Let V be an inner product space, let W be a finite-dimensional subspace of V. Prove that if x∈/W, then there exists y∈W⊥ with ⟨x,y⟩=0.
Solution.
Recall that V=W⊕W⊥ (via orthogonal projection), so we can write
x=xW+xW⊥,
where xW∈W and xW⊥∈W⊥. Note that xW⊥=0: if this were not the case, then x=xW∈W, which is impossible. Thus, if we set y=xW⊥, then
⟨x,y⟩=⟨xW+y,y⟩=⟨xW,y⟩+⟨y,y⟩=∥y∥2>0.
The penultimate inequality follows from the fact that y is orthogonal to every vector in W, and ∥y∥ is strictly larger than 0 since y=0.
Problem 4
Let V be a finite-dimensional inner product space, let W be a subspace of V. Prove that V/W is isomorphic to W⊥.
Solution.
There are two different ways to do this:
Every finite-dimensional vector space over a field F is isomorphic to Fn for some n, so we can try proving dim(V/W)=dim(W⊥).
We can try to find a linear isomorphism T:W⊥→V/W.
Method 1
Let dimV=n and dimW=m. Recall that because V,W are both finite-dimensional that dim(V/W)=n−m. To show dim(W⊥)=n−m, we need to find a basis for W⊥.
Let {v1,…,vm} be an orthonormal basis (it's very important that we pick an orthonormal basis for this to work) for W, and extend it into an orthonormal basis β={v1,…,vm,…,vn} for V. We will show that {vm+1,…,vn} is a basis for W⊥.
First, the set {vm+1,…,vn} is linearly independent since β was. Next, we need to show that it spans W⊥. Let v∈W⊥. Since β was an orthonormal basis for V, we can write
v=⟨v,v1⟩v1+⋯+⟨v,vm⟩vm+⋯+⟨v,vn⟩vn.
But v∈W⊥ and v1,…,vm∈W⊥, so ⟨v,v1⟩=⋯=⟨v,vm⟩=0. Thus,
v=⟨v,vm+1⟩vm+1+⋯+⟨v,vn⟩vn.
This means v∈span{vm+1,…,vn}, so the set spans. Thus, this set is a basis for W⊥ and has n−m vectors, so dimW⊥=n−m.
Method 2
Define T:W⊥→V/W by
T(v)=v+W.
This map is linear: given u,v∈W⊥ and a scalar c∈F, the definitions of addition and scalar multiplication in V/W tell us
T(cu+v)=(cu+v)+W=c(u+W)+(v+W)=cT(u)+T(v),
so T is linear. Let's show that T is one-to-one: let v∈N(T). Then
v+W=T(v)=W,
so v∈W. But v∈W⊥, so v∈W∩W⊥={0}, and so v=0. This shows that N(T)={0}, so T is one-to-one.
Now let's show that T is onto. Let v+W∈V/W. Note that we can't just say that
T(v)=v+W.
This is because v does not have to be in W⊥. However, if we could replace v with a vector y∈W⊥ such that v+W=y+W, then T(y)=v+W and we are done. The "natural" way to do this is to write
v=vW+vW⊥
where vW∈W and vW⊥∈W⊥ as in Problem 3. Thus, by definition of v+W, we have
v+W=(vW+vW⊥)+W=vW⊥+W,
since vW∈W. Hence, we have
T(vW⊥)=vW⊥+W=v+W,
so T is onto.
We have shown that T is an invertible linear map, so it's automatically an isomorphism.