We have been using matrices \([g]\) to represent the first fundamental form. The advantage of such a representation is that it is easier to understand. The drawback is that such a matrix would depend on the choice of local parametrizations \(F(u_1,\ldots,u_n)\). In differential geometry and physics, it is often the case that we use tensor notations to denote the first fundamental form of a regular (hyper)-surface, and tensors are independent of local parametrizations.
To begin, we take a local parametrization \(F(u_1,\ldots,u_n): \mathcal{U}\to F(\mathcal{U})\) of a regular hypersurface \(\Sigma^n\), and at each \(p\in F(\mathcal{U})\), the tangent space has a natural basis \(\left\{\pdF{u_i}\right\}_{i=1}^n\). We next define \((du^i)_p: \Tp{p}M \to \mathbb{R}\), which is a linear map taking a tangent vector to a scalar, as: \[ (du^i)_p\!\left(\pd{\mathbf{F}}{u_j}(p)\right) = \begin{cases} 1 & \text{if } i=j \\ 0 & \text{if } i\neq j \end{cases}, \] or simply, \(du^i\!\left(\pdF{u_j}\right) = \delta_{ij}\). Note that we often omit the point \(p\) if it is not important in the context. Take the surface case \(n=2\) as an example, in which we usually identify \((u_1, u_2) = (u,v)\). Then we have \[ du\!\left(\pdF{u}\right) = 1, \quad du\!\left(\pdF{v}\right) = 0, \quad dv\!\left(\pdF{u}\right) = 0, \quad dv\!\left(\pdF{v}\right) = 1. \] Given a tangent vector \(X = a\pdF{u} + b\pdF{v}\), we can then calculate by linearity that: \[ du(X) = du\!\left(a\pd{\mathbf{F}}{u}+b\pd{\mathbf{F}}{v}\right) = a\,du\!\left(\pdF{u}\right)+b\,du\!\left(\pdF{v}\right) = a. \] Similarly, \(dv(X)=b\).
\(\Span\{(du^i)_p\}_{i=1}^n\) is in fact the dual space of \(\Tp{p}M\). We often denote such a span by \(T^*_p M\) and is called the cotangent space at \(p\). The “vectors” \((du^i)_p\)’s are often called cotangent vectors.
Next we can then introduce tensor product \(\otimes\). In this course, it is a product between two cotangent vectors, and the outcome is a bilinear map taking two tangent vectors to a single scalar. It is defined precisely as follows. Given any tangent vectors \(X, Y \in \Tp{p}M\), and cotangent vectors \((du^i)_p, (du^j)_p \in T^*_p M\), we define \[ du^i \otimes du^j(X, Y) := du^i(X)\, du^j(Y). \] Note that both \(du^i(X)\) and \(du^j(Y)\) are real scalars, so it makes sense to multiply them. Again, take the surface case \(n=2\) as an example, given two vectors \(X = a\pdF{u}+b\pdF{v}\) and \(Y = c\pdF{u}+d\pdF{v}\), then \[ (du\otimes du)(X,Y) = ac, \quad (du\otimes dv)(X,Y) = ad, \quad (dv\otimes du)(X,Y) = bc, \quad (dv\otimes dv)(X,Y) = bd. \]
Recall that the first fundamental form can be represented by a matrix whose entries are \(g_{11} = \langle\pdF{u_1},\pdF{u_1}\rangle\), \(g_{12} = \langle\pdF{u_1},\pdF{u_2}\rangle\), \(g_{21} = \langle\pdF{u_2},\pdF{u_1}\rangle\), \(g_{22} = \langle\pdF{u_2},\pdF{u_2}\rangle\) under the identification \((u_1,u_2)=(u,v)\). One can verify that the bilinear map \(B = g_{ij}\, du^i \otimes du^j\) satisfies \(B(X,Y) = \langle X,Y\rangle\) for any tangent vectors \(X,Y\). Therefore, apart from using matrix representation, we can also represent the first fundamental form using tensor notations: \[ g = g_{ij}\, du^i \otimes du^j. \] In particular, for two-dimensional surfaces, it is given by: \[ g = g_{11}\,du^1\otimes du^1 + g_{12}\,du^1\otimes du^2 + g_{21}\,du^2\otimes du^1 + g_{22}\,du^2\otimes du^2. \] Recall from Example 3.3 that the first fundamental form of the local parametrization \(F(u,v)=(\sin u\cos v, \sin u\sin v, \cos u)\) of the unit sphere \(S^2\) is \([g]=\begin{bmatrix}1&0\\0&\sin^2 u\end{bmatrix}\). Using tensor notations, we can rewrite it as \[ g = du\otimes du + \sin^2 u\, dv\otimes dv. \] Physicists working on general relativity often abbreviate \(du\otimes du\) by simply \(du^2\), and write the above as \[ du^2 + \sin^2 u\, dv^2. \] Note that \(du^2\) is interpreted as \((du)^2 = du\otimes du\), not \(d(u^2)\).
Express the first fundamental forms in Exercises 3.1 and 3.2 using tensor notations.