Plancherel Theorem and Fourier Inversion Theorem

S. Kumaresan
School of Math. and Stat.
University of Hyderabad
Hyderabad 500046
kumaresa@gmail.com

1 Plancherel Theorem

Let f ∈ L1 (R). We define the Fourier transform Ff (x) := fˆ(x) := R f (t)e−ixt dt for x ∈ R.
R

The main results of these lectures are the Plancherel theorem which states that the linear
map F : L1 (R) ∩ L2 (R) → L2 (R) extends to an “isometry” of L2 (R) onto itself and that for
a continuous f ∈ L1 (R) with fˆ ∈ L1 (R) we have the Fourier inversion formula:
Z
f (x) = fˆ(y)eiyx dy for all x ∈ R.
R

We recall first the definition of a step function. A function f : R → C is called a step

function if it is a finite linear combination of characteristic (or indicator) functions of finite
intervals. Recall also that S, the space of step functions is dense in Lp (R) for all 1 ≤ p < ∞.
Thus it is natural to verify the assertion of the Plancherel theorem in the case of f := 1J , the
indicator function of a finite interval J = (a, b), [a, b), [a, b], (a, b]. Notice that whatever be the
form of J, the indicator functions 1J are all the same as elements of Lp . We now compute
the Fourier transform of f :
e−ibx − e−iax
Z Z b
−ixt
ˆ
f (x) = f (t)e dt = e−ixt dt = .
R a −ix
Also, we have
Z ∞ Z b Z b
kf k22 = 2
|f (t)| dt = 2
|1| dt = 1dt = b − a.
−∞ a a

We now check whether fˆ lies in L2 (R) and if so, compute its norm. Here we go:

e−ibx − e−iax 2 e−ibx − e−iax e−ibx − e−iax

|fˆ(x)|2 = | | = ( )
−ix −ix −ix
e−ibx − e−iax eibx − eiax 2 − (e−i(b−a)x + ei(b−a)x )
= =
−ix ix x2
2 − 2 cos(b − a)x 1 − cos(b − a)x
= 2
=2
x x2
2
sin ((b − a)x/2)
= 2·2 ,
x2

1
where in the last we have used a well-known trigonometric identity. We thus find:
Z ∞ Z ∞
ˆ
2
ˆ 2 sin2 ((b − a)x/2)
f (x) = |f (x)| dx = 4 dx.
2 −∞ −∞ x2

We put u := (b − a)x/2 so that the above becomes

sin2 u
Z
2 2 du
fˆ(x) = 4 2
(b − a)2
2 R 4u b −a
Z 2
sin u
= 2 2
(b − a) du
R u
= C(b − a).
sin2 u
R
Here we have let C stand for the real number 2 R u2
du.
2 2
Remark 1. Let us observe that ( C is a (finite) real number, i.e., sin u/u is integrable on
sin x/x if x 6= 0
R. For, the function g(x) := is continuous on R. Hence the continuous
1 if x = 0
function sin2 x/x2 is integrable over the finite interval [−1, 1] and it is dominated by the
continuous function 1/x2 on R \ [−1, 1] on which 1/x2 is integrable. Hence C is finite.

We now define F ∗ on S as follows:

Z
∗
F f (x) := f (t)eixt dt, for f ∈ S.
R

Proceeding as above, we find that kF ∗ f k22 = C kf k22 for f := 1(a,b) , for the same C.
If f := 1(a,b) and g := 1(c,d) , then we have:
Z d Z b Z b Z d
−ixt
hFf, gi = [ e dt]1dx = 1[ eixt dx] dt = hf, F ∗ gi .
c a a c

Thus on the indicator functions, F ∗ behaves like the adjoint of F. We now wish to extend
these results to f, g ∈ S. We observe that if f := 1(a,b) , and g := 1(b,c) , then
Z b Z c
−ixt
ˆ
(Ff + Fg)(x) = f (x) + ĝ(x) = e dt + e−ixt dt
a b
Z c
= e−ixt dt = F(f + g)(x).
a

Hence it follows that

Z Z c
kFf + Fg k22 = kF(f + g)k22 = | e−ixt dt|2 dx = C(c − a).
R a

But, since,
D E
kFf + Fg k22 = hFf + Fg, Ff + Fgi = kFf k22 + kFg k22 + 2 · Re fˆ, ĝ ,

2
 D E
we find that Re fˆ, ĝ = 0, for f and g as above. Similar result holds true also for F ∗ .
D E
Even if f := 1(a,b) and g := 1(c,d) with a ≤ b < c ≤ d we have Re fˆ, ĝ = 0. To see this,
D E D E
let h := 1(b,c) . Then using the earlier result, we have Re fˆ + ĥ, ĝ = 0 and Re ĥ, ĝ = 0.
D E
Subtracting the latter from the first, we get Re fˆ, ĝ = 0. Similarly for F ∗ .

We note that F1(a,b) satisfies: F1(a,b) (−x) = F1(a,b) (x), i.e., fˆ(−x) = fˆ(x):
Z b Z b
fˆ(−x) = e ixt
dt = e−ixt dt = fˆ(x).
a a

If f, g ∈ L2 (R) satisfy (∗), i.e., f (−x) = f (x) etc., then we have

Z Z Z
hf, gi = f (x)g(x)dx = f (x)g(x)dx = f (−x)g(−x)dx
Z
= f (x)g(x)dx = hf, gi .

where we have used the fact that the Lebesgue measure is invariant under x 7→ −x. This
observation, when applied to fˆ and ĝ for f := 1(a,b) and g := 1(c,d) , allows us to conclude
D E D E
fˆ, ĝ = 0 and hFf, Fgi = 0. That is, we can drop the prefix “Re” in Re fˆ, ĝ .

Now if f is any step function, say, of the form f = ni=1 ai 1Ji where Ji are finite intervals
P
PN
and ai ∈ C, we can write f = j−1 bj 1Ij , where Ij are pair-wise disjoint finite intervals. (It
is easier to convince yourself of this than writing down a formal verbose proof!) We then have
* +
2
X X
kFf k2 = bj F1Ij , bk F1Ik
j k
X
= bj bk C 1Ij , 1Ik
j,k
Z
2
X
= C |bj | 2
1Ij =C |f (x)|2 dx = C kf k22 .

Similarly, we have kF ∗ f k22 = C kf k22 , for f ∈ S. Also, by the bilinearity of the inner product
we have
hFf, gi = hf, F ∗ gi , for f, g ∈ S.

Thus we have linear maps F, F ∗ : S → L2 (R) such that i) kFf k22 = C kf k22 = kF ∗ f k22 ,
and ii) hFf, gi = hf, F ∗ gi for all f, g ∈ S. Since S is dense in L2 (R) and F, F ∗ are continuous
linear, we have unique extensions, denoted again by F and F ∗ , from L2 (R) to itself. This
follows from the following elementary result:

Lemma 2. Let T : D ⊂ E → F be a continuous linear map defined on a dense subspace D

of E to a Banach space F . Then T has a unique continuous linear extension T : E → F such
that T (E,F ) = kT k(D,F ) (operator norms).

3
Proof. We shall only sketch the proof.
For x ∈ E, take any xn ∈ D such that kx − xn k → 0. Then define T (x) := lim T xn which
exists since T xn is Cauchy in F (due to the uniform continuity of a continuous linear map!).
If yn ∈ D is such that kyn − xk → 0 then it can be easily seen that lim T yn = lim T xn so
that T x is well defined.

Hence we have kFf k22 = C kf k22 = kF ∗ f k22 and hFf, gi = hf, F ∗ gi for all f, g ∈ L2 (R),
by continuity of the inner product.
We also have
1
hf, gi = [kf + g k2 + i kf + ig k2 − kf − g k2 − i kf − ig k2 ]
4
1
= [kFf + Fg k2 + i kFf + iFg k2 − kFf − Fg k2 − i kFf − iFg k2 ]
4C
1 1
= hFf, Fgi = hF ∗ Ff, gi .
C C
The last equality is valid, as hh, Fgi = hF ∗ h, gi where h = Ff ∈ L2 (R).
We put g := F ∗ Ff − f ∈ L2 (R) in hf, gi = (1/C) hF ∗ Ff, gi to get
  2
1 1
0 = f − F ∗ Ff, g = f − F ∗ Ff = 0.
C C
That is, F ∗ Ff = Cf a.e. Similarly, FF ∗ f = Cf a.e. Thus we have proved the following
theorem:
Theorem 3 (Plancherel). Let S denote the dense subspace of the step functions in L2 (R).
Let F, F ∗ denote the Fourier and conjugate Fourier transforms defined as above. Then, for
C as above,
F and F ∗ map S into L2 (R); in fact, we have:

kFf k2 = C kf k2 = kF ∗ f k2 for f ∈ S.

F, F ∗ extend to an “isometry” of L2 (R) onto itself; that is, FF ∗ = C = F ∗ F on L2 (R).

2 Fourier Inversion Theorem

We may ask whether for f ∈ L1 (R) we have the formula fˆ(x) = R f (t)e−ixt dt and moti-
R

vated by the PlancherelR theorem whether for nice enough functions we can invert the Fourier
transform, i.e., f (t) = fˆ(x)eixt dx.
The first formula is not all obvious even if we assume that f ∈ L1 (R) ∩ L2 (R), as we have
extended F to L2 (R) by an abstract procedure. However, this is easy to justify: We start
with a non-negative f ∈ L1 (R) and take any sequence fn of step functions increasing to f .
We can now apply the monotone convergence theorem to conclude that Ff is given by the
above formula.
The proof of the second is given as the conclusion of the following theorem:

4
Theorem 4 (Fourier Inversion Theorem). Let f be a continuous function in L1 (R). Assume
that fˆ ∈ L1 (R). Then we have
Z
1
f (x) = fˆ(y)eiyt dy, for all x ∈ R.
2π R

Proof. The double integral fˆ(y)eixy dx = ( R f (t)e−iyt dt)eixy dx may not be absolutely
R R R

convergent (the trouble lies in the x-variable) and hence we cannot use Fubini to evaluate
it as an iterated integral. So, what we do, is to adopt a classical trick of introducing a
convergence factor in the x-variable. We take a “nice” function ψ such as a continuous
2
function with compact support with ψ̂ ∈ L1 (R), or ψ(y) := e−y or any function that “decays
2
rapidly at ∞”) with ψ(0) = 1. If you wish you may take ψ(y) = e−y in the following.
We have by dominated convergence theorem
Z Z
lim ψ(εy)f (y)e dy = fˆ(y)eixy dy.
ˆ ixy
(1)
ε→0

We unwind Eq. 1 and use Fubini on the LHS (of Eq. 1):
Z Z Z
lim ψ(εy)fˆ(y)eixy dy = lim ψ(εy)( f (t)e−iyt dt)eixy dy
ε→0 ε→0
Z Z R
= lim f (t)( ψ(εy)e−iy(t−x) dy) dt
ε→0
Z Z
iu
= lim f (t)( ψ(u)e− ε (t−x) du) dt where u = εy
ε→0
t − x dt
Z
= lim f (t)ψ̂( )
ε→0 ε ε
Z
= lim f (x + εv)ψ̂(v) dv where εv = t − x
ε→0
Z
= f (x) ψ̂(v)dv,

the last equality being in view of the continuity of f and dominated convergence theorem.
This completes the proof of the theorem, exceptR for an irritating but minor detail to be
attended to. For some ψ we need to compute ψ̂(v) dv, which in view of the conclusion of
the theorem should be nothing other than a constant times ψ(0). By computing the Fourier
2
transform of e−x , we can have satisfaction.
Remark 5. It is traditional to derive the Plancherel theorem from the Fourier inversion
theorem as follows:
Let f ∈ L1 (R) ∩ L2 (R). Take g(t) := f (−t). Then, f ∗ g is continuous and lies in L1 (R).
We have by the definition of convolution
Z Z
f ∗ g(0) = f (−t)g(t) dt = f (−t)f (−t) dt = kf k22 .

On the other hand, by the inversion formula, we have

Z Z Z
2
f ∗ g(0) = C f ∗ g(x) dx = C f (x)ĝ(x) dx = C fˆ(x)fˆ(x) dx = C fˆ
ˆ ˆ .
2

The Plancherel theorem follows from Eq. 5 and Eq. 5.

5
Acknowledgement: Lectures given at a Refresher Course for College teachers held in the
Department of Mathematics, University of Bombay in June 1991.