5.2. THE MATRIX OF A LINEAR TRANSFORMATION 103
where δ ik = 1 if i = k and 0 if i ̸= k. I will show that L ∈L (V,W ) is a linear combinationof these special linear transformations called dyadics, also rank one transformations.
Then let L ∈L (V,W ). Since {w1, · · · ,wm} is a basis, there exist constants, d jk suchthat
Lvr =m
∑j=1
d jrw j
Now consider the following sum of dyadics. ∑mj=1 ∑
ni=1 d jiw jvi. Apply this to vr. This yields
m
∑j=1
n
∑i=1
d jiw jvi (vr) =m
∑j=1
n
∑i=1
d jiw jδ ir =m
∑j=1
d jrw j = Lvr (5.1)
Therefore, L = ∑mj=1 ∑
ni=1 d jiw jvi showing the span of the dyadics is all of L (V,W ) .
Now consider whether these special linear transformations are a linearly independentset. Suppose
∑i,k
dikwivk = 0.
Are all the scalars dik equal to 0?
0 = ∑i,k
dikwivk (vl) =m
∑i=1
dilwi
and so, since {w1, · · · ,wm} is a basis, dil = 0 for each i = 1, · · · ,m. Since l is arbitrary, thisshows dil = 0 for all i and l. Thus these linear transformations form a basis and this showsthat the dimension of L (V,W ) is mn as claimed because there are m choices for the wi andn choices for the v j. ■
Note that from 5.1, these coefficients which obtain L as a linear combination of thediadics are given by the equation
m
∑j=1
d jrw j = Lvr (5.2)
Thus Lvr is in the span of the w j.
5.2 The Matrix of a Linear TransformationIn order to do computations based on a linear transformation, we usually work with itsmatrix. This is what is described here.
Theorem 5.1.4 says that the rank one transformations defined there in terms of twobases, one for V and the other for W are a basis for L (V,W ) . Thus if A ∈L (V,W ) , thereare scalars Ai j such that
A =n
∑i=1
m
∑j=1
Ai jwiv j
Here we have 1≤ i≤ n and 1≤ j≤m. We can arrange these scalars in a rectangular shapeas follows.
A11 A12 · · · A1(n−1) A1n
A21 A22 · · · A2(n−1) A2n...
......
...Am1 Am2 · · · Am(n−1) Amn