1: \begin{abstract}

2: We consider the solution $u(x,t)$ of the Fisher-KPP equation $\partial_t

3: u=\partial_x^2u+u-u^2$ centred around its $\alpha$-level

4: $\mu_t\uppar\alpha$ defined as $u(\mu_t\uppar\alpha,t)=\alpha$. It is well

5: known that for an initial datum that decreases fast enough, then

6: $u(\mu_t\uppar\alpha+x,t)$ converges as $t\to\infty$ to the critical

7: travelling wave. We study in this paper the speed of this convergence and

8: the asymptotic expansion of $\mu_t\uppar\alpha$ for large~$t$. It is known

9: from Bramson \cite{Bramson83} that for initial conditions that decay fast

10: enough, one has $\mu_t\uppar\alpha=2t-(3/2)\ln t+\text{Cste}+o(1)$. Work

11: is under way \cite{nrr} to show

12: that the $o(1)$ in the expansion is in fact

13: a $k\uppar\alpha/\sqrt t+\mathcal O(t^{\epsilon-1})$ for any $\epsilon>0$

14: for some $k\uppar\alpha$, where it is not clear at this point whether

15: $k\uppar\alpha$ depends or not on $\alpha$. We show that, unless

16: the time derivative of $\mu_t\uppar\alpha$ has a very unexpected behaviour

17: at infinity, the

18: coefficient $k\uppar\alpha$ does not, in fact, depend on $\alpha$.

19: 

20: We also conjecture that, for an initial condition that decays fast enough,

21: one has in fact $\mu_t\uppar\alpha=2t-(3/2)\ln

22: t+\text{Cste}-(3\sqrt\pi)/\sqrt t+g (\ln t)/t +o (1/t)$ for some

23: constant~$g$ which does not depend on $\alpha$.

24: 

25: \end{abstract}

26: