0

Let $ V $ be an inner product space. Assume $ dim(V)=\infty $

prove/disprove $ \left(U\cap W\right)^{\perp}=U^{\perp}+W^{\perp} $

Well, I proved that this statement is correct if $ dim\left(v\right)<\infty $. And I have seen a post here with the same question, but I could not understand the counterexample that was provided there, since it used the term "Hilbert space" and we have'nt learned about it yet. Is there a simple counter example for this statement? Thanks.

FreeZe
  • 3,735
  • Could you please provide the link for the post that you found? Note that a "Hilbert space" is just a type of infinite dimensional inner product space. – Ben Grossmann Jun 15 '20 at 14:17
  • https://math.stackexchange.com/questions/899273/does-s-bott-bot-s-cap-t-bot-hold-in-infinite-dimensional-spaces – FreeZe Jun 15 '20 at 14:31
  • I found an easier example in the answer here. Do you need anything from there explained? – Ben Grossmann Jun 15 '20 at 14:33
  • If we consider the space of functions $ f = A_0+\sum_{n=1}^\infty(A_n\cos nx + B_n\sin nx)$ with inner product $\langle f , g \rangle = \int_{-\pi}^{\pi}\bar{f}(x)g(x)dx$ and set $ U = span{\sin x}$, $V = span{ \cos x }$, does that not give a counter example? edit: no of course it does not. Check out @Jim 's answer. It's the simplest one I've seen. – Enforce Jun 15 '20 at 14:43

1 Answers1

1

Let $V$ be the vector space of sequences $x = (x_1, x_2, \ldots)$ of real numbers such that $\sum_ix_i^2$ converges, the usual dot product gives an inner product. Let $U$ be the set of sequences for which only finitely many of the $x_i$ are nonzero and let $W$ be the set of sequences of the form $\left(\frac{a}{1}, \frac{a}{2}, \frac{a}{3}, \ldots\right)$ for some $a$.

Now $U^\perp = 0$ because anything in $U^\perp$ must be orthogonal to all the standard basis vectors $(1, 0, 0, \ldots)$, $(0, 1, 0, \ldots)$, $\ldots$, hence it must be zero in every coordinate. Also $U \cap W = 0$ because every nonzero vector in $W$ is nonzero in infinitely many places. So $(U \cap W)^\perp = U^\perp + W^\perp$ reduces to $0^\perp = 0 + W^\perp$ and then $V = W^\perp$. But this last equation is not true, for example all those standard basis vectors are not in $W^\perp$.

Jim
  • 30,682