Suppose w1,,wnw_1,\dots,w_n is a basis of WW and VV is finite-dimensional. Suppose TL(V,W)T \in \mathcal L(V,W). Prove that there exists a basis v1,,vmv_1,\dots,v_m of VV such that all the entries in the first row of M(T)\mathcal M(T) (with repsect to the bases v1,,vmv_1,\dots,v_m and w1,,wnw_1,\dots,w_n) are 00 except for possibly a 11 in the first row, first column.

(In this exercise, unlike 3c/3, you are given the basis of WW instead of being able to choose a basis of WW.)


If w1range T1w_1 \in \text{range }T_1 set v1v_1 so Tv1=w1Tv_1 = w_1, otherwise (TODO)
then extend to a basis v1,,vnv_1,\dots,v_n of VV. For each k>1k > 1 we have

T(vk)=ckw1+T(v_k) = c_kw_1 + \dots

Meaning T(vkckv1)T(v_k - c_kv_1) has no w1w_1 component, so consider the list

(v1,v2c2v1,,vncnv1)(v_1,v_2-c_2v_1,\dots,v_n-c_nv_1)

It's linearly independent because a1v1++an(vncnv1)=0a_1v_1 + \dots + a_n(v_n - c_nv_1) = 0 implies

(a1a2c2ancn)v1+a2v2++anvn=0(a_1 - a_2c_2 - \dots - a_nc_n)v_1 + a_2v_2 + \dots + a_nv_n = 0

Since v1,,vnv_1,\dots,v_n are independent we have a2==an=0a_2 = \dots = a_n = 0 meaning the original equation is a1v1=0a_1v_1 = 0 thus a1=0a_1 = 0 as well. So our new list is a basis by 2.39, and we clearly have

T(a1v1+a2(v2c2v1)++an(vncnv1))=a1w1+b2w2++bmwmT(a_1v_1 + a_2(v_2 - c_2v_1) + \dots + a_n(v_n - c_nv_1)) = a_1w_1 + b_2w_2 + \dots + b_mw_m

As desired, meaning M(T)\mathcal M(T) with respect to the v1,,(vncnv1)v_1,\dots,(v_n-c_nv_1) basis will have the first row be (1,0,,0)(1,0,\dots,0).


This proof is effectively a matrix algorithm style proof, but written in terms of linear maps.