II. Differential Forms
1. Definitions
Convention. We say that a function $f$ is differentiable, or of class $\mathscr C^p$ on a compact set $K$ if there exists an open set $E$ containing $K$ and a map $g$ on $E$ such that on which $g$ is $\mathscr C^p$ on $E$ and $f(\mathbf x)=g(\mathbf x)$ for all $\mathbf x\in K$.
Definition ($k$-surface). Suppose $E\subset \mathbb R^n$ is open. A $k$-surface in $E$ is a $\mathscr C^1$ map $\Phi$ from a compact set $D\subset\mathbb R^k \to E$; here, we restrict ourselves to the case where $D$ is a $k$-cell or a $k$-simplex.
Definition (Differential form). Suppose that $E$ is an open subset of $\mathbb R^n$. A differential form of order $k\ge 1$ is a map $\omega$, symbolically represented by the formal sum
$$\omega = \sum_{1\le i_1, i_2,\dots,i_k \le n} a_{i_1i_2\dots i_k}(\mathbf x)~ dx_{i_1} \wedge dx_{i_2} \wedge \dots \wedge dx_{i_k},$$
where the all the real-valued functions $a_{i_1i_2\dots i_k}$ are continuous in $E$,
which assigns to each $k$-surface in $E$ $\Phi$ with domain $D$ a value $\omega(\Phi)$ given by
$$\omega(\Phi) = \int_\Phi \omega = \sum_{1\le i_1, i_2, \dots, i_k \le n} \int_D a_{i_1i_2\dots i_k}(\Phi(\mathbf u)) \frac{\partial (\Phi_{i_1}, \Phi_{i_2}, \dots, \Phi_{i_k})}{\partial (u_1,u_2,\dots, u_k)} d\mathbf u.$$
A $0$-form in $E$ is defined to be a continuous function on $E$.
Remark. At that point, the representation of $\omega$ is only symbolic and merely provides the information to compute the integral of $\omega$ on all the $k$-surfaces in $E$; one could as well give an array of $n^k$ entries containing the continuous functions $a_{i_1i_2\dots i_k}:E\to \mathbb R$ for $1\le i_1,i_2,\dots, i_k \le n$.
Definition (Basic $k$-form). Let $E\subseteq \mathbb R^n$ be open. The forms $dx_{i_1} \wedge dx_{i_2} \wedge \dots \wedge dx_{i_k}$ with $1\le i_1\le i_2 \le \dots \le i_k\le n$ are called basic $k$-forms.
In such a case, we call $I=\{i_1,i_2,\dots,i_k\}$ an increasing $k$-index and write $dx_I = dx_{i_1}\wedge \dots \wedge dx_{i_k}$.
Theorem. Let $E\subseteq \mathbb R^n$ be open. Let $\omega$ be a $k$-form in $E$. Then, there exists a unique family of functions $(b_I)_I$ continuous in $E$ such that
$$\omega = \sum_I b_I(\mathbf x) dx_I,$$
where the sum ranges over increasing $k$-indices in $\{1,2,\dots, n\}.$
This represention is called the standard presentation of $\omega$.
Remark. Seeing the $k$-forms as alternating $k$-tensors (as in HW 8), the previous theorem relates to the fact that the family $(dx_I)$ when $I$ ranges in the $k$-indices in $\{1,\dots, n\}$ is a basis of the vector space of alternating $k$-tensors.
Definition (Wedge product). Let $E\subseteq \mathbb R^n$ be open.
- If $I$ and $J$ are respectively $p$ and $q$ indices in $\mathbb R^n$, with
$$I=\{i_1, i_2,\dots, i_p\} \qquad \text{and} \qquad J=\{j_1,j_2,\dots, j_q\},$$
we define the $(p+q)$-form $dx_I \wedge dx_J$ by
$$dx_I \wedge dx_J = dx_{i_1} \wedge \dots \wedge dx_{i_p} \wedge dx_{j_1} \wedge \dots \wedge dx_{j_q}.$$
- If $\omega$ and $\lambda$ are $p$ and $q$-forms in $E$ with standard presentations
$$\omega = \sum_I b_I(\mathbf x) dx_I \qquad \text{and} \qquad \lambda = \sum_J c_J(\mathbf x) dx_J,$$
then we define the $(p+q)$-form $\omega \wedge \lambda$ by
$$\omega \wedge \lambda = \sum_{I,J} b_I(\mathbf x) c_J(\mathbf x) dx_I \wedge dx_J.$$
2. Computation rules
Proposition.
Let $\omega, \lambda, \eta$ be $k$, $\ell$ and $m$-forms in an open $E\subseteq \mathbb R^n$, respectively. Then
- $\omega \wedge (\lambda \wedge \eta) = (\omega \wedge \lambda ) \wedge \eta = \omega \wedge \lambda \wedge \eta$
- $\omega \wedge \lambda = (-1)^{k \ell } \lambda \wedge \omega$
- if $\ell=m$ then $\omega \wedge (\lambda + \eta) = \omega \wedge \lambda + \omega \wedge \eta$.
Definition. (Differentiation) Let $E\subset \mathbb R^n$ be open. Given $\omega$ a $\mathscr C^1$ differential form of order $k$ in $E$, we associate a $(k+1)$-form in $E$, denoted by $d\omega$ that is defined by
- if $k=0$, then $\omega = f\in \mathscr C^1(E)$, and
$$df = \sum_{i=1}^n D_if(\mathbf x) dx_i,$$
- if $k\ge 1$, then $\omega = \sum_{I} b_I(\mathbf x) dx_I$, for $b_I\in \mathscr C^1(E)$, for all $I$ increasing $k$-index in $\{1,2,\dots, n\}$ and
$$d\omega = \sum_{I} db_I \wedge dx_I.$$
Theorem (Product rule for differential forms). Let $E\subset \mathbb R^n$ be open.
- If $\omega$ and $\lambda$ are $\mathscr C^1$ differential forms in $E$, of order $k$ and $m$, respectively, we have
$$d(\omega \wedge \lambda) = d\omega \wedge \lambda + (-1)^k \omega \wedge d\lambda.$$
- If $\omega$ is of class $\mathscr C^2$, then $d^2 \omega := d(d\omega) = 0$.
3. Change of variables
Definition (Pull-back) Let $E\subseteq \mathbb R^n$ be open, $T\in \mathscr C^1(E,V)$ where $V\subset \mathbb R^m$ is open. Suppose that for $\mathbf x\in E$, $\mathbf y=T(\mathbf x) = \sum_{i=1}^m t_i(\mathbf x) \mathbf e_i \in V$. Let $\omega$ be a $k$-form in $V$ whose standard presentation is
$$\omega = \sum_I b_I(\mathbf y) dy_I.$$
We define the pull-back of $\omega$ by $T$ as the $k$-form in $E$ given by
$$\omega_T = \sum_I b_I(T(\mathbf x)) dt_{I}:=\sum_{I} b_I(T((\mathbf x))) dt_{i_1} \wedge \dots \wedge dt_{i_k},$$
where the sum ranges over increasing $k$-indices $I=\{i_1,i_2,\dots, i_k\}$ and we wrote $dt_I=dt_{i_1}\wedge dt_{i_2} \wedge \dots \wedge dt_{i_k}$.
Remark. In particular, when $\omega$ is a $0$-form given by a function $f$ continuous on $E$, we have $f_T(\mathbf x) = f(T(\mathbf x))$ for every $x\in E$.
Theorem. Let $E\subseteq \mathbb R^n$ be open, $T\in \mathscr C^1(E,V)$ where $V\subset \mathbb R^m$ is open. Suppose that $\omega$ and $\lambda$ are $k$ and $m$-forms in $V$, respectively. Then
- if $k=m$ then $(\omega + \lambda)_T = \omega_T + \lambda_T$;
- $(\omega\wedge \lambda)_T = \omega_T \wedge \lambda_T$;
- if $\omega$ is $\mathscr C^1$ and $T$ is $\mathscr C^2$, then $(d\omega)_T = d(\omega_T)$.
Theorem (Change of variable formula for differential forms). Suppose that $T$ is a $\mathscr C^1$ mapping from an open set $E\subseteq \mathbb R^n$ in an open set $V\subseteq \mathbb R^m$, and $\omega$ is a $k$-form in $V$. Then, if $\Phi$ is a $k$-surface in $E$, we have
$$\int_{T\Phi} \omega = \int_\Phi \omega_T.$$