The key ingredient of Gauss’s Theorema Egregium is the following Gauss–Codazzi equations, which hold for any regular hypersurface in any dimension. When the dimension is two, the RHS of the Gauss’s equation is similar to \(\det[h]\) while the LHS depends only on \(g\). This gives Theorema Egregium as a direct consequence of the Gauss’s equation.
On any regular hypersurface \(\Sigma\) in \(\mathbb{R}^{n+1}\), the following equations hold: \[ \pd{\Gamma^q_{jk}}{u_i} - \pd{\Gamma^q_{ik}}{u_j} + \Gamma^l_{jk}\Gamma^q_{il} - \Gamma^l_{ik}\Gamma^q_{jl} = g^{ql}(h_{jk}h_{li} - h_{ik}h_{lj}) \qquad (\text{Gauss}) \] \[ \pd{h_{jk}}{u_i} - \pd{h_{ik}}{u_j} + \Gamma^l_{jk}h_{il} - \Gamma^l_{ik}h_{jl} = 0 \qquad (\text{Codazzi}) \]
The key step is to start with the equality of mixed partials: \(\pd{^3\mathbf{F}}{u_i\partial u_j\partial u_k} = \pd{^3\mathbf{F}}{u_j\partial u_i\partial u_k}\), and rewrite both sides in terms of the tangent-normal basis of \(\mathbb{R}^{n+1}\). Gauss’s equation follows from equating the tangent coefficients, and Codazzi’s equation is obtained by equating the normal coefficient.
By (3.13): \(\pd{^2\mathbf{F}}{u_j\partial u_k} = \Gamma^l_{jk}\pdF{u_l} + h_{jk}\nu\). Differentiating with respect to \(u_i\): \begin{align*} \pd{^3\mathbf{F}}{u_i\partial u_j\partial u_k} &= \pd{\Gamma^l_{jk}}{u_i}\pdF{u_l} + \Gamma^l_{jk}\pd{^2\mathbf{F}}{u_i\partial u_l} + \pd{h_{jk}}{u_i}\nu + h_{jk}\pd{\nu}{u_i} \\ &= \pd{\Gamma^l_{jk}}{u_i}\pdF{u_l} + \Gamma^l_{jk}\left(\Gamma^q_{il}\pdF{u_q}+h_{il}\nu\right) + \pd{h_{jk}}{u_i}\nu - h_{jk}h^q_i\pdF{u_q} \\ &= \left(\pd{\Gamma^q_{jk}}{u_i} + \Gamma^l_{jk}\Gamma^q_{il} - h_{jk}h^q_i\right)\pdF{u_q} + \left(\pd{h_{jk}}{u_i} + \Gamma^l_{jk}h_{il}\right)\nu. \end{align*} Here we used \(\pd{\nu}{u_i} = -h^q_i\pdF{u_q}\) (from the shape operator computation).
Switching \(i\) and \(j\): \[ \pd{^3\mathbf{F}}{u_j\partial u_i\partial u_k} = \left(\pd{\Gamma^q_{ik}}{u_j} + \Gamma^l_{ik}\Gamma^q_{jl} - h_{ik}h^q_j\right)\pdF{u_q} + \left(\pd{h_{ik}}{u_j} + \Gamma^l_{ik}h_{jl}\right)\nu. \] Equating tangent and normal coefficients gives Gauss’s and Codazzi’s equations respectively. \(\square\)
We derived the Gauss–Codazzi equations (Theorem 3.28). It is worthwhile to note that the LHS of the Gauss’s equation \[ \underbrace{\pd{\Gamma^q_{jk}}{u_i} - \pd{\Gamma^q_{ik}}{u_j} + \Gamma^l_{jk}\Gamma^q_{il} - \Gamma^l_{ik}\Gamma^q_{jl}}_{\text{depends only on }\Gamma^k_{ij}\text{’s}} = g^{ql}(h_{jk}h_{li} - h_{ik}h_{lj}) \] involves only Christoffel symbols and their derivatives. From (3.14) we also know that the Christoffel symbols depend only on the first fundamental form \(g\) but not on \(h\). For simplicity, we denote: \[ R^q_{ijk} := \pd{\Gamma^q_{jk}}{u_i} - \pd{\Gamma^q_{ik}}{u_j} + \Gamma^l_{jk}\Gamma^q_{il} - \Gamma^l_{ik}\Gamma^q_{jl}. \] The lower and upper indices for \(R^q_{ijk}\) are chosen so as to preserve their positions in the RHS expression (\(q\) being upper, and \(i,j,k\) being lower). We will see later that \(R^q_{ijk}\)’s are the components of the Riemann curvature tensor.
We are now in a position to give a proof of Gauss’s Theorema Egregium as a direct consequence of the Gauss’s equation.
On any regular surface \(\Sigma^2\) in \(\mathbb{R}^3\), the Gauss curvature \(K\) depends only on its first fundamental form \(g\). In other words, \(K\) is intrinsic for regular surfaces.
Consider the Gauss’s equation, which asserts that for any \(i,j,k,q\): \[ R^q_{ijk} = g^{ql}(h_{jk}h_{li} - h_{ik}h_{lj}). \] Multiplying both sides by \(g_{pq}\), and summing up all \(q\)’s: \[ g_{pq}R^q_{ijk} = g_{pq}g^{ql}(h_{jk}h_{li}-h_{ik}h_{lj}) = h_{jk}h_{pi} - h_{ik}h_{pj}. \]
The above result is true for any \(i,j,k,p\). In particular, when \((i,j,k,p)=(1,2,2,1)\): \[ g_{1q}R^q_{122} = h_{22}h_{11} - h_{12}^2 = \det[h]. \] This shows \(\det[h]\) depends only on \(g\) since \(R^q_{122}\) does so.
Finally, recall that the Gauss curvature is given by \(K = \dfrac{\det[h]}{\det[g]}\). Therefore, we have completed the proof that \(K\) depends only on \(g\). \(\square\)
It is important to note that Theorem 3.29 holds for surfaces (i.e. \(\dim\Sigma = 2\)).
The long \(R^q_{ijk}\)-term can be interpreted in a nicer way using covariant derivatives. For simplicity, we denote \(\partial_i := \pdF{u_i}\) and \(\nabla_i := \nabla_{\partial_i} = \nabla_{\pdF{u_i}}\). By direct computations: \[ \nabla_j\partial_k = \Gamma^l_{jk}\partial_l, \] \[ \nabla_i(\nabla_j\partial_k) = \nabla_i(\Gamma^l_{jk}\partial_l) = \pd{\Gamma^l_{jk}}{u_i}\partial_l + \Gamma^l_{jk}\nabla_i\partial_l = \pd{\Gamma^q_{jk}}{u_i}\partial_q + \Gamma^l_{jk}\Gamma^q_{il}\partial_q. \] Similarly: \[ \nabla_j(\nabla_i\partial_k) = \pd{\Gamma^q_{ik}}{u_j}\partial_q + \Gamma^l_{ik}\Gamma^q_{jl}\partial_q. \] Hence, the term \(R^q_{ijk}\) is the commutator of \(\nabla_i\) and \(\nabla_j\): \[ [\nabla_i,\nabla_j]\partial_k = \nabla_i(\nabla_j\partial_k) - \nabla_j(\nabla_i\partial_k) = \left(\pd{\Gamma^q_{jk}}{u_i} - \pd{\Gamma^q_{ik}}{u_j} + \Gamma^l_{jk}\Gamma^q_{il} - \Gamma^l_{ik}\Gamma^q_{jl}\right)\partial_q = R^q_{ijk}\partial_q. \] In summary, we see that once \(g_{ij}\) is fixed, the covariant derivative \(\nabla\) and hence the curvature term \(R^q_{ijk}\) are uniquely determined. It motivates the development of intrinsic geometry, a branch of geometry that plays no reference to the ambient space but only the “surface” itself. This branch is called Riemannian Geometry!
Gauss’s Theorema Egregium can be further generalized to higher dimensional regular hypersurfaces \(\Sigma^n\subset\mathbb{R}^{n+1}\). The determinant of the second fundamental form \(h\) is much more complicated in higher dimensions, and is not just a quadratic expression of \(h_{ij}\)’s as in the surface case. More algebraic manipulation will be needed in order to prove that \(\det[h]\) depends only on \(g\), and the argument is not as direct as the proof in the surface case.