I am studying differential topology and I have some trouble understanding how coordinates are induced on the tangent space at any point.

Let $M$ be an $n$-dimensional smooth manifold, and let $p \in M$. Let $(U,\varphi)$ be a coordinate chart around $p$, with coordinates $x_1,\dots,x_n$. The tangent space $T_p M$ is defined to be $(\mathfrak{m}_p/\mathfrak{m}_p^2)^*$, where $\mathfrak{m}_p$ is the $\Bbb{R}$-algebra of germs at $p$ that vanish at $p$.

Now, we have the following lemma (whose proof I omit).

Lemma. If $x_1,\dots,x_n$ are coordinates around $p \in M$ and $f \in \mathfrak{m}_p$, then there exist germs $g_1,\dots,g_n$ at $p$ such that $f = \sum_i x_i g_i$.

Using this we can show that the images of the coordinates in $\mathfrak{m}_p/\mathfrak{m}_p^2$ form a basis for this vector space. Indeed, we have the following:

  1. $f \in \mathfrak{m}_p^2 \iff g_i \in \mathfrak{m}_p$ for all $i = 1,\dots,n$, since clearly $x_i \in \mathfrak{m}_p$ for all $i = 1,\dots,n$.
  2. If $f = \sum x_i g_i \in \mathfrak{m}_p$, then $f = \sum x_i (g_i - g_i(0)) + \sum g_i(0) x_i \implies \bar{f} = \sum g_i(0) \bar{x}_i$. Here, the bar indicates the image in $\mathfrak{m}_p/\mathfrak{m}_p^2$.

Let $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_n}$ be the dual basis of $\bar{x}_1,\dots,\bar{x}_n$. Then, we say that the coordinates $x_1,\dots,x_n$ around $p$ induce the coordinates $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_n}$ on $T_p M$.

If my understanding so far is correct, what this means is that I can talk about the coordinates $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_n}$ on $T_p M$ induced by $\varphi$ only when the coordinates around $p$ given by $\varphi$ are such that $\varphi(p) = 0$. Otherwise, in the above construction it is no longer true that the coordinates $x_1,\dots,x_n$ define germs at $p$ that vanish at $p$.

So, in particular this means that although $(U,\varphi)$ defines coordinates around every point $p' \in U$, $\varphi$ does not induce coordinates on $T_{p'} M$ if $\varphi(p') \neq 0$.

Is this true? I would appreciate any helpful comments and answers in this regard.

Further motivation.

This line of thinking emerged while I was trying to solve the following problem.

Exercise. Let $f : M \to N$ be a smooth map between smooth manifolds. Show that the set of critical points of $f$ form a closed subset of $M$.

The natural method in my opinion is to show that the complement, that is the set of regular points, is open in $M$. So if $p \in M$ is a regular point, then $df_p : T_p M \to T_{f(p)} N$ is of maximal rank. If suitable coordinate charts $(U,\varphi)$ and $(V,\psi)$ are chosen around $p$ and $f(p)$, respectively, defining coordinates $x_1,\dots,x_m$ and $y_1,\dots,y_n$, respectively, then $df_p$ is represented by the Jacobian of $\tilde{f} \equiv \psi \circ f \circ \varphi^{-1}$ at the point $\varphi(p)$, with respect to the bases $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_m}$ on $T_p M$ and $\frac{\partial}{\partial y_1},\dots,\frac{\partial}{\partial y_n}$ on $T_{f(p)} N$. Now, I know that for every $\varphi(p')$ in a small enough neighbourhood around $\varphi(p)$, $\mathrm{Jac}(\tilde{f})_{\varphi(p')}$ is also of maximal rank.

At this point, I would like to be able to say that $\mathrm{Jac}(\tilde{f})_{\varphi(p')}$ represents the map $df_{p'} : T_{p'} M \to T_{f(p')} N$ with respect to the bases $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_m}$ on $T_{p'} M$ and $\frac{\partial}{\partial y_1},\dots,\frac{\partial}{\partial y_n}$ on $T_{f(p')} N$.

If I could, then I would be done. But, I went back to the definition of $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_m}$ as I've understood it above, and I'm not so sure that I am correct in saying so.


There are several equivalent definitions of the tangent space. Consider $C_p(M)$ the $\mathbb{R}$-algebra of germs at $p$ and define the tangent space of $M$ at $p$ as the dual space of the quotient $C_p(M)/(\mathfrak{m}_p^2+\mathbb{R})$, where the copy of $\mathbb{R}$ in the denominator is the vector subspace of constant functions. With this definition, the value of $\varphi(p)$ does not matter.


Let $G_p$ be the set of germs at $p$. Let $(U,\varphi)$ and $(V,\psi)$ be coordinate charts around $p$, with coordinates $x_1,\dots,x_n$ on $U$ and $y_1,\dots,y_n$ on $V$. If $f : M \to \Bbb{R}$ is a smooth function such that $$ \frac{\partial f}{\partial x_i}\bigg|_{p} = 0 \quad \text{ for all } \quad i = 1,\dots,n,\tag{$*$} $$ then $$ \frac{\partial f}{\partial y_i}\bigg|_{p} = 0 \quad \text{ for all } \quad i = 1,\dots,n. $$ Also, if $f_1,f_2 : M \to \Bbb{R}$ are smooth functions that agree on a small neighbourhood around $p$, then we have $$ \frac{\partial f_1}{\partial x_i}\bigg|_p = \frac{\partial f_2}{\partial x_i}\bigg|_p \quad \text{ for all } \quad i = 1,\dots,n. $$ Let $S_p$ be the subspace of stationary germs at $p$, that is, those germs at $p$ for which ($*$) hold. By the above remarks, $S_p$ is a well-defined subspace of $G_p$.

Now, suppose $f \in G_p$ and $x_1,\dots,x_n$ are coordinates around $p$ with $(\alpha_1,\dots,\alpha_n)$ the coordinates of $p$. Then, by the above Lemma, we have that $$ f(x) = f(p) + \sum_{i=1}^n (x_i - \alpha_i) g_i(x) = f(p) - \sum_{i=1}^n \alpha_i g_i(p) + \sum_{i=1}^n (x_i - \alpha_i) (g_i(x) - g_i(p)) + \sum_{i=1}^n g_i(p) x_i. $$ The first two terms of the right-most expression are constant germs and hence stationary germs. The third term is also a stationary germ, as can be seen by differentiating and applying the chain rule. Hence, if $\bar{f}$ denotes the image of $f$ in $G_p / S_p$, we have that $$ \bar{f} = \sum_{i=1}^n g_i(p) \bar{x}_i. $$

So, we can still make sense of $\bar{x}_1,\dots,\bar{x}_n$ and its dual basis $\frac{\partial}{\partial x_1},\dots,\frac{\partial}{\partial x_n}$ when $p$ does not have coordinates $(0,\dots,0)$. Note that $G_p/S_p \cong \mathfrak{m}_p/(\mathfrak{m}_p \cap S_p)$ by the second isomorphism theorem, because $G_p = \mathfrak{m}_p + S_p$. Moreover, $\mathfrak{m}_p \cap S_p = \mathfrak{m}_p^2$. So, there is no trouble in defining the tangent space $T_p M$ to be $(G_p / S_p)^*$.

I obtained some supplementary course notes that briefly mention $T_p^* M := G_p/S_p = \mathfrak{m}_p / \mathfrak{m}_p \cap S_p$, and $T_p M := (T_p^* M)^*$. I have only elaborated this in greater detail in my answer.

Also, the expression for $f(x)$ that I have obtained here shows that $S_p = \mathfrak{m}_p^2 + \Bbb{R}$, where $\Bbb{R}$ denotes the subspace of constant germs. So, this answer can also be considered an elaboration of the answer by @DanteGrevino.

  • $\begingroup$ It is more or less the proof that I have in mind. Good job filling the details! Let me know if something remains unclear. $\endgroup$ – Dante Grevino Dec 11 '18 at 16:02
  • $\begingroup$ Thank you @DanteGrevino :) I think I've finally got the hang of the definitions now. $\endgroup$ – Brahadeesh Dec 11 '18 at 16:03
  • 1
    $\begingroup$ I think there is a typo. $m_p \cap S_p = {m_p}^2 $ You have written $m_p + S_p = {m_p}^2$ $\endgroup$ – Error 404 Dec 12 '18 at 16:21
  • $\begingroup$ @Error404 Thanks, corrected :) $\endgroup$ – Brahadeesh Dec 12 '18 at 16:25
  • 1
    $\begingroup$ @Error404 In most contexts, yes, but in the problem I described in the "Further motivation" section I want to be able to say that $\frac{\partial}{\partial x_1}\big|_{q},\dots,\frac{\partial}{\partial x_n}\big|_{q}$ is a basis of $T_q M$ for all points $q \in U$, where $(U,\varphi)$ is a chart around $p$, and $\varphi$ describes coordinates $x_1,\dots,x_n$ in $U$ with $p \equiv 0$. I cannot do this without the general definition of $T_p M$ as $(G_p / S_p)^*$. $\endgroup$ – Brahadeesh Dec 13 '18 at 10:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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