Annex
Complements on Hilbert spaces
1. Trace of a quadratic form with respect to another(1)
In this No., we denote by E a real vector space and by Q,H two positive quadratic forms on E. There exist two symmetric bilinear forms (x,y)↦(x∣y)Q and (x,y)↦(x∣y)H characterized by
Q(x)=(x∣x)Q,H(x)=(x∣x)H for all x∈E.
One calls trace of Q with respect to H and one denotes by Tr(Q/H) the positive real number, finite or not, defined as follows:
a) If there exists an x∈E such that H(x)=0 and Q(x)=0, one sets Tr(Q/H)=+∞.
b) In the contrary case, Tr(Q/H) is the supremum of the set of numbers of the form ∑i=1pQ(ei), where (e1,…,ep) runs over the set of finite sequences of elements of E orthonormal for H.
Let E be a real Hilbert space and Q a positive quadratic form on E. Set H(x)=∥x∥2 for all x∈E; then H is a positive quadratic form on E. One says that Q is nuclear if Tr(Q/H) is finite. For every x∈E of norm 1, one has Q(x)≤Tr(Q/H), whence Q≤Tr(Q/H)⋅H; in particular, every nuclear form Q is continuous.
Proposition 1
— *Assume that E is finite-dimensional and H is nondegenerate.
a) There exists an endomorphism u of E characterized by (x∣y)Q=(u(x)∣y)H for x, y in E.
b) Tr(Q/H)=Tr(u).
c) Tr(Q/H)=∑i=1mQ(ei) for every basis (e1,…,em) of E orthonormal for H.
a) follows from the fact that the bilinear form (x,y)↦(x∣y)H is non-degenerate. Every sequence in E orthonormal for H may be completed to a basis of E orthonormal for H. Consequently, Tr(Q/H) is the supremum of the set of numbers of the form ∑i=1mQ(ei) over all bases (e1,…,em) of E orthonormal for H. To prove b) and c), it suffices to show that ∑i=1mQ(ei)=Tr(u) for every basis of this kind. Set aij=(u(ei)∣ej)H=(ei∣ej)Q for 1≤i,j≤m; then u(ei)=∑j=1maijej for 1≤i≤m, whence
Tr(u)=i=1∑maii=i=1∑m(ei∣ei)Q=i=1∑mQ(ei).
Q.E.D.
Proposition 2
— Assume that E is finite-dimensional. There exist a basis (e1,…,en) of E and an integer m with 0≤m≤n *such that
H(i=1∑ntiei)=i=1∑mti2
for t1,…,tn real. If, moreover, the relation H(x)=0 implies Q(x)=0 for x∈E, then Tr(Q/H)=∑i=1mQ(ei).
There exists a basis (e1′,…,en′) of E orthogonal for H. One can assume the basis to be indexed in such a way that H(ei′)>0 for 1≤i≤m and H(ei′)=0 for m<i≤n. Then set ei=ei′/H(ei′)1/2 for 1≤i≤m and ei=ei′ for m<i≤n; the relation (1) is satisfied.
Let F be the subspace of E generated by em+1′,…,en′; it is the set of x∈E such that H(x)=0. Denote by π the canonical mapping of E onto E1=E/F. Since Q and H are zero on F, there exist two positive quadratic forms Q1 and H1 on E1 such that Q=Q1∘π and H=H1∘π. Moreover, (π(e1),…,π(em)) is a basis of E1 orthonormal for H1, and therefore H1 is nondegenerate.
By Prop. 1 and Remark 2,
Tr(Q/H)=Tr(Q1/H1)=i=1∑mQ1(π(ei))=i=1∑mQ(ei).
Q.E.D.
2. Hilbert–Schmidt mappings(2)
Let E be a real Hilbert space, in which the scalar product is denoted by (x∣y). There exists an isometry jE of E onto its dual, characterized by the formula
(3)
(x∣y)=⟨x,jE(y)⟩for x,y in E (TVS, V, §1, No. 7, Th. 3).
Let E1 and E2 be two real Hilbert spaces and u a continuous linear mapping of E1 into E2. One calls adjoint of u the continuous linear mapping u∗=jE1−1∘tu∘jE2 of E2 into E1. The mapping u∗ is characterized by the relation
(4)
(u(x1)∣x2)=(x1∣u∗(x2))for x1∈E1, x2∈E2.
If v is a continuous linear mapping of E2 into a Hilbert space E3, one has (v∘u)∗=u∗∘v∗.
Let E1 and E2 be two real Hilbert spaces and u a linear mapping of E1 into E2. One defines two positive quadratic forms H and Q on E1 by the formulas
H(x)=∥x∥2,Q(x)=∥u(x)∥2(x∈E1).
Proposition 3
Suppose u is continuous. Let (ei)i∈I be an orthonormal basis of E1 and (fj)j∈J an orthonormal basis of E2. Then
Tr(Q∣H)=i∈I∑∥u(ei)∥2=j∈J∑∥u∗(fj)∥2=i∈I∑j∈J∑(u(ei)∣fj)2.
(2) Cf. TVS, V, §4, No. 7.
For every x∈E1, we have ∥x∥2=∑i∈I(x∣ei)2, and similarly ∥y∥2=∑j∈J(y∣fj)2 for every y∈E2. Consequently,
i∈I∑∥u(ei)∥2=i∈I∑j∈J∑(u(ei)∣fj)2=j∈J∑i∈I∑(ei∣u∗(fj))2=j∈J∑∥u∗(fj)∥2.
In particular, the number ∑i∈I∥u(ei)∥2 is independent of the orthonormal basis (ei)i∈I of E1.
Set t=Tr(Q∣H). For every finite subset I′ of I, by definition
i∈I′∑∥u(ei)∥2=i∈I′∑Q(ei)≤t,
whence ∑i∈I∥u(ei)∥2≤t. Let (e1′,…,ep′) be a finite orthonormal sequence in E. This sequence can be completed to an orthonormal basis (eα′)α∈A of E1. Then
α=1∑p∥u(eα′)∥2≤α∈A∑∥u(eα′)∥2=i∈I∑∥u(ei)∥2
and, passing to the supremum over all (e1′,…,ep′), one obtains the inequality t≤∑i∈I∥u(ei)∥2. We have thus established the equality t=∑i∈I∥u(ei)∥2.
Q.E.D.
One says that u is a Hilbert–Schmidt mapping of E1 into E2 if the positive quadratic form Q:x↦∥u(x)∥2 on E1 is nuclear. When this is so, one has Q≤Tr(Q/H)⋅H, therefore u is continuous and
∥u∥≤Tr(Q/H)1/2.
Let u:E1→E2 be a continuous linear mapping. By Prop. 3, u is a Hilbert–Schmidt mapping if and only if there exists an orthonormal basis (ei)i∈I of E1 such that ∑i∈I∥u(ei)∥2<+∞. When this is so, every orthonormal basis of E1 has the same property. Moreover, if u is a Hilbert–Schmidt mapping, then so is its adjoint u∗ by virtue of the formula ∑i∈I∥u(ei)∥2=∑j∈J∥u∗(fj)∥2 (Prop. 3).
Exercises
Exercises
See the exercises for Appendix 0.