SearcharxivSearch

arXiv subjects

Mayukh Mukherjee

Publications and source records attributed to Mayukh Mukherjee.

At least 19 recordsLinked to original sources

A sharp covering theorem and Solyanik estimates for Euclidean balls

For every finite family of Euclidean balls in $\mathbb{R}^n$, $n\ge2$, and every $0<\delta<1/2$, we select a subfamily whose $(1+\delta)$-dilations cover the original union and whose undilated balls have multiplicity at most $C_n\delta^{-(n-1)/2}$. This proves the covering estimate conjectured by Han and Lu \cite{HL}. As an application, we determine the optimal Solyanik asymptotic \[ \mathcal C_n(\alpha)-1\asymp_n(1-\alpha)^{2/(n+1)} \qquad(\alpha\uparrow1) \] for the uncentered Hardy--Littlewood maximal operator over Euclidean balls, which is Conjecture~1(b) of Hagelstein and Parissis \cite{HP14}. For the modified uncentered maximal operators, we determine the optimal weak $(1,1)$ growth rate $(k-1)^{-(n-1)/2}$ as $k\downarrow1$, uniformly over Radon measures. We also obtain the corresponding $L^p$ and Fefferman--Stein bounds.

math.CA

Wavelength-scale optional stopping, critical Feynman-Kac gauges, and capacitary spectral inequalities

We develop two Brownian stopping methods: one uses hitting probabilities and capacity, the other moments of weighted exit distributions. On a closed $m$-dimensional Riemannian manifold, $m\geq 2$, the first gives low-energy spectral inequalities for equilibrium measures supported on sets of zero volume. Combining scales gives heat observability and null controllability from a full-support measure carried by a dense set of Hausdorff dimension $m-2$, minimal under a bounded $H^1$-trace condition. For every nondecreasing unbounded $\Phi$ with $\Phi(4E)\leq D\Phi(E)$, the measure can be chosen so that the optimal spectral constant is comparable to $\Phi(E)$ and the small-time control cost to $T^{-1/2}\Phi(T^{-1})^{1/2}$. Squaring the Feynman-Kac martingale doubles the potential: for radial $V$, the boundary second moment is log-convex in $\log r$ if $r^2V(r)$ is nondecreasing, wherever the regular solution of $(\tfrac12\Delta+2V)g=0$ stays positive; the factor $2$ is sharp when $V(0)>0$. On small geodesic balls, weighted second moments give an almost-monotone frequency and a weighted three-radius inequality; bounds for its logarithmic derivative give doubling estimates, which, after a ground-state transform, feed the Remez theorem of Logunov-Malinnikova. The normalised first-moment exit law gives an averaged boundary-variance identity for Laplace eigenfunctions. At $r=c\lambda^{-1/2}$, outside a set carrying $O(c^{2/m})$ of the $\varphi^2$-mass, the component of $\{\varphi(x)\varphi>0\}\cap B(x,r)$ through $x$ contains a concentric ball of nearly full radius, fills nearly all of $B(x,r)$, and nearly matches its first Dirichlet eigenvalue. First-moment stopping also controls positive superlevel components. Reflected Brownian motion and local time give boundary-to-collar $L^\infty$ estimates for Steklov eigenfunctions on bounded Lipschitz domains, and an $L^2$ estimate for $C^2$ domains.

math.AP

Spectral Dehn functions and a characterisation of word-hyperbolicity

We introduce a \emph{spectral Dehn function} \[ \Lambda_{\mathcal{P}}(n):=\inf \lambda_1(\Delta), \] where $\lambda_1(\Delta)$ is the first Dirichlet eigenvalue of the random-walk Laplacian on a van Kampen diagram $\Delta$, and the infimum runs over area-minimising diagrams with boundary length at most $n$. We prove a spectral-isoperimetric inequality relating $\Lambda_{\mathcal{P}}$ to the Dehn function, and show that its degree-free face-dual variant $\Lambda^\ast_{\mathcal P}$ characterises word-hyperbolicity: a finitely presented group is word-hyperbolic if and only if \[ \inf_n \Lambda^\ast_{\mathcal{P}}(n)>0. \] Every disk diagram satisfies a diagramwise filling-length bound \[ \mathrm{FL}_b(\Delta)\cdot \operatorname{Area}(\Delta) \ge c/\lambda_1(\Delta); \] combined with a discrete Faber-Krahn inequality, this yields the sharp exponent $1/2$ in the quadratic case, attained by rectangular commutator grids over $\mathbb Z^2$. By passing to the free completion and introducing a hole-free-ancestor hereditary quasi-minimality condition, we obtain a spectral filling profile whose positivity criterion is a quasi-isometry invariant of finitely presented groups and again characterises word-hyperbolicity. The resulting profile carries finer information than the Dehn function: it separates presentations within the linear Dehn class.

math.GR

An Approximate Inverse Spectral Theorem for Manifolds of Constant Negative Curvature

A classical theorem of Colin de Verdi\`ere shows that on a closed manifold of fixed topology one can prescribe an arbitrary finite portion of the Laplace-Beltrami spectrum (including multiplicities, subject to the usual topological constraints) by choosing a sufficiently heterogeneous smooth metric. In this paper, we study the same inverse problem under the rigid geometric constraint of \emph{constant negative sectional curvature}. Allowing the topological complexity to vary, we prove that any finite strictly increasing target list can be approximated to arbitrary precision by a closed manifold of constant negative curvature in any dimension $d\ge2$. In $d=2$ the construction uses hyperbolic collar degeneration and the discrete spectral limit theorems of Burger, building on the collar estimates of Buser; in $d\ge3$ we build macroscopically heterogeneous hyperbolic covering manifolds assembled from ``heavy'' vertex clusters and ``long'' corridor chains whose low-energy limit is a prescribed \emph{discrete} graph Laplacian. We also record the universal obstructions at curvature normalization $\kappa\equiv -1$: Yang-Yau in $d=2$ and Kazhdan-Margulis combined with Bishop--Gromov volume comparison in $d\ge3$. In particular, $\lambda_1$ is universally bounded at $\kappa=-1$, so target lists whose first positive eigenvalue exceeds this bound cannot be approximated within the class $\kappa\equiv -1$, and accommodating arbitrarily large prescribed $\lambda_1^*$ forces $|\kappa|\to\infty$. A corollary on the arbitrarily precise prescription of scale-invariant eigenvalue ratios at $\kappa\equiv -1$ and an explicit worked example are included.

math.SP

Sharp thresholds for the Escobar functional: the Escobar-Willmore mass, geometric selection, and compactness trichotomy

We study the hemisphere threshold for the conformally covariant Escobar functional on compact Riemannian manifolds $(M^n,g)$ with boundary. The near-threshold landscape is organized by boundary invariants: the first-order coefficient $\rho_n^{\mathrm{conf}}H_g$ vanishes identically, so the leading obstruction is a renormalized boundary mass $\mathfrak R_g$ (second order, $n\ge5$), followed by a cubic invariant $\Theta_g$ (third order, $n\ge6$), with a Green kernel interaction $\mathsf G_\partial$ in the multi-bubble regime. Exact evaluation of weighted profile moments yields $\kappa_1=\kappa_2=0$: the coefficients of $\operatorname{Ric}_g(\nu,\nu)$ and $\mathrm{Scal}_{\bar g}$ in the bare mass vanish. On $\{H_g=0\}$ the mass reduces to $\mathfrak R_g^{\mathrm{bare}}=\frac{6-n}{2(n-1)(n-3)(n-4)}|\mathring{\mathrm{II}}|^2$. The Lyapunov--Schmidt correction gives $\mathfrak R_g^{\mathrm{red}}\le\mathfrak R_g^{\mathrm{bare}}\le0$ for $n\ge6$; for $n=5$ the nonlocal back-reaction overcomes the positive bare coefficient. In every dimension $n\ge5$, non-umbilic boundaries are automatically subcritical: $C^*_{\mathrm{Esc}} 0$ yields compactness and hemispherical rigidity. In the multi-bubble regime we establish global compactness at Escobar multiples with equal-mass quantization and conditional exclusion of pure multi-bubbling.

math.DG

The Lipschitz Liouville Property, Affine Rigidity, and Coarse Harmonic Coordinates on Groups of Polynomial Growth

We develop a quantitative theory of Lipschitz harmonic functions (LHF) on finitely generated groups, with emphasis on the Lipschitz Liouville property, affine rigidity, and quasi-isometric invariance for groups of polynomial growth. On finitely generated nilpotent groups we prove an affine rigidity theorem: for any adapted, smooth, Abelian-centered probability measure $\mu$, every Lipschitz $\mu$-harmonic function is affine, $f(x)=c+\varphi([x])$. For any finite generating set $S$ this yields a canonical isometric identification $$ \mathrm{LHF}(G,\mu)/\mathbb{C} \cong \mathrm{Hom}(G_{\mathrm{ab}},\mathbb{C}),\qquad \|\nabla_S f\|_\infty=\max_{s\in S}|\varphi([s])|, $$ independent of the choice of centered measure. In addition, we prove an identification of HF$_1$ with LHF on polynomial growth groups for adapted, smooth, Abelian-centered measures. Next, for any finite-index subgroup $H\le G$ and adapted smooth $\mu$ we prove a quantitative induction-restriction principle: restriction along $H$ and an explicit averaging operator give a linear isomorphism $\mathrm{LHF}(G,\mu)\cong\mathrm{LHF}(H,\mu_H)$, where $\mu_H$ is the hitting measure, with two-sided control of the Lipschitz seminorms. For groups of polynomial growth equipped with $\mathrm{SAS}$ measures we then show that $\mathrm{LHF}$ is a quasi-isometry invariant as a seminormed affine space, via choice-dependent Shalom--Sauer transport on virtual first cohomology. Separately, for quasi-isometries with bounded Abelian defect, we construct coarse harmonic coordinates that straighten them up to bounded error. Finally, within the Lyons-Sullivan/Ballmann-Polymerakis discretization framework, we prove a quantitative discrete-to-continuous extension theorem: Lipschitz harmonic data on an orbit extend to globally Lipschitz $L$-harmonic functions on the ambient manifold, with gradient bounds controlled by the background geometry.

math.GR

Wulff Isoperimetry on Cayley Graphs: Submodular BV, Tempered F{\o}lner, and Profile Ratio Bounds

We develop a BV framework on Cayley graphs, which yields a sharp discrete Wulff isoperimetric inequality with the best constant tied to the generating set/stencil $S$, a quantitative $\Gamma$-convergence of discrete to continuum anisotropic perimeter, and a gauge construction in Heisenberg group that neutralizes shear, yielding an BV+shear identity and scale-sharp compactness. As an application we revisit a question of Gromov (2008) on the ratio between the isoperimetric profile and its greatest nondecreasing minorant. We show that this is uniformly bounded on non-amenable and two-ended groups, and our Wulff inequality and $\Gamma$-convergence give a constant-tracked proof for virtually nilpotent groups. We isolate a Tempered F{\o}lner criterion (TF) (exhaustion principle with controlled increment), which forces bounded ratio in general. We verify (TF) in two families: finite-lamp wreath products over (TF) bases and lamplighters over amenable bases. For semidirect products $\mathbb Z^d\rtimes_A\mathbb Z$ we construct layer-nested sets of logarithmic height that are $A$-covariantly nested, F{\o}lner, and satisfy (TF)(ii) with a constant independent of $A\in\mathrm{GL}(d,\mathbb Z)$; and when $A$ is hyperbolic (no eigenvalue on the unit circle) we have full (TF). We also formulate ``gap conjectures'' that would settle the question for all amenable Cayley graphs. The BV viewpoint has spectral and analytic consequences: we derive constant-tracked Cheeger-type, Faber-Krahn, Nash inequalities, etc. The (TF) control further yields a tempered Property A, leading to explicit coarse embeddings into Hilbert space with compression $\rho(t) \gtrsim t^{1/2}/\log t$. Finally, we show that (TF) is a robust reflection of bi-equivariant geometry, and in virtually nilpotent classes this \emph{doubles} the sharp Wulff constant asymptotically-refining and answering another question of Gromov.

math.GR

Nodal Domains on Surfaces under Perturbation: Upper Semicontinuity, Courant-Sharpness, and Boundary Intersections

We study how the number of nodal domains of eigenfunctions of Schr\"odinger operators $-\Delta_{g_t}+V_t$ on closed surfaces changes under smooth perturbations of $(g_t,V_t)$ along convergent eigenbranches. Locally, near each nodal critical point of the limit eigenfunction, we give a sector/graph count showing that no new local domains can be created and that vanishing orders cannot increase. Globally, we prove upper semicontinuity of the nodal domain count; in the noncritical case the count is stable. The result is branch-free on spectral clusters. At the wavelength scale, new closed nodal loops cannot be created. We also treat localised (topology-changing) perturbations: the count inside the unperturbed core cannot increase. As applications, we construct metrics on any closed surface that are Courant-sharp up to an arbitrary finite level and prescribe $2n_i$ boundary intersections on each boundary component. An appendix records a uniform (wavelength-scale) lower bound on the inner radius of nodal domains along the branch.

math.SP

Green geometry, Martin boundary and random walk asymptotics on groups

We identify a single computationally checkable analytic quantity interlacing Martin boundary collapse, Green geometry, and linear escape for transient random walks on finitely generated groups: the Green-variation functional \[ \Delta(S;a,b):=\max_{x\in\partial S}\frac{|G(a,x)-G(b,x)|}{G(a,x)}. \] We prove that $\Delta\to0$ along exhaustions characterises the strong Liouville property (under mild, verifiable hypotheses on the ``strong Liouville $\Rightarrow \Delta\to0$'' direction), turning boundary oscillation estimates for Green kernels into potential-theoretic rigidity. We then give two general criteria for $\Delta$-vanishing. The first one derives quantitative bounds on $\Delta$ from coarse heat-kernel envelopes at an intrinsic scale together with a Tauberian comparability, covering Gaussian/sub-Gaussian and stable-like regimes; and the second one is purely elliptic: an ``elliptic H\"older exhaustion'' criterion. Conversely, on groups of exponential growth, $\Delta$ fails to decay along balls already under stretched-exponential on-diagonal upper bounds, yielding a quantitative obstruction to strong Liouville. As consequences, trivial Martin boundary forces linear-scale collapse of Green geometry ($d_G(e,x)=o(|x|)$) and vanishing Green speed (in probability), without any entropy hypothesis. On the non-Liouville side we prove an abundance principle: the existence of a single minimal positive harmonic function at a prescribed growth scale forces infinitely many. Finally, we clarify the role of moment assumptions in speed theory: any linear-speed law of large numbers on a set of positive probability forces $\mathbb E|X_1|<\infty$, while on torsion-free nilpotent groups one can have $\mathbb E|X_1|=\infty$ yet $|X_n|/n\to0$ in probability.

math.GR

On the Spectrum of Certain Hadamard Manifolds

We show the absolute continuity of the spectrum and determine the spectrum as a set for two classes of Hadamard manifolds and for specific domains and quotients of one of the classes.

math.DG

On the effects of small perturbation on low energy Laplace eigenfunctions

We investigate several aspects of the nodal geometry and topology of Laplace eigenfunctions, with particular emphasis on the low frequency regime. This includes investigations in and around the Payne property, opening angle estimates of nodal domains, saturation of (fundamental) spectral gaps etc., and behaviour of all of the above under small scale perturbations. We aim to highlight interesting aspects of spectral theory and nodal phenomena tied to ground state/low energy eigenfunctions, as opposed to asymptotic results.

math.SP

Heat profile, level sets and hot spots of Laplace eigenfunctions

We use probabilistic tools based on Brownian motion and Feynman-Kac formulae to investigate the heat profile for the ground state Dirichlet and second Neumann eigenfunctions. Among other topics, we comment on supremum norm bounds for ground state Dirichlet eigenfunctions and look at the corresponding Neumann problem, namely the comparison of maximum temperatures on the interior and the boundary, the latter being partially motivated by the hot spots problem. We also investigate the proximity/distance of level sets of ground state Dirichlet eigenfunctions, some with analogous statements for Neumann eigenfunctions. Domains with bottlenecks make occasional appearances as an illuminating example as well as testing ground for our theory.

math.AP

On the Impact of Stable Ranks in Deep Nets

A recent line of work has established intriguing connections between the generalization/compression properties of a deep neural network (DNN) model and the so-called layer weights' stable ranks. Intuitively, the latter are indicators of the effective number of parameters in the net. In this work, we address some natural questions regarding the space of DNNs conditioned on the layers' stable rank, where we study feed-forward dynamics, initialization, training and expressivity. To this end, we first propose a random DNN model with a new sampling scheme based on stable rank. Then, we show how feed-forward maps are affected by the constraint and how training evolves in the overparametrized regime (via Neural Tangent Kernels). Our results imply that stable ranks appear layerwise essentially as linear factors whose effect accumulates exponentially depthwise. Moreover, we provide empirical analysis suggesting that stable rank initialization alone can lead to convergence speed ups.

cs.LG

Mass non-concentration at the nodal set and a sharp Wasserstein uncertainty principle

We prove $L^p$-mass concentration properties of Laplace eigenfunctions away from their nodal sets, extending a recent result in \cite{GM3} to all dimensions, and giving a slight refinement of a result in \cite{JN}. As a consequence, we are able to derive a sharp Wasserstein uncertainty principle that holds uniformly in the high frequency regime, proving a conjecture in \cite{St1}.

math.AP

Some applications of heat flow to Laplace eigenfunctions

We consider mass concentration properties of Laplace eigenfunctions $φ_λ$, that is, smooth functions satisfying the equation $-Δφ_λ= λφ_λ$, on a smooth closed Riemannian manifold. Using a heat diffusion technique, we first discuss mass concentration/localization properties of eigenfunctions around their nodal sets. Second, we discuss the problem of avoided crossings and (non)existence of nodal domains which continue to be thin over relatively long distances. Further, using the above techniques, we discuss the decay of Laplace eigenfunctions on Euclidean domains which have a central "thick" part and "thin" elongated branches representing tunnels of sub-wavelength opening. Finally, in an Appendix, we record some new observations regarding sub-level sets of the eigenfunctions and interactions of different level sets.

math.AP

Heating up decision boundaries: isocapacitory saturation, adversarial scenarios and generalization bounds

In the present work we study classifiers' decision boundaries via Brownian motion processes in ambient data space and associated probabilistic techniques. Intuitively, our ideas correspond to placing a heat source at the decision boundary and observing how effectively the sample points warm up. We are largely motivated by the search for a soft measure that sheds further light on the decision boundary's geometry. En route, we bridge aspects of potential theory and geometric analysis (Mazya, 2011, Grigoryan-Saloff-Coste, 2002) with active fields of ML research such as adversarial examples and generalization bounds. First, we focus on the geometric behavior of decision boundaries in the light of adversarial attack/defense mechanisms. Experimentally, we observe a certain capacitory trend over different adversarial defense strategies: decision boundaries locally become flatter as measured by isoperimetric inequalities (Ford et al, 2019); however, our more sensitive heat-diffusion metrics extend this analysis and further reveal that some non-trivial geometry invisible to plain distance-based methods is still preserved. Intuitively, we provide evidence that the decision boundaries nevertheless retain many persistent "wiggly and fuzzy" regions on a finer scale. Second, we show how Brownian hitting probabilities translate to soft generalization bounds which are in turn connected to compression and noise stability (Arora et al, 2018), and these bounds are significantly stronger if the decision boundary has controlled geometric features.

cs.LG

Nodal sets of Laplace eigenfunctions under small perturbations

We study the stability properties of nodal sets of Laplace eigenfunctions on compact manifolds under specific small perturbations. We prove that nodal sets are fairly stable if said perturbations are relatively small, more formally, supported at a sub-wavelength scale. We do not need any assumption on the topology of the nodal sets. As an indirect application, we are able to show that a certain "Payne property" concerning the second nodal line remains stable under controlled perturbations of the domain.

math.AP

Polyhedral billiards, eigenfunction concentration and almost periodic control

We study dynamical properties of the billiard flow on convex polyhedra away from a neighbourhood of the non-smooth part of the boundary, called ``pockets''. We prove there are only finitely many immersed periodic tubes missing the pockets and moreover establish a new quantitative estimate for the lengths of such tubes. This extends well-known results in dimension $2$. We then apply these dynamical results to prove a quantitative Laplace eigenfunction mass concentration near the pockets of convex polyhedral billiards. As a technical tool for proving our concentration results on irrational polyhedra, we establish a control-theoretic estimate on a product space with an almost-periodic boundary condition. This extends previously known control estimates for periodic boundary conditions, and seems to be of independent interest.

math.AP