The product \(P\) of two linear transformations \(A\) and \(B\) , \(P=A B\) , is defined by the equation \(P x=A(B x)\) .

The notion of multiplication is fundamental for all that follows. Before giving any examples to illustrate the meaning of transformation products, let us observe the implications of the symbolism, \(P=A B\) . To say that \(P\) is a transformation means, of course, that given a vector \(x\) , \(P\) does something to it. What it does is found out by operating on \(x\) with \(B\) , that is, finding \(B x\) , and then operating on the result with \(A\) . In other words, if we look on the symbol for a transformation as a recipe for performing a certain act, then the symbol for the product of two transformations is to be read from right to left. The order to transform by \(A B\) means to transform first. by \(B\) and then by \(A\) . This may seem like an undue amount of fuss to raise about a small point; however, as we shall soon see, transformation multiplication is, in general, not commutative, and the order in which we transform makes a lot of difference.

The most notorious example of non-commutativity is found on the space \(\mathcal{P}\) . We consider the differentiation and multiplication transformations \(D\) and \(T\) , defined by \((D x)(t)=\frac{d x}{d t}\) and \((T x)(t)=t x(t)\) ; we have \[(D T x)(t)=\frac{d}{d t}(t x(t))=x(t)+t \frac{d x}{d t}\] and \[(T D x)(t)=t \frac{d x}{d t}.\] In other words, not only is it false that \(D T=T D\) (so that \(D T-T D=0\) ), but, in fact, \((D T-T D) x=x\) for every \(x\) , so that \(D T-T D=1\) .

On the basis of the examples in Section: Linear transformations , the reader should be able to construct many examples of pairs of non-commutative transformations. Those who are used to thinking of linear transformations geometrically can, for example, readily convince themselves that the product of two rotations of \(\mathbb{R}^{3}\) (about the origin) depends in general on the order in which they are performed.

Most of the formal algebraic properties of numerical multiplication (with the already mentioned notable exception of commutativity) are valid in the algebra of transformations. Thus we have

\begin{align} A 0 & =0 A=0, \tag{1}\\ A 1 & =1 A=A, \tag{2}\\ A(B+C) & =A B+A C, \tag{3}\\ (A+B) C & =A C+B C, \tag{4}\\ A(B C) & =(A B) C. \tag{5} \end{align} 

The proofs of all these identities are immediate consequences of the definitions of addition and multiplication; to illustrate the principle we prove (3), one of the distributive laws. The proof consists of the following computation:

\begin{align} \big(A(B+C)\big) x &= A\big((B+C) x\big)\\ &= A(B x+C x) \\ &= A(B x)+A(C x) \\ &= (A B) x+(A C) x \\ &= (A B+A C) x \end{align}