This question is a kind of local version of a previous post (MO224171).

Let $\ f_n \ $ be a sequence of Morse functions on $\mathbb{R}^d$, converging, in the $C^\infty$-topology, to a limit Morse function $\ f$: $$ f_n \to f \ .$$

At any critical point $\ p\ $ of the limit function $\ f$, it can be proved that there exists an open neighborhood $\ U\ $ such that, for $n \geq n_0$, $$ f_{n|U} \mbox{ has a unique critical point, $p_n$, of the same Morse index than } p. $$

Moreover, the sequence of these critical points $p_n$ converges to $p$.

On the other hand, let $\ W^s(p,f)\ $ stand for the stable manifold associated to $p$: if $\tau_t(x)$ denotes the integral curve of the gradient of $f$ passing through $x$ at $t=0$, then

$$ W^s(p , f) := \left[ \ x \in \mathbb{R}^d \ \colon \ \lim_{t \to \infty} \tau_t (x) = p \ \right] . $$

If $V$ is a neighborhood of $p$, we can also consider the local stable manifold:

$$ W^s_V(p , f) := W^s(p , f_{|V}) . $$

My question is:

Does there exist a neighborhood $V$ of $p$ such that the local stable manifolds $\ W^s_V(p_n , f_n)\ $ converge (in some adequate sense) to $W^s_V(p , f)$?

For example, such that the Hausdorff distance $d_H (W^s_V(p_n , f_n) , W^s_V(p , f))$ converge to zero?

I've read in Wikipedia that local stable manifolds vary continuously in a neighborhood of $f$, but couldn't find a proof. Probably this is not difficult for the well versed in dynamical systems or differential topology, but I cannot arrange the proof of the Stable Manifold Theorem to work for a family.

  • 1
    $\begingroup$ I think the answer is yes, and $C^2$ convergence is enough.There is a box nbd of $p$ of the form $D^s\times D^u$ such that $W^s( p; f )\cap D^s\times D^u$ and $W^s( p_n; f_n )\cap D^s\times D^u$ for $n$ large enough are graphs of maps $F: D^s\to D^u$ resp. $F_n: D^s\to D^u$; moreover $F_n\to F$ uniformly with second order derivatives on $D^s$. The same local convergence of $W^s$ as graphs happens then at any $x\in W^s(p;f)$. (Of course, no hope of global convergence, e.g. Hausdorff distance, as shown by easy 2 dimensional examples). $\endgroup$ Nov 24, 2015 at 20:44
  • 1
    $\begingroup$ Thanks! But, how can you assure that all the $F_n$ are defined in the same $D^s$? (I looked at a proof of the Stable Manifold Theorem, and tried to arrange precisely this idea, but I couldn't avoid the $D^s_n$ of the $F_n$ from becoming arbitrary small!). $\endgroup$ Nov 25, 2015 at 10:34
  • $\begingroup$ This is a key point of course. If you keep track of the radius of $D^s$ in the proof of the stable manifold theorem, e.g. via the graph transform, you can check there is a uniform radius. $\endgroup$ Nov 25, 2015 at 13:02
  • $\begingroup$ Or here, via transversality. link.springer.com/chapter/10.1007/1-4020-4266-3_01#page-1 $\endgroup$ Nov 25, 2015 at 13:06
  • $\begingroup$ ops, sorry the latter is not free; here's a free version dm.unipi.it/~abbondandolo/preprints/montreal.pdf $\endgroup$ Nov 25, 2015 at 13:12

1 Answer 1


Yes, $p$ is a hyperbolic fixed point of the gradient flow, and local (un)stable manifolds of a hyperbolic fixed persist smoothly under small perturbations (which your functions $f_n$ are, relative to $f$ for $n$ large). I don't have a basic reference at hand, but this is a special case of a normally hyperbolic invariant manifold $M$ consisting of a single point, $M = \{p\}$, see for the NHIM result theorem 4.1(f) in Hirsch, Pugh, Shub: "Invariant manifolds", Springer LNM 538 (1977). The $C^r$ (here $r = \infty$) closeness should be viewed in the way already commented by Pietro Majer: you view $W^{u,s}(p;f|_V)$ as graphs, and then the graphs of $W^{u,s}(p;f_n|_V)$ converge to these in $C^r$ norm.

Moreover, this local result can be extended to any compact neighborhood of $p$ that does not contain other fixed points: simply note that the extended (un)stable manifold is obtained under the gradient flow for a fixed, finite time $t$. Smooth dependence of flows on parameters implies that the flows $\Phi_n^t$ converge to $\Phi^t$, hence the (un)stable manifolds converge too. (Near another fixed point $q$ the problem is that the flow time to pass it blows up.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.