Suppose VV and WW are finite-dimensional and T1,T2L(V,W)T_1,T_2 \in \mathcal L(V,W). Prove that there exist invertible operators RL(V)R \in \mathcal L(V) and SL(W)S \in \mathcal L(W) such that T1=ST2RT_1 = ST_2R if and only if dimnull T1=dimnull T2\dim \text{null }T_1 = \dim \text{null }T_2.


Suppose dimnull T1=dimnull T2\dim \text{null }T_1 = \dim\text{null }T_2. Let R:null T1null T2R : \text{null }T_1 \to \text{null }T_2 be an isomorphism then extend RR to all of VV in such a way that it remains invertible (see 3d/3). This gives

null T1=null T2R\text{null }T_1 = \text{null }T_2R

Now let v1,,vrv_1,\dots,v_r be independent vectors in VV such that wk=T1vkw_k = T_1v_k is a basis for range T1\text{range }T_1.
Define wk=T2Rvkw_k' = T_2Rv_k, w1,,wrw_1',\dots,w_r' are independent since

a1w1++arwr=0    T2R(a1v1++arvr)=0a_1w_1' + \dots + a_rw_r' = 0 \iff T_2R(a_1v_1 + \dots + a_rv_r) = 0

Define Swk=wkSw_k' = w_k

T1=ST2RT_1 = ST_2R

as desired, RR is an isomorphism between nullspaces and SS is an isomorphism between ranges.

Now suppose T1=ST2RT_1 = ST_2R. let u1,,unu_1,\dots,u_n be a basis for null T1\text{null }T_1 and notice how Ru1,,RunRu_1,\dots,Ru_n is a basis for null T2\text{null }T_2 (SS can be ignored as it's injective so ST2Rv=0ST_2Rv = 0 iff T2Rv=0T_2Rv = 0.)