Professional Documents
Culture Documents
A PANORAMA OF
HARMONIC
ANALYSIS
Steven G. Krantz
A Panorama
of Harmonic Analysis
Originally published by
The Mathematical Association of America, 1999.
ISBN: 978-1-4704-5112-7
LCCN: 99-62756
VOL 27
A Panorama
of Harmonic Analysis
Stephen G. Krantz
THE
CARUS MATHEMATICAL MONOGRAPHS
Published by
THE MATHEMATICAL ASSOCIATION OF AMERICA
Committee on Publications
William Watkins, Chair
Preface xiii
0 Overview of Measure Theory and Functional Analysis 1
0.1 Pre-Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
0.2 A Whirlwind Review of Measure Theory . . . . . . . . . . . . . . 2
0.3 The Elements of Banach Space Theory . . . . . . . . . . . . . . . . 13
0.4 Hilbert Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
0.5 Two Fundamental Principles of Functional Analysis . . . . 26
Index 347
Preface
∂ 2u ∂ 2u
− =0
∂x2 ∂t 2
u(x, 0) = sin j x or cos j x
The 1930s and 1940s were a relatively quiet time for Fourier anal-
ysis but, beginning in the 1950s, the focus of Fourier analysis became
singular integrals. To wit, it rapidly developed that, just as the Hilbert
transform is the heart of the matter in the study of Fourier series of one
variable, so singular integrals usually lie at the heart of any nontrivial
problem of several-variable linear harmonic analysis. The Calderón-
Zygmund theory of singular integrals blossomed into the Fefferman-
Stein-Weiss theory of Hardy spaces; Hardy spaces became the focus of
Fourier analysis.
In the 1980s, two seminal events served to refocus Fourier anal-
ysis. One was the David-Journé-Semmes T (1) theorem on the L 2
boundedness of (not necessarily translation-invariant) singular inte-
grals. Thus an entirely new perspective was gained on which types of
singular integrals could induce bounded operators. Calderón commu-
tators, Hankel operators, and other classical objects were easy pickings
using the powerful new tools provided by the T (1) theorem, and more
generally the T (b) theorem.
The other major event of the 1980s was the development of
wavelets by Yves Meyer in 1985. Like any good idea, wavelet the-
ory has caused us to “reinvent” Fourier analysis. Now we are no longer
bound to model every problem on sine waves and cosine waves. In-
stead, we can invent a Fourier analysis to suit any given problem. We
have powerful techniques for localizing the problem both in the space
variable and the phase variable. Signal processing, image compression,
and many other areas of applied mathematics have been revolutionized
because of wavelet theory.
The purpose of the present book is to give the uninitiated reader
an historical overview of the subject of Fourier analysis as we have
just described it. While this book is considerably more polished than
merely a set of lectures, it will use several devices of the lecture: to
prove a theorem by considering just an example; to explain an idea by
considering only a special case; to strive for clarity by not stating the
optimal form of a theorem.
We shall not attempt to explore the more modern theory of Fourier
analysis of locally compact abelian groups (i.e., the theory of group
Preface xv
to ensure that this project came out as it should. The assistance provided
by these three scholars has been so extraordinary, and so extensive, that
I sometimes feel as though this manuscript has four authors.
Of course responsibility for the extant manuscript lies entirely
with me. I am always happy to receive reports of errors or suggestions
for improvement.
Steven G. Krantz
St. Louis, Missouri
Appendices and Ancillary Material
Appendix I
The Existence of Testing Functions and
Their Density in L p
Consider functions on R1 . The real function
( 2
e−1/x if x > 0
φ(x) =
0 if x ≤ 0
is C ∞ and is identically zero outside the set (−1, 1). We say that ψ
is a “C ∞ function with compact support.” The support of ψ—that is,
the closure of the set on which ψ is nonzero—is the interval [−1, 1].
Given any compact interval [a, b], it is clear that this construction may
be adapted to produce a C ∞ function whose support is precisely [a, b].
The C ∞ functions with compact support (usually denoted Cc∞ or
Cc∞ (R) for specificity) form a dense subset of L p (R), 1 ≤ p < ∞.
To see this, first note that the characteristic function of an interval I
may be approximated in L p norm by a Cc∞ function (see Figure 1). We
315
316 Appendices and Ancillary Material
lim k f − τa f k L p = 0.
a→0
k f − τa f k L p ≤ k f − φk L p + kφ − τa φk L p + kτa φ − τa f k L p
= k f − φk L p + kφ − τa φk L p + kτa (φ − f )k L p
Appendices and Ancillary Material 317
≤ + kφ − τa φk L p + kφ − f k L p
≤ + kφ − τa φk L p + .
Appendix II
Schwartz Functions and the Fourier Transform
A Schwartz function φ is an infinitely differentiable function such that,
for any multi-indices α and β, the expression
∂
ρα,β (φ) ≡ sup x α φ(x)
x ∂xβ
Appendix III
The Interpolation Theorems of Marcinkiewicz
and Riesz-Thorin
The simplest example of an interpolation question is as follows. Sup-
pose that the linear operator T is bounded on L 1 and bounded on L 2 .
Does it follow that T is bounded on L p for 1 < p < 2? [The space
L p here is an instance of what is sometimes called an “intermediate
space” between L 1 and L 2 .] Note that this question is similar to (but
not precisely the same as) one that we faced when considering the L p
boundedness of Calderón-Zygmund singular integral operators. Here
we record (special) versions of the Riesz-Thorin Theorem (epitomizing
the complex method of interpolation) and the Marcinkiewicz Interpo-
lation Theorem (epitomizing the real method of interpolation) that are
adequate for the applications in the present book.
kT f k L p0 ≤ C0 · k f k L p0
and
kT f k L p1 ≤ C1 · k f k L p1 .
If 0 ≤ t ≤ 1 and
1 1 1−t t
= = + ,
p pt p0 p1
then we have
kT f k L p ≤ C01−t · C1t · k f k L p .
kT f k L p ≤ C p · k f k L p .
1 1 1
= (1 − t) · +t ·
pt p0 p1
and
1 1 1
= (1 − t) · +t · .
qt q0 q1
Appendix IV
Hausdorff Measure and Surface Measure
If ⊆ R N has C 1 boundary, then we use the symbol dσ to denote
(N − 1)-dimensional area measure on ∂. This concept is funda-
mental; we discuss, but do not prove, the equivalence of several def-
initions for dσ. A thorough consideration of geometric measures on
lower-dimensional sets may be found in the two masterpieces [FE1]
and [WHI].
First we consider a version of a construction due to Hausdorff. Let
S ⊆ R N and δ > 0. Let U = {Uα }α∈A be an open covering of S. Call
U a δ-admissible covering if each Uα is an open Euclidean N -ball of
radius 0 < rα < δ. If 0 ≤ k ∈ Z, let Mk be the usual k-dimensional
Lebesgue measure of the unit ball in Rk (e.g., M1 = 2, M2 = π, M3 =
4π/3, etc.). Define
(
Hδk (S) Mk rαk : U = {Uα }α∈A
X
= inf
α∈A
is a δ-admissible cover of S} .
Clearly, Hδk (S) ≤ Hδk0 (S) if 0 < δ 0 < δ. Therefore limδ→0 Hδk (S)
exists in the extended real number system. The limit is called the k-
dimensional Hausdorff measure of S and is denoted by Hk (S). The
function Hk is an outer measure.
IV.4. Define Mα = 0(1/2)α / 0(1 + α/2), and α > 0 (note that this is
consistent with the preceding definition of Mk ). Then define H α
for any α > 0 by using the Hausdorff construction. Let S be a
subset of R N . Set α0 = sup{α > 0 : Hα (S) = ∞}. Also com-
pute α1 = inf{α > 0 : Hα (S) = 0}. Then α0 ≤ α1 . This number
is called the Hausdorff dimension of S. What is the Hausdorff
dimension of the Cantor ternary set? What is the Hausdorff di-
mension of a regularly imbedded, k-dimensional, C 1 manifold
in R N ? In fact (see [FOL, p. 325], [FE1]), any rectifiable set
S ⊆ R N has the property that α0 = α1 .
L N (E )
σk (E) = lim sup ,
→0+ M N −k N −k
8 = (φ1 , . . . , φ N ) : W P → U P ∩ M
is C 1 , one-to-one, and onto, and the Jacobian of this mapping has rank
k at each point of W P . In this circumstance, we define
Z
τk (U P ∩ M) = |Mx | dLk (x),
x∈W P
Hk = σk = m k = τk .
Appendices and Ancillary Material 323
Appendix V
Green’s Theorem
Here we record the standard form of Green’s theorem that is used
in harmonic analysis. A derivation of this particular formula from
Stokes’s theorem appears in [KRA4, Section 1.3]; that reference also
contains applications to the theory of harmonic functions. See also
[BAK], [KRP].
∂v ∂v
Z Z
u −v dσ = (u1v − v1u) d V.
∂ ∂ν ∂ν
Appendix VI
The Banach-Alaoglu Theorem
Let X be a Banach space and X ∗ its dual. Assume for the moment that
X is separable. For φ j , φ elements of X ∗ , we say that φ j → φ in the
weak-∗ topology if, for each x ∈ X , φ j (x) → φ(x). Thus weak-∗ con-
vergence is pointwise convergence for linear functionals. It induces the
weakest topology on X ∗ under which the point evaluation functionals
are continuous.
Appendix VII
Expressing an Integral in Terms of the
Distribution Function
Let f be a measurable function on R N . For α > 0 we set
Appendix VIII
The Stone-Weierstrass Theorem
Let X be a compact metric space. Let C(X ) be the algebra of con-
tinuous functions on X equipped with the supremum norm. Let A ⊆
C(X ) be an algebra of continuous functions that contains the con-
stant function 1. [Here an algebra is a vector space with a notion
of multiplication—see [LAN].] The Stone-Weierstrass theorem gives
conditions under which A is dense in C(X ).
Appendices and Ancillary Material 325
Appendix IX
Landau’s O and o Notation
Sometimes a good piece of notation is as important as a theorem. Lan-
dau’s notation illustrates this point.
Let f be a function defined on a neighborhood in R N of a point
P. We say that f is O(1) near P if
| f (x)| ≤ C,
| f (x)| ≤ C · 1,
f (x) ≤ C · g(x) ,
f (x) ≤ · g(x)
327
328 Table of Notation
339
340 Bibliography
347
348 Index
Steve Krantz has written a remarkable book that leads the reader on a tour of
Euclidean harmonic analysis from its genesis to the frontiers of recent research. It
can be warmly recommended to professionals, graduate students, and adventurous
undergraduates who want to know what this beautiful subject is about.
—Gerald Folland, University of Washington, Seattle
This book treats the subject of harmonic analysis, from its earliest
beginnings to the latest research. Following both a historical and a
conceptual genesis, the book discusses Fourier series of one and several
variables, the Fourier transform, spherical harmonics, fractional
integrals, and singular integrals on Euclidean space.
The climax of the book is a consideration of the earlier ideas from the
point of view of spaces of homogeneous type. The book culminates with
a discussion of wavelets—one of the newest ideas in the subject.
CAR/27