Let k=dimker(T) and choose a basis
u1,…,ukfor ker(T). Extend it to a basis of V:
u1,…,uk,v1,…,vr.The vectors T(v1),…,T(vr) span im(T). Indeed, if x∈V, write
x=a1u1+⋯+akuk+b1v1+⋯+brvr.Applying T gives
T(x)=b1T(v1)+⋯+brT(vr),because T(ui)=0 for each i.
The vectors T(v1),…,T(vr) are linearly independent. If
c1T(v1)+⋯+crT(vr)=0,then
T(c1v1+⋯+crvr)=0,so c1v1+⋯+crvr∈ker(T). It is therefore a linear combination of u1,…,uk. Since the full list
u1,…,uk,v1,…,vris linearly independent, c1=⋯=cr=0. Thus T(v1),…,T(vr) is a basis of the image, so rank(T)=r. The basis of V has k+r vectors, giving
rank(T)+nullity(T)=r+k=dim(V).