Notion of a basis that allows infinite sums

78 Views Asked by At

So with no other structure it is clear that a vector space has no notion of convergence and thus infinite sums make no sense. However, suppose $V$ is an inner product space, then we have a norm which induces a topology on $V$ so we can talk about convergence. Is there then a notion of a basis which allows infinite sums?

My question is mainly motivated by statements I see by physicists in introductory quantum mechanics textbooks. For example, I have seen physicist claim that the vector space of solutions to the time independent Schroedinger equation confined to an infinite square well of width $a$ has a countable basis given by: $$\psi_n(x)=\sqrt{\frac{2}{a}}\sin\left(\frac{n\pi}{a}x\right)$$ and that every solution can be written as an infinite sum of these functions for certain constants $c_n\in \mathbb{R}$. So is this just abuse of language or is there actually a way of defining a basis that allows for such infinite sums?

Moreover, I have seen physicists talk about an uncountably infinite basis of functions parameterized by $\mathbb{R}$ where any solution to some PDE can be written as an integral over this basis for some choice of square integrable function $f$. Generally this is talked about in the context of Fourier transformations, which I am not crazy familiar with outside of what was covered in my undergraduate real analysis course. Does it also make sense to call this a basis in this context or is it again abuse of language?

Edit: I see in the tags that my first question is yes, and that it is called a Schauder Basis. So my main question is can we extend this notion to an uncountable basis when it makes sense to talk of integration.

1

There are 1 best solutions below

0
On BEST ANSWER

In Hilbert spaces we have available the notion of an orthonormal or Hilbert basis, which is not a basis in the usual algebraic sense (sometimes called a Hamel basis in this context to distinguish it) but allows infinite sums which converge with respect to the norm. The $\psi_n$ are an orthonormal basis in this sense. These figure prominently in the spectral theorem for compact operators. This is not an abuse of language, just a generalization of the meaning of "basis" suitable to this context.

The uncountable stuff is formalized mathematically by the concept of a direct integral. I think most mathematicians would consider the use of the word "basis" here an abuse of language but maybe physicists don't, I don't know. This sort of thing is needed for the spectral theorem for non-compact operators.