2

Given $\nabla$ a torsionless connection, the Ricci identity for co-vectors is $$\nabla_a \nabla_b \lambda_c - \nabla_b \nabla_a \lambda_c = -R^d_{\,\,cab}\lambda_d.$$ Prove $R^a_{[bcd]} = 0$ by considering the co-vector field $\lambda_c = \nabla_c f$.

By definition, $$R^a_{[bcd]} = 0 = \frac{1}{3!} \left(R^a_{\,\,bcd} + R^a_{\,\,cdb} + R^a_{\,\,dbc} - R^a_{\,\,bdc} - R^a_{\,\,cbd} - R^a_{\,\,dcb}\right)\,\,\,\,(1)$$

Attempt:

Input the given form for the covector into the Ricci identity in the question. Then since $\nabla_c f = e_c(f),$ we have $$\nabla_a \nabla_b e_c(f) - \nabla_b \nabla_a e_c(f) = -R^d_{\,\,cab}e_d(f).$$ True for all functions $f$, so $$\nabla_a \nabla_b e_c - \nabla_b \nabla_a e_c = -R^d_{\,\,cab}e_d.$$ Then since $\nabla_a e_b = \Gamma^d_{ba} e_d$ we can simplify the above to give $$\nabla_a \Gamma^d_{cb}e_d - \nabla_b \Gamma^d_{ca}e_d = -R^d_{\,\,cab}e_d$$ which can then be further rewritten like $$\nabla_a \Gamma^d_{cb} + \Gamma^{\alpha}_{cb}\Gamma^d_{\alpha a} - \nabla_b \Gamma^d_{ca} - \Gamma^{\alpha}_{ca}\Gamma^d_{\alpha b} = -R^d_{\,\,cab}.$$ I was then going to relabel all indices to get terms like that in the equation in $(1)$ and sum them all up and I hoped to get zero, but it is not. Have I made an error in the above somewhere? Thanks!

CAF
  • 2,850
  • Hint: try to use the fact that the connection $\nabla$ is torsionless if and only if $\nabla_a \nabla_b f = \nabla_b \nabla_a f$ for any smooth scalar $f$. – Yuri Vyatkin Mar 11 '15 at 11:30
  • One more suggestion: see how we establish the more complicated Bianchi identity here, which may give you a clue. – Yuri Vyatkin Mar 11 '15 at 11:44
  • So I can write $$2\nabla_{|a}\nabla_{b|} \lambda_c = -R^d_{,,cab}\lambda_d \Rightarrow 2\nabla_{|a}\nabla_{b|} \nabla_c f = -R^d_{,,cab} \nabla_d f$$ I could then take another covariant derivative wrt $e$ but I am not sure how to progress really. I think the aim is to show $R^{d}{,,cab}$ has bianchi symmetry in the indices $c,a,b$. I am new to this material so please bear with me :) – CAF Mar 11 '15 at 13:38
  • You are on the way to the solution. Just observe that if you now antisymmetrize the indices $a$, $b$ and $c$, you must have all the terms cancel out, because $\nabla_b \nabla_c f$ is symmetric. – Yuri Vyatkin Mar 11 '15 at 13:50
  • Thanks! so when I do that I get $R^a_{[bcd]}\nabla_a f=0$ and since $f$ was arbritary, I get the result. Is there anything in particular wrong with what I had done in my first attempt, where I did more work with the christoffel symbols? – CAF Mar 11 '15 at 15:03
  • Your $\nabla_c f = e_c(f)$ gives you a twist of the notation, which causes the problem. – Yuri Vyatkin Mar 12 '15 at 07:40
  • Ah so you mean $\nabla_c f $ is notation for the $c$th component of $\nabla f$ not $\nabla_c$ acting on $f$ which would be denoted by $\nabla_c (f) = e_c(f)$? – CAF Mar 12 '15 at 08:32

0 Answers0