Вы находитесь на странице: 1из 11

Some research in analysis

Terence Tao
Abstract: I describe some analysis in my research area
which could be a starting point for a graduate thesis.

1
Possible areas of research:
• The Kakeya problem (geometric combinatorics, arith-
metic combinatorics, geometric measure theory, har-
monic analysis)
• Lp bounds for oscillatory integrals (harmonic analy-
sis, linear PDE; closely related to the Kakeya prob-
lem)
• Korteweg de Vries, non-linear Schrödinger, wave maps,
and other nonlinear dispersive equations (nonlinear
PDE, harmonic analysis, possible use of numerics)
• Multilinear operators, Schrodinger eigenfunctions (har-
monic analysis, spectral theory; Thiele works more
heavily in this area than I do)
• ...
• I’ll only talk about the first two topics today.

2
The Kakeya problem
• The Kakeya problem is not a single problem, but is
really a class of problems, all trying to address the
following issue: to what extent can lines in different
directions overlap each other?
• The original question of Kakeya was the Kakeya nee-
dle problem posed in 1917: Given a needle of unit
length and zero thickness, how much area is needed
in the plane in order to turn the needle completely
around? Besicovitch gave the surprising answer in
1928 that one can turn this needle using arbitrarily
small area. (On the other hand, if the needle has
non-zero thickness 0 < δ ≪ 1, the area needed is
∼ 1/log(1/δ), which goes to zero only very slowly as
δ → 0).
• There is a corresponding problem in three dimen-
sions: given a tube of unit length and thickness δ,
how much volume is needed in order that the tube
can be oriented in each direction in space? If δ = 0,
one can use arbitrarily small area. When δ > 0, the
answer is unknown. In 1995 Wolff showed that one
needs & δ 1/2 volume; in 1999 Keich showed that one
can use ∼ 1/ log(1/δ). The conjecture is that one
3
needs & δ ε volume for any fixed 0 < ε ≪ 1.
• There are more fancy variants of this conjecture in-
volving Hausdorff dimension, or a maximal function
(similar to the Hardy-Littlewood maximal function),
which I won’t detail here. These fancier variants
have application to oscillatory integrals (more on this
later).
• There is a “toy” version of this problem which is
cleaner to state, and takes place in a finite field F .
Suppose that E is a subset of the n-dimensional fi-
nite field geometry F n which contains a line in each
direction. (A line in F n is a finite collection of points,
|F | to be exact). What is the least possible cardi-
nality of E? In two dimensions n = 2 this is known
exactly (basically, |E| can be as low as 21 |F |2 but not
much lower). In three and higher dimensions it is
conjectured that |E| has to essentially be as large as
all of |F |n, but this is not known. Numerous partial
results (lower bounds in |E|, mostly) are known.

4
Methods used:

• Extremal graph theory. One can use geometric facts,


such as “every two lines intersect in at most one
point”, to limit the number of times lines can overlap.
This requires some basic combinatorics (pigeonhole
principle, Cauchy-Schwartz, etc.). Also the geome-
try required can get a little fancy, verging on alge-
braic geometry (example: in F 3, the number of lines
which intersect three given lines in general position
is O(|F |)).
• Arithmetic combinatorics. Lines are very similar to
arithmetic progressions, for instance if x, y belong to
a line then so does (x + y)/2. However, the differ-
ence x − y is parallel to the direction of the line.
Since our lines all point in different directions, we
thus are faced with a situation where E − E is large
but E + E is small. This leads to arithmetic combi-
natorics - the combinatorics of sums, differences, and
arithmetic progressions of finite sets. These are of
course very simple concepts, but the problems here
are quite challenging - the Fields medallists Bour-
gain and Gowers have worked in this area recently,
for instance.

5
• Geometric measure theory. Once you leave the finite
field setting and go into Euclidean space, some other
technical issues arise regarding Hausdorff dimension,
covering lemmas, etc. - in short, a lot of 245AB gets
used here.
• Many, many open problems, some more difficult than
others. Many of the tools used are new and not fully
developed, and we need people to extend and refine
them to more general contexts. One nice thing about
this problem is that one can focus on one aspect at a
time; for instance, one can deal exclusively with the
combinatorial aspects, or with the measure theory
aspects, etc.

6
Oscillatory integrals
• I’m interested in understanding oscillatory linear op-
erators which have a form such as

Tλf (x) = eiλϕ(x,y)a(x, y)f (y) dy

where a is some fairly smooth amplitude function,


ϕ is a real smooth phase function, and λ ≫ 1 is a
large parameter. A typical example can come from
appropriate modification of the Fourier transform

fˆ(ξ) = e2πix·ξ f (ξ) dξ.

• More examples come from solving PDE. For instance


the Schrödinger equation
iut(t, x)+∆u(t, x) = 0 for all (t, x) ∈ R×Rn; u(0, x) = f (x)
has the explicit solution

1 2 /4t
u(t, x) = ei|x−y| f (y) dy.
(4πt)n/2
• The basic question arises: to what extent does the
size of f control the size of T f ? For the Fourier
transform, this is answered by the Plancherel and
Hausdorff-Young inequalities. For the Schrödinger
7
equation, the question is (partially) answered by the
Strichartz inequalities, which play a major role in the
theory of non-linear Schrödinger equations.
• For more general T , the question is to obtain the best
possbile estimates of the form
∥Tλf ∥Lq ≤ Cλ−α ∥f ∥Lp .
• These estimates have many applications to harmonic
analysis, PDE, spectral theory, and may possibly
have application to number theory. When p = 2
or q = 2 the theory is fairly well understood, but
the general case is not. Model cases include some
long-standing conjectures in harmonic analysis, no-
tably the restriction conjecture, the Bochner-Riesz
conjecture, and the local smoothing conjecture.
• Interestingly, this problem is closely related to the
Kakeya problem. This comes from a “wave packet
decomposition” of f , which decomposes f as a su-
perposition of oscillating waves supported on tubes.
(From a Schrödinger point of view, we are trying
to isolate the particle trajectories inside a quantum
wave function). To control Tλf one then needs to
control the overlap of these tubes, which leads to
Kakeya issues.
8
• However, knowing the Kakeya problem is not enough;
one also needs to understand how various oscilla-
tory functions superimpose on each other. This re-
quires harmonic analysis techniques (notably, use of
the Fourier transform, uncertainty principle, and or-
thogonality). A little bit of symplectic geometry also
comes in, in order to handle both space and frequency
simultaneously.
• I’m mostly interested in clarifying the interconnec-
tion between the oscillatory integral problems and
the Kakeya-type problems, but there are many, many
interesting and open problems in oscillatory integrable
problems which can be looked at.

9
Skills

• Level of mathematical background is not so impor-


tant to me; one nice thing about analysis is that one
can already get into some interesting unsolved prob-
lems without needing too much formalism. But being
able to pick up ideas quickly is a plus. So is maintain-
ing a broad interest in analysis, and mathematics as a
whole; one can get quite far by narrowly focusing on
one particular sub-field, but often the breakthrough
comes from keeping an ear open on developments
in nearby fields (or sometimes fields not so nearby).
Please pester me with crazy ideas as to how to solve
a problem!
• The nature of these problems can change over time.
A problem can be too difficult, and we move to a
simpler model problem; or it can be too easy, and
we look for generalizations. Also it may turn out
that a technique we thought would be effective is
not, and so we have to try something rather differ-
ent. Expect some early disappointments and some
unexpected successes; the problems we’ll look at will
not just be routine applications of known techniques
(at least, not completely).

10
• I can travel for periods of up to a month, though
usually I’m reachable by e-mail. So a certain amount
of self-sustainability is desirable.

11

Вам также может понравиться