definitions, and Proposition 5.3's regularity proof is sketched, with a density step that is not obvious. Inequality (26) relies on a private communication from Freitas; if it is essential for comparing the Dirac bounds, it should be proved or replaced by a checkable reference.\n\nBottom line: good paper, load-bearing but fixable error. Send it to peer review, keep it in the pipeline conditional on the even-dimensional constants being corrected.","headline":"The Bessel-function comparison in Theorem 3.1 is a genuine advance, but the even-dimensional proof has a concrete constant error that needs fixing before the paper can be accepted.","tokens_in":34031,"tokens_out":7602,"would_cite":true,"duration_ms":77109,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C27","53C21","58J60","35P15","34B09","33C10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Bessel functions give a sharp lower bound for boundary-to-bulk integrals on manifolds with nonnegative Ricci curvature and positive boundary mean curvature, with equality only on Euclidean geodesic balls; the estimate drives Dirichlet…","keywords":["Bessel functions","eigenvalue estimates","Dirac operator","Robin Laplacian","differential forms","boundary mean curvature","Yamabe operator","isoperimetric eigenvalue inequalities"],"falsifier":"Compute the first Robin eigenvalue $\\lambda_1(\\tau,\\Omega)$ numerically for a smooth, strictly convex Euclidean domain $\\Omega$ that is not a disk, with $H_0$ the minimum of its boundary curvature, and compare it with the first Robin eigenvalue of the Euclidean disk of radius $1/H_0$. Corollary 3.8 asserts $\\lambda_1(\\tau,\\Omega)\\ge \\lambda_1(\\tau,B_{1/H_0})$ with equality only for the disk; a computed violation, for any $\\tau>0$, would disprove the main theorem, while equality for a non-disk would disprove the rigidity statement.","tokens_in":32788,"feed_emoji":"📐","tokens_out":20226,"duration_ms":192153,"temperature":0.7,"pith_summary":"The paper establishes a sharp comparison for the ratio between the boundary integral and the bulk integral of a positive function on a compact manifold with boundary. Under the assumptions that the Ricci curvature is nonnegative and the boundary mean curvature is at least a positive constant $H_0$, any positive smooth $f$ with $\\Delta f \\le \\lambda f$ and $\\sqrt{\\lambda}/H_0 < j_{n/2,1}$ satisfies $\\int_{\\partial M} f \\ge \\sqrt{\\lambda}\\, \\frac{J_{n/2-1}(\\sqrt{\\lambda}/H_0)}{J_{n/2}(\\sqrt{\\lambda}/H_0)} \\int_M f$. Equality forces the manifold to be a Euclidean geodesic ball of radius $1/H_0$. Because no boundary condition is imposed on $f$, this one inequality feeds directly into eigenvalue problems: it gives a direct proof of the isoperimetric eigenvalue inequalities for the Dirichlet and Robin Laplacians, new Dirac eigenvalue bounds that involve a Bessel zero alongside scalar curvature, and a Robin-type Laplacian on differential forms with a positive discrete spectrum bounded below by Bessel functions.","feed_headline":"Bessel quotient sharpens eigenvalue bounds on curved manifolds","feed_subtitle":"A boundary-to-bulk bound forces equality only on Euclidean balls, sharpening Dirichlet, Robin, Dirac, and form spectra.","key_machinery":"The load-bearing mechanism is the comparison of a boundary-distance accumulated function $F(r)=\\int_{\\{\\rho>r\\}} f\\,d\\mu_g$ with the explicit solution of a Bessel equation. The mean value lemma expresses $F''$ through $\\Delta f$ and the Laplacian of the distance function, and the volume comparison inequality bounds that Laplacian by $-\\Theta'/\\Theta\\circ\\rho$; under $K=0$ and $H\\ge H_0$, $\\Theta(r)=(1-rH_0)^{n-1}$. The substitution $s=1-rH_0$ turns the differential inequality $F''-\\frac{\\Theta'}{\\Theta}F'+\\lambda F\\ge 0$ into a transformed Bessel equation, so the model solution is $(1-rH_0)^{n/2}$ times $J_{n/2}\\left(\\frac{\\sqrt{\\lambda}}{H_0}(1-rH_0)\\right)$ (with $J_{-n/2}$ or $Y_{n/2}$ as the second independent solution). Comparing the first zero of that solution with the inner radius of $M$ yields the quotient bound in terms of $\\frac{J_{n/2-1}}{J_{n/2}}$.","core_discovery":"The central claim is Theorem 3.1: with $\\mathrm{Ric}\\ge 0$ and boundary mean curvature $H\\ge H_0>0$, for every positive smooth $f$ satisfying $\\Delta f\\le \\lambda f$, $\\lambda>0$, and with $\\sqrt{\\lambda}/H_0$ below the first positive zero of $J_{n/2}$, the quotient $\\int_{\\partial M} f\\,/\\,\\int_M f$ is bounded below by the Bessel quotient displayed above. Equality holds exactly when $M$ is isometric to the Euclidean ball of radius $1/H_0$, and for that ball with $\\Delta f=\\lambda f$ the inequality is an equality. The proof runs through the distributional second derivative of $F(r)=\\int_{\\{\\rho>r\\}} f$, a distance-function comparison that under the curvature hypotheses gives $\\Theta(r)=(1-rH_0)^{n-1}$, and the change of variable $s=1-rH_0$, which turns the associated differential inequality into a Bessel-type equation whose first zero dominates the geometry. All subsequent Dirichlet, Robin, Dirac, Yamabe, and form-eigenvalue estimates are applications or direct corollaries of this quotient bound.","pith_inferences":["The paper leaves implicit that, because the quotient bound needs no boundary condition on $f$, the same differential-inequality scheme should apply to eigenfunctions of any operator whose Bochner identity produces a pointwise inequality $\\Delta |u|^2 \\le \\mu |u|^2$, not only to Laplacians and spinors.","The paper's restriction to $K=0$ is a convenience rather than a necessity: replacing $\\Theta(r)=(1-rH_0)^{n-1}$ by the corresponding function for a nonzero Ricci lower bound should replace Bessel functions by the analogous special solutions, so the theorem should extend to manifolds with a lower Ricci bound $K\\ne 0$.","The strictness of the Dirac and form estimates suggests the Bessel term acts as a genuine spectral gap; a numerical computation of the first Robin eigenvalue on a smooth non-spherical convex Euclidean domain should show a positive gap relative to the ball, and measuring how that gap scales with $\\tau$ would test the sharpness of the constants $\\tau_0$ and $\\tau_1$."],"forward_implications":["The first Dirichlet eigenvalue on such a manifold satisfies $\\lambda_1^D \\ge H_0^2\\, j_{n/2-1,1}^2$, with equality only for the Euclidean ball of radius $1/H_0$.","The first Robin eigenvalue satisfies $\\lambda_1(\\tau,M)\\ge \\lambda_1(\\tau,B_{H_0})$, so the Euclidean ball minimizes the Robin spectrum among all manifolds with the same curvature and boundary-mean-curvature bounds.","Dirac eigenvalues under the CHI, gAPS, or mgAPS boundary conditions obey $\\lambda^2 > \\frac{n}{4(n-1)}\\min_M S + \\frac{n H_0^2}{2(n-1)}\\tau_0^2$, where $\\tau_0$ is a Bessel zero; the estimate remains nontrivial even when the scalar curvature vanishes at a point.","The Robin Laplacian on $p$-forms defined by $\\iota^*(\\nu\\lrcorner d\\omega)=\\tau\\,\\iota^*\\omega$ and $\\iota^*(\\nu\\lrcorner\\omega)=0$ is elliptic, self-adjoint, and has purely positive discrete spectrum, with first eigenvalue bounded below by a Bessel expression depending on the $p$-curvature of the boundary.","The same comparison yields gap and Gallot-Meyer-type estimates: $\\lambda_{1,p}(\\tau)-\\lambda_{1,p-1}(\\tau)\\ge \\frac{1}{p}\\inf_M (W_M^{[p]}-T^{[p]})$, and under a positive curvature operator $\\lambda_{1,p}(\\tau)\\ge p(n-p)\\frac{c}{c-1}\\gamma$ when $\\tau\\ge -\\frac{c}{c-1}\\sigma_p$."],"supporting_citations":[{"why":"It supplies the mean value lemma expressing $F''$ in terms of $\\Delta f$ and the distance-to-boundary Laplacian, the starting point of the main inequality.","marker":"[39]"},{"why":"It gives the distributional bound on $\\Delta\\rho$ under the Ricci and mean-curvature assumptions, producing the coefficient $-\\Theta'/\\Theta$.","marker":"[20]"},{"why":"It characterizes when the inner radius equals the first zero of $\\Theta$, a step used in the equality case of Theorem 3.1.","marker":"[24]"},{"why":"It provides the comparison principle and the result used to deduce $R_0\\ge R$ and the geodesic-ball rigidity.","marker":"[36]"},{"why":"It is the earlier constant-coefficient reduction for the same quotient problem; the paper's Bessel treatment is the next step and its corollaries are compared with it.","marker":"[19]"},{"why":"It supplies the boundary-condition framework and eigenvalue estimates for the Dirac operator used in Theorem 4.1.","marker":"[22]"},{"why":"It gives the conformal lower bound for Dirac eigenvalues combined with the Yamabe estimate in Corollary 4.4.","marker":"[35]"},{"why":"It provides the spectral theory for boundary value problems used to prove self-adjointness, ellipticity, and the variational characterization of the Robin Laplacian on forms.","marker":"[42]"},{"why":"It supplies the Bessel identities and Lommel formulas needed to solve the differential equation and compare the quotient $J_{n/2-1}/J_{n/2}$.","marker":"[43]"}],"fun_headline_variants":["Bessel roots sharpen Dirac and Robin eigenvalue bounds","Bessel functions yield sharp spectral bounds on curved spaces","Bessel quotient forces sharp eigenvalue inequalities","Sharp manifold eigenvalue bounds from Bessel zeros"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole chain rests on the combined assumption that the manifold's Ricci curvature is nonnegative and the boundary's inward mean curvature has a positive lower bound $H_0$; if either fails, the distance-function comparison that produces the Bessel equation no longer holds, and the bound collapses.","fun_headline_variants_meta":{"raw":{"variants":["Bessel roots sharpen Dirac and Robin eigenvalue bounds","Bessel functions yield sharp spectral bounds on curved spaces","Bessel quotient forces sharp eigenvalue inequalities","Sharp manifold eigenvalue bounds from Bessel zeros"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000818,"raw_usage":{"total_tokens":3603,"prompt_tokens":986,"completion_tokens":2617,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":2560}},"tokens_in":602,"tokens_out":2617,"duration_ms":17716,"temperature":1.0,"reasoning_tokens":2560,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:43:49.890722+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the first Robin eigenvalue $\\lambda_1(\\tau,\\Omega)$ numerically for a smooth, strictly convex Euclidean domain $\\Omega$ that is not a disk, with $H_0$ the minimum of its boundary curvature, and compare it with the first Robin eigenvalue of the Euclidean disk of radius $1/H_0$. Corollary 3.8 asserts $\\lambda_1(\\tau,\\Omega)\\ge \\lambda_1(\\tau,B_{1/H_0})$ with equality only for the disk; a computed violation, for any $\\tau>0$, would disprove the main theorem, while equality for a non-disk would disprove the rigidity statement.","supporting_citations":[{"cited_title":"Savo, A mean value lemma and applications , Bull","cited_arxiv_id":null,"evidence_quote":"It supplies the mean value lemma expressing $F''$ in terms of $\\Delta f$ and the distance-to-boundary Laplacian, the starting point of the main inequality."},{"cited_title":"Heintze and H","cited_arxiv_id":null,"evidence_quote":"It gives the distributional bound on $\\Delta\\rho$ under the Ricci and mean-curvature assumptions, producing the coefficient $-\\Theta'/\\Theta$."},{"cited_title":"Kasue, Ricci curvature, geodesics and some geometric properties o f Riemannian manifolds with boundary, J","cited_arxiv_id":null,"evidence_quote":"It characterizes when the inner radius equals the first zero of $\\Theta$, a step used in the equality case of Theorem 3.1."},{"cited_title":"Raulot and A","cited_arxiv_id":null,"evidence_quote":"It provides the comparison principle and the result used to deduce $R_0\\ge R$ and the geodesic-ball rigidity."},{"cited_title":"Gu´ erini and A","cited_arxiv_id":null,"evidence_quote":"It is the earlier constant-coefficient reduction for the same quotient problem; the paper's Bessel treatment is the next step and its corollaries are compared with it."},{"cited_title":"Hijazi, S","cited_arxiv_id":null,"evidence_quote":"It supplies the boundary-condition framework and eigenvalue estimates for the Dirac operator used in Theorem 4.1."},{"cited_title":"Raulot, The Hijazi inequality of manifolds with boundary , J","cited_arxiv_id":null,"evidence_quote":"It gives the conformal lower bound for Dirac eigenvalues combined with the Yamabe estimate in Corollary 4.4."},{"cited_title":"Taylor, Partial differential equations I","cited_arxiv_id":null,"evidence_quote":"It provides the spectral theory for boundary value problems used to prove self-adjointness, ellipticity, and the variational characterization of the Robin Laplacian on forms."},{"cited_title":"Watson, A treatise on the theory of Bessel functions , 2nd edn","cited_arxiv_id":null,"evidence_quote":"It supplies the Bessel identities and Lommel formulas needed to solve the differential equation and compare the quotient $J_{n/2-1}/J_{n/2}$."}],"review_version":1}