I recently started studying pullbacks and differential forms. As far as I understand, one can define, for a smooth map $f:M\rightarrow N $, the pullback of $f$ to be the map $$ f^{*}:\Omega(N)\rightarrow\Omega(M) $$ such that, for all $\omega\in\Omega^{k}(N)$, $p\in M$ and $(v_{1},...,v_{k})\in T_{p}M\times...\times T_{p}M $, $$ f^{*}\omega(p)(v_{1},...,v_{k})=\omega(f(p))\left(f_{*}(v_{1}),...,f_{*}(v_{k})\right) $$ (Notice that $f^{*}\omega(p)\in\bigwedge^{k}\left(T_{p}^{*}M\right)$).

I am also familiar with the following properties of the pull-back:

$\bullet$ $f^{*}(\omega\wedge\alpha)=f^{*}(\omega)\wedge f^{*}(\alpha) $
$\bullet$ $(f\circ g)^{*}=g^{*}\circ f^{*} $
$\bullet$ $f$ is $\mathbb{R}$-linear.

Now, I have seen this done: (assume $\phi:M\supseteq U\rightarrow \phi(U)\subseteq\mathbb{R}^m$ and $\omega'\in\Omega^{k}(\phi(U))$) $$\phi^{*}(\omega')=\phi^{*}(\underset{I}{\sum}a_{I}\text{dx}_{I})=\underset{I}{\sum}\phi^{*}(a_{I})\cdot\phi^{*}(\text{dx}_{I})=\underset{I}{\sum}(a_{I}\circ\phi)\cdot\text{d}\phi_{I} $$

My problems with this:

1) I think I understand the first equality: $\{dx_{i_1}\wedge...\wedge dx_{i_k}\,:\,1\le i_1 <...< i_k \le m\}$ is a basis for $\Omega^k(\phi(U))$ (and $dx_I:=dx_{i_1}\wedge...\wedge dx_{i_k}$). The second one, however, I do not: why is $\phi^*(a_I dx_I)=\phi^*(a_I)\cdot \phi^*(dx_I)$? This does not seem to come from the properties of the pullback above. Also, how can $\phi(a_I)$ make sense? $a_I\in\mathbb{R}$, and only the pullback of differential forms is defined. Finally: what operation is the $\cdot$ ?
2) About the third equality: how is $\phi(a_I)=a_I\circ \phi$?


1 Answer 1


One can think of a smooth function $ g: N \rightarrow \mathbb{R} $ as an element of $ \Omega^0(N) $. In this case, the pullback of $ f^* g $ is well-defined, and tracing through the definition, $ f^*g(p) = g(f(p)) = (g \circ f)(p) $. Moreover, given $ g \in \Omega^0 $ and $ \omega \in \Omega^k $, $ f^* (g \omega) = f^*g \wedge f^*\omega $ by the first property, and since $ f^* g $ is a function, $ f^*g \wedge f^*\omega = (g \circ f ) \cdot f^* \omega $.

This is what they are doing in that example; think of $ dx_I $ as a $ k$-form, and $ a_I $ as a 0-form.

  • $\begingroup$ You said: "since $ f^* g $ is a function, $ f^*g \wedge f^*\omega = (g \circ f ) \cdot f^* \omega $." Can you please tell me why that follows? $\endgroup$
    – Soap
    Mar 16, 2017 at 16:48
  • 1
    $\begingroup$ Sure, but I'm not sure which part is confusing you. Is it clear why $ f^* g = g \circ f $? Otherwise, is it clear why for example $ 2 \wedge dx = 2 dx $? It is the same thing. $\endgroup$
    – asdjfksj
    Mar 16, 2017 at 17:40
  • $\begingroup$ Just looking at the definition and properties that I have for the pullback, it is not clear why $f^*g=g\circ f$. Is it defined to be like this in the particular case of $g$ being a zero-form? $\endgroup$
    – Soap
    Mar 16, 2017 at 17:54
  • 1
    $\begingroup$ The general definition is: $ f^*\omega(p)( v_1 \cdots v_k ) = \omega(f(p))(f_* v_1, \cdots f_* v_k) $. If $ \omega $ is a zero form, it takes no arguments, so that the definition reduces to $ f^* \omega(p)() = \omega(f(p)) () = \omega(f(p)) = (\omega \circ f) (p) $ $\endgroup$
    – asdjfksj
    Mar 16, 2017 at 17:56

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.