Composition algebras and \(PG(m,2)\) (Q1339782)

The three-dimensional vector space \(V(3,2)\) over the field \(GF(2)= \mathbb{F}_2\) may be used to label an orthonormal basis \(\{e_x\}\) of the Cayley algebra such that \(e_x e_y= \pm e_{x+y}\) for all \(x,y\in V(3,2)\). Here the seven non-zero elements of \(V(3,2)\) are represented by the seven points of the Fano plane \(PG(2,2)\) which is equipped with a specific orientation. Then the relation \(e_x e_y= \pm e_{x+y}\) can be written as \(e_x e_y= (-1)^{f(x, y)} e_{x+y}\), where \(f\) is an \(\mathbb{F}_2\)-valued function of two variables from \(V(3,2)\). Now let \(\{e_x\mid x\in V(m+ 1,2) \}\) be a basis of a real \(2^{m+1}\)-dimensional algebra \({\mathcal A}_f\) with multiplication \(e_x e_y= (-1)^{f (x,y)} e_{x+y}\), \(x,y\in V(m+ 1,2)\), where \(f\) is some \(\mathbb{F}_2\)-valued function on \(V\times V\). Assume \({\mathcal A}_f\) is endowed with a quadratic form \(Q\). Then certain properties of the algebra \({\mathcal A}_f\) correspond to certain properties of the function \(f\). E.g., \({\mathcal A}_f\) is commutative if and only if \(f\) is a symmetric function. The authors investigate the conditions under which \({\mathcal A}_f\) is a composition algebra. They determine all functions which give rise to a split or non-split Cayley algebra. The properties of these functions are expressed in terms of the left and right ``linearizers'' of \(f\), which are defined by \(l(x, y,z)= f(x+ y,z)+ f(x, z)+ f(y, z)\), and \( r(x, y,z)= f(x, y+z)+ f(x, y)+ f(x, z)\) for all \(x, y, z\in V(m+ 1,2)\). With a certain function \(f_0\), \({\mathcal A}_f\) is isomorphic to a Cayley algebra if and only if \(f= f_0+s\) or \(f= f^t_0+ s\), where \(s\) is a symmetric bilinear function. If \(s\) is alternating, then \({\mathcal A}_f\) is a non-split Cayley algebra, and \({\mathcal A}_f\) is a split Cayley algebra if \(s\) is not alternating. In the case that \({\mathcal A}_f\) is non-split the invariance group of \(f\) is a flag-transitive subgroup of \(GL(3,2)\) isomorphic to \(\mathbb{Z}_7\times \mathbb{Z}_3\). The results of the paper yield also a new proof of Hurwitz's theorem.