{"id":"3a6b64ef-da67-4a9f-abe3-f025be3d25ce","arxiv_id":"2608.08816","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a non-perturbing monitor of diffusing molecules, the optimal unbiased linear estimator weighs only the boundary of the sensing volume, improving concentration precision by 5/6 and gradient precision by 7/10 relative to standard uniform estimators.","lead":"For a perfect sensor that watches individual molecules without disturbing them, the best way to measure concentration or gradient is to count molecules only in a thin shell at the boundary of the sensing region, not throughout its volume. This result, derived by mapping the estimation problem to electrostatics, slightly beats the classic Berg-Purcell sensing limit and shows how sensor shape affects precision.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-T correction to the long-time covariance (Eq. 10 to Eq. 11) is unquantified; the boundary-localization result is asymptotic but the abstract does not say so.","rationale":"The reader's weakest assumption is the same one I find load-bearing: Eq. (10) to Eq. (11) replaces the finite-time integral of the diffusion propagator by the Laplacian Green's function. I checked the algebra of the covariance and the factors 5/6 and 7/10; they are internally consistent in the stated long-time limit. The finite-T correction is O(a/sqrt(DT)) relative to the asymptotic variance, so the central boundary-localization claim is not wrong, but it is asymptotic. The paper does not quantify the correction, and the abstract's unqualified statement goes beyond the proven regime. The permeable-shell regularization to the surface delta is a secondary gap; a direct Fourier-space calculation for a thin shell gives the claimed finite limit, so I do not treat it as fatal. The CONDITIONAL verdict is appropriate, and my stress test does not move it.","tokens_in":20056,"tokens_out":37806,"duration_ms":432381,"concrete_test":"Perform a numerical variational calculation on a sphere at T = 1, 10, and 100 a^2/D: discretize the radial weight w(r), minimize the exact finite-T variance from Eq. (10) subject to Eq. (5), and compare the minimizer and the minimal variance with the surface-delta weight and Eq. (14). If already at T = 10 a^2/D the minimizer is strongly boundary-localized and the variance is within a few percent of c0/(2 pi D a T), the concern is minor; if the minimizer is not concentrated at the boundary at T = 10 a^2/D, the abstract and conclusion need an explicit long-time qualifier and a finite-T analysis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central electrostatic mapping rests on replacing the exact time-averaged density covariance in Eq. (10) by the Coulomb kernel in Eq. (11): the integral over the diffusion propagator from 0 to infinity replaces the finite-T integral with a 1/|r-r'| interaction. This is the step that turns variance minimization into an electrostatic problem, and hence the step that forces the optimal weight onto the boundary. The paper states the required T >> a^2/D, but it does not compute the leading finite-T correction. In Fourier space the exact kernel is C~_T(k) = (2 c0/(D k^2 T)) [1 - (1 - e^{-D k^2 T})/(D k^2 T)], which agrees with the long-time form only for D k^2 T >> 1. Modes with k less than about 1/sqrt(DT) contribute corrections of relative order a/sqrt(DT); at T = 10 a^2/D this is roughly 30%. For finite T the optimal weight minimizing the exact quadratic form need not be the surface delta, and the variance need not equal the Coulomb self-energy. The 5/6 and 7/10 improvement factors are therefore long-time limiting statements, while the abstract presents them as unqualified physical limits. The low-dimensional discussion in Appendix D correctly identifies where the approximation fails, but it does not supply the finite-T error estimate needed to judge when the main result applies.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper considers an idealized 'perfect monitor' that observes the instantaneous molecular density in a bounded region without perturbing the diffusing field, and asks for the minimum-variance unbiased estimator within the class of spatially weighted, time-averaged linear estimators, for both a uniform concentration and a weak linear gradient. It shows that, under a long-time approximation that replaces the time-integrated diffusion propagator by the Coulomb Green's function, the variational problem mapping variance to electrostatic self-energy yields a surface-localized optimal weight: for a sphere, w*(r)=δ(r-a)/(4πa^2) with Var=c0/(2πDaT), which is 5/6 of the Berg-Purcell variance, and for a gradient, w*(r)=3/(4πa^3)δ(r-a) r-hat with Cov=c0/(2πDa^3T)I, which is 7/10 of the Endres-Wingreen covariance. The results are extended to arbitrary conductor shapes via capacitance and polarizability, and to d dimensions, with a discussion of the recurrent case d≤2.","tokens_in":20263,"tokens_out":19960,"duration_ms":195335,"significance":"The electrostatic mapping is elegant, parameter-free, and the special cases are computed cleanly; the explicit 5/6 and 7/10 ratios are falsifiable predictions, and the capacitance/polarizability formulation provides a unifying view across shapes and dimensions. If the long-time limit is accepted, the finding that a perfect monitor should ignore bulk molecules and weight the boundary is a genuine conceptual correction to the standard reading of the Berg-Purcell bound. The manuscript ships self-contained derivations in the appendices and is careful about the d≤2 recurrence issue. The significance is tempered by the fact that the optimality proof is asymptotic and the finite-T error is not controlled; the abstract and conclusion currently present the result without this caveat.","major_comments":[{"comment":"The key replacement of the finite-T time integral by the infinite-time Coulomb integral is only justified for modes with D k^2 T >> 1. For the exact covariance kernel, in Fourier space C~_T(k) = (2 c0/(D k^2 T))[1 - (1 - e^{-D k^2 T})/(D k^2 T)], so modes with k ≲ (D T)^(-1/2) are not yet in the Coulomb regime and contribute corrections of relative order a/sqrt(DT). At T = 10 a^2/D this is roughly 30%. Since the electrostatic mapping is the basis for the boundary-localization theorem, the statement that the optimal unbiased linear estimator assigns all weight to the boundary is only shown in the long-time limit, yet the abstract and conclusion present it without qualification. Please either compute the leading finite-T correction to the optimal weight and to Var(bc*) and Cov(bg*), or explicitly state throughout that the results are the leading-order long-time (T >> a^2/D) limits. In its current form, the proof does not establish exact boundary localization at finite T.","section":"Concentration Sensing, Eqs. (10)-(11); also Eq. (19)"},{"comment":"The d≤2 concentration result is stated in the main text as Var(bc*) ~ c0/sqrt(DT) for d=1 and c0 ln(DT/a^2)/(DT) for d=2, but Appendix D only gives the general expression Var(bc*) = C_T(R0) + 2 c0 V/(DT) + o(T^{-1}) without computing C_T(R0) and V for the ball in d=1 and d=2. Since these constants are needed to reproduce Eq. (32), and since the validity of the o(T^{-1}) statement for all admissible weights is not analyzed, the low-dimensional extension is not fully supported by the provided derivation. Please supply the missing calculation or a precise reference.","section":"Appendix D, Eq. (D1)-(D3) and main text Eq. (32)"}],"minor_comments":[{"comment":"The abstract and conclusion should explicitly state that the optimal surface-weight result and the 5/6 and 7/10 factors are long-time (T >> a^2/D) asymptotic results; currently they are presented as unqualified physical limits.","section":"Abstract and Conclusion"},{"comment":"The sentence 'One can show by explicit calculations that for ellipsoids, w_n is indeed the optimal weight' needs the calculation or a reference, since the saturation of the bound tr α^{-1}_Ω ≤ 1/|Ω| is used to conclude that ellipsoids maximize the MSE among shapes of fixed volume.","section":"Appendix C"},{"comment":"The claim that 'The trace of α^{-1}_Ω is the same for all ellipsoids of the same volume' is stated without derivation; a short calculation using the depolarizing factors would make the argument self-contained.","section":"Nonspherical Geometries"},{"comment":"The blue and orange dots in the schematic figures may be difficult to distinguish in grayscale or print; consider using different marker shapes or hatching.","section":"Figures 1 and 2"},{"comment":"Reference [26] is cited as a recent permeable-shell calculation that agrees with the 5/6 factor; the text should clarify what new element the present variational proof adds beyond that result.","section":"References, Ref. [26]"}],"recommendation":"major_revision","confidential_remarks":"The paper is well-organized and the derivations are mostly careful. The main risk is the finite-T issue: if the corrections are significant at biologically relevant T, which often is not much larger than a^2/D, the claimed physical limits may not be realized. I would ask the authors to either quantify the leading correction or clearly reframe the results as asymptotic; this is feasible within the scope of a major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Farshid's paper is worth a serious referee. The core result is clear and mostly solid: among unbiased linear estimators based on the full instantaneous density in a region, the optimal concentration and gradient weights live on the boundary. The electrostatic mapping is elegant and the variational proofs in Appendix B are careful. The 5/6 concentration factor is already in McCusker and Lubensky (Ref. [26]), and the paper says so; the genuinely new pieces are the general variational optimality proof, the 7/10 gradient improvement over Endres-Wingreen, and the arbitrary-geometry results expressed through capacitance and polarizability. The shape-dependence result (tr alpha^{-1} <= 1/|Omega|, saturated by ellipsoids) is a nice touch, and the factor-of-two relation to the perfect absorber is a clean, non-obvious statement.\n\nThe main soft spot is exactly what the stress-test note points to: the step from Eq. (10) to Eq. (11) replaces the finite-T time integral of the diffusion propagator by the full Coulomb kernel, and the paper does not quantify the finite-T corrections. That is the load-bearing step for the boundary-localization result, because for finite T the optimal weight is not necessarily a surface delta. The paper states the long-time assumption and correctly flags the d<=2 failure, but a reader cannot tell from the abstract how asymptotic the claim is. At T = 10 a^2/D, modes with k below about 1/sqrt(DT) give corrections of order 30%, so the 5/6 and 7/10 factors are long-time statements, not exact finite-time limits. This should be fixed with an explicit finite-T error estimate or at least a sentence in the abstract saying the results are asymptotic in T >> a^2/D.\n\nA second, minor point: the permeable-shell regularization to the surface-delta weight is asserted rather than derived. The statement that the instantaneous shot noise diverges as 1/epsilon while the correlation time vanishes as a epsilon/D is plausible but the cancellation deserves a few lines. That is minor and fixable.\n\nOverall the math is sound, the literature is cited fairly (including the overlap with Ref. [26]), and the paper is parameter-free and self-contained. The central argument holds up as a long-time limit. I would bring it to a reading group and would cite it for the general variational result. A serious referee should see it; with the finite-T correction addressed it would be a solid contribution.","headline":"A clean variational proof that optimal non-perturbing monitors put all sensing weight on the boundary, improving the Berg-Purcell and Endres-Wingreen limits by modest factors; the main caveat is that the finite-time corrections to the central electrostatic mapping are not quantified.","tokens_in":20831,"tokens_out":644,"would_cite":true,"duration_ms":8608,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A perfect monitoring instrument achieves its optimal unbiased concentration and gradient estimates by weighting only the boundary of its sensing region, which for a sphere lowers the variance to $\\frac{5}{6}$ and the covariance to…","keywords":["concentration sensing","gradient sensing","perfect monitor","electrostatic analogy","boundary weighting","capacitance","polarizability","diffusion noise"],"falsifier":"Run a Brownian-dynamics simulation of a non-perturbing spherical monitor of radius $a$ over a range of finite measurement times $T$, computing the variance of the surface-shell estimator and of the uniform-volume estimator for identical molecule fields; if for any $T \\gg a^2/D$ the surface-shell variance exceeds the predicted $c_0/(2\\pi D a T)$, or a numerical variational search finds a non-boundary weight with lower variance, the central claim is falsified.","tokens_in":19814,"feed_emoji":"🧪","tokens_out":6429,"duration_ms":65457,"temperature":0.7,"pith_summary":"Cells can detect chemical gradients from just a few diffusing molecules, and the classic model of such sensing averages molecular counts over an instrument's whole volume. This paper asks whether a perfect, non-perturbing monitor can do better by weighting positions unequally. It establishes that the optimal unbiased weighting is entirely concentrated on the instrument's boundary: for a sphere, only molecules infinitesimally close to the surface should be counted. That reduces the concentration variance to $\\frac{5}{6}$ of the classic value and the gradient covariance to $\\frac{7}{10}$ of the standard value, and it makes the shape dependence of precision exactly the capacitance and polarizability of the region. If correct, the result sharpens the physical limit of diffusion-limited sensing and changes which geometries a designer would choose.","feed_headline":"Optimal molecular sensing weights concentrate on the boundary","feed_subtitle":"For a sphere, surface weighting cuts the classic volume-average variance to 5/6 and gradient covariance to 7/10.","key_machinery":"The electrostatic mapping is the central device: the time integral of the diffusion propagator, $\\int_0^\\infty G(|r-r'|,\\tau)\\,d\\tau$, equals the Green's function of the Laplacian, so the long-time covariance of the time-averaged density becomes the Coulomb kernel $1/|r-r'|$ (or $1/R^{d-2}$ in $d$ dimensions). This turns variance minimization into minimizing the electrostatic self-energy of a charge distribution $w(r)$ under constraints: charge conservation for concentration, charge neutrality plus fixed dipole moment for gradient. The equilibrium solution on a conductor is a surface charge, which is why the optimal weight is a boundary delta-shell; for general shapes the solution is the surface charge density $\\sigma(r)$ of the conductor, so capacitance and polarizability carry all the shape dependence.","core_discovery":"The paper considers a perfect monitoring instrument that observes the instantaneous molecular density everywhere inside a finite region without disturbing the field or tracking molecular identities, and it studies unbiased estimators that are linear in the time-averaged density with a spatial weight $w(r)$. In the long-time limit, the variance of any such estimator is proportional to the Coulomb self-energy of the weight viewed as a charge density, so minimizing variance is the electrostatic problem of a conductor's equilibrium charge. The optimal concentration weight is a uniform surface charge, $w^*(r)=\\delta(r-a)/(4\\pi a^2)$ for a sphere, with variance $c_0/(2\\pi D a T)$; the optimal gradient weight is $w^*(r)=3/(4\\pi a^3)\\delta(r-a)\\,\\hat{r}$, with covariance $c_0/(2\\pi D a^3 T)\\,I$. Hence a perfect monitor should ignore the interior and count only molecules in an infinitesimal boundary shell. For arbitrary shapes the same argument gives concentration variance $2c_0/(D T C_\\Omega)$ and gradient covariance $2c_0/(D T \\alpha_\\Omega^{-1})$, so geometry enters only through capacitance and polarizability; in $d\\le 2$, recurrence modifies the concentration variance but not the boundary-localization structure.","pith_inferences":["A practical consequence the paper leaves implicit: a thin permeable shell near the surface should approximate the optimal estimator, and the divergent instantaneous noise of an infinitesimally thin shell is canceled by its vanishing correlation time, so finite shells may realize most of the gain.","The exact factor-of-two cost of non-perturbation relative to an absorber suggests a general trade-off for any unbiased linear estimator of diffusion-limited signals; testing it with finite molecule numbers or nonlinear estimators could show whether the factor survives beyond the perfect-monitor class.","Because shape enters only through capacitance and polarizability, engineered nonspherical detection volumes could measurably outperform spheres of equal volume in gradient sensing; a quantitative extension would optimize shape under a fixed-volume or fixed-accessibility constraint.","The $d\\le 2$ result that gradient sensing retains $T^{-1}$ scaling while concentration sensing does not implies that in effectively one- or two-dimensional environments, gradient detection may be more favorable relative to concentration sensing than in three dimensions—a testable prediction for confined cell-sensing experiments."],"forward_implications":["A perfect monitor should ignore the interior and count only molecules at the boundary; for a sphere this lowers concentration variance by $\\frac{5}{6}$ and gradient covariance by $\\frac{7}{10}$ relative to uniform volume averaging.","For arbitrary shapes, the optimal concentration precision is set entirely by the shape's capacitance and the gradient precision by its polarizability tensor; the monitor's variance is exactly twice that of a perfect absorber, independent of shape.","Among instruments of equal volume, spheres are the worst concentration sensors, ellipsoids saturate the gradient bound, and non-ellipsoidal shapes can strictly improve total gradient precision.","In higher dimensions the benefit of optimizing grows: the concentration variance ratio is $(d+2)/(2d)$ for a $d$-dimensional ball, and the gradient covariance ratio is $(d+4)/(2(d+2))$.","In one and two dimensions, recurrent returns to the instrument dominate concentration sensing at long times, with variance decaying as $T^{-1/2}$ or as $\\ln(DT/a^2)/(DT)$, but gradient sensing still reaches $T^{-1}$ because charge neutrality cancels the recurrence term."],"supporting_citations":[{"why":"Defines the perfect-monitor model and the uniform volume estimator whose variance is the concentration baseline.","marker":"[1]"},{"why":"Defines the standard position-weighted gradient estimator and supplies its covariance baseline.","marker":"[4]"},{"why":"Supplies the permeable-shell calculation that matches the factor $\\frac{5}{6}$ improvement for a sphere.","marker":"[26]"}],"fun_headline_variants":["Perfect monitors should ignore the interior for sensing","Boundary-only sensing beats volume averaging","Optimal sensing uses only the surface, not the bulk","For perfect monitors, the best weight is a surface charge"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the measurement time $T$ is long enough that time-integrated diffusion correlations equal the static Coulomb Green's function ($T \\gg a^2/D$), and in one and two dimensions this requires an additional recurrence subtraction; if that long-time replacement fails, the variance is no longer exactly the Coulomb energy and boundary localization need not be optimal.","fun_headline_variants_meta":{"raw":{"variants":["Perfect monitors should ignore the interior for sensing","Boundary-only sensing beats volume averaging","Optimal sensing uses only the surface, not the bulk","For perfect monitors, the best weight is a surface charge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1345,"prompt_tokens":907,"completion_tokens":438,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":523,"completion_tokens_details":{"reasoning_tokens":378}},"tokens_in":523,"tokens_out":438,"duration_ms":5161,"temperature":1.0,"reasoning_tokens":378,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:23:46.890584+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a Brownian-dynamics simulation of a non-perturbing spherical monitor of radius $a$ over a range of finite measurement times $T$, computing the variance of the surface-shell estimator and of the uniform-volume estimator for identical molecule fields; if for any $T \\gg a^2/D$ the surface-shell variance exceeds the predicted $c_0/(2\\pi D a T)$, or a numerical variational search finds a non-boundary weight with lower variance, the central claim is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the standard position-weighted gradient estimator and supplies its covariance baseline."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the permeable-shell calculation that matches the factor $\\frac{5}{6}$ improvement for a sphere."}],"review_version":1}