Let A be a commutative field endowed with an automorphism J, Ei,Fi finite-dimensional vector spaces over A, and Φi a sesquilinear form for J on Ei×Fi (1⩽i⩽m). If the forms Φi are non-degenerate, the same is true of their tensor product Φ. In this case the inverse form Φ of Φ is the tensor product of the inverse forms Φi.
Indeed, since A is a field, it follows from props. 6 and 7 of chap. III, § 1, no. 3 that a tensor product of injective (resp. surjective) linear mappings of A-modules is an injective (resp. surjective) linear mapping. Since the sΦi are bijective by assumption (prop. 6, no. 6), the same is therefore true of their tensor product. On the other hand the canonical mapping js of ⨂iFi∗ into (⨂iFi)∗ is bijective (chap. III, § 1, no. 5, prop. 11). Therefore, by virtue of (36), sΦ is bijective, and this establishes our first assertion (prop. 6, no. 6). Analogously dΦ is bijective.
In the second assertion we have implicitly identified ⨂iFi∗ with (⨂iFi)∗ and ⨂iEi∗ with (⨂iEi)∗ by means of the mappings js and jd, which are here isomorphisms. The inverse forms in the statement exist since the sΦi, the dΦi, sΦ and dΦ are bijective (no. 7). Let us then put x′=x1′⊗⋯⊗xm′, y′=y1′⊗⋯⊗ym′ (xi′∈Ei∗, yi′∈Fi∗, i=1,…,m). By definition of the inverse forms, and viewed (36), we have
of Em×(Fj)m into A is A-multilinear. It therefore defines a bilinear form Φ′ on (⨂mE)×(⨂mFj) characterized by
Φ′(x1⊗⋯⊗xm,y1⊗⋯⊗ym)=det(Φ(xi,yk)).
Since the first member is zero when xi=xk or when yi=yk (i=k), Φ′ defines, by passing to the quotients, a bilinear form on (∧mE)×(∧mFj), or again, since ∧mFj is identified with (∧mF)j, a sesquilinear form Φ(m) for J on (∧mE)×(∧mF). This one is characterized by