Recall that in the proof of Theorem 3.29, the Gauss curvature can be written as \[K=\frac{g_{1q}R^q_{122}}{\det[g]}=\frac{g\!\left(R^q_{122}\frac{\partial F}{\partial u^q},\frac{\partial F}{\partial u^1}\right)}{\det[g]}.\] The Riemann curvature tensor \(R^q_{122}\) is related to covariant derivatives by \(R^q_{122}\partial_q=\nabla_1\nabla_2\partial_2-\nabla_2\nabla_1\partial_2\) (where \(\partial_i:=\frac{\partial F}{\partial u^i}\) and \(\nabla_i=\nabla_{\frac{\partial F}{\partial u^i}}\)). We then have \[K=\frac{g(\nabla_1\nabla_2\partial_2-\nabla_2\nabla_1\partial_2,\partial_1)}{\det[g]}.\] This motivates defining the “4-tensor” \(T(X,Y,Z,W):=g(\nabla_X\nabla_Y Z-\nabla_Y\nabla_X Z,W)\), so \(K=\frac{T(\partial_1,\partial_2,\partial_2,\partial_1)}{\det[g]}\). Unfortunately \(T\) is not a tensor in general, but the following clever modification is:
Consider a regular hypersurface \(\Sigma^n\subset\mathbb{R}^{n+1}\) locally parametrized by \(F(u^i)\). The Riemann curvature 4-tensor is defined as: \[\mathrm{Rm}(X,Y,Z,W):=g(\nabla_X\nabla_Y Z-\nabla_Y\nabla_X Z-\nabla_{\nabla_X Y-\nabla_Y X}Z,\,W),\] for any vector fields \(X,Y,Z,W\).
The modification is “clever” because when we take \(X,Y\) to be \(\partial_i\)'s, we have \(\nabla_X Y-\nabla_Y X=\Gamma^k_{ij}\partial_k-\Gamma^k_{ji}\partial_k=0\). Therefore we still have \(K=\frac{\mathrm{Rm}(\partial_1,\partial_2,\partial_2,\partial_1)}{\det[g]}\).
Verify that \(\mathrm{Rm}\) is a 4-tensor, i.e. for any scalar functions \(f_i\)'s and vector fields \(X,Y,Z,W\), we have \(\mathrm{Rm}(f_1 X,f_2 Y,f_3 Z,f_4 W)=f_1 f_2 f_3 f_4\,\mathrm{Rm}(X,Y,Z,W)\).
On a regular hypersurface \(\Sigma^n\subset\mathbb{R}^{n+1}\) locally parametrized by \(F(u^i)\), denote \(R_{ijkl}=\mathrm{Rm}\!\left(\frac{\partial F}{\partial u^i},\frac{\partial F}{\partial u^j},\frac{\partial F}{\partial u^k},\frac{\partial F}{\partial u^l}\right)\). Then:
(i) is trivial from the definition of \(\mathrm{Rm}\). (ii) would follow from (i) and (iii) once (iii) is proved. To prove (iii), recall that \(R_{ijkl}=g(\nabla_i\nabla_j\partial_k-\nabla_j\nabla_i\partial_k,\partial_l)=R^q_{ijk}g_{ql}\) (since \(\nabla_i\partial_j-\nabla_j\partial_i=0\)).
By Gauss's equation (Theorem 3.28), \(R^q_{ijk}=g^{qs}(h_{jk}h_{si}-h_{ik}h_{sj})\), and so \(R_{ijkl}=\delta_{sl}(h_{jk}h_{si}-h_{ik}h_{sj})=h_{jk}h_{li}-h_{ik}h_{lj}\). Since \(h\) is a symmetric 2-tensor, (iii) \(R_{ijkl}=R_{klij}\) is obvious. \(\square\)
The alternating properties of \(\mathrm{Rm}\) imply \(\mathrm{Rm}(X,X,Z,W)=0\) (taking \(Y=X\)) and \(\mathrm{Rm}(X,Y,Z,Z)=0\).
Let \(\{e_1,e_2\}\) be an orthonormal frame of a regular surface \(\Sigma^2\subset\mathbb{R}^3\). Show that \(K=\mathrm{Rm}(e_1,e_2,e_2,e_1)\). Hint: Results from Exercise 4.10 will be helpful.