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 ring $C^\infty(M)$ of smooth functions on a smooth manifold $M$ is a topological ring with respect to the Whitney topology and the usual ring operations. Is it possible to describe, maybe under some conditions on $M$, the ideals and the closed ideals of $C^\infty(M)$?

share|cite|improve this question
J'imagine il sera utile d'ajouter quelque detail supplementaire. En plus, j'ai vu que avant vous avez ecrit votre questions en Anglais. Pour la plupart des particpants dans ce site-oueb (je crois) il est plus agreable de lire en Anglais. Alors, si ce n'est pas tres penible pour vous, il sera gentil d'ecrire en Anglais. Merci beaucoup par avance. – user9072 Oct 5 '11 at 18:44
Cette question m'ennerve moins que les questions affreuses deja posees en Anglais par l'auteur... mais enfin, elle n'est pas une bonne question – Yemon Choi Oct 5 '11 at 19:53
Je crois que cette question a ete fermee plutot a cause des questions precedentes posees par son auteur que parce qu'elle est mauvaise en soi. C'est vrai qu'elle n'est pas tres bien formulee, mais je pense qu'on peut la sauver. C'est ce que je vais essayer de faire ; cependant, je la reecrirai en anglais pour la rendre plus accessible aux participants de ce site. – algori Oct 5 '11 at 20:19
unknowngoogle: now that it has been edited to include some precise definitions, the question looks much better. The original version was in itself inoffensive, but the lack of detail was not encouraging – Yemon Choi Oct 5 '11 at 20:50

In 1948 Whitney proved his ideal (spectral) theorem [1] describing the closed ideals

Let $M$ be an $n$-dimensional manifold. for each point $p \in M$ and each natural $k$ we define $N(k)$ to be the number of (up to $n$) tuples $m$ such that $|m| \leq k$. Define the map $J_p^k: C^\infty(M) \rightarrow \mathbb{R}^{N(k)}$ by assigning to $f$ the $m$-jets of $f$ at $p$ up to $|m|=k$.

If $I$ is an ideal of $C^\infty(M)$ then its closure is the ideal of functions $f$ such that for each $p$ in $M$ and $k \geq 0$ then $J^k_p f \in J^k_p(I)$.

So in some sense the closed ideals are like $I_\infty$ in Neil's answer.

[1] H.Whitney. On ideals of differentiable functions. American Journal of Mathematics. Vol. 70, No. 3, pp. 635-658 (1948)

share|cite|improve this answer
I was about to upvote this question when I saw the edit history.. I guess I should upvote algori :) EDIT: well, couldn't resist, it's the question we vote, not who asks it. – Reimundo Heluani Oct 6 '11 at 1:21

Here are some examples, for the case $M=\mathbb{R}$.

For each $n\geq 0$ we have a closed ideal $$ I_n=\{f: f^{(i)}(0)=0 \text{ for } 0\leq i < n\} $$ We can write $I_\infty$ for the intersection of these, which is again closed. We can also put $$ J = \{ f : f(x)=0 \text{ for all } x \leq 0\} $$ and note that this is closed and contained in $I_\infty$.

Next, for $n,a>0$ with $n\in\mathbb{Z}$ we can let $K_{n,a}$ be the principal ideal generated by the function $\exp(-a/x^{2n})$. These are all different and contained in $I_\infty$. I am not sure whether they are closed.

For another kind of example, let $\mathcal{U}$ be a free ultrafilter on $\mathbb{R}$ and put $$ L = \{f : f^{-1}\{0\} \in \mathcal{U} \}. $$ This is a non-closed maximal ideal.


Now let $A$ be an arbitrary closed ideal in $C^\infty(\mathbb{R})$. Put $$ X_n = \{ x\in\mathbb{R} : f^{(i)}(x)=0 \text{ for all } i \leq n \text{ and } f\in A\}. $$ Specialising Reimundo's answer to the case $M=\mathbb{R}$, we see that $$ A = \{ f : f^{(i)}=0 \text{ on } X_n \text{ for all } i\leq n \}. $$ The sets $X_n$ are closed, with $X_n\supseteq X_{n+1}$. Moreover, if $x$ is a non-isolated point of $X_n$ (so it is in the closure of $X_n\setminus\{x\}$) then it is easy to see that $x\in X_{n+1}$. I would guess that the closed ideals biject with chains of subsets with these properties, but I have not tried to prove that.

share|cite|improve this answer
Am I missing something here? Take $A:=\{ f\in C^{\infty}(\mathbb{R}):f(0)=0\}$. Then, $X_0=\{ 0\}$ and $X_n=\emptyset$ for $n\geq 1$. Hence, while we do have $A=\{ f\in C^{\infty}(\mathbb{R}):f^{(i)}=0\text{ on }X_n\text{ for all }i\leq n\}$ if $n=0$, do we not have $\{ f\in C^{\infty}(\mathbb{R}):f^{(i)}=0\text{ on }X_n\text{ for all }i\leq n\} =C^{\infty}(\mathbb{R})\neq A$ for $n\geq 1$? – Jonathan Gleason Dec 1 '15 at 22:48

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.