For a∈T(R), write a=b+c where b∈radR and c∈[R,R]. We have for every m:bpm=(a−c)pm≡apm−cpm≡apm (mod [R,R]). Choosing m to be large enough, we have bpm=0 (since radR is nil). Therefore, the above congruence shows that apm∈[R,R]. Now assume k is a splitting field for R, and let a∈R be such that apm∈[R,R] for some m. Let R′=R/radR∼∏iAi where Ai=Mni(k). Our job is to show that a=a+radR belongs to [R,R]. Using the direct product decomposition above, we are reduced to showing that, for any i, the image ai′∈Ai of a′ belongs to [Ai,Ai] (given that ai′pm∈[Ai,Ai] for some m). Therefore, we may as well assume that R=Mn(k). Here, let us compare T′(R):={a∈R:apm∈[R,R] for some m≥1} with T(R)=[R,R]. Then T′(R) is a k-subspace of R containing T(R). But, T(R) has codimension 1 in R. Since a=diag(1,0,…,0)∈/T′(R), we must have T(R)=T′(R).