MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The following question is probably classic in Morse theory, so a reference to an existing result should be sufficient. I don't know much about Morse theory and I am dealing with the following situation. I have a compact manifold $X$ and I have a Morse function $f$ on it with a saddle point at $x_0$. I don't like the properties of $f$(for some "mysterious" reason) so I take a parametrization around $x_0$, taking $x_0$ to the origin and containing no additional critical points, and add to $f$ a function of the type $\epsilon \rho Q$, where $Q$ is a quadratic polynomial, $\epsilon$ is a small number and $\rho$ is a bump function living in the coordinate patch that is identically equal to 1 in a neighborhood of the origin.

Clearly this new function still has a critical point at $x_0$. My question is: for small enough $\epsilon$ is this new function still Morse, with the same critical points as the original function $f$?

share|cite|improve this question
up vote 4 down vote accepted

Sure. For any sequence $\epsilon_n\to 0$ a sequence of critical points of $f_n:=f+\epsilon_n \rho Q$ would have a subsequence which converges to a critical point of $f$, so in particular would eventually lie in a neighborhood on which the function equals either the original $f$ (if the limit isn't $x_0$) or $f+\epsilon_n Q$ (if the limit is $x_0$). But in such neighborhoods the only critical points are the critical points of the original function $f$, at least if $\epsilon_n$ is small enough.

If the answer to your question was "no" then you could find $\epsilon_n\to 0$ contradicting the above.

share|cite|improve this answer
Many thanks for your time. – The Common Crane Jul 1 '12 at 16:52
@Mike: The critical point might shift when you perturbed the function. Just think about quadratic functions of one variable. – Misha Jul 1 '12 at 16:55
@Misha: My interpretation of the question is that we are using a coordinate system centered at $x_0$, and the unique critical point of $Q$ is the center of the coordinate system. So, as noted in the original question, $x_0$ will continue to be a critical point of the perturbation. (It goes without saying that the critical points would shift under generic perturbations, but the question asked about a perturbation of a rather particular kind.) – Mike Usher Jul 1 '12 at 17:11
"But in such neighborhoods the only critical points are the critical points of the original function f, at least if $\epsilon_n$ is small enough". Why is this obvious? – The Common Crane Jul 1 '12 at 17:25
Working in a (suitably small) coordinate system with $x_0$ identified with the origin and using the standard metric on this coordinate system, the nondegeneracy of the Hessian of $f$ at $x_0$ gives you an estimate $|(\nabla f)(v)|\geq C_1|v|$ for $v$ in some fixed neighborhood of the origin. Meanwhile there's also an estimate $|(\nabla Q)(v)|\leq C_2|v|$. So $|\nabla(f+\epsilon Q)(v)|\geq (C_1-\epsilon C_2)|v|$, showing that, in this fixed neighborhood and for $\epsilon$ sufficiently small, the only critical point of $f+\epsilon Q$ is the origin. – Mike Usher Jul 1 '12 at 17:55

There is classic result (I think going back to H. Whitney) that states that if $f:M\to \mathbb{R}$ is a Morse function on a compact smooth manifold $M$, then there exist a neighborhood $\mathscr{N}$ of $f$ in $C^\infty(M)$ sucht that if $g\in \mathscr{N}$, then $g$ is equivalent to $f$, i.e., one can obtain $g$ from $f$ by conjugating $f$ with a diffeomorphism $\Phi$ of $M$ and a diffeomorphim $\varphi$ of $\mathbb{R}$. More precisely, this means that

$$ g = \varphi \circ f \circ \Phi. $$

For a proof see the book of Golubitski and Guillemin Stability of smooth mappings. Loosely speaking this result says that if $g$ is not too far from $f$, then $g$ is obtained from $f$ via a change of "coordinates" on $M$ and a change in the target space $\mathbb{R}$.

share|cite|improve this answer
intersting. Thanks! – The Common Crane Jul 7 '12 at 5:48

Lemma B on page 12 of Milnor's Lectures on h-cobordism theorem. I am sure it is also in his book on Morse functions, I just do not have it with me now.

share|cite|improve this answer
thanks. A reference is just what I wanted. – The Common Crane Jul 1 '12 at 16:52
I think I declared victory too soon :(. How does it follow that for small $\epsilon$ the new function will have the same critical points as the original $f$. – The Common Crane Jul 1 '12 at 17:23

Are you sure you cannot achieve what you want by simply changing coordinates? Answers are about adding general small function but you want to add a quadratic polynomial only on a neighborhood of $x_0$.

share|cite|improve this answer
No. I am dealing with quantities that are coordinate independent! – The Common Crane Jul 1 '12 at 17:12
You add a "quadratic polynomial Q". This only make sense in a choosen local coordinates. – Maciej Starostka Jul 1 '12 at 21:34

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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