6 1. MATHEMATICAL SHAPES OF UNCERTAINTY

of them launching whole areas of Harmonic and Complex Analysis. For instance,

it is natural to consider the support of a measure and pronounce it small if it is

bounded, or has finite volume (length) or if it is thin (porous) in some sense. The

most basic statement of that type, the one that Wiener relays to non-mathematical

audience using musical analogies in his memoir [149], is that the supports of f and

ˆ

f cannot both be compact. Further statements in the same direction are given in

the next subsection.

Another natural way to understand the smallness of f is in terms of its decay

near infinity. In many situations the Fourier transform

ˆis

f an analytic function that

can be considered to be small if it has a large zero set. Of course, the smallness of

f and

ˆ

f does not have to be understood in the same way and many deep problems

appear when the criteria for smallness applied to these two objects is different.

As was mentioned above, the Fourier transform does not have to be a part of

the statement and similar questions can be formulated about other transforms or,

more generally, a variety of pairs of ‘dual’ properties of f. Despite rather extensive

literature, many of such questions remain open.

In the rest of this section we give further examples of particular problems of UP,

with smallness in the general statement understood in terms of support, decay or

the size of the zero set. We then discuss an immediate connection of such questions

with completeness problems.

2.2. Support/Support. As was mentioned above, perhaps the simplest

statement of that type is the observation that the supports of f and

ˆ

f on the

real line cannot both be bounded. A more advanced statement is a corollary of a

theorem by F. and M. Riesz that says that if μ ∈ M has semi-bounded support

then the support of ˆ μ contains the whole line R.

The next significant step in the same direction begins with results of Amrein-

Berthier [5] and Benedicks [12] that imply that the Lebesgue measures of the

supports of f and

ˆ

f cannot both be finite. These results were later improved by

Nazarov [114] who showed that for any two sets T, S ∈

Rn

||f||L2(Rn) C exp (C|S||T |)(||f||L2(Sc) + ||

ˆ||L2(T

f

c)

),

where T c,Sc denote the complements of T, S correspondingly. If one of the sets is

convex, the product |S||T| in the above inequality may be raised into the power

1/n. It is unknown if the same is true in general.

As we can see, even in this case of UP there are subcases corresponding to

different meanings of smallness of the supports of f and

ˆ.

f Another set of problems

appears if one understands such smallness in terms of porosity. One of the classical

results in this group is the following theorem by Beurling from 1961 [14].

A collection of disjoint intervals {In} on R (or the union of these intervals,

when it is more convenient) is called long if

n

|In|2

1 +

dist2(In,

0)

= ∞

or short if the sum is finite. Here |In| denotes the length of In. Such collections of

intervals appear in many results in the area of UP including those discussed further

in this text.