In this subsection, we introduce an important concept in differential geometry – the first fundamental form. Loosely speaking, it is just the dot product between the tangent vectors of a regular (hyper-) surface, yet the way to express it is worth some discussion as it is a quantity that geometers and physicists call tensors. It captures and encodes intrinsic geometric information (such as curvature) about the hypersurface.
Let \(\langle \cdot, \cdot \rangle\) be the dot product in \(\mathbb{R}^{n+1}\), i.e. \(\langle u, v \rangle := u \cdot v\) for any \(u, v \in \mathbb{R}^{n+1}\). It is then a bilinear map \(\langle \cdot, \cdot \rangle : \mathbb{R}^{n+1} \to \mathbb{R}^{n+1} \to \mathbb{R}\) taking a pair of vectors in \(\mathbb{R}^{n+1}\) to a scalar. The first fundamental form of a regular hypersurface \(\Sigma^n \subset \mathbb{R}^{n+1}\) is restriction of \(\langle \cdot, \cdot \rangle\) on \(\Tp{p}\Sigma^n \times \Tp{p}\Sigma^n\), where \(p \in \Sigma^n\). In modern notations, it is commonly denoted by \(g\), i.e. for any \(X, Y \in \Tp{p}\Sigma^n\), we have \[ g(X, Y) := \langle X, Y \rangle. \]
Apparently, the first fundamental form is nothing but just the dot product, restricting the inputs to be tangent vectors to \(\Sigma^n\). What makes it distinctive from the dot product in \(\mathbb{R}^{n+1}\) is when we represent \(g\) using matrices. Recall from linear algebra that a bilinear map \(b : V \times V \to \mathbb{R}\) on a vector space \(V\) with basis \(\{e_i\}_{i=1}^n\) is uniquely determined by its product on basis vectors, i.e. \(b(e_i, e_j)\). It is because for any \(v, w \in V\), one can write \[ v = \sum_{i=1}^n x_i e_i, \qquad w = \sum_{j=1}^n y_j e_j, \] for some \(x_i, y_j \in \mathbb{R}\), then \[ b(v, w) = b\!\left(\sum_{i=1}^n x_i e_i,\, \sum_{j=1}^n y_j e_j\right) = \sum_{i,j=1}^n x_i y_j\, b(e_i, e_j). \] Therefore, the value of \(b(v,w)\) is completely determined by \(b(e_i, e_j)\)'s. In other words, the “data” of the inner product \(b\) is encoded by the matrix \[ \begin{bmatrix} b(e_1,e_1) & b(e_1,e_2) & \cdots & b(e_1,e_n) \\ b(e_2,e_1) & b(e_2,e_2) & \cdots & b(e_2,e_n) \\ \vdots & \vdots & & \vdots \\ b(e_n,e_1) & b(e_n,e_2) & \cdots & b(e_n,e_n) \end{bmatrix}, \] which is often called the matrix representation of the bilinear map \(b\).
Now back to the discussion of the first fundamental form \(g\). Recall that \(g\) is the restriction of the dot product on tangent vectors. The tangent space at \(p \in \Sigma^n\) has a basis \(\left\{\pdF{u_i}\right\}_{i=1}^n\), whereas \(\mathbb{R}^{n+1}\) has the standard basis \(\{e_i\}_{i=1}^{n+1}\). Therefore, to represent the dot product in \(\mathbb{R}^{n+1}\) with respect to the standard basis, we may use the matrix \[ \begin{bmatrix} \langle e_1,e_1\rangle & \langle e_1,e_2\rangle & \cdots & \langle e_1,e_{n+1}\rangle \\ \langle e_2,e_1\rangle & \langle e_2,e_2\rangle & \cdots & \langle e_2,e_{n+1}\rangle \\ \vdots & \vdots & & \vdots \\ \langle e_{n+1},e_1\rangle & \langle e_{n+1},e_2\rangle & \cdots & \langle e_{n+1},e_{n+1}\rangle \end{bmatrix}, \] which is simply the \((n+1)\times(n+1)\) identity matrix \(I\). However, to represent \(g\) using a matrix, we should consider the basis \(\left\{\pdF{u_i}\right\}_{i=1}^n\) instead. Its matrix representation is then the \(n\times n\)-matrix: \[ [g] := \begin{bmatrix} g\!\left(\pd{\mathbf{F}}{u_1},\pd{\mathbf{F}}{u_1}\right) & g\!\left(\pd{\mathbf{F}}{u_1},\pd{\mathbf{F}}{u_2}\right) & \cdots & g\!\left(\pd{\mathbf{F}}{u_1},\pd{\mathbf{F}}{u_n}\right) \\ g\!\left(\pd{\mathbf{F}}{u_2},\pd{\mathbf{F}}{u_1}\right) & g\!\left(\pd{\mathbf{F}}{u_2},\pd{\mathbf{F}}{u_2}\right) & \cdots & g\!\left(\pd{\mathbf{F}}{u_2},\pd{\mathbf{F}}{u_n}\right) \\ \vdots & \vdots & & \vdots \\ g\!\left(\pd{\mathbf{F}}{u_n},\pd{\mathbf{F}}{u_1}\right) & g\!\left(\pd{\mathbf{F}}{u_n},\pd{\mathbf{F}}{u_2}\right) & \cdots & g\!\left(\pd{\mathbf{F}}{u_n},\pd{\mathbf{F}}{u_n}\right) \end{bmatrix}, \] which may not be the identity matrix. We will give more examples later. It turns out that the above matrix encodes a lot of information about a surface, including area, curvature, etc. Because of this, we will denote the entries of the above matrix by: \[ g_{ij} := g\!\left(\pd{\mathbf{F}}{u_i},\pd{\mathbf{F}}{u_j}\right) = \left\langle \pd{\mathbf{F}}{u_i},\pd{\mathbf{F}}{u_j} \right\rangle, \] where \(\langle \cdot, \cdot \rangle\) denotes the usual dot product on \(\mathbb{R}^{n+1}\).
From now on we will omit the summation sign whenever there are double indices. For example, what we mean by \(a_i \pdF{u_i}\) is precisely: \[ \sum_{i=1}^n a_i \pd{\mathbf{F}}{u_i}, \] where \(n\) is typically the dimension of the object. This is commonly called Einstein’s summation convention, which was extremely useful in the days when type-writers were used to produce math and physics journal papers.
Given any tangent vectors \(X, Y \in \Tp{p}\Sigma^n\), one can express them as \[ X = a^i \pdF{u_i}, \qquad Y = b^j \pdF{u_j}, \] and so we have \[ g(X,Y) = g\!\left(a^i \pdF{u_i},\, b^j \pdF{u_j}\right) = a^i b^j\, g\!\left(\pdF{u_i},\pdF{u_j}\right) = a^i b^j g_{ij}. \] Therefore, the bilinear form \(g: V \times V \to \mathbb{R}\) is uniquely determined by the entries \(g_{ij}\)’s.
Let \(S^2\) be the unit 2-sphere and \(F\) be the following local parametrization: \[ F(u,v) = (\sin u \cos v,\, \sin u \sin v,\, \cos u), \quad (u,v) \in (0,\pi)\times(0,2\pi). \] By direct computations, we have: \[ \pd{\mathbf{F}}{u} = (\cos u \cos v,\, \cos u \sin v,\, -\sin u), \] \[ \pd{\mathbf{F}}{v} = (-\sin u \sin v,\, \sin u \cos v,\, 0). \] \[ g\!\left(\pdF{u},\pdF{u}\right) = \pd{\mathbf{F}}{u}\cdot\pd{\mathbf{F}}{u} = 1, \qquad g\!\left(\pdF{u},\pdF{v}\right) = \pd{\mathbf{F}}{u}\cdot\pd{\mathbf{F}}{v} = 0, \] \[ g\!\left(\pdF{v},\pdF{u}\right) = \pd{\mathbf{F}}{v}\cdot\pd{\mathbf{F}}{u} = 0, \qquad g\!\left(\pdF{v},\pdF{v}\right) = \pd{\mathbf{F}}{v}\cdot\pd{\mathbf{F}}{v} = \sin^2 u. \] Therefore, its first fundamental form can be represented by the matrix \[ [g] = \begin{bmatrix} 1 & 0 \\ 0 & \sin^2 u \end{bmatrix}. \] If one treats \(u_1 = u\) and \(u_2 = v\), then we may say: \[ g_{11} = 1, \quad g_{12} = 0, \quad g_{21} = 0, \quad g_{22} = \sin^2 u. \]
The following are all local parametrizations of the unit sphere. Express their first fundamental forms using matrix:
Consider the parametrization \[ F(u,v) = (v(2+\cos u),\, v(2+\sin u),\, v), \] where \(u\in(0,2\pi)\) and \(v>0\). How does the surface look like? Express the first fundamental form in matrix form.
We will see in subsequent sections that \(g\) “encodes” crucial geometric information such as curvatures. There are also some familiar geometric quantities, such as length and area, which are related to the first fundamental form \(g\).
Consider a curve \(\gamma\) on a regular hypersurface \(\Sigma^n \subset \mathbb{R}^{n+1}\) parametrized by \(F(u_i)\). Suppose the curve can be parametrized by \(\gamma(t)\), \(a < t < b\), then from calculus we know the arc-length of the curve is given by: \[ L(\gamma) = \int_a^b |\gamma'(t)|\, dt. \] In fact one can express this above in terms of \(g\). The argument is as follows: Suppose \(\gamma(t)\) has local coordinates \((\gamma^i(t))\) such that \(F(\gamma^i(t)) = \gamma(t)\). Using the chain rule, we then have: \[ \gamma'(t) = \pdF{u_i}\frac{d\gamma^i}{dt}. \] Recall that \(|\gamma'(t)| = \sqrt{\langle \gamma'(t),\gamma'(t)\rangle}\) and that \(\gamma'(t)\) lies on \(\Tp{p}M\), we then have: \[ |\gamma'(t)| = \sqrt{g(\gamma'(t),\gamma'(t))}. \] We can then express it in terms of the matrix components \(g_{ij}\)’s: \[ \tag{3.1} g(\gamma'(t),\gamma'(t)) = g\!\left(\frac{d\gamma^i}{dt}\pdF{u_i},\,\frac{d\gamma^j}{dt}\pdF{u_j}\right) = \frac{d\gamma^i}{dt}\frac{d\gamma^j}{dt}\,g\!\left(\pdF{u_i},\pdF{u_j}\right) = g_{ij}\frac{d\gamma^i}{dt}\frac{d\gamma^j}{dt}. \] Therefore, the arc-length can be expressed in terms of the first fundamental form by: \[ L(\gamma) = \int_a^b \sqrt{g(\gamma'(t),\gamma'(t))}\, dt = \int_a^b \sqrt{g_{ij}\frac{d\gamma^i}{dt}\frac{d\gamma^j}{dt}}\, dt. \]
Another familiar geometric quantity which is also related to \(g\) is the area of a surface. For simplicity, we focus on dimension 2 first. Suppose a regular surface \(\Sigma\) can be parametrized by \(F(u,v)\) with \((u,v)\in D\subset\mathbb{R}^2\) where \(D\) is a bounded domain, the area of this surface is given by: \[ A(\Sigma) = \iint_D \left|\pd{\mathbf{F}}{u}\times\pd{\mathbf{F}}{v}\right|\, du\,dv. \] It is also possible to express \(\left|\pd{\mathbf{F}}{u}\times\pd{\mathbf{F}}{v}\right|\) in terms of the first fundamental form \(g\). Let \(\theta\) be the angle between the two vectors \(\pdF{u}\) and \(\pdF{v}\), then from elementary vector geometry, we have: \[ \left|\pd{\mathbf{F}}{u}\times\pd{\mathbf{F}}{v}\right|^2 = \left|\pd{\mathbf{F}}{u}\right|^2\left|\pd{\mathbf{F}}{v}\right|^2\sin^2\theta = \left|\pd{\mathbf{F}}{u}\right|^2\left|\pd{\mathbf{F}}{v}\right|^2 - \left(\pd{\mathbf{F}}{u}\cdot\pd{\mathbf{F}}{v}\right)^2 = g_{11}g_{22} - (g_{12})^2 = \det[g]. \] Therefore, \[ \tag{3.2} A(\Sigma) = \iint_D \sqrt{\det[g]}\, du\,dv. \] In fact, for higher dimensional regular hypersurfaces parametrized by \(F(u_i): \mathcal{U}\to\Sigma\), we also have its area equals \[ \int_{\mathcal{U}} \sqrt{\det[g]}\, du^1\cdots du^n. \]
Let \(\Sigma_f\) be the graph of a smooth function \(f: \mathcal{U}\to\mathbb{R}\) defined on an open subset \(\mathcal{U}\) of \(\mathbb{R}^n\), then \(\Sigma_f\) has a globally defined smooth parametrization: \[ F(u_1,\ldots,u_n) = (u_1,\ldots,u_n,\, f(u_1,\ldots,u_n)). \] By straight-forward computations, we can get: \[ g_{ij} = \delta_{ij} + \pd{f}{u_i}\pd{f}{u_j}. \] Therefore, the length of the curve \(\gamma(t):= F(\gamma^1(t),\ldots,\gamma^n(t))\) can be computed by integrating the square root of: \[ g_{ij}\frac{d\gamma^i}{dt}\frac{d\gamma^j}{dt} = \sum_{i=1}^n\left(\frac{d\gamma^i}{dt}\right)^2 + \left(\frac{d}{dt}f(\gamma(t))\right)^2. \] To compute the surface area of a region \(F(\Omega)\subset\Sigma_f\) where \(\Omega\) is a bounded domain on \(\mathcal{U}\), we first compute \(\det[g]\). The matrix \([g]\) can be written as \[ [g] = I_{n\times n} + (\nabla f)(\nabla f)^T. \] By standard linear algebra, we know that its eigenvalues are \(1+|\nabla f|^2, 1, \ldots, 1\), and so \(\det[g] = 1+|\nabla f|^2\). According to (3.2), we have: \[ A(F(\Omega)) = \iint_\Omega \sqrt{1+|\nabla f|^2}\, du^1\cdots du^n, \] which is exactly the same as what you have seen in multivariable calculus.
Find the surface area of the part of sphere covered by the local parametrizations in each of Example 3.3 and Exercise 3.1.