Back to Blog

Heat Kernels: Building Intuition

An accessible introduction to heat kernels on Riemannian manifolds — what they are, why they matter, and how to think about them geometrically.

Mathematics Geometry Analysis

The heat kernel is one of the most beautiful objects in geometric analysis. It connects partial differential equations, probability, and the shape of spaces in a single, elegant package. In this post, I want to build intuition for what the heat kernel is and why geometers care about it so deeply.

The Setup

Imagine placing a point source of heat on a curved surface — say, the north pole of a sphere. As time progresses, the heat diffuses across the surface. The heat kernel $K(t, x, y)$ tells you exactly how much heat arrives at point $y$ after time $t$, given that the source was at point $x$.

On flat Euclidean space $\mathbb{R}^n$, this is the familiar Gaussian:

$$K(t, x, y) = \frac{1}{(4\pi t)^{n/2}} \exp!\left(-\frac{|x - y|^2}{4t}\right)$$

But on a curved manifold, the geometry bends the heat flow. Curvature acts like a lens — positive curvature focuses the heat, while negative curvature disperses it.

Why It Matters

The heat kernel encodes an astonishing amount of geometric and topological information:

  1. Short-time asymptotics reveal local geometry (scalar curvature, Ricci tensor).
  2. The trace $\int_M K(t, x, x), dx$ connects to the spectrum of the Laplacian via the heat trace expansion.
  3. Long-time behaviour detects topology — the number of connected components, Betti numbers, and $L^2$-invariants.

This last point was the starting point for my Master’s thesis on the analytic aspects of $L^2$-invariants.

A Probabilistic Perspective

There’s a beautiful probabilistic interpretation: $K(t, x, y)$ is the transition density of Brownian motion on the manifold. In other words, if you release a random walker at $x$, the heat kernel tells you the probability density of finding them at $y$ after time $t$.

This connection between probability and geometry is one of the deepest themes in modern mathematics, and it continues to inspire new research directions — from stochastic analysis on singular spaces to applications in machine learning and graph neural networks.


In a future post, I’ll explore how these ideas extend to non-compact manifolds and how the heat kernel is used to define $L^2$-Betti numbers.