Papers

By addressing the problem of describing the differences between the saturation monoids of the Prandtl and Leray-$\alpha$ boundary layers, and algebraically distinguishing the action of the Helmholtz operator, we can show that, when the modification of the vector field is halted at a minimum scale determined by the differential filter, the paradox of constructing a solution that reaches infinite velocity in a finite time is avoided. Furthermore, it allows us to obtain the coefficients of the Taylor expansion near the wall for boundary layers where the Helmholtz operator has been applied.

We establish a probabilistic Lie-Trotter-Kato formula for the gradient flow generated by an integral function over suitable separable Hadamard spaces. This result leverages Stojković's Lie-Trotter-Kato formula for gradient flows of sums of convex functions and Bačák's result relating resolvent and gradient flow convergence, along with a strong law of large numbers for normal convex integrands. For separable Hilbert spaces and locally compact Hadamard spaces, these strong laws follow from results of King and Wets and Artstein and Wets, respectively, but we crucially go beyond these examples by establishing a new such strong law for normal convex integrands on all Hadamard spaces whose tangent cones are isometric to Hilbert spaces, a class that includes non-Hilbertian, non-locally compact examples such as infinite-dimensional hyperbolic space. This new strong law arises as a combination of a strong law of large numbers for integrable random monotone vector fields on Hilbert-Hadamard spaces, that is, Hadamard spaces whose tangent cones all embed isometrically into Hilbert spaces, extending a result of Salim from Hilbert spaces, together with a new result on the interchangeability of the subdifferential and the integral, which we establish for $L^2$-Lipschitz integrands. This latter ingredient provides, to our knowledge, the first nonlinear version of (a particular case of) a previous result due variously to Ioffe and Tikhomirov, Levin, Hiriart-Urruty, Thibault as well as Rockafellar and Wets. On the way to this strong law for random monotone vector fields, we also establish a distribution-uniform generalization of Sturm's strong law of large numbers for inductive means of $L^1$-sequences of i.i.d. random variables on Hadamard spaces via a new effective non-asymptotic concentration inequality. Throughout the paper, we highlight various remaining open problems.

We give an explicit example of a $W^{*}$-bundle $M$ over a compact, metrizable space $K$, which has each fiber $M_{p}$ a $\textrm{II}_{1}$-factor with separable predual, and which satisfies the following negation of the generator problem: given any finite family $a_{1},\cdots,a_{n}\in M$ of continuous sections, there is a $p\in K$ (depending upon that family) so that $a_{1,p},\cdots,a_{n,p}$ do not generate $M_{p}$ as a von Neumann algebra. More generally, if $N$ is a sub-bundle of $M$ with the property that each fiber is hyperfinite, or has a Cartan, or is generated by two commuting diffuse subalgebras, or has diffuse central sequence algebra, or is generated by a single sequential commutation orbit, then given any finite family $a_{1},\cdots,a_{n}\in M$ of continuous sections, there is a $p\in K$ (depending upon that family) so that $a_{1,p},\cdots,a_{n,p}$ together with $N_{p}$ do not generate $M_{p}$. We discuss implications for the generator problem for von Neumann algebras: e.g. there is no “continuous" way to take countably many generators for a tracial von Neumann algebra with separable predual and produce a single generator, at least if such a procedure works for all von Neumann algebras simultaneously.

In this short note, we show that the loss of order $N^{c_p/\log\log N}$ in the version of the Ionescu–Wainger multiplier theorem for canonical fractions, recently established by Kosz, Mirek, Słomian, and Zhang in arXiv:2606.12663, is optimal.

Jets in stratified flows are ubiquitous in geophysics. A fundamental question is whether the instabilities and coherent structures generated by the flow interact with the physical boundaries and, if so, how. To this end, a simple canonical model flow is employed: a parallel Bickley jet confined between walls in a uniformly stratified fluid. We first show that short wavelength Kelvin-Helmholtz instability can generate asymmetry in the mean flow. Although the fluctuations decay rapidly away from the jet, the asymmetric mean flow does not. We then examine long-wavelength coherent structures that can emit internal gravity waves, which decay slowly towards the walls or, in some cases, radiate away from the jet. These waves transport momentum and distort the mean flow, reducing the jet centreline velocity significantly. When radiation occurs, whether the instability persists at high Reynolds numbers depends strongly on the boundary conditions imposed at the walls. Moreover, we found an instability and associated nonlinear coherent structures that persist at Richardson numbers above the Miles-Howard criterion. The existence of this viscous mode does not rely on the presence of the walls.

We study the semilinear heat equation ${}^{S}D_0^{\alpha(t)}u=\Delta u+u^p$ on $\RR^n$, where $p>1$ and the Scarpi order changes exponentially from $\alpha_1$ to $\alpha_2$, with $0<\alpha_1,\alpha_2<1$. We establish local and maximal mild well-posedness for abstract Scarpi–Volterra equations. For the whole-space heat problem, positive resolvent families yield nonnegative solutions, comparison, a mass identity, and an $L^\infty$ blow-up alternative. To study finite-time growth, we use a Gaussian version of Kaplan's weighted-moment method. It reduces the PDE to a scalar nonlinear Volterra inequality and requires no pointwise lower estimate for the non-self-similar Scarpi heat kernel. Consequently, every nontrivial solution blows up in finite time when $1<p<1+2/n$, and sufficiently large data blow up for every $p>1$. For $u_0=A\varphi$ with fixed nonzero $0\le\varphi\in L^1\cap L^\infty$, the maximal lifespan satisfies \[ T_A\asymp A^{-(p-1)/\alpha_1} \qquad(A\to\infty), \] while a subcritical small-amplitude upper bound is governed by the long-time order $\alpha_2$. The lifespan bounds are expressed through the inverse of the integrated memory. In the numerical section, we complement these estimates by comparing the Scarpi dynamics with both Caputo endpoint models. Continuous Laplace inversion shows that the logarithmic slope of the integrated memory varies nonmonotonically across the transition. At large amplitudes, the growth thresholds approach those of the Caputo $\alpha_1$ model. A nonlinear space–time rescaling probes the long-time regime and reveals nonmonotone threshold-time ratios relative to Caputo $\alpha_2$. For fixed-width initial profiles, increasing the transition rate delays the prescribed growth threshold at larger tested amplitudes but advances it at smaller ones.

Machine learning-based partial differential equations (PDEs) solvers have attracted significant attention in recent years. Most progress in this area has been driven by deep neural networks such as physics-informed neural networks (PINNs) and kernel method (such as physics-informed Gaussian Processes). We introduce a physics-informed random feature method for countering part of the spectral bias which PINN-based solvers are facing for a certain class of PDEs. Random feature method was originally proposed to approximate large-scale kernel machines and can be viewed as a specialized randomized neural network. Compared to other state-of-the-art PINN-based solvers which require a large number of collocation points, our proposed method reduces the computational complexity. In this paper, we develop a rigorous approximation error analysis and derive high-probability error bounds on the $H^1$ norm. We provide extensive numerical tests for verifying our theoretical guarantees on error decay rates, as well as several comparison tests to showcase our claimed capability for combating spectral bias in these deep learning based methods.

On a connected closed smooth Riemannian manifold, the algebraic structure of the Laplace eigenfunctions, as described by eigenfunction triple products, uniquely determines the geometry. We refine this correspondence by introducing the notion of an $N$-product eigenbasis, which consists of eigenfunctions whose pairwise products may be written as linear combinations of at most $N$ basis elements. We prove that a manifold admits a $2$-product eigenbasis if and only if it is a flat torus. We also prove an analogous result for Laplace eigenvectors of bounded-degree graphs.

We prove that every $n$-vertex $15$-tough graph contains a cycle of length $\Omega(n^{1/20})$.

The Skolem Problem asks whether a given linear recurrence sequence (LRS) has a zero term, and is equivalent to proving an effective version of the Skolem-Mahler-Lech theorem, which states that a non-degenerate LRS has finitely many zeros. Decidability of the Skolem Problem however, has remained open for many decades. Bilu et al. (2022) showed that the Skolem Problem for simple LRS is decidable subject to the weak $p$-adic Schanuel Conjecture and the Skolem Conjecture (also known as the exponential local-global principle), the latter of which states that an LRS has an integer zero if and only if it has a zero modulo every integer $m$. This paper works towards understanding the Skolem Conjecture. We provide an example showing that a natural strengthening of the Skolem Conjecture, where we restrict $m$ to be a prime power, is false. This failure is accounted for by the relationship between $p$-adic zeros of LRS (which arise naturally in the proof of the Skolem–Mahler–Lech theorem and have been studied algorithmically by Bacik et al. (2026)) and twisted rational zeros (introduced by Bilu et al. (2025)). By studying this relationship further, we are able to prove the main result of this paper: a local-global principle for simultaneous zeros of two coprime LRS, subject to the $p$-adic Schanuel Conjecture. A fundamental step in proving this result is characterising when twisted rational zeros are (or are not) $p$-adic zeros for infinitely many primes $p$, which we do unconditionally. This also answers two open questions of Bilu et al. (2025). Finally, we conjecture that the existence of twisted rational zeros is the only way that the aforementioned strengthened Skolem Conjecture may fail, which is supported by a heuristic argument.

We construct a metric 2-sphere of finite Hausdorff 2-measure that is doubling, linearly locally connected and Loewner but cannot be mapped to the standard sphere by a quasisymmetric mapping. Our result gives a negative answer to a question of Ntalampekos.

This is the second of two papers dedicated to the explicit computation of the Fargues–Scholze correspondence. We compute the Tate cohomology of cuspidal representations arising from Yu's construction. Combining this with modular functoriality in the Local Langlands Correspondence, independence of $\ell$, and the partial characterization of the Local Langlands Correspondence established in the first paper, we compare the Fargues–Scholze and Kaletha parametrizations. Among the consequences, we deduce that the Fargues–Scholze correspondence has finite fibers and is surjective onto inertial L-parameters.

This is the first of two papers dedicated to the explicit computation of the Fargues–Scholze correspondence. In this first part, we construct (inspired by, and extending, Kaletha's explicit Local Langlands Correspondence for non-singular supercuspidal representations) semisimple inertial L-parameters for all cuspidal representations arising from Yu's constructions, with arbitrary coefficients. We then establish a characterization of this correpondence in terms of certain functoriality properties. This characterization will be used in the sequel paper in order to compute the Fargues–Scholze correspondence after restriction to inertia.

In their previous work devoted to the distribution of rational points near Brownian motion, the authors conjectured the existence of an oscillation principle governing the asymptotic behavior of the number of rational points with bounded denomi\-nators near the graph of a monofractal curve. In this note, a weaker form of this conjecture is shown to hold for a broad class of deterministic fractal curves. These include the classical Takagi and Weierstrass nowhere differentiable functions, and indeed a prevalent (i.e.~"large") class of functions among those which are Hölder continuous with a given exponent of regularity. This constitutes the first instance of deterministic fractal curves for which a precise count of the rational points under consideration is established.

In 2015, Bringmann, Dousse, Lovejoy, and Mahlburg defined the function $\overline{t}(n)$ to be the number of overpartitions of weight $n$ where (i) the difference between two successive parts may be odd only if the larger part is overlined and (ii) if the smallest part is odd then it is overlined. In their work, they proved that $\overline{t}(n)$ satisfies an elegant congruence modulo 3. Since then, a number of authors have studied arithmetic properties satisfied by $\overline{t}(n)$. In particular, in 2023, Hanson and Smith utilized the theory of modular forms to prove the following two congruences modulo 5: For all $n\geq 0$, \begin{align*} \overline{t}(80n+40)\equiv \overline{t}(80n+60)\equiv 0 \pmod{5}. \end{align*} Our goal in this work is to provide a truly elementary proof of this pair of congruences.

We construct a computable irrational number that simultaneously has small continued fraction entropy and base $b$ entropy for every integer base $b\geq 2.$ This addresses a question raised by Bugeaud.

We study function fields in one variable over the field with one element $\mathbf{F}_1$. It is proved that the Galois extensions of $\mathbf{Q}$ are isomorphic to the curves over $\mathbf{F}_1$ being understood as the Deitmar schemes. Specifically, one gets explicit formulas linking the genus and the number of cusps of an algebraic curve over the extension $\mathbf{F}_{1^m}$ of order $m\ge 1$ of the field $\mathbf{F}_1$ and the $m$-th roots of unity of the corresponding number field. It follows that elliptic curves with one cusp over $\mathbf{F}_1$ are either the cyclotomic fields or the maximal abelian unramified extensions of the quadratic number fields. Our proof depends on representaion of the Drinfeld modules by the bounded linear operators on a Hilbert space and the crossed product structure of the Cuntz-Krieger algebras.

A partial matrix has entries in $\{0,1,\star\}$, and a disambiguation replaces each $\star$ by $0$ or $1$. We construct partial matrices whose fully specified submatrices satisfy strong restrictions, yet every disambiguation contains every binary matrix of a prescribed size. Our first result answers a question of Alon, Hanneke, Holzman and Moran on the disambiguation of linear classifiers with margin. For $0<\varepsilon<\pi/2$, let $M_\varepsilon^d$ be the partial matrix indexed by points of the unit sphere $\mathbb S^d$, with entry $0$ for pairs at spherical distance at most $\varepsilon$, $1$ for pairs at distance at least $\pi-\varepsilon$, and $\star$ otherwise. Although these matrices have VC-dimension bounded independently of $d$, we prove that every disambiguation contains every binary $k\times k$ matrix once $d$ is sufficiently large. This also yields a partial concept class of Littlestone dimension $1$ with no disambiguation of finite VC-dimension. We also construct, for every $k$, a finite partial matrix whose fully specified $2\times2$ submatrices are all constant, while every disambiguation contains every binary $k\times k$ matrix. A symmetric analogue holds for partial graphs: for every $k$, there exists a partial graph of VC-dimension at most $1$ whose fully specified induced subgraphs are all cliques or stable sets, yet every disambiguation contains every $k$-vertex graph as an induced subgraph. A disambiguation can be viewed as a $2$-coloring of the unspecified entries, making Ramsey theory a natural framework for forcing prescribed patterns. Our proofs draw on two recent Ramsey theorems: the geometric argument uses Pálvölgyi's Dense Block theorem, while the combinatorial constructions rely on the girth Ramsey theorem of Reiher and Rödl, a suitable strengthening of the induced Ramsey theorem.

Let $G$ be a graph with vertex set $V(G)$. A set $I\subseteq V(G)$ is an independent dominating set of $G$ if no two vertices in $I$ are adjacent and every vertex in $V(G)\setminus I$ is adjacent to at least one vertex in $I$. The independent domination number of $G$ is the minimum cardinality among all independent dominating sets of $G$. The aim of this article is to obtain tight bounds and closed formulas for the independent domination number of central graphs. The results are expressed in terms of parameters of the original graph from which the central graph is constructed.

Visible dynamics and mean dissipation need not determine the collective phases of rare histories. We construct local Markov models with identical ordinary visible path distributions, stationary states, mean entropy production, mean jump activity and driving affinities, yet different phases when histories are weighted by their full entropy production. Visible configurations modulate hidden driven cycles, whose exact elimination turns kinetic correlations into interactions. A positive spectral reduction separates hidden relaxation from collective modes. Periodic binary models retain opposite phases with entropy variances matched at every observation time. For conserved particles, a pair matches the first two joint cumulants of entropy and hidden activity at every observation time, yet selects segregation or a fluctuating state with suppressed long-wavelength density noise. Bound-state transport and fermionic evolution determine their contrasting dynamics. At a clean Ising critical midpoint, the entropy source couples quadratically to the thermal perturbation, making the fourth cumulant the first to detect the critical susceptibility. Its rate per site grows logarithmically, with an exact cutoff set by system size and observation time. For two quenched disorder ensembles, exact sample-response formulas and numerical evidence support a finite fourth-cumulant density despite very slow relaxation; a weak-disorder continuum response admits a controlled Borel sum. A common positive hidden eigenvector characterizes autonomous selected dynamics. When this condition fails, a finite network develops visible memory while preserving its ordinary visible process and mean diagnostics. These results identify kinetic information missed by aggregate thermodynamic observations.

Operator splitting methods are a fundamental class of algorithms for solving structured monotone inclusion problems arising in optimization, control, and game theory. We consider the case, common in stochastic regimes, where the forward evaluation of one of the constituent operators is either unavailable in closed form or computationally expensive to evaluate, and is therefore approximated using a finite number of noisy oracle samples. We establish distribution-free, finite-sample certificates for the quality of the output produced by data-driven Davis-Yin splitting algorithms. Unlike previous works, our analysis directly controls the residual error via martingale inequalities instead of relying on algorithmic stability arguments for a tailored surrogate loss, yielding the first a priori certificates whose statistical excess provably vanishes with the sample size. We further show that, under linear convergence of the Davis-Yin splitting algorithm, the dependence of our bounds on the iteration count improves from linear growth to exponential decay. We validate our theoretical results on a stochastic portfolio optimization problem with uncertain asset returns.

In this paper, we give an explicit formula for the rank and generators (up to finite index) over $\mathbb{Q}(t)$ of all non-trivial elliptic curves of the form $y^2=x^3+At^6+Bt^3+C$, which is a larger class of elliptic surfaces than the one in Kloosterman's paper (2026) and Desjardins and Naskrecki's paper (2024), namely $y^2=x^3+At^6+C$. Our proof provides a new method to find this formula using generators of the geometric Mordell-Weil group while this geometric Mordell-Weil group (which is a $\mathbb{Z}[\omega]$-module) no longer decomposes into rank-one submodules by the methods in Kloosterman's paper (2026) and Desjardins and Naskrecki's paper (2024). Moreover, our proof uses only a few elements in the Galois group of the coordinates of the generators and only light computations by hand.

We propose a factorization of the graph isomorphism problem for Bestvina-Brady groups (BBGs) through matroid theory. In particular, we show that finitely presented BBGs depend on their defining graphs only through their cycle matroids. On the other hand, we construct graphs of arbitrarily high connectivity such that they have non-isomorphic cycle matroids but their BBGs are isomorphic. To do so, we prove that if a graph admits a tree clique-spanner, then the Dicks-Leary presentation of its BBG can be explicitly simplified to a right-angled Artin group presentation. In particular, we show that BBGs defined by dually chordal graphs are right-angled Artin groups.

Let $k$ be an algebraically closed field of characteristic $p>0$, $n\ge1$, and let $X\subset \mathbb P^n_k$ be a finite nonempty set of distinct points with the defining ideal $I=I(X)$. We give a proof of Demailly's inequality in positive characteristic. The argument is based on the Frobenius–Hasse derivative method used in this context by Hà and Sivakumar. The key additional observation is a strict-growth lemma for the initial degrees of symbolic powers of a finite set of affine points: $$ \alpha(J^{(t)})\ge \alpha(J^{(t-1)})+1\qquad(t\ge1). $$ In characteristic $p$, if a minimum-degree polynomial has a nonzero first ordinary derivative, this follows by differentiation; if all first ordinary derivatives vanish, perfectness gives a $p$th root and an induction on the symbolic exponent. The remainder of the proof uses the $q=p^e$ Frobenius decomposition and a maximal Hasse derivative.

We establish a general mechanism that turns qualitative smoothness into uniform, quantitative regularity estimates for fully nonlinear elliptic equations. The central conclusion is that, for compact classes preserved by the natural rescalings of the equation, qualitative and quantitative regularity have the same critical threshold. At second order, mere twice differentiability throughout the corresponding centered hull already forces a uniform $C^{2,\alpha}$-theory for some $\alpha>0$. A one-sided tangent version gives a new partial regularity criterion for a single operator: every point at which a solution is twice differentiable is regular, whereas the singular set has universal positive codimension. The proof combines compactness of the full renormalized class with a scale-adaptive improvement of flatness, converting information available separately at each profile into estimates uniform across the entire class. The argument requires no quantitative control of the assumed smoothness and opens a path in problems where smoothness is visible, but estimates remain out of reach.

Locally decodable codes (LDCs) enable recovery of any message symbol by probing only a small number of positions in a possibly corrupted codeword. The central parameters of an LDC are its rate, locality, and error tolerance. Ideally, one would like all three parameters to be constant. However, classical lower bounds show that such codes cannot exist. A recent line of work introduced amortized locally decodable codes (aLDCs), in which the decoder is tasked with recovering an entire block of consecutive message symbols rather than a single symbol. While prior work obtained ideal aLDCs with constant rate, constant error tolerance, and constant amortized locality, those constructions relied on either shared randomness hidden from the channel or computational assumptions restricting the channel. Another well-studied relaxation is the notion of a relaxed locally decodable code (RLDC), in which the decoder may output a special failure symbol $\bot$ rather than risk decoding incorrectly. In this work, we introduce the notion of an amortized relaxed locally decodable code (aRLDC), combining amortized decoding with the relaxed decoding paradigm. Unlike prior ideal aLDC constructions, our model is fully information-theoretic and makes no assumptions about shared randomness or computational limitations of the adversarial channel. We construct the first aRLDC with constant rate, constant error tolerance, and constant amortized locality. Moreover, for any block of length $\Omega(\mathrm{polylog}(k))$, our decoder achieves amortized locality $1+\delta^{1 - o(1)}$, where $\delta$ is the error tolerance parameter. Thus, asymptotically, recovering a long block requires essentially less than two codeword probe per message symbol recovered. By contrast, without amortization no RLDC can simultaneously achieve constant rate, constant error tolerance, and constant locality.

The completely bounded Bohnenblust–Hille inequality of Arunachalam, Dutt, Escudero Gutiérrez and Palazuelos controls coefficient summability with optimal constant one, independently of the ambient dimension. We classify all nonzero equality cases over both the real and complex fields: they are precisely scalar multiples of Hadamard chains supported on equal-sided Cartesian boxes. Thus equality determines both the support and the factorization of the coefficient tensor. For each fixed degree $d\ge2$, we also establish sharp structural stability. A form with completely bounded norm one and deficit $\varepsilon$ lies within $C_d\sqrt\varepsilon$, in the critical coefficient $\ell_{2d/(d+1)}$ norm, of both a flat unimodular path and a normalized unitary chain on the same box. The corresponding unimodular and scaled unitary edges satisfy a matching normalized Frobenius estimate. At zero deficit both approximants coincide with the original Hadamard chain. All constants are independent of the ambient dimension, and the exponent $1/2$ is optimal for both approximations over either field.

This work presents the first steps for simulating kinetic plasmas with collisions using the characteristic mapping method (CMM). The CMM is a semi-Lagrangian method that explores a semi-group structure to store diffeomorphic flow maps efficiently. Using the semi-group structure, individual submaps can be composed to relate the flow backward in time to its initial food point. The novelty of the presented work is handling the source term by storing sub-integrals that correspond to the individual sub-maps and allow efficient integration. Furthermore, we use the Lagrangian structure to avoid implicit time integration in the hydrodynamic regime, which is known to be stiff. We benchmark our method on the Boltzmann-BGK and Vlasov-Poisson-BGK equations and consider different test cases, the Sod shock tube problem and nonlinear Landau damping for different Knudsen numbers. We show third-order spatial and temporal convergence and illustrate the fine-scale zoom property of CMM for the bump-on-tail instability.

We study quantum systems modeled by Liouville-Von Neumann equations, with a finite number of energy levels. The system is coupled to a wave equation, intended to reproduce environmental effects on the quantum system. The whole dynamic is energy conservative, but damping effects are expected to be embodied into the energy exchanges with the environment. We analyse how these effects can effectively damp quantum coherences. We exhibit key mechanisms and structural properties for such a damping to arise, with a neat distinction between pure and mixed cases

A classical consequence of Caffarelli's fully nonlinear regularity theory [L. A. Caffarelli, Ann. of Math. (2) 130 (1989), no. 1, 189-213] is that viscosity solutions of uniformly elliptic equations $F(D^2u)=f$, with $f\in L^p$, $p > n$, are locally $C^{1,\gamma}$ for every $\gamma < \min\{\alpha_H,\sigma_p\}$. Here $\alpha_H=\alpha_H(n,\lambda,\Lambda)\in(0,1)$ denotes the universal Hölder exponent for gradient regularity of $F(D^2h)=0$, while $\sigma_p=1-n/p$ is the scaling exponent of the source term. In the source-limited regime $\sigma_p < \alpha_H$, the singularity of $f$ is the decisive obstruction and the endpoint $\gamma=\sigma_p$ is attainable. In the homogeneous-limited regime $\alpha_H\le\sigma_p$, however, classical theory only yields $\gamma < \alpha_H$, leaving the limiting differentiability estimate unquantified. This is the endpoint gap addressed here. When $\alpha_H < \sigma_p$, we prove that solutions admit pointwise Taylor expansions satisfying $|u(x)-u(x_0)-Du(x_0)\cdot(x-x_0)|\lesssim |x-x_0|^{1+\alpha_H}\left(1+\log\frac{1}{|x-x_0|}\right)^m$. Thus the homogeneous differentiability scale is reached up to an explicit logarithmic defect. At the critical threshold $\alpha_H=\sigma_p$, finite logarithmic powers no longer close the iteration; nevertheless, a slower selection of scales yields $|u(x)-u(x_0)-Du(x_0)\cdot(x-x_0)|=O\left(|x-x_0|^{1+\alpha_H}\exp\left(A\sqrt{1+\log\frac{1}{|x-x_0|}}\right)\right)$. Both estimates improve the full family of classical sub-endpoint $C^{1,\gamma}$ bounds, $\gamma < \alpha_H$, by quantifying differentiability at the limiting homogeneous exponent. The proof introduces a new scale-selection mechanism for endpoint Campanato-type recurrences, suggesting a flexible tool whenever the limiting smoothness is dictated by the homogeneous theory itself. We also discuss the role and possible optimality of the resulting logarithmic defects.