MA40254 Differential and geometric analysis : Exercises 6
Hand in answers by 1:15pm on Wednesday 15 November for the Seminar of Thursday 16 November Homepage: http://moodle.bath.ac.uk/course/view.php?id=57709 ex06
  • 0 (Warmup). Let \(\gamma = \dfrac {-x_2\,dx_1 + x_1\,dx_2}{x_1^2 + x_2^2}\in \Omega ^1(\R ^2\setminus \{0\})\). Show that \(d\gamma =0\).

    [Solution: By the product rule

    \[ d\gamma = \frac {d(-x_2\,dx_1+x_1\,dx_2)}{x_1^2 + x_2^2} - \frac {2x_1\,dx_1 + 2x_2\, dx_2}{(x_1^2 +x_2^2)^2} \wedge (-x_2\,dx_1+x_1\,dx_2) = 0, \]

    since \(d(-x_2\,dx_1+x_1\,dx_2)=2dx_1\wedge dx_2\) and \((x_1\,dx_1 + x_2\, dx_2) \wedge (-x_2\,dx_1+x_1\,dx_2)=(x_1^2+x_2^2) dx_1\wedge dx_2\).]

  • 1. Let \(U = \{ p \in \R ^4 : x_2(p) \neq 0\}\), and \(\varphi = \left (x_1, \, x_2^2 x_3, \, x_4/x_2 \right ): U \to \R ^3\), where \(x_1,x_2,x_3,x_4\) are the coordinate functions on \(U\sub \R ^4\). Let \(\alpha = y_1 dy_2 {\wedge } dy_3 \in \Omega ^2(\R ^3)\) where \(y_1, y_2, y_3 : \R ^3 \to \R \) denote the coordinate functions.

    • (i) Express \(\varphi ^*\alpha \) and \(\varphi ^*d\alpha \) in standard form, i.e., as a sum of terms \(f dx_I\).

      [Hint: First expand \(\varphi ^*\alpha \) as \((\varphi ^*y_1) d(\varphi ^*y_2) \wedge d(\varphi ^*y_3) \) and similarly \(\varphi ^*d\alpha \).]

    • (ii) Compute directly \(d(\varphi ^*\alpha )\). What do you observe?

      [Hint: To save work, just compute the terms that don’t appear in \(\varphi ^*d\alpha \).]

  • 2. For \(U\) open in \(\R ^n\), \(\alpha \in \Omega ^k(U)\), \(p\in U\), and \(v_1,\ldots ,v_k\in \R ^n\), show that

    \[ d\alpha _p(v_0,\ldots ,v_k)=\sum _{i=0}^k (-1)^i D\alpha _p(v_i)(v_0,\ldots , \widehat v_i,\ldots , v_k) \]

    where \(v_0,\ldots ,\widehat v_i,\ldots , v_k\) denotes the list obtained from \(v_0,\ldots , v_k\) by omitting \(v_i\). Equivalently

    \[ d\alpha _p=\sum _{i=0}^k\sgn (\sigma _i) \, \sigma _i\cdot D\alpha _p^\vee , \]

    where \(\sigma _i = (0\;1\;\cdots \; i)^{-1}\in G:=\sym (\{0,1,\ldots k\})\cong S_{k+1}\), and for \(\sigma \in G\) and \(\beta \in \Alt ^{k+1}(\R ^n)\), \((\sigma \cdot \beta )(v_0,\ldots , v_k) =\beta (v_{\sigma (0)},\ldots , v_{\sigma (k)})\) (which is \(\beta (v_i,v_0,\ldots ,\widehat v_i,\ldots , v_k)\) when \(\sigma =\sigma _i\)).

    [Hint: One approach, using the second formula, is to let \(H\cong S_k\) be the subgroup of \(G\) fixing \(0\), and observe that \(\sigma _iH:i=0,\ldots k\) is a left coset partition of \(G\). Now split the sum into sums over each coset: what is \(\tau \cdot D\alpha _p^\vee \) for \(\tau \in H\) ?]

  • 3 (Less essential). Let \(U \subseteq \R ^3\) an open subset. Given a vector-valued function \(v = (v_1, v_2, v_3) : U \to \R ^3\), define \(v^\flat \in \Omega ^1(U)\) and \(v \inner \Det \in \Omega ^2(U)\) by applying the corresponding operations on vectors pointwise.

    • (i) Let \(\div (v) : U \to \R \) be defined by

      \[ \div (v) = \pd {v_1}{x_1} + \pd {v_2}{x_2} + \pd {v_3}{x_3} . \]

      Show that

      \[ d(v \inner \Det ) = \div (v) \Det \in \Omega ^3(U) . \]

    • (ii) Let \(\curl (v) : U \to \R ^3\) be defined by

      \[ \curl (v) = \left ( \pd {v_3}{x_2} {-} \pd {v_2}{x_3},\; \pd {v_1}{x_3} {-} \pd {v_3}{x_1},\; \pd {v_2}{x_1} {-} \pd {v_1}{x_2}\right ) . \]

      Show that

      \[ d(v^\flat ) = \curl (v) \inner \Det \in \Omega ^2(U) . \]

    [Hint: Write all the forms in standard form, e.g., \(v \inner \mathrm {Det} \) can be expressed as \(v_1 dx_2 {\wedge } dx_3 - v_2 dx_1 {\wedge } dx_3 + v_3 dx_1 {\wedge } dx_2\).]

  • 4. Let \(x_1,x_2\colon \R ^2\setminus \{0\}\to \R \) be the coordinate functions. Which of the following elements of \(\Omega ^1(\R ^2 \setminus \{0\})\) are closed? Which are exact?

    • (i) \(\alpha = -2x_1x_2\,dx_1 + x_1^2dx_2\)

    • (ii) \(\beta = x_2dx_1 + x_1dx_2\)

    • (iii) \(\gamma = \dfrac {-x_2dx_1 + x_1dx_2}{x_1^2 + x_2^2}\)

      [Hint: For (iii), consider \(\varphi : \mathbb {R} \to \R ^2\setminus \{0\}\) defined by \(t \mapsto (\mathrm {cos} \, t, \mathrm {sin} \, t) \). Suppose that \(\gamma = df \) for some \(f : \mathbb {R}^2 \setminus \{0\} \to \mathbb {R}\). What can you say about \(\varphi ^*f\) and \(\varphi ^*\gamma \)? Or about \(\int _0^{2\pi } \frac {d(f\circ \varphi )}{dt} dt\) ?]

MA40254 Differential and geometric analysis : Solutions 6

  • 1.

    • (i)

      \begin{align*} \varphi ^*(y_1dy_2\wedge dy_3) &= (\varphi ^*y_1) d((\varphi ^*y_2)\wedge d(\varphi ^*y_3))) = x_1 d\left (x_2^2x_3 \frac {x_2dx_4 - x_4 dx_2}{x_2^2}\right ) \\ &= x_1 (d(x_3x_2)\wedge dx_4 - d(x_3x_4) \wedge dx_2) \\ &= 2x_1x_3 dx_2 \wedge dx_4 + x_1x_4dx_2\wedge dx_3 + x_1x_2dx_3 \wedge dx_4 \end{align*} Similarly, \(\varphi ^*d\alpha =d(\varphi ^*y_1)\wedge d((\varphi ^*y_2)\wedge d(\varphi ^*y_3)))=2x_3 dx_1\wedge dx_2 \wedge dx_4 + x_4 dx_1\wedge dx_2\wedge dx_3 + x_2dx_1\wedge dx_3\wedge dx_4\)

    • (ii) Observe that \(2x_1dx_3\wedge dx_2 \wedge dx_4 + x_1dx_4\wedge dx_2\wedge dx_3 + x_1dx_2\wedge dx_3 \wedge dx_4=0\) so \(d(\varphi ^*\alpha )=\varphi ^*d\alpha \).

  • 2. Let \(H\cong S_k\) be the subgroup of \(G\) fixing \(0\); then \(\sigma _iH\) consists of elements of \(G\) which send \(0\) to \(i\), so these cosets are disjoint, and there are \(k+1\) of them, which is the index of \(H\) in \(G\). Hence

    \begin{align*} \alt (D\alpha _p^\vee )&=\sum _{\sigma \in G} \sgn (\sigma ) \sigma \cdot D\alpha _p^\vee =\sum _{i=0}^k \sum _{\tau \in H}\sgn (\sigma _i)\sgn (\tau ) \sigma _i\cdot \tau \cdot D\alpha _p^\vee = k! \sum _{i=0}^k\sgn (\sigma _i) \, \sigma _i\cdot D\alpha _p^\vee , \end{align*} since \(\tau \cdot D\alpha _p^\vee =\sgn (\tau ) D\alpha _p^\vee \) and \(|H|=k!\). Dividing by \(k!\) proves the result.

  • 3.

    • (i) \(v \inner \Det = v_1 dx_2 {\wedge }dx_3 + v_2 dx_3 {\wedge } dx_1 + v_3 dx_1 {\wedge }dx_2\), so

      \begin{align*} d(v \inner \Det ) &= dv_1 {\wedge } dx_2 {\wedge }dx_3 + dv_2 {\wedge } dx_3 {\wedge } dx_1 + dv_3 {\wedge } dx_1 {\wedge }dx_2\\ &= \left (\pd {v_1}{x_1} + \pd {v_2}{x_2} + \pd {v_3}{x_3}\right ) dx_1 {\wedge }dx_2 {\wedge } dx_3 . \end{align*}

    • (ii) \(v^\flat = v_1 dx_1 + v_2 dx_2 + v_3 dx_3\), so

      \begin{align*} d(v^\flat ) &= dv_1 {\wedge } dx_1 + dv_2 {\wedge } dx_2 + dv_3 {\wedge } dx_3 \\ &= \left (\pd {v_1}{x_2}dx_2 + \pd {v_1}{x_3}dx_3\right ) {\wedge } dx_1 + \left (\pd {v_2}{x_1}dx_1 + \pd {v_2}{x_3}dx_3\right ) {\wedge } dx_2 + \left (\pd {v_3}{x_1}dx_1 + \pd {v_3}{x_2}dx_2\right ) {\wedge } dx_3 \\ &= \left (\pd {v_3}{x_2} {-} \pd {v_2}{x_3}\right )dx_2{\wedge }dx_3 + \left (\pd {v_1}{x_3} {-} \pd {v_3}{x_1}\right )dx_3{\wedge }dx_1 + \left (\pd {v_2}{x_1} {-} \pd {v_1}{x_2}\right )dx_1{\wedge }dx_2 = \curl v \inner \Det . \end{align*}

  • 4.

    • (i) \(d(-2x_1x_2dx_1 + x_1^2 dx_2) = -2x_1dx_2 {\wedge } dx_1 + 2x_1dx_1{\wedge }dx_2 = 4x_1 dx_1{\wedge }dx_2\) is not zero everywhere, so not closed. Hence also not exact.

    • (ii) \(x_2dx_1 + x_1dx_2 = d(x_1x_2)\), so exact, hence also closed.

    • (iii) We’ve seen in the warmup that \(\gamma \) is closed. However, suppose that \(\gamma = df\) for some \(f \in \Omega ^0(U)\).

      Let

      \[ \varphi : \R \to \R ^2, \; \theta \mapsto (\cos \theta , \sin \theta ) . \]

      Then \(\varphi ^*\gamma = d\theta \). So \(\varphi ^*f = f \circ \varphi : \R \to \R \) is a function such that \(d(\varphi ^*f) = d\theta \). Then \(\varphi ^*f - \theta \) is a constant function on \(\R \). In particular, \((\varphi ^*f)(2\pi ) = (\varphi ^*f)(0) + 2\pi \). But that contradicts \((\varphi ^*f)(0) = f(\varphi (0)) = f(\varphi (2\pi )) = (\varphi ^*f)(2\pi )\).