In this video, Felix Otto says that the correct way to count dimensions for parabolic equations is $2+\text{number of space dimensions}$. He said nothing about this. In the accompanying notes it is stated, again with no explanation.

When I searched the literature the only place I could find any reference is again more SPDE/regularity structures people. Here it is stated with little explanation. (Page 10)

Other PDE people usually say that time counts as one dimension. Why in SPDE/regularity structures does time "count for" 2 dimensions?

Does it have something to do with complex spacetime and Wick rotation? If we consider time complex it is 2D. However in all discussed SPDEs, time is strictly real.

Edit: I thought about it some more. Does it have to do with parabolic scaling? I.e. we scale $u(t,x_1,...x_n)$ as $u(c^2 t, c x_1,...,c x_n)$. The $t$ term picks up $c^2$ so it "counts twice"?

  • $\begingroup$ Yes, scaling is the key word. Consider the equation $(\partial_t {-}\Delta) u = 0.$ Then $u^{\lambda} = u(\lambda^2t, \lambda x)$ solves the equation. The dimension they refer to can be considered the "dimension of scaling" (to see why it's a dimension, try to compute the $L^1$ norm of the rescaled solution). Maybe to probabilists this is well known: in time $t$ Brownian motion usually reaches a box of size $\sqrt{t}$. But I think in PDE theory in general scaling plays a crucial role. $\endgroup$ – Kore-N Sep 8 '18 at 15:10

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.