From the definition of Riemann tensor we have:

$$ \mathbf{R}\left( \mathbf{z},\mathbf{v},\mathbf{w}\right)=\nabla_{\mathbf{[v}}\nabla_{\mathbf{w}]}\mathbf{z}-\nabla_{[\mathbf{v},\mathbf{w}]}\mathbf{z} \label{riemannnew} $$

and computing the coordinates of $\mathbf{R}$ in a coordinate basis we obtain:

$$ R^a_{\hphantom{a}bcd}=\partial_c\Gamma^a_{\hphantom{a}bd}-\partial_{d}\Gamma^a_{\hphantom{a}bc}+\Gamma^a_{\hphantom{a}\mu c}\Gamma^\mu_{\hphantom{a}bd}-\Gamma^a_{\hphantom{a}\mu d}\Gamma^\mu_{\hphantom{a}bc} $$

I find another way to compute the coefficient fo Riemann tensor with not vanishing torsion:

$$ [\nabla_c,\nabla_d]V^a=2\nabla_{[c}\nabla_{d]}V^a = 2\partial_{[c}\nabla_{d]}V^a-2\Gamma^e_{\hphantom{e}[dc]}\nabla_eV^a+2\Gamma^a_{\hphantom{a}e[c}\nabla_{d]}V^e \nonumber \\ = 2\partial_{[c}(\partial_{d]}V^a+\Gamma^a_{\hphantom{a}|e|d]}V^e)+2S^e_{\hphantom{a}cd}\nabla_eV^a+2\Gamma^a_{\hphantom{a}e[c}(\partial_{d]}V^e+\Gamma^e_{\hphantom{a}|b|d]}V^b) \nonumber \\ = 2 \partial_{[c}\Gamma^a_{\hphantom{a}|b|d]}V^b-2 \Gamma^a_{\hphantom{a}e[c}\partial_{d]}V^e+2S^e_{\hphantom{a}cd}\nabla_eV^a+2\Gamma^a_{\hphantom{a}e[c}\partial_{d]}V^e+2\Gamma^a_{\hphantom{a}e[c}\Gamma^e_{\hphantom{a}|b|d]}V^b= \nonumber \\ =2(\partial_{[c}\Gamma^a_{\hphantom{a}|b|d]}+\Gamma^a_{\hphantom{a}e[c}\Gamma^e_{\hphantom{a}|b|d]})V^b + 2S^e_{\hphantom{a}cd}\nabla_eV^a \tag{1} $$

where the first bracket is the Riemann-Cartan tensor and second term is the part due to the non vanishing torsion tensor.

My question is:

The first term of the first definition $\nabla_{\mathbf{[v}}\nabla_{\mathbf{w}]}\mathbf{z}$ is the second equation (1) but only the first term of the second equation is the Riemann tensor. How can I solve this problem? Is the definition of the Riemann tensor incomplete?


1 Answer 1


In the invariant notation $\nabla_X\nabla_Y$ corresponds to $X^a\nabla_a(Y^b\nabla_b)$, not $X^a Y^b\nabla_a\nabla_b$, eg. the vector field $Y$ also gets differentiated.

We can define $\nabla^2_{X,Y}Z=i_Xi_Y\nabla\nabla Z$, where here $i$ means "insert into the last empty argument", then we have $$ X^a\nabla_a(Y^b\nabla_b)Z^c=X^a\nabla_aY^b\nabla_bZ^c+X^aY^b\nabla_a\nabla_bZ^c, $$ so $$ \nabla^2_{X,Y}Z=\nabla_X\nabla_YZ-\nabla_{\nabla_XY}Z. $$

This gives then $$ R(X,Y)Z=\nabla_X\nabla_YZ-\nabla_Y\nabla_XZ-\nabla_{[X,Y]}Z=\nabla^2_{X,Y}Z+\nabla_{\nabla_XY}Z-\nabla^2_{Y,X}Z-\nabla_{\nabla_YX}Z-\nabla_{[X,Y]}Z \\ =\nabla^2_{X,Y}Z-\nabla^2_{Y,X}Z+\nabla_{\nabla_XY-\nabla_YX-[X,Y]}Z=\nabla^2_{X,Y}Z-\nabla^2_{Y,X}Z+\nabla_{T(X,Y)}Z. $$

As you can see the $[\nabla_a,\nabla_b]X^c=R^c_{\ dab}X^d$ Ricci-identity corresponds to $R(X,Y)Z=\nabla^2_{X,Y}Z-\nabla^2_{Y,X}Z$, which is certainly true in absence of torsion.

In the presence of torsion, this gets modified to $$ R(X,Y)Z=\nabla^2_{X,Y}Z-\nabla^2_{Y,X}Z+\nabla_{T(X,Y)}Z, $$ but the definition of the curvature tensor, $$ R(X,Y)=\nabla_X\nabla_Y-\nabla_Y\nabla_X-\nabla_{[X,Y]} $$ doesn't depend on torsion at all.

  • $\begingroup$ Ok, I can see that from the definition in presence of torsion we get that the curvature tensor is $$ R(X,Y)Z=\nabla^2_{X,Y}Z-\nabla^2_{Y,X}Z+\nabla_{T(X,Y)}Z, $$ but only the first two terms correspond to the Riemann tensor? Curvature tensor and Riemann tensor differs in presence of torsion? $\endgroup$ Commented Jul 2, 2017 at 10:47
  • $\begingroup$ @raskolnikov No, all three terms correspond to the curvature tensor. The more fundamental definition of the curvature tensor is $ R(X,Y)Z=\nabla_X\nabla_YZ-\nabla_Y\nabla_XZ-\nabla_{[X,Y]}Z $, if you write this in terms of the second covariant derivative, you get the expression you wrote in the comment. $\endgroup$ Commented Jul 2, 2017 at 11:22
  • 1
    $\begingroup$ Ok but $ R^a_{\ bcd} $ is equal to which terms in presence of torsion? $\endgroup$ Commented Jul 2, 2017 at 15:29
  • $\begingroup$ @raskolnikov $R^a_{bcd}X^b=[\nabla_c,\nabla_d]X^a+T^b_{\ cd}\nabla_bX^a$. $\endgroup$ Commented Jul 2, 2017 at 15:49
  • $\begingroup$ Sorry but I struggle with this problem again. In your last comment you said: $R^a_{\ bcd}X^b=[\nabla_c,\nabla_d]X^a+T^b_{\ cd}\nabla_b X^a$ but this is incompatible with my calculation of $[\nabla_c, \nabla_d]$ I wrote in my question since I can't eliminate the torsion tensor $T^b_{\ cd}$ putting my formula in your one. $\endgroup$ Commented Jul 6, 2017 at 13:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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