1. $\phi^* \left( \sum_{i = 1}^n f_i dx_i \right) = \sum_{i = 1}^n f_i \circ \phi \frac{\partial \phi^i}{\partial t} dt$

  2. Let $\omega = \sum_{i = 1}^n f_i x_1 \cdots \hat{x_i} \cdots x_n$ be a $n-1$ for m on $\mathbb{R}^n$ (an element of $\Omega^{n-1}( \mathbb{R}^n)$. Let $\phi : \mathbb{R}^{n-1} \rightarrow \mathbb{R}^{n}$ be a smooth map. Then $$ \phi^* (\omega) = \text{det} \left( \begin{bmatrix} f_1 \circ \phi & \cdots & f_n \circ \phi \\ \frac{\partial \phi^1}{\partial x_1} & \cdots & \frac{\partial \phi^1}{\partial x_{n-1} } \\ \vdots & & \vdots\\ \frac{\partial \phi^n}{\partial x_1} & \cdots & \frac{\partial \phi^n}{\partial x_{n-1} } \\ \end{bmatrix} \right) $$

  3. Let $\phi : \mathbb{R}^n \rightarrow \mathbb{R}^n$ be a map. The pullback of $f dx_1 \cdots dx_n \in \Omega^n ( \mathbb{R}^n)$ by $\phi$ is $f \circ \phi \text{det}(D(\phi))$. $$ \phi^* (\omega) = f \circ \phi \ \text{det} \left( \begin{bmatrix} \frac{\partial \phi^1}{\partial x_1} & \cdots & \frac{\partial \phi^1}{\partial x_{n} } \\ \vdots & & \vdots\\ \frac{\partial \phi^n}{\partial x_1} & \cdots & \frac{\partial \phi^n}{\partial x_{n} } \\ \end{bmatrix} \right) $$

Can we generalize these formulas to find a formula for the pullback of a $k$ form in $\Omega^k (\mathbb{R}^n)$ to $\mathbb{R}^k$? I want a formula involving only multilinear operations that take $\frac{\partial \phi^i}{\partial x_j}$ and $f_i \circ \phi$ as inputs.


If I understood your question correctly, we can discuss first how to pullback a single wedge of $dx$s, then generalize by linearity. Let $\omega = dx_I = dx_{i_1}\wedge \dots \wedge dx_{i_k}$ and $\phi(u_1,\dots,u_m):\mathbb{R}^m\rightarrow\mathbb{R}^n$ the smooth map, then

$$\phi^*\omega =\bigwedge_{1\leq j\leq k} \sum_{q_j=1}^m \frac{\partial\phi_{i_j}}{\partial u_{q_j}}du_{q_j}=\sum_{1\leq q_1,\dots,q_k\leq m} \prod_{j=1}^k \frac{\partial\phi_{i_j}}{\partial u_{q_j}}du_{q_1}\wedge \dots du_{q_k}$$

Now for every $Q=(q_1,\dots,q_k)$ if two $q$s are the same we get $0$. Otherwize, there exists a unique increasing multi-index $J$ and a permutation $\sigma$ such that $\sigma(J)=Q$. If we sum according to $J$s and $\sigma$s, we get

$$\sum_{J}\sum_{\sigma \in S_k} \prod_{j=1}^k \frac{\partial\phi_{i_j}}{\partial u_{q_{\sigma(j)}}}du_{q_{\sigma(1)}}\wedge \dots du_{q_{\sigma(k)}} =\sum_{J}\sum_{\sigma \in S_k} \prod_{j=1}^k \frac{\partial\phi_{i_j}}{\partial u_{q_{\sigma(j)}}}\text{sgn}\sigma \ du_J$$ Note that by the permutation property of determinant we get the determinant created by the $i_1,\dots,i_k$ rows and $j_1,\dots,j_k$ columns of $D\phi$. We can write this as

$$\sum_J \det \frac{\partial \phi_I}{\partial u_J} du_J$$ Remember $J$ runs over all increasing multi-indexes of length $k$ from $1,\dots,m$.

  • $\begingroup$ You understood perfectly. This is exactly what I was looking for. $\endgroup$ Aug 25 '19 at 14:10

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.