Suppose $f:M\to N$ is a map between Riemannian manifolds. In the literature on Harmonic maps, we often see the Dirichlet energy $$E[f]=\int |df|^2 dV.$$ Everywhere I've been reading, in papers and textbooks alike, the text has jumped straight to the equation for $|df|^2$ in local coordinates $$g^{\alpha\beta}\frac{\partial f^i}{\partial x^\beta}\frac{\partial f^i}{\partial x^\alpha}$$ where $g_{\alpha\beta}$ is the metric tensor of $M$.
I know a bit about Riemannian geometry but haven't done any explicit computations with local coordinates before, so I'm struggling to get a derivation for this on my own. To see a derivation would be very instructive to me in how to use local coordinates in future. The sources I've been looking at are
- Variational Methods (Struwe) Chapter 6
- A Regularity Theory for Harmonic Maps (Schoen, Uhlenbeck)
- Harmonic Maps to Spheres (Solomon)
Summary: Could someone please give a detailed explanation of how to express $|df|^2$ in local coordinates?
Let $f \colon (M,\gamma) \to (N,g)$ be a map between Riemannian manifolds. Choose local coordinates $(x^\alpha)$ on $N$ and $(y^i)$ on $N$. We shall regard $df$ as a section of the vector bundle $T^*M \otimes f^*(TN) \to M$, so that in our chosen coordinates: $$df = \frac{\partial f^i}{\partial x^\alpha}\,dx^\alpha \otimes \frac{\partial}{\partial y^i}$$ There are several metrics in play here:
Since $\gamma$ is a metric on $M$, each $\gamma|_x$ is an inner product on $T_xM$, which induces an inner product on $T_x^*M$. This latter inner product has components $$(\gamma^{\alpha \beta}).$$ Note that $(\gamma^{\alpha \beta})$ is the inverse matrix of $(\gamma_{\alpha \beta})$.
Similarly, since $g$ is a metric on $N$, each $g|_{f(x)}$ is an inner product on $T_{f(x)}N$, which induces (via the map $f \colon M \to N$) an inner product on $(f^*TN)_x$. This latter inner product has components $$(g_{ij} \circ f).$$
Using these two metrics, we get a metric --- which I'll denote $\langle \cdot, \cdot \rangle$ --- on $T^*M \otimes f^*(TN)$.
Putting this all together, we can now compute: \begin{align*} |df|^2 = \left\langle df, df \right\rangle & = \left\langle \frac{\partial f^i}{\partial x^\alpha}\,dx^\alpha \otimes \frac{\partial}{\partial y^i}, \, \frac{\partial f^j}{\partial x^\beta}\,dx^\beta \otimes \frac{\partial}{\partial y^j} \right\rangle \\ & = \frac{\partial f^i}{\partial x^\alpha}\, \frac{\partial f^j}{\partial x^\beta}\,\left\langle dx^\alpha \otimes \frac{\partial}{\partial y^i}, \,dx^\beta \otimes \frac{\partial}{\partial y^j} \right\rangle \\ & = \frac{\partial f^i}{\partial x^\alpha}\, \frac{\partial f^j}{\partial x^\beta}\,\gamma^{\alpha \beta}\, (g_{ij} \circ f). \end{align*} If the metric $g$ on $N$ is flat, then we can choose our coordinates $(y^i)$ so that $g_{ij} = \delta_{ij}$. In that case, our formula reduces to the one you wrote: $$|df|^2 = \frac{\partial f^i}{\partial x^\alpha}\, \frac{\partial f^i}{\partial x^\beta}\,\gamma^{\alpha \beta}$$