Sharp Kernel: Precise Heat Kernel Bounds

• Sharp Kernel is a precise pointwise estimate for integral kernels that matches both upper and lower bounds with optimal exponents and exact boundary behavior.
• The Jacobi heat kernel result extends sharp estimates to the entire parameter range α,β > -1 using analytic continuation and reduction techniques.
• These sharp kernel estimates provide a foundational framework for harmonic analysis, impacting studies of Fourier–Bessel, spectral fractional, and oscillator kernels.

In analysis, PDE, and harmonic analysis, a sharp kernel estimate is a pointwise description of an integral kernel whose leading behavior is captured up to multiplicative constants, with the decisive exponents, boundary factors, and off-diagonal decay fixed in their optimal form. In the Jacobi setting, this notion is realized by the 2024 result of Nowak–Sjögren–Szarek, which establishes genuinely sharp two-sided bounds for the Jacobi heat kernel for the full admissible parameter range $\alpha,\beta>-1$, completing earlier work that had left the singular ranges $\alpha<-\tfrac12$ or $\beta<-\tfrac12$ unresolved (Nowak et al., 2024).

1. Sharpness as a kernel-theoretic notion

In the literature represented here, “sharp” does not mean merely that a kernel admits Gaussian-type control. It means that the kernel is bounded above and below by expressions of the same functional form, with no loss in the decisive exponents and with boundary singularities described by the correct polynomial factors. In the Jacobi heat-kernel result, sharpness is stated explicitly as the coincidence of the Gaussian exponent $\tfrac14$ in both lower and upper bounds, together with prefactors in $t$ and boundary factors that are exact up to constants independent of $\theta,\phi,t$ (Nowak et al., 2024).

This usage is consistent across several adjacent settings. For the harmonic-oscillator potential kernel, the sharp statement is a uniform two-sided estimate with the precise power-law, logarithmic, or $(1+\|x+y\|)^{d-2\sigma}$ behavior depending on the threshold $\sigma=d/2$, combined with the mixed off-diagonal exponential factor $\exp(-c\,\|x-y\|(\|x\|+\|y\|))$ (Nowak et al., 2011). In $\mathrm{RCD}(0,N)$ metric-measure spaces with maximum volume growth, sharp Gaussian heat-kernel bounds mean matching $\alpha<-\tfrac12$0 scaling and asymptotically optimal volume constants (Li, 2016). For spectral fractional Laplacians perturbed by a gradient, sharpness refers to preservation of the exact boundary-decay profile encoded by the factors involving $\alpha<-\tfrac12$1 and $\alpha<-\tfrac12$2 (Song et al., 2017).

A common misconception is that sharp kernel estimates require an explicit Mehler-type or closed-form summation formula. The Jacobi and Fourier–Bessel settings show otherwise: both kernels are defined by oscillatory series that cannot be computed explicitly, yet one can still obtain sharp pointwise bounds through product formulas, transference, subordination, or perturbative representations [(Nowak et al., 2024); (Nowak et al., 2012)].

2. The Jacobi heat kernel as the central model

For $\alpha<-\tfrac12$3, the Jacobi heat kernel is

$\alpha<-\tfrac12$4

with the usual normalizations

$\alpha<-\tfrac12$5

Writing $\alpha<-\tfrac12$6 and $\alpha<-\tfrac12$7 for $\alpha<-\tfrac12$8, the 2024 theorem states that for every fixed $\alpha<-\tfrac12$9 there is $\beta<-\tfrac12$0 such that for all $\beta<-\tfrac12$1 and $\beta<-\tfrac12$2,

$\beta<-\tfrac12$3

where

$\beta<-\tfrac12$4

The result holds in the full range $\beta<-\tfrac12$5, for all $\beta<-\tfrac12$6, and for all $\beta<-\tfrac12$7; the long-time regime $\beta<-\tfrac12$8 follows by elementary semigroup arguments (Nowak et al., 2024).

The proof analysis identifies, up to comparable constants, the decisive short-time structure as

$\beta<-\tfrac12$9

This exhibits the two endpoint singularities separately. The factors involving $\tfrac14$0 control behavior near $\tfrac14$1, while the factors involving $\tfrac14$2 control behavior near $\tfrac14$3. The Gaussian term captures the off-diagonal concentration.

The significance of the 2024 theorem is that it resolves precisely the parameter regime in which previous methods did not deliver genuinely sharp two-sided bounds. Earlier sharp short-time estimates were available only for $\tfrac14$4 (Nowak et al., 2011), whereas the new theorem includes the previously open subcases with $\tfrac14$5 or $\tfrac14$6, including the delicate region $\tfrac14$7 (Nowak et al., 2024).

3. Analytic continuation of the Dijksma–Koornwinder formula

The key structural advance is an extension of the Dijksma–Koornwinder product formula from the classical range $\tfrac14$8 to the full range $\tfrac14$9 by analytic continuation (Nowak et al., 2024). This is the step that converts a formerly inaccessible singular-parameter problem into one amenable to reduction arguments.

For each integer $t$0, one defines

$t$1

with

$t$2

One also introduces

$t$3

together with the signed measures

$t$4

In the simplest case $t$5, the product formula takes the form

$t$6

For the other parameter regions, the representation becomes a finite linear combination of repeated integrals of $t$7 against the corresponding signed measures. This extension is the technical novelty that makes it possible to pass to heat kernels by inserting the factors $t$8 and summing over $t$9 (Nowak et al., 2024).

A plausible implication is that the phrase “sharp kernel” in this context is inseparable from a structural representation theorem. The sharp estimate is not obtained by direct asymptotic manipulation of the oscillatory series; it is obtained by replacing the series with a superposition of simpler kernels whose sharp behavior is already understood.

4. Reduction to one-dimensional kernels and assembly of the bounds

After inserting the heat weights and using Fubini, the extended product formula yields reduction formulas expressing $\theta,\phi,t$0 as a superposition of simpler heat kernels, ultimately ultraspherical heat kernels in one variable (Nowak et al., 2024). The reduction depends on the position of $\theta,\phi,t$1 and $\theta,\phi,t$2 relative to $\theta,\phi,t$3.

In the mixed case $\theta,\phi,t$4, the paper gives

$\theta,\phi,t$5

where

$\theta,\phi,t$6

and

$\theta,\phi,t$7

with $\theta,\phi,t$8 and $\theta,\phi,t$9 (Nowak et al., 2024).

The one-dimensional building blocks are controlled by explicit asymptotic estimates. For any $(1+\|x+y\|)^{d-2\sigma}$0,

$(1+\|x+y\|)^{d-2\sigma}$1

uniformly for $(1+\|x+y\|)^{d-2\sigma}$2 and $(1+\|x+y\|)^{d-2\sigma}$3. A corresponding odd-part extension satisfies a similar estimate when $(1+\|x+y\|)^{d-2\sigma}$4 (Nowak et al., 2024).

The remaining step is an integration problem over Dirichlet-type measures. In each parameter region, one arrives at finitely many terms involving

$(1+\|x+y\|)^{d-2\sigma}$5

where

$(1+\|x+y\|)^{d-2\sigma}$6

Lemmas 5.4–5.5 show that each such integral is comparable to

$(1+\|x+y\|)^{d-2\sigma}$7

and summing the finitely many contributions yields Theorem A (Nowak et al., 2024).

This proof architecture clarifies why the result is stronger than a qualitative Gaussian bound. The reduction preserves enough information to recover not only off-diagonal decay, but also the exact polynomial boundary behavior.

5. Historical development and related sharp-kernel results

The 2024 Jacobi theorem completes a program that began with sharp heat-kernel estimates in the non-singular Jacobi range. In 2011, Nowak–Sjögren proved sharp two-sided short-time bounds for the Jacobi heat kernel when $(1+\|x+y\|)^{d-2\sigma}$8, with the characteristic factors

$(1+\|x+y\|)^{d-2\sigma}$9

and used the upper bound to derive a weak type $\sigma=d/2$0 inequality for the maximal operator of the multidimensional Jacobi heat semigroup (Nowak et al., 2011). The later extension to $\sigma=d/2$1 shows that the limitation to $\sigma=d/2$2 was not intrinsic to the kernel itself, but to the range of the classical product formula (Nowak et al., 2024).

The Jacobi kernel also serves as a transfer model for the Fourier–Bessel setting. In that framework, for $\sigma=d/2$3 one considers

$\sigma=d/2$4

and sharp two-sided bounds were derived by relating the generator to a Jacobi operator with parameters $\sigma=d/2$5 and using Trotter’s formula (Nowak et al., 2012). Earlier half-integer results in the Fourier–Bessel setting had already shown the special role of $\sigma=d/2$6, where one can identify the problem with the radial Dirichlet Laplacian on the Euclidean unit ball (Nowak et al., 2011).

Beyond orthogonal-polynomial systems, the same sharp-kernel paradigm appears in other operator settings. The harmonic oscillator admits qualitatively sharp potential-kernel estimates with the exact trichotomy at $\sigma=d/2$7 (Nowak et al., 2011). In bounded $\sigma=d/2$8 domains, the heat kernel of the spectral fractional Laplacian perturbed by a time-dependent gradient satisfies sharp two-sided bounds preserving the boundary profile of the unperturbed kernel (Song et al., 2017). In metric-measure spaces satisfying $\sigma=d/2$9 and maximum volume growth, sharp Gaussian-type heat-kernel estimates lead to exact large-time asymptotics for both the heat kernel and the minimal Green function (Li, 2016).

A useful distinction emerges from these examples. Some results are genuinely sharp in the sense of exact matching exponent and boundary structure, while older literature sometimes provided only qualitative bounds with an “exponential gap” in exponent constants. The 2024 Jacobi theorem explicitly removes that gap in the singular-parameter range (Nowak et al., 2024).

6. Analytical consequences and broader significance

The Jacobi heat kernel underlies analysis on Euclidean spheres and compact rank-one symmetric spaces, Fourier–Bessel and more general Laguerre settings, and analysis on the ball, simplex, cone, and their boundaries (Nowak et al., 2024). Because of this, genuinely sharp Jacobi bounds immediately imply precise Gaussian bounds for the corresponding heat semigroups, sharp estimates of associated maximal functions, Littlewood–Paley-type operators and functional calculi, and frame and sampling constructions in Dirichlet spaces in the full admissible parameter range $\exp(-c\,\|x-y\|(\|x\|+\|y\|))$0 (Nowak et al., 2024).

The broader significance of the sharp-kernel viewpoint is methodological as much as quantitative. In the Jacobi case, analytic continuation of a product formula is the enabling mechanism. In the Fourier–Bessel case, the bridge is transference to Jacobi or to Euclidean balls [(Nowak et al., 2012); (Nowak et al., 2011)]. In the spectral-fractional setting, Duhamel’s formula and a perturbation series preserve the kernel’s leading order (Song et al., 2017). In degenerate parabolic problems, weighted volume terms of the form $\exp(-c\,\|x-y\|(\|x\|+\|y\|))$1 encode precisely the boundary degeneracy, and the two-sided Gaussian form is again optimal (Negro et al., 2024).

This suggests that “sharp kernel” is best understood not as a single object, but as a standard of description. A kernel estimate is sharp when it resolves the exact interaction between diffusion scale, geometry, boundary singularity, and spectral structure. In the recent Jacobi theory, that standard is met on the full parameter domain $\exp(-c\,\|x-y\|(\|x\|+\|y\|))$2, and the result now functions as a foundational input for several adjacent areas of harmonic analysis and semigroup theory (Nowak et al., 2024).