The Heat Equation and the Fourier transform: a paradox

Paradox (noun) a statement or proposition that seems self-contradictory or absurd, but in reality expresses a possible truth.

Solution by Fourier transform

To solve the heat equation \[ u_t = \Delta u, \qquad u(x, 0) = g(x) \] we consider the Fourier transform \[ \hat u(\xi, t) = (2\pi)^{-n/2} \int e^{-ix\cdot \xi} u(x, t) \, dx. \] It satisfies \[ \hat u_t = -|\xi|^2 \hat u(\xi, t), \qquad \hat u(\xi, 0) = \hat g(\xi), \] so that $\hat u(\xi, t) = e^{-t|\xi|^2} \hat g(\xi)$. The solution of the heat equation can be recovered by applying the inverse Fourier transform: \[ u(x, t) = (2\pi)^{-n/2} \int e^{ix\cdot \xi} e^{-t|\xi|^2} \hat g(\xi)\, d\xi. \]
Since $e^{-t|\xi|^2}$ is a smooth function which vanishes very rapidly as $|\xi|\to\infty$, multiplication with $e^{-t|\xi|^2}$ only improves the function $ e^{-t|\xi|^2} \hat g(\xi)$, so the above solution is defined for all $t\ge0$.
Conclusion: solutions to the heat equation never become singular for $t\gt0$.

A singular solution

The function \[ u(x, t) = \frac{1} {\bigl(4\pi(1-t)\bigr)^{n/2}} e^{|x|^2/4(1-t)} \] is a smooth solution to the heat equation which becomes singular as $t\nearrow1$.
Conclusion: solutions to the heat equation can become singular at some $t\gt0$.