582 CHAPTER 22. HILBERT SPACES

If U is separable, then letting D be a countable dense subset, it follows from the conti-nuity of the operators L,L−1 discussed above that L(D) is separable in L(U). To see this,note that

∥Lxn−Lx∥L(U) =∥∥L(L−1 (Lxn−Lx)

)∥∥≤ ∥L∥L (U,H)

∥∥L−1 (L(xn− x))∥∥

U

≤ ∥L∥L (U,H) ∥xn− x∥U

As before, L−1 (L(xn− x)) is the smallest vector which maps onto L(xn− x) and so itsnorm is no larger than ∥xn− x∥U .

Consider the last claim. If L is one to one, then for y ∈ L(U) , there is only one vectorwhich maps to y. Therefore, L−1 (L(x)) = x.Hence for y ∈ L(U) ,∥y∥L(U) ≡

∥∥L−1 (y)∥∥

U .Also,

∥Lu∥L(U) ≡∥∥L−1 (L(u))

∥∥U ≡ ∥u∥U

Thus when L is one to one, ∥L∥L (U,L(U)) = 1. ■

22.3 Approximations in Hilbert SpaceThe Gram Schmidt process applies in any vector space which has an inner product.

Theorem 22.3.1 Let {x1, · · · ,xn} be a basis for M a subspace of H a Hilbert space.Then there exists an orthonormal basis for M, {u1, · · · ,un} which has the property that foreach k ≤ n, span(x1, · · · ,xk) = span(u1, · · · ,uk) . Also if {x1, · · · ,xn} ⊆ H, then the finitedimensional subspace span(x1, · · · ,xn) is a closed subspace.

Proof: Let {x1, · · · ,xn} be a basis for M. Let u1 ≡ x1/ |x1| . Thus for k = 1, span(u1) =span(x1) and {u1} is an orthonormal set. Now suppose for some k < n, u1, · · · , uk havebeen chosen such that (u j ·ul) = δ jl and span(x1, · · · ,xk) = span(u1, · · · ,uk). Then define

uk+1 ≡xk+1−∑

kj=1 (xk+1 ·u j)u j∣∣∣xk+1−∑kj=1 (xk+1 ·u j)u j

∣∣∣ , (22.11)

where the denominator is not equal to zero because the x j form a basis and so

xk+1 /∈ span(x1, · · · ,xk) = span(u1, · · · ,uk)

Thus by induction,

uk+1 ∈ span(u1, · · · ,uk,xk+1) = span(x1, · · · ,xk,xk+1) .

Also, xk+1 ∈ span(u1, · · · ,uk,uk+1) which is seen easily by solving 22.11 for xk+1 and itfollows

span(x1, · · · ,xk,xk+1) = span(u1, · · · ,uk,uk+1) .

If l ≤ k,

(uk+1 ·ul) =C

((xk+1 ·ul)−

k

∑j=1

(xk+1 ·u j)(u j ·ul)

)

=C

((xk+1 ·ul)−

k

∑j=1

(xk+1 ·u j)δ l j

)=C ((xk+1 ·ul)− (xk+1 ·ul)) = 0.

582 CHAPTER 22. HILBERT SPACESIf U is separable, then letting D be a countable dense subset, it follows from the conti-nuity of the operators L,L~! discussed above that L(D) is separable in L(U). To see this,note thatI|Lxn — Lx||7(y) IL (L7! (Lan — Lx)) || < El. guy |B (n—X))|lylAILI gH) IlXn — llyAs before, L~! (L(x, —x)) is the smallest vector which maps onto L(x, —x) and so itsnorm is no larger than |x; —x!||y.Consider the last claim. If L is one to one, then for y € L(U), there is only one vectorwhich maps to y. Therefore, L~! (L(x)) = x.Hence for y € L(V), llyllzuy = ||Z7!Also,ylLelie) = WE" Ly = lel1Thus when L is one to one, i(U,L(U)) ~22.3. Approximations in Hilbert SpaceThe Gram Schmidt process applies in any vector space which has an inner product.Theorem 22.3.1 Lez {x1,°++ Xn} be a basis for M a subspace of H a Hilbert space.Then there exists an orthonormal basis for M, {u,,-++ ,Un} which has the property that foreach k <n, span(x1,--+ ,x~) = span(uy,--- ug). Also if {x1,-++ ,xXn} CH, then the finitedimensional subspace span (x,,--+ ,Xn) is a closed subspace.Proof: Let {x),--- ,x,} be a basis for M. Let uy =x, / |x|. Thus for k = 1, span(u) =span (x;) and {u;} is an orthonormal set. Now suppose for some k <n, uy, ---, ug havebeen chosen such that (uj; - uj) = 6 ;) and span (x,--- ,x,) = span(u),--- ,u,). Then definekNet ~ DL jay (hk 1 Uz) UjWey) = Djinn (ev Ms) Mi (22.11)Net — Diy ee uj) ujwhere the denominator is not equal to zero because the x; form a basis and soXe € span (x1,--+ xXx) = span (uy,--- , uz)Thus by induction,Ugy-1 € Span (Uy ,++* Me, Xk +1) = Span (X1,-+* Xk, ARH) +Also, xg41 © span (u1,-++ ,Ug,Ue41) which is seen easily by solving 22.11 for xz, and itfollowsspan (x1 oe Xk; Xk+1) = Span (uy ce Uk, Uk+1) .If i <k,k(Ugp1-uy) =C (\ Xky1 Uy) ~ hI Xe41 + Uj) ora)j=lk=e (tu ur) — (Xen Uj 13) =C((Xer1 Wr) — epi ur) = 0.j=!