Reason: Initial publication. Existence proof by induction on the length, subtracting the orthogonal projection at each step and normalising; non-vanishing of the new vector comes from the predecessor claim on linear independence.
Normalisation. Let xβV with xξ =0Vβ, where 0Vβ is the zero vector. By the positivity of a norm, β₯xβ₯ is a real number with 0β€β₯xβ₯ and β₯xβ₯ξ =0, so Ξ»=β₯xβ₯β1 exists in the field of complex numbers. By the absolute homogeneity of a norm, then (M8) applied to the nonnegative real number β₯xβ₯, then (M4),
By (M4) again, β£1β£=β£1β£β£1β£, and β£1β£ξ =0 by (M3) because 1ξ =0; multiplying by the inverse of β£1β£ gives β£1β£=1. Hence Ξ»x is a unit vector. Note also Ξ»ξ =0, since Ξ»β₯xβ₯=1ξ =0.
The induction. Let A be the set of those qβN such that for every linearly independent cβVq there is an orthonormal fβVq with span(fβ£[k]β)=span(cβ£[k]β) for every kβ[q]. We show A=N by Principle of Induction for the Natural Numbers.
Base case. Let cβV1 be linearly independent. By (L2), c1βξ =0Vβ. Put Ξ»=β₯c1ββ₯β1 and let fβV1 have component f1β=Ξ»c1β. By normalisation f1β is a unit vector, and f is orthonormal because [1] contains no two distinct indices. Since f1β=Ξ»c1β lies in span(c) and c1β=Ξ»β1f1β lies in span(f), and both spans are linear subspaces by (P1), two applications of (P3) give span(f)=span(c). As [1] has the single element 1 and fβ£[1]β=f, cβ£[1]β=c, this is the required equality, so 1βA.
Induction step. Let qβA and let bβ²βVq+1 be linearly independent. By (L1) the restriction bβ²β£[q]β is linearly independent, so there is an orthonormal fβVq with
The vector w is not 0Vβ. By the definition of the span, pβspan(f)=span(bβ²β£[q]β). If w=0Vβ, then u=p, so bq+1β²ββspan(bβ²β£[q]β), contradicting (L2) applied to bβ² with q+1β[q+1].
Put ΞΌ=β₯wβ₯β1 and define eβVq+1 by ekβ=fkβ for kβ[q] and eq+1β=ΞΌw.
The tuple e is orthonormal. Each ekβ with kβ[q] is a unit vector because f is orthonormal, and eq+1β is a unit vector by normalisation. Let i,jβ[q+1] be distinct. If both lie in [q], then eiβ and ejβ are orthogonal because f is orthonormal. Otherwise exactly one of them equals q+1; for jβ[q], condition 3 of Complex Inner Product Space gives
The span equalities. For kβ[q] we have eβ£[k]β=fβ£[k]β, so those equalities are exactly the ones already obtained. There remains k=q+1, that is, span(e)=span(bβ²). By (P1) both spans are linear subspaces of V, and by (P1) each component of a tuple lies in its own span.
First, span(f)βspan(e): each fkβ=ekβ lies in span(e), so (P3) applies. Every component bkβ²β with kβ[q] lies in span(bβ²β£[q]β)=span(f)βspan(e), and
bq+1β²β=u=w+p=ΞΌβ1eq+1β+p,
where pβspan(f)βspan(e) and ΞΌβ1eq+1ββspan(e) by condition 3 of Linear Subspace; hence bq+1β²ββspan(e) by condition 2 of that definition. So every component of bβ² lies in span(e), and (P3) gives span(bβ²)βspan(e).
Conversely, every component bkβ²β with kβ[q] lies in span(bβ²), so (P3) gives span(bβ²β£[q]β)βspan(bβ²), whence span(f)βspan(bβ²) and in particular ekβ=fkββspan(bβ²) for kβ[q], and pβspan(bβ²). Since bq+1β²ββspan(bβ²) and span(bβ²) is a linear subspace, conditions 2 and 3 of Linear Subspace together with claim 5 of Elementary Identities in a Vector Space give w=bq+1β²β+(β1)pβspan(bβ²) and then eq+1β=ΞΌwβspan(bβ²). So every component of e lies in span(bβ²), and (P3) gives span(e)βspan(bβ²).