I am looking to calculate the Riemann curvature on an orthonormal frame for the tangent bundle of the surface parametrized by the equation $\phi: \mathbb{R}^2\to \mathbb{R}^3$ given by $$ (x,y) \mapsto (x,y,f(x,y))$$ for some continuous $f:\mathbb{R}^2 \to \mathbb{R}$. I would like to do this by calculating the Christoffel symbols with respect to the orthogonal tangent vectors $$ \begin{align} \frac{\partial}{\partial \varphi_1} &= -f_y\frac{\partial}{\partial x} + f_x\frac{\partial}{\partial y} \\ \frac{\partial}{\partial \varphi_2} &= f_x\frac{\partial}{\partial x} + f_y\frac{\partial}{\partial y} + (f_x^2 + f_y^2)\frac{\partial}{\partial z} \end{align} $$ (following the approach of Orthonormal basis for a tangent plane) where the metric is the restriction of the Euclidean metric from $\mathbb{R}^3$. I would then calculate $\nabla_\frac{\partial}{\partial \varphi_i}\frac{\partial}{\partial \varphi_j}$ for all $i,j$ and hence the curvature on the normalised frame. However, I am unclear on a few points:

  1. If $z = f(x,y)$, then should I calculate $\frac{\partial}{\partial z} = f_x \frac{\partial}{\partial x} + f_y \frac{\partial}{\partial y}$ (where $f_x = \frac{\partial f}{\partial x}$, etc.)? Doesn't this mess up orthogonality?
  2. Should I somehow try to rewrite $x,y$ in terms of $\varphi_1, \varphi_2$ and then do the Christoffel symbol calculation? My only previous example was calculating the Riemann curvature for a sphere where the parametrization was in terms of spherical polar coordinates, which made the calculus cleaner.
  3. There is an expression for Christoffel symbols in terms of $\phi_x, \phi_y, \phi_{xx}$, etc. as per Christoffel Symbols on a Surface. However, here $\phi_x$ and $\phi_y$ are not orthogonal, so would it still be possible for me to calculate Christoffel symbols using that method and then apply those to the orthogonal frame I wish to use?

EDIT MADE: Corrected the choice of tangent vectors in the orthonormal frame, as the original frame was not tangent.

  • $\begingroup$ If you see your definition as a pullback from $\mathbb{R}^3$ to $\mathbb{R}^2$ you can start from the eucledian metric of $\mathbb{R}^3$ $ds^2=dx^2+dy^2+dz^3=\left(1+\left(\frac{df}{dx}\right)^2\right) dx^2+\left(1+\left(\frac{df}{dy}\right)^2\right) dy^2 + 2\frac{df}{dy}\frac{df}{dx}dxdy$ to get the metric on the surface f. Thats all you need to calculate the Christoffel symbols. $\endgroup$ – Bort Feb 15 '18 at 18:07
  • $\begingroup$ Hello @Bort, thank you for your comment. My confusion is arising from the fact that I want to calculate the Christoffel symbols to derive the curvature tensor and then apply that tensor to an orthonormal frame. However, if I calculate the $\Gamma^{i}_{jk}$ with respect to one frame, will the curvature matrix need to be transformed as well? $\endgroup$ – An Coileanach Feb 16 '18 at 19:31
  • $\begingroup$ To be more specific, can I say something like $ds^2 = d\varphi_1^2 + d\varphi_2^2$ and then compute the corresponding Christoffel symbols just in terms of those? $\endgroup$ – An Coileanach Feb 17 '18 at 11:00

I am currently struggling with surfaces as well at the moment, so I'm no expert, take this answer with a grain of salt. As you're correclty stating $\phi_x , \phi_y$ are NOT an orthonormal frame for $T_p S$, though I think that it is much easier to express Christoffel symbols in terms of those basis vectors. Using the system $$\phi_{xx} = \Gamma_{11}^1 \phi_x + \Gamma_{11}^2 \phi_y + L^1 N$$ $$\phi_{xy} = \Gamma_{12}^1 \phi_x + \Gamma_{12}^2 \phi_y + L^2 N$$ $$\phi_{yy} = \Gamma_{22}^1 \phi_x + \Gamma_{22}^2 \phi_y + L^3 N$$ and multiplying each equation scalarly by $\phi_x$ and $\phi_y$ the normal vector $N$ always cancels out (being orthgonal to $T_p S$), and you're left with a linear system of six equations from which you can derive the six Christoffel Symbols.

  • $\begingroup$ In simpler terms, in my opinion you should follow the third route you suggested. The fact that $\phi_x,\phi_y$ are not orthogonal just means that you won't be able to extract $\Gamma_{ij}^k$ right away, and you'll have to solve a linear system. No big deal. $\endgroup$ – barmanthewise Feb 17 '18 at 11:23
  • $\begingroup$ Thank you for your comment @barmanthewise. The thing that is confusing me here is whether the $\Gamma^{k}_{ij}$ coming from this will be the same as the $\Gamma^{k}_{ij}$ from the orthonormal coordinate system, or will I need to apply some linear transformation? I appreciate that you're in the same boat as me, as someone who is just learning this stuff, but you seem to have a better intuition for it than I do. $\endgroup$ – An Coileanach Feb 17 '18 at 11:33
  • $\begingroup$ If by 'the' orthonormal coordinate system you mean the standard basis of $\Bbb R^3$ as the tangent space to $S$, that is $\partial_x,\partial_y,\partial_z$, then, yes, you'd need to apply a linear transformation, for exapmle the one that takes $\phi_x \mapsto \partial_x$, $\phi_y \mapsto \partial_y$, $N \mapsto \partial_z$. $\endgroup$ – barmanthewise Feb 17 '18 at 14:04
  • $\begingroup$ I should have been clearer: by "the orthonormal coordinate system", I meant the normalised version of the frame corresponding to the tangent vectors $\frac{\partial}{\partial\varphi_1}$, $\frac{\partial}{\partial\varphi_2}$. I don't think $\partial_x$, $\partial_y$, $\partial_z$ would be a basis for the tangent space to $S$ in general. $\endgroup$ – An Coileanach Feb 17 '18 at 19:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.