= Backward heat equation
{title2=$u_t=-D u_{xx},\quad D>0$}
The backward heat equation $u_t=-D u_{xx}$, $D>0$, amplifies a spatial <Fourier mode> by $e^{Dk^2t}$. On a periodic domain, fix $T>0$ and take initial data $f_n=e^{-Dn^2T/2}\sin(nx)$. Their <L2 norms> tend to zero, while the corresponding solutions at $T$ have <L2 norms> growing without bound. Thus the solution map is not <continuous> in the ordinary <L2 norm>; this is an <ill-posed problem>.<Partial Differential Equations by John K. Hunter>[§5.1.3, printed pp131–132, irreversibility and Example5.6]{ref}
Real analyticity alone does not resolve the issue. The periodic <analytic function> $f(x)=\sum_{n\geq1}e^{-n}\cos(nx)$ has exponentially decaying coefficients, but after any positive time the proposed coefficients $e^{Dn^2t-n}$ are not square summable. The required data restriction must control growth on the Gaussian frequency scale. For example, a fixed-horizon norm
$$
\|f\|_{X_T}^2=\sum_{n\in\mathbb Z}e^{2DTn^2}|\widehat f_n|^2
$$
gives $\|u(T)\|_2=\|f\|_{X_T}$ when the Fourier basis is normalized. This very strong data norm changes the stability question. In inverse heat reconstruction, an explicit <regularization> or frequency cutoff is needed when measured data carry ordinary small-scale noise.
Back to article page