Bernstein’s inequality and Nikolsky’s inequality for
1 Complex Borel measures and the Fourier transform
Let be the set of complex Borel measures on . This is a Banach algebra with the total variation norm, with convolution as multiplication; for , we denote by the total variation of , which itself belongs to , and the total variation norm of is .
For , it is a fact that the union of all open sets such that itself satisfies . We define , called the support of .
For , we define by
It is a fact that belongs to , the collection of bounded uniformly continuous functions . For ,
(1) |
Let be Lebesgue measure on . For , let
which belongs to . We define by
The following theorem establishes properties of the Fourier transform of a complex Borel measure with compact support.11 1 Thomas H. Wolff, Lectures on Harmonic Analysis, p. 3, Proposition 1.3.
Theorem 1.
If and is compact, then and for any multi-index ,
For , if , then
Proof.
For , let be the th coordinate vector in , with length 1. Let , and define
We can write this as
For any ,
Because has compact support, . Furthermore, for each we have
Therefore, the dominated convergence theorem tells us that
On the other hand, for for and otherwise,
so
and in particular, . (The Fourier transform of a regular complex Borel measure on a locally compact abelian group is bounded and uniformly continuous.22 2 Walter Rudin, Fourier Analysis on Groups, p. 15, Theorem 1.3.3.) Because has compact support so does , hence we can play the above game with , and by induction it follows that for any ,
and in particular, .
Suppose that . The total variation of the complex measure is the positive measure
hence
Then using (1),
But we have already established that , which with the above inequality completes the proof. ∎
2 Test functions
For an open subset of , we denote by the set of those such that is a compact set. Elements of are called test functions.
It is a fact that there is a test function satisfying: (i) for , (ii) for , (iii) , and (iv) is radial. We write, for ,
For any multi-index ,
hence
(2) |
We use the following lemma to prove the theorem that comes after it.33 3 Thomas H. Wolff, Lectures on Harmonic Analysis, p. 4, Lemma 1.5.
Lemma 2.
Suppose that and for each . Then for each , in as .
Proof.
Let . In the case ,
Because , this tends to as .
Suppose that . The Leibniz rule tells us that with , we have, for each ,
For ,
Let . By (2), for we have
Thus
which tends to as . For any ,
and because , this tends to as . But
which completes the proof. ∎
Now we calculate the Fourier transform of the derivative of a function, and show that the smoother a function is the faster its Fourier transform decays.44 4 Thomas H. Wolff, Lectures on Harmonic Analysis, p. 4, Proposition 1.4.
Theorem 3.
If and for each , then for each ,
(3) |
There is a constant such that
Proof.
If , then for any , integrating by parts,
It follows by induction that if , then for each ,
Let . For , let . For each we have , hence
On the one hand,
and Lemma 2 tells us that this tends to as . On the other hand, for ,
which by Lemma 2 tends to as . Therefore, for ,
and because the right-hand side tends to as , we get
If then there is at least one with , from which we get
The function is continuous , so there is some such that
For nonzero , write , with which . Therefore
For , because the Fourier transform of an element of belongs to , we have by (3) that belongs to , and in particular is bounded. Then for ,
On the one hand, for we have
hence
giving
On the other hand, for we have
and so
Thus, for
we have
completing the proof. ∎
3 Bernstein’s inequality for L2
For a Borel measurable function , let be the union of those open subsets of such that for almost all . In other words, is the largest open set on which almost everywhere. The essential support of is the set
The following is Bernstein’s inequality for .55 5 Thomas H. Wolff, Lectures on Harmonic Analysis, p. 31, Proposition 5.1.
Theorem 4.
If , , and
(4) |
then there is some such that for almost all , and for any multi-index ,
Proof.
Let be the indicator function for . By (4), the Cauchy-Schwarz inequality, and the Parseval identity,
so . The Plancherel theorem66 6 Walter Rudin, Real and Complex Analysis, third ed., p. 187, Theorem 9.14. tells us that if and , then
for almost all . Thus, for defined by
we have for almost all . Because almost everywhere,
Applying Theorem 1 to , we have and for any multi-index ,
By Parseval’s identity,
proving the claim. ∎
4 Nikolsky’s inequality
Nikolsky’s inequality tells us that if the Fourier transform of a function is supported on a ball centered at the origin, then for , the norm of the function is bounded above in terms of its norm.77 7 Camil Muscalu and Wilhelm Schlag, Classical and Multilinear Harmonic Analysis, volume I, p. 83, Lemma 4.13.
Theorem 5.
There is a constant such that if , ,
and , then
Proof.
Let , i.e.
Then for ,
showing that . Let with for , with which
Then , and using Young’s inequality, with ,
(5) |
Moreover,
so (5) tells us
i.e.
Now, , so because , namely, . By the log-convexity of norms, for we have
Thus with
we have proved the claim. ∎
5 The Dirichlet kernel and Fejér kernel for R
The function defined by
and , is called the Dirichlet kernel. Let be the indicator function for the set . We have, for ,
For , . Thus,
For and , we define
It is straightforward to check that
For , , and ,
We define the Fejér kernel by
and . Thus, because is an even function,
One proves that is an approximate identity: ,
and for any ,
The fact that is an approximate identity implies that for any , in as .
We shall use the Fejér kernel to prove Bernstein’s inequality for .88 8 Mark A. Pinsky, Introduction to Fourier Analysis and Wavelets, p. 122, Theorem 2.3.17.
Theorem 6.
If , , and
then
Proof.
For , let . has the same support has , and
It follows that to prove the claim it suffices to prove that .
Write . Define by
We calculate, for ,
so
Then for ,
On the one hand, the integral of the left-hand side with respect to is
On the other hand, the integral of the right-hand side with respect to is
Hence
giving
proving the claim. ∎