MA40254 Differential and geometric analysis : Exercises 8
Hand in answers by 1:15pm on Wednesday 29 November for the Seminar of Thursday 30 November Homepage: http://moodle.bath.ac.uk/course/view.php?id=57709 ex08
  • 0 (Warmup). Let \(M\) be an \(n\)-dimensional submanifold of \(\R ^s\). Prove (from the definition of \(d\) on submanifolds) that if \(\alpha \in \Omega ^k(M)\) is exact, i.e., \(\alpha =d\beta \), then \(\alpha \) is closed, i.e., \(d\alpha =0\).

    Solution
    Suppose \(\alpha =d\beta \). We have seen in lectures that \(d^2=0\) so \(d\alpha =d^2\beta =0\), but let us recall how to prove this. Note that it suffices to prove \(d\alpha =0\) locally on \(M\), i.e., on sufficiently small open neighbourhoods \(U\cap M\) of each \(p\in M\). There are two ways to proceed.

    First by definition, we may assume that on \(U\cap M\), \(\beta =\iota ^*\gamma \), where \(\gamma \in \Omega ^k(U)\) and \(\iota \colon U\cap M \to U\) is the inclusion. Thus \(d\alpha =d^2(\iota ^*\gamma )=d(\iota ^*d\gamma ) =\iota ^*(d^2\gamma )=0\) on \(U\cap M\). Alternatively, we may assume that there is a parametrisation \(\varphi \colon \wt U\to U\cap M\). Then \(\varphi ^*d\alpha =\varphi ^*d^2\beta = d\varphi ^*d\beta =d^2\varphi ^*\beta =0\), so \(d\alpha =0\) on \(U\cap M\) because \(\varphi ^*\colon \Omega ^k(U\cap M)\to \Omega ^k(\wt U)\) is a bijection with inverse \((\varphi ^{-1})^*\).

  • 1. Let \(M \subseteq \R ^s\) be an \(n\)-dimensional submanifold and \(\alpha \in \Omega ^k(M)\) for \(k > 0\).

    • (i) Suppose that \(M\) is diffeomorphic to \(\R ^n\) and \(\alpha \) is closed; prove that \(\alpha \) is exact.

      Hint
      Let \(\varphi : \mathbb {R}^n \to M \) be a diffeomorphism, and consider the pullback \(\varphi ^*\alpha \in \Omega ^k(\mathbb {R}^n) \).

    • (ii) Suppose \(k=n\); show that \(\alpha \) is closed.

      Hint
      What is the dimension of \(\mathrm {Alt}^{n+1}(T_xM)\)?

    • (iii) If \(k=n\), does \(\alpha \) have to be exact?

      Hint
      Look for examples on \(M = S^1\). A previous exercise may help.

  • 2. Give an example of an \(n\)-dimensional submanifold \(M \subseteq \R ^s\) and an \(\alpha \in \Omega ^k(M)\) such that \(\alpha \) is not the pullback by the inclusion of any \(\beta \in \Omega ^k(\R ^s)\).

    Hint
    The simplest examples have \(n = s = 1\) and \(k = 0\).

  • 3. Let \(S^2 = \{z \in \R ^3 : \norm {z} = 1\}\), and \(U = S^2 \setminus \{ (0,0,1) \}\), and consider the parametrisation \(\varphi \colon \R ^2\to U\) defined by

    \[ (x_1, x_2) \mapsto \frac {1}{1+\norm {x}^2} \left (2x_1, \, 2x_2, \norm {x}^2-1 \right ).\]

    Let \(\alpha = i^* dz_3\) where \(i\colon S^2\to \R ^3\) is the inclusion and \(z_1,z_2,z_3\) denote the coordinate functions on \(\R ^3\). Compute \(\varphi ^*\alpha \).

    Hint
    \(\varphi ^*i^*dz_3=(i\circ \varphi )^*dz_3=\cdots \).

MA40254 Differential and geometric analysis : Solutions 8

  • 1.

    • (i) Suppose \(\alpha \) is closed. Let \(\varphi : \R ^n \to M\) be a diffeomorphism. Then \(\varphi ^*\alpha \in \Omega ^k(\R ^n)\) is closed, so exact by the Poincaré lemma, i.e, there is some \(\gamma \in \Omega ^{k-1}(\R ^n)\) such that \(d\gamma = \varphi ^*\alpha \). If we let \(\beta = (\varphi ^{-1})^*\gamma \in \Omega ^{k-1}(M)\) then \(d\beta = (\varphi ^{-1})^*d\gamma = (\varphi ^{-1})^*(\varphi ^*\alpha ) = (\varphi \circ \varphi ^{-1})^*\alpha = \alpha \).

    • (ii) For each \(p \in M\), \(d\alpha _p\) is an element of \(\mathrm {Alt}^{n+1}(T_pM)\). Since \(n+1 > \dim T_p M\), we have \(\mathrm {Alt}^{n+1}(T_p M) = \{0\}\), so \(d\alpha \) is zero everywhere.

    • (iii) No. Consider \(M = S^1 \subseteq \R ^2\), and let \(\alpha \in \Omega ^1(S^1)\) be the pullback by \(i\colon S^1\to \R ^2 \setminus \{0\}\) of the non-exact 1-form

      \[ \gamma = \dfrac {-x_2\,dx_1 + x_1\,dx_2}{x_1^{\,2} + x_2^{\,2}} \in \Omega ^1(\R ^2 \setminus \{0\}) \]

      from Exercises 6 Q 4(iii). The same argument as there shows that \(\alpha \) is not exact either.

  • 2. We can for instance take \(M = (0,\infty ) \subset \R \), and \(k = 0\). Then an element of \(\Omega ^k(M)\) is just a smooth function \((0,\infty ) \to \R \) and pullback reduces to restriction. If we take it to be \(x \mapsto \frac {1}{x}\) then that is not the restriction of any smooth function \(\R \to \R \).

  • 3. We have \(\varphi ^*i^*dz_3=(i\circ \varphi )^*dz_3=d ((i\circ \varphi )^*z_3)\) and \((i\circ \varphi )^*z_3= (x_1^2+x_2^2-1)/(1+x_1^2+x_2^2)\). The exterior derivative of this is

    \[ \frac {4x_1dx_1 + 4x_2dx_2}{(1+x_1^2+x_2^2)^2}. \]