Matrix Product of Vectors

0.- Consider a three-dimensional vector space with

A,B 3-d vectors,

and orthonormal basis: {e1, e2, e3}.

1.- Given B, neither A·B nor A×B define A in a unique way. Let us combine both defining the non-commutative "Clifford product" of two vectors, with the following two axioms:

Axiom 1: The Clifford (or matrix) product of a vector with itself is the same as the dot product:

DD = D·D = D2 a scalar.

Axiom 2: The Clifford product is associative:

(AB)C = A(BC).

2.- Working directly with the basis, let us choose

D = e1 + e2, to find immediately that:

e1e2 = -e2e1,

and similarly, e2e3 = -e3e2, and e3e1 = -e1e3. In general,

AB + BA = 2 A·B,

for the symmetric part of the Clifford product.

On the other hand, the Clifford product e1e2e3 (= e3e1e2 = e2e3e1) is such that:

(e1e2e3)2 = e1e2e3e1e2e3 = -e1e2e1e3e2e3 = e2e3e2e3 = -1,

and it commutes with e1, e2, and e3:

(e1e2e3)e1 = e1e2e3e1 = -e1e2e1e3 = e1e1e2e3 = e1(e1e2e3), and so on.

Hence, we can identify e1e2e3 = i, the imaginary unit.

3.- We can thus rewrite e1e2 = ie3, e2e3 = ie1, and e3e1 = ie2, since, e.g.

e1e2 = e1e2(e3e3) = (e1e2e3)e3 = ie3.

All possible products yield the basis

{1, e1, e2, e3, i, ie1, ie2, ie3}

of an eight-dimensional linear space of "cliffors" allowing for linear combinations of complex scalars and complex vectors:

cliffor = a + ib + A + iB,           a,b real numbers,     A,B real 3-d vectors,

with i = √-1.

Since e1e2 = ie1×e2 = -e2e1, the antisymmetric part of the Clifford product of A and B is given by iA×B, so finally:

AB = A·B + iA×B.

In general, the Clifford product of two cliffors is:

(u + U)(v + V) = uv + uV + vU + U·V + iU×V,

where u, v are complex numbers and U, V are complex vectors.

Using this definition, a product of two cliffors is again a cliffor, i.e. the Clifford algebra is closed with respect to the product.

This product allows us to find the inverse of a vector, and in general, the inverse of a cliffor:

AA = A2, so A-1 = A(A2)-1 = (A2)-1A, and

(u + U)-1 = (u - U)/(u2U2),

whenever the (numerical) denominator does not vanish.