Pseudo-Differential Operators on S1

Shahla Molahajloo and M. W. Wong∗

Department of Mathematics and Statistics

York University
4700 Keele Street
Toronto, Ontario M3J 1P3
Canada

Abstract Pseudo-differential operators on the unit circle S1 with center at

the origin are defined in terms of symbols on S1 × Z. Results on the bound-
edness and compactness of these pseudo-differential operators on L2 (S1 )
are presented. In addition, we prove a result on the Lp -boundedness for
1 < p < ∞.

1 Introduction
For m ∈ (−∞, ∞), let S m be the set of all functions σ in C ∞ (Rn × Rn ) such
that for all multi-indices α and β, there exists a positive constant Cα,β for
which
|(Dxα Dξβ σ)(x, ξ)| ≤ Cα,β (1 + |ξ|)m−|β| , x, ξ ∈ Rn .
A function σ in S m is called a symbol. Let σ be a symbol. Then we define
the pseudo-differential operator Tσ on the Schwartz space S by
Z
−n/2
(Tσ ϕ)(x) = (2π) eix·ξ σ(x, ξ)ϕ̂(ξ) dξ, x ∈ Rn ,
Rn
∗
This research has been supported by the Natural Sciences and Engineering Research
Council of Canada.

1
for all functions ϕ in S, where
Z
−n/2
ϕ̂(ξ) = (2π) e−ix·ξ ϕ(x) dx, ξ ∈ Rn .
Rn

It is easy to prove that Tσ maps S into S continuously. A much deeper fact

is that if σ ∈ S 0 , then Tσ can be extended to a bounded linear operator
from Lp (Rn ) into Lp (Rn ) for 1 < p < ∞. See, for instance, Theorem 10.7 in
the book [13] by Wong.
Although the phase space Rn ×Rn is used in defining symbols of pseudo-
differential operators on Rn , a more appropriate setting is, loosely speaking,
the phase space G × G b when we are interested in pseudo-differential oper-
ators on G, where G is a Lie group and G b is the set of all irreducible and
unitary representations of G. In the case of Rn , R cn can be identified with
n n n
R and hence the phase space R × R used for pseudo-differential opera-
tors on Rn is consistent with this big picture. In this paper the focus is on
pseudo-differential operators on S1 with symbols defined on the phase space
S1 × Z, where S1 is the unit circle centered at the origin and Z is the dual
group Sb1 . To be more precise, let σ be a measurable function on S1 × Z.
Then for every function f in L1 (S1 ), we define the function Tσ f on S1 by
X
(Tσ f )(θ) = σ(θ, n)fˆ(n)einθ , θ ∈ [−π, π],
n∈Z

where Z π
1
fˆ(n) = e−inθ f (θ) dθ, n ∈ Z.
2π −π

Among the most interesting problems are the ones on finding good condi-
tions on σ to produce bounded and compact pseudo-differential operators
Tσ : Lp (S1 ) → Lp (S1 ), 1 < p < ∞. Some results in this direction are given
in this paper. Related works can be found in [1, 2, 3, 5, 6, 9] among others.
We begin in Section 2 with an Lp formula for 1 ≤ p < ∞. This is then
used in Section 3 to prove that a pseudo-differential operator Tσ : L2 (S1 ) →
L2 (S1 ) is Hilbert–Schmidt if and only if σ ∈ L2 (S1 × Z). In Section 4,
necessary conditions and sufficient conditions on symbols σ are given for
the corresponding pseudo-differential operators Tσ : L2 (S1 ) → L2 (S1 ) to be
bounded linear operators. Sufficient conditions on symbols σ are given in
Section 5 to guarantee that the corresponding pseudo-differential operators
Tσ : L2 (S1 ) → L2 (S1 ) are compact. We give in Section 6 sufficient conditions

2
on symbols σ to give bounded linear operators Tσ : Lp (S1 ) → Lp (S1 ), 1 <
p < ∞.
The results in this paper are valid for the multi-dimensional torus using
multiple Fourier series in, e.g., the book [10] by Stein and Weiss. For the
sake of simpler notation and greater transparency, we have chosen to work
only with the unit circle S1 with center at the origin.

2 An Lp Formula, 1 ≤ p < ∞
For j ∈ Z, we let ej be the function defined by

ej (θ) = eijθ , θ ∈ [−π, π].

Then we have the following formula.

Theorem 2.1 Let σ ∈ Lp (S1 × Z), 1 ≤ p < ∞. Then

X
kTσ ej kpLp (S1 ) = kσkpLp (S1 ×Z) .
j∈Z

Proof For j ∈ Z, we get

X
(Tσ ej )(θ) = σ(θ, n)ebj (n)einθ , θ ∈ [−π, π]. (2.1)
n∈Z

But

Z π Z π
 1, n = j,
1 1
ebj (n) = e−inθ eijθ dθ = ei(j−n)θ dθ = (2.2)
2π −π 2π −π  0, n 6= j.

So, by (2.1) and (2.2),

(Tσ ej )(θ) = σ(θ, j)eijθ = σ(θ, j)ej (θ), θ ∈ [−π, π]. (2.3)

Hence π
X XZ
kTσ ej kpLp (S1 ) = |σ(θ, j)|p dθ = kσkpLp (S1 ×Z) ,
j∈Z j∈Z −π

as asserted. 

3
3 Hilbert–Schmidt Operators
Let us begin with the fact that a bounded linear operator A on a complex
and separable Hilbert space X is a Hilbert–Schmidt operator ⇔ there exists
an orthonormal basis {ϕk }∞k=1 for X such that

∞
X
kAϕk k2X < ∞,
k=1

where k kX denotes the norm in X. Moreover, if A is a Hilbert–Schmidt

operator on X, then the Hilbert–Schmidt norm kAkHS of A is given by
∞
X
kAk2HS = kAϕk k2X ,
k=1

where {ϕk }∞k=1 is any orthonormal basis for X. So, using Theorem 2.1 and
the fact that {(2π)−1/2 ek }∞ 2 1
k=−∞ is an orthonormal basis for L (S ), we have
the following result.

Theorem 3.1 The pseudo-differential operator Tσ : L2 (S1 ) → L2 (S1 ) is a

Hilbert–Schmidt operator ⇔ σ ∈ L2 (S1 × Z). Moreover, if Tσ : L2 (S1 ) →
L2 (S1 ) is a Hilbert–Schmidt operator, then

kTσ kHS = (2π)−1/2 kσkL2 (S1 ×Z) .

4 L2-Boundedness
The proof of Theorem 2.1 gives a necessary condition on a measurable
function σ on S1 × Z for Tσ : L2 (S1 ) → L2 (S1 ) to be a bounded linear
operator. To wit, let σ be a measurable function on S1 × Z and let f ∈
L2 (S1 ). Then for n ∈ Z, we can use (2.3) to obtain

(Tσ en )(θ) = σ(θ, n)en (θ), θ ∈ [−π, π],

and hence Z π
kTσ en k2L2 (S1 ) = |σ(θ, n)|2 dθ.
−π

4
So, if Tσ : L2 (S1 ) → L2 (S1 ) is a bounded linear operator, we can get a
positive constant C such that
kTσ en k2L2 (S1 ) ≤ Cken k2L2 (S1 ) , n ∈ Z,
which is then the same as
Z π
|σ(θ, n)|2 dθ ≤ 2πC, n ∈ Z.
−π

Therefore a necessary condition for Tσ : L2 (S1 ) → L2 (S1 ) to be a bounded

linear operator is Z π
sup |σ(θ, n)|2 dθ < ∞. (4.1)
n∈Z −π
The following example shows that (4.1) is not sufficient for Tσ : L2 (Rn ) →
L2 (Rn ) to be a bounded linear operator.
Example 4.1 Let σ be the function on S1 × Z defined by
σ(θ, n) = e−inθ , θ ∈ [−π, π], n ∈ Z,
and let f be the function on S1 defined by
∞
X 1 inθ
f (θ) = e , θ ∈ [−π, π].
n=1
n

Then it is easy to see that f ∈ L2 (S1 ), but Tσ f ∈

/ L2 (S1 ).
The following theorem gives sufficient conditions for a pseudo-differential
operator Tσ : L2 (S1 ) → L2 (S1 ) to be a bounded linear operator.
Theorem 4.2 Let σ be a measurable function on S1 × Z. Suppose that we
can find a positive constant C and a function w in L1 (Z) such that
|σ̂(m, n)| ≤ C|w(m)|, m, n ∈ Z, (4.2)
where Z π
1
σ̂(m, n) = e−imθ σ(θ, n) dθ.
2π −π
2 1 2 1
Then Tσ : L (S ) → L (S ) is a bounded linear operator and
kTσ k∗ ≤ CkwkL1 (Z) ,
where k k∗ is the norm in the C ∗ -algebra of all bounded linear operators on
L2 (S1 ).

5
Proof Let f ∈ C 2 (S1 ). Then we have
Z π X 2

kTσ f k2L2 (S1 ) = σ(θ, n)fˆ(n)einθ  dθ
 

−π  n∈Z 

Z π X X 2

=

 ˆ
σ̂(m, n)f (n)e i(m+n)θ 
 dθ.
−π  n∈Z m∈Z


So,
 2
Z π
X X 
kTσ f k2L2 (S1 ) = σ̂(k − n, n)fˆ(n)eikθ  dθ
 

−π  n∈Z k∈Z 
Z π X (X ) 2

=

 ˆ ikθ 
σ̂(k − n, n)f (n) e  dθ. (4.3)
−π  k∈Z n∈Z


Using (4.3) and the orthogonality of the functions {ej }∞ 2 1

j=−∞ in L (S ),
 2
X X 
kTσ f k2L2 (S1 ) = 2π σ̂(k − n, n)fˆ(n)


 
k∈Z n∈Z
!2
X X
≤ 2π |σ̂(k − n, n)||fˆ(n)| . (4.4)
k∈Z n∈Z

Using (4.2), we get

!2
X X
kTσ f k2L2 (S1 ) ≤ 2πC 2 |w(k − n)| |fˆ(n)|
k∈Z n∈Z
X
≤ 2πC 2
|(|w| ∗ |fˆ|)(k)|2 ,
k∈Z

where |w| ∗ |fˆ| is the convolution of |w| and |fˆ|. Finally, using Young’s
inequality, we have
kTσ f k2L2 (S1 ) ≤ 2πC 2 kwk2L1 (Z) kfˆk2L2 (Z) ≤ C 2 kwk2L1 (Z) kf k2L2 (S1 ) . (4.5)

Since C 2 (S1 ) is dense in L2 (S1 ), it follows that (4.5) holds for all functions
f in L2 (S1 ). 

6
Remark 4.3 In order to justify the interchange of the two sums in (4.3),
we note that by Fubini’s theorem, it is sufficient to prove that
XX
|σ̂(k − n, n)||fˆ(n)| < ∞.
k∈Z n∈Z
2 1
Since f ∈ C (S ), it follows that
|fˆ(n)| ≤ O(n−2 )
as |n| → ∞. Hence fˆ ∈ L1 (Z). Using an argument in the proof of Theorem
4.2, we have
XX
|σ̂(k − n, n)||fˆ(n)| ≤ CkwkL1 (Z) kfˆkL1 (Z) < ∞.
k∈Z n∈Z

5 L2-Compactness
Let σ be a measurable function on S1 × Z such that Tσ : L2 (S1 ) → L2 (S1 )
is compact. Since en → 0 weakly in L2 (S1 ) as |n| → ∞, it follows from the
compactness of Tσ : L2 (S1 ) → L2 (S1 ) that kTσ en kL2 (S1 ) → 0 as |n| → ∞.
By (2.3), we see that Z π
|σ(θ, n)|2 dθ → 0 (5.1)
−π
as |n| → ∞. That the condition (5.1) is not enough for Tσ : L2 (S1 ) → L2 (S1 )
to be compact can be illustrated by the following example.
Example 5.1 Let σ be the function on S1 × Z defined by

 1 e−inθ , n > 1,
ln n
σ(θ, n) =
 0, n ≤ 1,
for all θ in [−π, π]. If we let f be the function on S1 defined by
∞
X 1 inθ
f (θ) = e , θ ∈ [−π, π],
n=1
n
then f ∈ L2 (S1 ). But
∞
X 1
(Tσ f )(θ) = , θ ∈ [−π, π].
n=2
n ln n
So, Tσ f is not even in L2 (S1 ).

7
 The following theorem gives a sufficient condition for the L2 -compactness
of pseudo-differential operators on L2 (S1 ).
Theorem 5.2 Let σ be a measurable function on S1 × Z. Suppose that we
can find a function w in L1 (Z) and a function C on Z such that
lim C(n) = 0
|n|→∞

and
|σ̂(m, n)| ≤ C(n)|w(m)|, m, n ∈ Z. (5.2)
Then Tσ : L2 (S1 ) → L2 (S1 ) is compact.
Proof For all positive integers N , we define the function σN on S1 × Z by

 σ(θ, n), |n| ≤ N,
σN (θ, n) =
 0, |n| > N,

for all θ in [−π, π] and n in Z. Then for N = 1, 2, . . . ,

XZ π N
X Z π
2
|σN (θ, n)| dθ = |σ(θ, n)|2 dθ < ∞
n∈Z −π n=−N −π

and hence by Theorem 3.1, TσN : L2 (S1 ) → L2 (S1 ) is a Hilbert–Schmidt

operator. Let τN = σ − σN . Then by the definition of σN , we have

 0, |n| ≤ N,
τN (θ, n) =
 σ(θ, n), |n| > N,

for all θ in [−π, π]. Let ε be a given positive number. Then there exists a
positive integer N0 such that
|C(n)| < ε
whenever |n| > N0 . So, for N ≥ N0 ,
Z π
k(TσN − Tσ )f k2L2 (S1 ) = |(TτN f )(θ)|2 dθ
−π
Z π X 2

= τN (n, θ)fˆ(n)einθ  dθ.
 

−π  n∈Z


8
Now, we use the same argument in the derivation of (4.4) to get
!2
X X
k(TσN − Tσ )f k2L2 (S1 ) ≤ 2π |τ̂ (k − n, n)||fˆ(n)|
k∈Z n∈Z
 2
X X
= 2π  |σ̂(k − n, n)||fˆ(n)| .
k∈Z |n|>N

Using (5.2), we get

 2
X X
k(TσN − Tσ )f k2L2 (S1 ) ≤ 2π  C(n)|w(k − n)| |fˆ(n)|
k∈Z |n|>N
!2
X X
≤ 2πε2 |w(k − n)| |fˆ(n)| .
k∈Z n∈Z

Using the same argument in the the derivation of (4.5), we have

k(TσN − Tσ )f k2L2 (S1 ) ≤ ε2 Bkf k2L2 (S1 ) ,

where B = kwk2L1 (Z) . Thus, for N ≥ N0 ,

√
kTσ − TσN k∗ ≤ Bε.

In other words, Tσ is the limit in norm of the sequence {TσN }∞

N =1 and so
must be compact. 

6 Lp-Boundedness
The following lemma gives sufficient conditions on a function σ on Z to
ensure that the corresponding pseudo-differential operator Tσ : Lp (S1 ) →
Lp (S1 ), 1 < p < ∞, is a bounded linear operator. Pseudo-differential op-
erators Tσ , where σ is a function on Z only, are often known as Fourier
multipliers.

9
Lemma 6.1 Let σ be a measurable function on Z and let k be the smallest
integer greater than 12 . Suppose that there exists a positive constant C such
that
|(∂ j σ)(n)| ≤ Chni−j , n ∈ Z,
for 0 ≤ j ≤ k, where ∂ j is the difference operator given by
j  
j
X
j−l j
(∂ σ)(n) = (−1) σ(n + l), n ∈ Z,
l=0
l

and
hni = (1 + |n|2 )1/2 , n ∈ Z.
Then for 1 < p < ∞, Tσ : Lp (S1 ) → Lp (S1 ) is a bounded linear operator
and there exists a positive constant B, depending on p only, such that

kTσ f kLp (S1 ) ≤ BCkf kLp (S1 ) , f ∈ Lp (S1 ). (6.1)

Remark 6.2 The condition on the number k of “derivatives” in Lemma

6.1 can best be understood in the context of the multi-dimensional torus
in which k should be the smallest integer greater than d/2, where d is the
dimension of the torus.

The proof of Lemma 6.1 entails the use of the Littlewood–Paley theory
in Fourier series in, e.g., Chapter XV of the book [14] by Zygmund. See,
in particular, Theorem 4.14 in Chapter XV of [14] in this connection. Ex-
tensions of Lemma 6.1 to the context of compact Lie groups are attributed
to Weiss [11, 12] and the Littlewood–Paley theory for compact Lie groups
can be found in Stein [8]. Analogs of Lemma 6.1 for Rn can be found in the
works of Hörmander [4] and Stein [7].
The following theorem gives sufficient conditions for the Lp -boundedness
of pseudo-differential operators on S1 . The ideas for the result and its proof
come from Theorem 10.7 in the book [13] by Wong.

Theorem 6.3 Let σ be a measurable function on S1 × Z and let k the

smallest integer greater than 21 . Suppose that we can find a positive constant
C and a function w in L1 (Z) such that

|(∂nj σ̂)(m, n)| ≤ C|w(m)|hni−j , m, n ∈ Z, (6.2)

10
for 0 ≤ j ≤ k, where ∂nj is the partial difference operator with respect to the
variable n in Z. Then for 1 < p < ∞, Tσ : Lp (S1 ) → Lp (S1 ) is a bounded
linear operator. Moreover, there exists a positive constant B depending only
on p such that
kTσ k∗ ≤ BCkwkL1 (Z) kf kLp (S1 ) ,
where k k∗ here denotes the norm in the Banach algebra of all bounded linear
operators on Lp (S1 ).

Proof Let f ∈ Lp (S1 ). Then by Fubini’s theorem,

X
(Tσ f )(θ) = fˆ(n)σ(θ, n)einθ
n∈Z
( )
X X
= fˆ(n) σ̂(m, n)eimθ einθ
n∈Z m∈Z
( )
X X
= eimθ σ̂(m, n)fˆ(n)einθ
m∈Z n∈Z
X
imθ
= e (Tσm f )(θ)
m∈Z

for all θ in [−π, π], where

Z π
1
σm (n) = σ̂(m, n) = e−imθ σ(θ, n) dθ, m, n ∈ Z.
2π −π

By (6.2),

|(∂ j σm )(n)| = |(∂nj σ̂)(m, n)| ≤ C|w(m)|hni−j , m, n ∈ Z,

for 0 ≤ j ≤ k. Therefore by Lemma 6.1, there exists a positive constant B

depending only on p such that

kTσm f kLp (S1 ) ≤ BC|w(m)| kf kLp (S1 ) , m ∈ Z.

Then by (6.1) and Minkowski’s inequality, we get

(Z  p ) 1
π
X  p
imθ
kTσ f kLp (S1 ) = e (Tσm f )(θ) dθ
 

−π  m∈Z


11
 X Z π  p1
p
≤ |(Tσm f )(θ)| dθ
m∈Z −π
X
= kTσm f kLp (S1 )
m∈Z
X
≤ BC kwkL1 (Z) kf kLp (S1 ) ,
m∈Z

and this completes the proof. 

Acknowledgment We are grateful to Professor M. S. Agranovich for very

useful comments on the results and the presentation of this paper, and to
Professor Ville Turunen for pointing out the reference [6] to us.

References
[1] M. S. Agranovich, Spectral properties of elliptic pseudodifferential op-
erators on a closed curve, (Russian) Funktsional. Anal. i Prilozhen. 13
(1979), 54–56.

[2] M. S. Agranovich, Elliptic pseudodifferential operators on a closed

curve, (Russian) Trudy Moskov. Mat. Obshch. 47 (1984), 22–67, 246.

[3] B. A. Amosov, On the theory of pseudodifferential operators on the

circle, (Russian) Uspekhi Mat. Nauk. 43 (1988), 169–170; Translation
in Russian Math. Surveys 43 (1988), 197–198.

[4] L. Hörmander, Estimates for translation invariant linear operators in

Lp spaces, Acta Math. 104 (1960), 93–139.

[5] M. Ruzhansky and V. Turunen, On the Fourier analysis of operators on

the torus, in Modern Trends in Pseudo-Differential Operators, Editors:
J. Toft, M. W. Wong and H. Zhu, Birkhäuser, 2007, 87–105.

[6] M. Ruzhansky and V. Turunen, Quantization of pseudo-differential

operators on the torus, preprint, arXiv:0805.2892

[7] E. M. Stein, Singular Integrals and Differentiability Properties of Func-

tions, Princeton University Press, 1970.

12
 [8] E. M. Stein, Topics in Harmonic Analysis Related to Littlewood–Paley
Theory, Third Printing with Corrections, Princeton University Press,
1985.

[9] J. Saranen and G. Vainikko, Periodic Integral and Pseudodifferential

Equations with Numerical Approximation, Springer-Verlag, 2002.

[10] E. M. Stein and G. Weiss, Introduction to Fourier Analysis on Eu-

clidean Spaces, Princeton University Press, 1971.

[11] N. J. Weiss, Multipliers on compact Lie groups, Proc. Nat. Acad. Sci.
U.S.A. 68 (1971), 930–931.

[12] N. J. Weiss, Lp estimates for bi-invariant operators on compact Lie

groups, Amer. J. Math. 94 (1972), 103–118.

[13] M. W. Wong, An Introduction to Pseudo-Differential Operators, Sec-

ond Edition, World Scientific, 1999.

[14] A. Zygmund, Trigonometric Series, Third Edition, Volumes I & II

Combined, Cambridge University Press, 2002.

13