A more complicated example is provided by spaces of square integrable complex functions
on some intervalS(or all of) the real lineR, defined by
L^2 (S)≡{f:S→C; (f,f)<∞} (f,g)≡
∫
S
dxf(x)∗g(x) (3.20)
These spaces will be ubiquitous in quantum mechanics. The Fourier transform gives a
convenient way of describing L^2 (S). For example, on an interval S = [−πℓ,+πℓ] with
periodic boundary conditions (or equivalently a circe of radiusℓ), we have
f(x) =
∑
m∈Z
fm
eimx/ℓ
√
2 πℓ
(f,g) =
∑
m∈Z
fm∗gm (3.21)
wheregmare the Fourier components ofg. This shows thatL^2 andL^2 (S) are isomorphic.
The basis forL^2 (S) used here corresponds to
|m〉∼
eimx/ℓ
√
2 πℓ
(3.22)
and is orthonormal in view of the relation,
∫+πℓ
−πℓ
dx
(
eimx/ℓ
)∗
einx/ℓ= 2πℓδm,n (3.23)
which is a standard relation of Fourier analysis.
The Hilbert spaceL^2 (R) must be handled with additional care. Fourier analysis (as well
as the limitℓ→ ∞of the above example) suggests a basis given by exponentials exp(ikx)
withk∈R. The problem is that these basis vectors are not square integrableand form a
basis which isnot countable. Nonetheless, countable bases forL^2 (R) do exist. A familiar
example is provided by the orthonormal basis of wave functions forthe harmonic oscillator,
Hn(x)e−x
(^2) / 2
whereHn(x) are the Hermite polynomials. Equivalently, one can work with a
non-orthonormal but simpler basis given byxne−x
(^2) / 2