Only write out one index summation

Question

I have a basic question based on https://cadabra.science/qa/1851/manipulate-matrix. I have seen that you have updated some codes related to the above question a few days ago, but I still have the following question: How to just expand the dummy indices a,b and don't touch the indices $\mu,\nu$ in the following expression?

$$L:=D_{\mu} B^a D_{\nu}B^b D_{\mu} B^a D_{\nu}B^b,$$

where a,b and $\mu,\nu$ are the different index types, and $B^a$ is SelfNonCommuting.

Comments

This is a variant of your earlier question; once there is a solution I will let you know.

---

Sorry for my repetition. I have an expedient solution. First, substitute $\mu,\nu$ to $t,r$. Second, expand dummy indices. Last, substitute $t,r$ to $\mu,\nu$. But this is not the optimal solution obviously. I look forward to your good news.