Given $f\in C^\infty(U)$, $U$ open set of $\mathbb{R}^n$, we define the differential of $f$ at $p$ $$ (df)_p:T_p\mathbb{R}^n\to\mathbb{R}\\ (df)_p(v):=v(f) $$ and the differential of $f$ $$ df:U\to T^*U\\ p\mapsto (p,(df)_p) $$ where $T^*U$ is the cotagent bundle.

I can understand that $(df)_p$ is the map which associates for every point $p$ any derivative $v$ to $f$. While $df$ is just obtained by gluing together all these local differentials.

Now I'm asked to compute the differential of the i-th-coordinate map $x_i:\mathbb{R}^n\to\mathbb{R}$.

Here is my (unsuccesful) reasonement. I have to start by calculating, for every $p$, $(dx_i)_p$. By definition $(dx_i)_p\in (T_p\mathbb{R^n})^*$ and the latter set is spanned by $\{(dx_i)_p:1\le i\le n\}$. Now, I am a bit confused. My professor defined the basis of $(T_p\mathbb{R}^n)^*$ as the dual basis of the tangent plane and the symbols $(dx_i)_p$ are just formal symbols, nothing to do with the differential. (Right?) I would proceed by applying the differential of $x_i$ at $p$ to a general element of the tangent space $v=\sum_j v_j(\frac{\partial}{\partial x_j})_p$ $$ (dx_i)_p(v):=v_i $$ Then how shall I go on?

  • $\begingroup$ By doing some calculations I obtained $(dx_i)_p=\sum_k\delta_{ik}(dx_k)_p$. Is it correct? Then, how can I conclude that $dx_i=\sum_k\delta_{ik}(dx_k)_p$? $\endgroup$
    – avati91
    Jul 16 '15 at 15:20
  • $\begingroup$ I thought about it. I think I made confusion because my professor had to define first what the differential of a point is and then prove that the differential of the coordinates maps at a point form a basis of the cotangent space, while he first defined the dual basis of the tangent space (as formal symbols) and then he introduced the differentials. $\endgroup$
    – avati91
    Jul 16 '15 at 15:44

To remove notational confusion, let $\lambda^1,\ldots,\lambda^n$ be the dual basis of $\frac{\partial}{\partial x^1},\ldots,\frac{\partial}{\partial x^n}$. Then, your goal is to show that $$d(x^i)=\lambda^i.$$ i.e. $d(x^i)=dx^i$, where the later is the formal symbol usually used for $\lambda^i$. Note that this justifies the notation $dx^i$.

Now, compute: $$d(x^i)\left(\frac{\partial}{\partial x^j}\right)=\frac{\partial}{\partial x^j}x^i=\delta^i_j=\lambda^i\left(\frac{\partial}{\partial x^j}\right).$$ Hence, $d(x^i)$ agree with $\lambda^i$ on a basis, so they must be equal.


By whatever arguments, you yourself arrived at the correct answer: $dx_i$ computes at each point $p\in{\mathbb R}^n$ the $i$th coordinate of any tangent vector $v$ attached at $p$.

A simple way to see this is as follows: $x_i(\cdot)$ can be viewed as a real-valued function on ${\mathbb R}^n$. In order to compute the differential $dx_i(p)$ we have to look at increments $$\Delta x_i(v):=x_i(p+v)-x_i(p)$$ and to determine an approximation for $\Delta x_i(v)$ which is linear in $v$ when $v\to0$. Now in this special case an approximation is not needed at all, since $$\Delta x_i(v)=v_i$$ is already linear in $v$. This implies that the differential $dx_i(p)$ is given by $$dx_i(p).v= v_i\ ,$$ independently of the point $p$ under consideration.


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.