This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
(4.59)
0, 2. 4
—d. Then
(R°'f)(x) =
(4.60)
Proof. By Theorem I and Lemma 4. / is well defined, and almost all x. Hence (4.60) holds in view of Corollary 3.
Remark. If I a > 0, a + d —
is finite for
then, by Corollary 3. the equality (4.60) holds for all The limit case a = 0 gives (4.53). Furthermore, by (4.52), K°f can be regarded as the identity operator. Thus (4.60) means that the n
0. 2.4
inverse off is just
the analytic continuation of (4.59) at the point a = —d (at least formally). Various applications of the remarkable equality (4.60) are given in lRlOJ.
5 The convolution-backprojection method and wavelet-like transforms The general scheme of the convolution-backprojection method in the hyperbolic setup is as follows. Given a measurable function w(s) on (0, let
=
XE X.
(5.61)
Owing to G-invariance, we have
(Wf)(x) =
f
By (3.37). a formal application of (2.31) yields
(5.62)
C.A. Berenstein and B. Rubin
52
(Wf)(x)
— 1)dy,
=
=
c
=
(5.63)
where
j(t
= is
—
the Riemann-.Liouville fractional integral. Thus
(Wf)(x)
=
f
y]2
—
(5.64)
1)dy.
lf fx f(y)ifr([x. yJ2 — l)dy is an approximate identity on X, then (5.64) enables us to recover f from!. Now we discuss details.
Lemma 5. Let! E w(s)
p < (n — 1)/(d
I
ifs
1.
o(s_n_1/P'—E2) ifs
I.
I
=
for sufficiently small
— 1).
Suppose that
I/p + lip' = 1.
(5.65)
e2 > 0. Then (5.64) holds for almost all x.
Proof. We split (5.61) into two integrals over the sets
> I}.
< 1),
By (5.65), WI I
C
(Ii + 12) where c = c(a) = const,
i= 1.2;
I =
l—d—e2+(n—I)/p.
flu
By Theorem 2. is finite a.e. Thus the integral (5.62) with f replaced by Ill is finite a.e., and therefore (2.31) is applicable. This gives (5.64). I
We denote
(W,ço)(x) =
!
: >0,
(5.66)
where the function w will be at our disposal. This operator is a scaled version of (5.61). Suppose that w(s) obeys (5.65). By Lemma 5,
(W,f)(x)
=
f
fO')*(Ex. y]2 — l)dy.
=
CT I—n/2
A(.),
(5.67)
Radon Transform
=
53
(5.68)
c being the constant from (5.63). The right-hand side of the first equality in (5.67) represents an approximate identity of the form
(K6f)(x)
y])f(y)dy,
= Jx
=
s>
koI
C
0.
Theorem 3. Let f be a measurable function on X. We introduce a maximal operator
(K*f)(x) = sup I(K€f)(x)I. The following statements hold.
I. If ko(s) has a decreasing integrable majoran:, then Kf
f*(x) = sup
r>O IB(x,r)l
jf
B(x.r)
X : dist(x, y) = cosh'[x, yJ
B(x, r) = {y
II. 1ff E LP(X), I
JTJETJ 11'T,' which can be handled by orthogonality arguments), so we turn to the problem of estimating the L2 quantity: 11/TI
II
(I
—
T1
As is customary, we subdivide the large ball B(O. R) into cubes q of size 'Ii. The contribution of each cube q is 11'T1*r2(l —
II
T1ET1 T2ET2
Roughly speaking, we only need to consider pairs Ti, of tubes which pass through q (because of the localization of 11'r1 and and such that q is not contained in both and Br,). For the sake of argument, suppose that we only consider the terms BT1. Then we can rewrite the above expression as where q II
T1ET;(q)
T2ET2(q)
where T2(q) denotes all the tubes T2 in T2 which intersect q, and denotes all the tubes Ti in T1 which intersect q and for which q Note that the tubes in (q) must point in essentially different directions (since they all go through q and are essentially distinct tubes), and similarly for The function 11'r1 behaves roughly like the function where (q) is the subset of the paraboloid whose unit normals lie within I / R of the directions of one of the tubes in (q). This can be seen by recalling the origin of these wave packets 1/IT1 as Fourier transforms of functions on S (or more precisely on N,/R(S); the discrepancy between the two explains the factor). The function is similarly comparable to the expression for a suitThus one is faced with an expression of the form able
240
T. Tao
(22) where we have discarded some powers of R for the sake of exposition, as well as the localization to q. As observed previously, the restriction estimate (2 x 2 —* 2) allows us to bound this quantity by something proportional to Actually we can do a little better and refine this to, say, this comes from not discarding the constraint E in the argument immediately preceding (18), and by exploiting the localization to q more; we omit the details. This is the type of bound used in Wolff's argument, in combination with the combinatorial arguments controlling the multiplicity of the tubes in T1 and 12 mentioned in the previous section, to obtain a sharp bilinear estimate in the case of the cone. However, this bound is insufficient for the paraboloid case because of the failure of the tubes T1 through a point to lie on a hypersurface. Fortunately, this can be rectified by exploiting the gain inherent in (18). Indeed, by refusing to use (18) one can obtain a bound on (22) which is proportional to
sup
-
(q) 2 mentioned earlier, but it is a little is restricted to a hyperplane. When improved because one of the factors of one inserts this bound back into the coarse-scale combinatorial analysis of Wolff, this This is similar to the bound of ici'1 (q) I
effectively allows us to restrict the tubes T2 passing through a cube q to be incident to a hyperplane. This turns out to be a good substitute for the hypersurface localization property used in Wolff's argument, and it is the key new ingredient which permits us to generalize the bilinear cone estimate to paraboloids (and by similar reasoning to other positively curved surfaces, such as the sphere). One interesting feature of this argument is that it introduces a non-trivial correlation between the fine-scale analysis and the coarse-scale analysis; one may speculate that future developments on these problems will deal with the fine-scale and coarsescale aspects of the restriction operator in a more unified manner.
References I. Banner, A.. Restriction of the Fourier transform to quadratic submanifolds, Princeton University Thesis, 2002. 2. Barcelo, B.. On the restriction of the Fourier transform to a conical surface, Trans. Amer Math. Soc. 292 (1985). 32 1—333. 3. Barceló, J.A., Bennett, 3.M., and Carbery, A., A bilinear extension inequality in two dimensions, J. Funci. AnaL 201 (2003), no. 1. 57—77. 4. Barcelo, J.A., Ruiz, A., and Vega, L., Weighted estimates for the Helmholtz equation and some applications,). Funct. Anal. 150, no. 2 (1997), 356—382. 5. Beckner, W., Carbery, A., Semmes, S., and Soria, F., A note on restriction of the Fourier transform to spheres, BulL London Math. Soc. 21, no. 4 (1989), 394—398. 6. Bourgain, .J., Sesicovitch-type maximal operators and applications to Fourier analysis, Geom. and Funct. Anal. 22(1991), 147—187.
Recent Progress on Restriction
241
7. Bourgain, J., On the Restriction and Multiplier Problem in R3, Lecture Notes in Mathematics, no. 1469, Springer-Verlag, 1991. 8. Bourgain, J.. On the dimension of Kakeya sets and related maximal inequalities, Geom. Func:. AnaL 9, no. 2 (1999), 256—282.
9. Bourgain, J., A remark on Schrodinger operators. Israel). Math. 77(1992), 1—16. 10. Bourgain, J.. Fourier restriction phenomena for certain lattice subsets and applications to nonlinear evolution equations, Part I, Geometric and Funct. AnaL 3 (1993), 107—156. 11. Bourgain, J., Estimates for cone multipliers. Opei Theory: App!. 77(1995). 41—60. 12. Bourgain, J., Some new estimates on oscillatory integrals, in Essays in Fourier Analysis in Honor of E. M. Stein. Princeton University Press, 1995, pp. 83—112. 13. Bourgain, I., Harmonic analysis and combinatorics: How much may they contribute to Math. Society (2000), each other?, Mathematics: Frontiers and Perspectives, 13—32.
14. Carbery, A., The boundedness of the maximal Bochner-Riesz operator on L4(R2). Duke Math.). 50(1983),409—416. 15. Carbery, A., Restriction implies Bochner-Riesz for paraboloids, Math. Proc. Cambridge Philos. Soc. 111, no. 3 (1992), 525—529. 16. Carleson, L. and SjOlin, P., Oscillatory integrals and a multiplier problem for the disc, Studia Math. 44(1972), 287—299. 17. Christ, M., Restriction of the Fourier transform to submanifoids of low codimension, Thesis, U. Chicago, 1982. 18. Cérdoba, A., Maximal functions, covering lemmas and Fourier multipliers, Harmonic analysis in Euclidean spaces (Proc. Sympos. Pure Math., Williams Coil., Williamstown, Mass., 1978), Part 1, pp. 29—50, Proc. Sympos. Pure Math., XXXV, Part, Amer. Math. Soc., Providence, RI, 1979. 19. Córdoba. A., The Kakeya maximal function and the spherical summation multipliers, J. Math. 99(1977), 1—22. 20. Dc Carli, L. and losevich, A., Some sharp restriction theorems for homogeneous manifolds,). Fourier Anal. AppL 4(1998). 105—128. 21. Fefferman, C., Inequalities for strongly singular convolution operators. Acta Math. 124 (1970). 9—36.
22. Fefferman, C., The multiplier problem for the ball, Ann. of Math. 94(1971), 330—336.
23. Fefferman, C., A note on spherical summation multipliers. Israel J. Math. 15 (1973). 44—52.
24. Foschi, D. and Klainerman, S., Bilinear space-time estimates for homogeneous wave equations, Ann. Sci. Ecole Norm. Sup. (4) 33, no. 2, 211—274. 25. Ginibre, .1., Le probléme de Cauchy pour des EDP semi-linéaires pénodiques en variables d'espace, Séminaire Bourbaki 1994/1995, Astérisque 237(1996), Exp. 796, 163—187. 26. Green, B., Roth's theorem for the primes, preprint. 27. Katz. N.. Labs, I., and Tao, T., An improved bound on the Minkowski dimension of Besicovitch sets in R3, Annals of Math. 152 (2000), 383—446. 28. Katz. N. and Tao. T.. Recent progress on the Kakeya conjecture, Publicacions Matematiques. Proceedings of the 6th International Conference on Harmonic Analysis and Partial Differential Equations, U. Barcelona, 2002, pp. 161—180.
29. Kenig, C.. Ponce, G., and Vega, L., A bilinear estimate with applications to the KdV equation,). Math. Soc. 9(1996), 573—603. 30. Klainerman, S. and Machedon, M., Space-time estimates for null forms and the local existence theorem, Comm. Pure AppL Math. 46(1993), 1221—1268.
242
T. Tao
34. Klainerman, S., Rodnianski, L and Tao, T., A physical space approach to wave equation bilinear estimates. Dedicated to the memory of Thomas EL Wolff. J. Anal. Moth. (2002). 299—336.
32. Laba. L and Tao. 1., An x-ray estimate in R", Revista Mat. iberoam. 11(2001), 375—407. 31 Laba, L and Tao T.. An improved bound for the Minkowski dimension of Besicovitch sets in medium dimension, Geom. Funet. Anal. 11 (2001). 773—806. 34. Laba, L and Wolff, T.. A local smoothing estimate in higher dimensions. Dedicated to the memory of Thomas EL Wolff. J. Anal. Math. Xg(2002). 149—171. 31 Lee, S., Improved bounds for Bochner-Riesz and maximal Bochner-Riesz operators, Duke Math. J. 122(2004), no. L 205—232. 36. Mockenhaupt, G., A restriction theorem for the Fourier transform, Bull. Amer Math. Soc. 25(1991). 31—36. 31.
Mockenhaupt. G., A note on the cone multiplier. Proc. Amer Math. Soc. 112 (1993), 14c—1S2
Mockenhaupt, G., Bounds in Lebesgue spaces of oscillatory integrals, Habilitationsschrift. U. of Siegen, 1996. 39. Mockenhaupt. 0.. Salem sets and restriction properties of Fourier transforms, GAFA 10 35.
(2000), 1579—1587. 49. Mockenhaupt. 0. and Tao. T., Restriction and Kakeya phenomena for finite fields, Duke Math. J. 12.1 (2004). no. L 4J. Mockenhaupt. 0.. Seeger. A.. and Sogge, C.. Wave front sets, local smoothing and Bour207—218. gain's circular maximal theorem. Ann. of Math. 136 42. Moyua, A.. Vargas. A., and Vega, L., Schrodinger maximal function and restriction properties of the Fourier transform, intema:. Math. Res. Notices, no. i.6j1996), 793—815. 43. Moyua. A.. Vargas, A.. and Vega, L., Restriction theorems and maximal operators related no. 3 (1999), 547—574. to oscillatory integrals in R3, Duke Math. J. 44. Presini, E., Restriction theorems for the Fourier transform to some manifolds in K'7, in Harmonic Anal rsi.s in Euclidean Spaces. (Proc. Sympos. Pure Math. part L 1979, pp. 101—109.
Sjölin, P. and Soria, F.. Some remarks on restriction of the Fourier transform for general no. 2 (1999), 655—664. measures. Publ. Ma:. 46. Stein. E.M.. Some problems in harmonic analysis, in Harmonic Analysis in Eudidean Spaces (Proc. Sympos. Pure Math.. Williams CoIl.. Williamstown. Mass.. 1978). Part L 45..
pp. 320. 41 Stein. E.M., Harmonic Analysis, Princeton University Press. 1993. 4.& Stein, EM.. On limits of sequences of operators, Ann. of Math. 74(1961). 140—170 49. Strichartz, R.S., Restriction of Fourier transform to quadratic surfaces and decay of solutions of wave equations, Duke Math. J. 44 (1977), 705—774. 511. Sunouchi, 0.. On the summability almost everywhere of the multiple Fourier series at the critical index. Kodai Math. J. g(1985), 5.1.. Tao, T., Weak-type endpoint bounds for Riesz means. Proc. Amer Math. Soc. 124(1996), 2797—2805.
Tao, 1.. The Bochner—Riesz conjecture implies the Restriction conjecture, Duke Math J. 96 (1999. 363—376. 51 Tao, T.. The weak-type endpoint Bochner—Riesz conjecture and related topics, indiana Math. J. 41(1998). 1097—1124. 54. Tao, T.. On the maximal Bochner—Riesz conjecture in the plane, for p < L Trans. Amer 52.
Math. Soc. 354(2002), 1947—1959.
55. Tao. T., Endpoint
bilinear
restriction theorems for the cone, and some sharp null form
estimates, Math Z. 238 (2001), 215—268.
Recent Progress on Restriction
243
56. Tao. 1.. Non-endpoint bilinear cone restriction theorems, unpublished. 57. Tao, T.. Multilinear weighted convolution of L2 functions, and applications to non-linear dispersive equations. I. Math. 123 (2001), 839—908. 58. Tao, T.. From rotating needles to stability of waves: emerging connections between cornbinatorics, analysis, and POE, Notices Math. Soc. no. 3(2001), 294—303. 59. Tao, 1., A sharp bilinear restriction estimate on paraboloids. GeonL Fund. Anal. 13, no. 6(2(X)3), 1359—1384.
60. Tao, T., Vurgas. A., and Vega. L.. A bilinear approach to the restriction and Kakeya conMath. Soc. 11(1998). 967—1000. jectures. J. 61. Tao, T. and Vargas, A.. A bilinear approach to cone multipliers I. Restriction Estimates, GAFA
10(2000),
185—215.
62. Tao. 1. and Vargas. A., A bilinear approach to cone multipliers II. Applications. GAFA JO (2000). 216—258.
63. Tomas. P.. A restriction theorem for the Fourier transform. Bull.
Math. Soc. 81
(1975). 477—478.
64. Wolff, T., An improved bound for Kakeya type maximal functions. Revista Mat. Iben,americuna. 11(1995), 651—674. 65. Wolff. T., Recent work connected with the Kakeya problem, in Prospects in Matheinaiks (Princeton. NJ. 1996). pp. 129—162, Amer. Math. Soc.. Providence. RI. 1999. 66. Wolff. T.. A mixed norm estimate for the x-ray transform, Revisia Mat. Ibernamericana 14(1998), 561—600. 67. Wolff, T.. Maximal averages and packing of one-dimensional sets, in Proceedings of the International congress of Mathematics. Vol. Il, Berlin, 1998, pp. 755—764. 68. Wolff, T., Decay of circular means of Fourier transforms of measures, Internat. Math.
Res. Notices 10 (1999), 547—567. 69. Wolff, 1.. A sharp bilinear cone restriction estimate, Annals of Math. 153 (2001). 661— 698.
70. Wolff, T., Local smoothing estimates in for large p. GAFA 10(2000). 1237—1288. 71. Zygmund, A., On Fourier coefficients and transforms of functions of two variables, Studia
Math. 50(1974),
189—201.
Average Decay of the Fourier Transform Giancarlo Travaglini Dipartimento di Matematica e Applicazioni, Universiti di Milano-Bicocca, Edificio U5, Via R. Cozzi 53. 20125 Milano, Italy
giancarlo. travag].ini@unintib.
it http: / /www.matap. unizaib. it/"travaglini/
aim of this paper is to describe some recent results and applications of the average decay of the Fourier transform of measures or characteristic functions spherical of convex bodies. We first focus on the different results depending on the indices p and on the shapes of the bodies. We then consider applications to geometric number theory and to irregularities of distributions, as well as to generalized Radon transforms. The results are mostly
Summary. The
known and they are described in the 2-dimensional case, even when extensions to several variables are known to be true.
1 Introduction Consider a convex body (a convex compact set with nonempty interior) B C R2 and the Fourier transform of its characteristic function
= A classical problem concerning
is given by the study of its decay as
I
+00 (see e.g. [22] and [21], see also [44] and [43] as general references on Fourier analysis). As a first example, let us consider the unit disc D. In this case
=
J1 (2ir III)
irt
— 3,r/4)
+00, where ii is the Bessel function of order 1. Of course, for an arbitrary 0) and depends on the direction oft. Consider e.g. = B, the decay of note that as
=
ff
dxjdx2
246
G. Travaglini
where f(s) is the length of the chord in Figure 1. We call f the parallel section function (with respect to a given direction).
Let us now replace B by a square Q. If we consider (Figure 2) the Fourier transform in a direction orthogonal to a side of Q, then the graph of the parallel section function f(s) is discontinuous (see Figure 3), and its Fourier transform f decays of
0
8
e
order 1. On the other hand, for most directions (Figure 4) f(s) is piecewise smooth and continuous (Figure 5), and f decays of order 2. Of course, the study of in a given direction can be very difficult. On the other hand, several problems do not require precise estimates in every direction, but
Average Decay of the Fourier Transform
Fig. 3
Fig. 4
I
247
248
G. Travaglini
rather a global information. This is the case e.g. when one needs to integrate the Fourier transform in a suitable sense. The content of this paper is as follows. First we will introduce the circular average decay of the Fourier transfonns of characteristic functions of convex planar bodies. Then we will consider several applications: to geometric number theory, to irregularities of distribution and to generalized Radon transforms. Almost all the results are known, and some of them have been extended to several variables. We will try to give precise references.
2 Average decay of Fourier transfonns 2.1 L1 circular means for Fourier transforms of characteristic functions of polygons Given a convex planar body B we first consider the following L1 circular average: f22T
IXB(P®)IdO.
I
Jo
(9 = (cos 9, sin 9). When B is a disc the average is meaningless (and essentially the same happens when a B is smooth with positive curvature). The case of a polygon P is different. By the divergence theorem we have where
I Jp
dt
= 2ir
I
.
v(t)dS,.
is the measure on a P. We split the last integral according to the sides of P and we consider one of them (so that the vector v is constant), which we can assume to have extremes (± 1,0). In this way, the Fourier transform (p®) is controlled by a finite sum of terms of the form
where v is the outward normal vector and
p
f
—
sin(22rpcos 9)
—
so that ,2,r I
Jo
,2,r
Ixp(p®)IdO_
1
Jo
cosO
dO
Using an induction argument it is possible to extend (2) to several variables (see [7]).
There is a more geometrical way to get (2). Consider a convex planar body B (Figure 6) and denote by A(S, 9) the chord at (small) distance S from the boundary of B, with respect to the direction 0. The following geometric bound for the Fourier transform is well known (see e.g. [131, [34], [12], [36)).
Average Decay of the Fourier Transform
249
FIg. 6
Theorem 1. Let B be a convex planar body. For every direction 8 we have
+
cp
lX(p.9 +r)I)
(3)
where IAI is the length of the chord and c depends only on B.
The above right-hand side (RHS) can be written through the paraHel section func-
tion f. Since B is a convex body there is no restriction assuming the support of f to be the interval [—.1, 1]; we also note that f is concave on its support. Then (3) reduces to
(f(_L +
II(t)I
+ fU
—
Itl'))
.
(4)
Assume for simplicity f continuous. Integrating by parts, we have to bound a term
of the form
j
—f'(s) sin (2jrst)
(5)
where —f'(s) is increasing in [0, IJ. The graphs of the functions —f(s) and
—f'(s) sin (2,rst) are therefore as in Figure 7, and the integral in (5) is essentially a Leibniz sum, bounded by its largest term
f
I—Ill—'
A. Podkorytov ([35]) has proved that (4) is sharp, i.e. that there exist a positive constant c and a real diverging sequence such that the parallel section function f satisfies
II(tj)I
c
(f(—1 +
+ f(' — ItJI))
250
G. Travaglini
1
Fig. 7
We recall that (3) cannot be extended in this form to convex bodies in more than two variables. Indeed, consider iiV and a cube, the side of which has length Choose the direction of a main diagonal. The corresponding parallel section function f(s) is even with support (—I, I], and it represents the area of the "slice" of the cube at distance si from the origin. Observe that f(s) is piecewise smooth but not everywhere differentiable in (— 1. 1), moreover it vanishes of second order at ± 1. Then,forlargei > If(flI > so that the analogue of (3) fails. The easy argument of this last example does not work if we consider a convex body B with smooth boundary. This problem has been studied in (131 and 111. where it has been shown that an analogue of (3) exists in 1R3 only under additional geometric assumptions. We go back to the case of a polygon, for which it is easy to estimate the lengths of the chords A(p , 0) defined in Figure 6 and then deduce (2). As another application of (3), one easily obtains the well-known bound
Ixü(UI
c•IEI
-.
which is true for every convex planar body B, the boundary of which is smooth with everywhere positive curvature. It can be shown that (2) is best possible. Indeed, for any convex planar body B, we have
Jo
>0.
(7)
The proof of (7) depends on a modification of an argument used by Yudin (1481) in the study of Lebesgue constants (see 1111). Note that a connection between dO and f12
dx seems to be natural (see also 132)).
The bound in (2) is also essentially best possible in a perhaps more subtle sense. If we assume P = Pp.j to be a polygon with N sides, contained in the unit disc, then the above argument (divergence theorem + splitting the integration according to the sides) shows that
Average Decay of the Fourier Transform
251
Fig. 8
,2,r
cNp2logp.
J0
where now c is independent of N. We show that (8) is essentially best possible, since for any e > Owe cannot replace N in the above RHS by NI_t. Indeed, assume PN is a regular N-gon inscribed in the unit disc D and consider XD\PN (p8). Here the parallel section function f(s) is the sum of the lengths of the two small segments obtained after removing the chord of the polygon from the chord of the disc in Figure 8. The set D\ PN is the union of N "lunes." Let us number them counterclockwise and let us consider only the first [N/21 lunes for simplicity. Figure [N/21) to the total variation 9 shows that the contribution of the kth lune (1 k
of the parallel section function f(s) is
Adding on k and using the theorem on the Fourier transform of a function of bounded variation we get, uniformly in 9,
1logN On the other hand (l)unplies
Jo
IXD(PNO)I dO =
for a suitable diverging sequence (e small). Then
—I
(2lrpN)I
?
—3/2
C2PN
We can choose PN so that PN
G. Travaglini
252
e
0' Fig. 9
f2Jf
t2,r
J0
IIPNpNO)Idoa
IXD(PN8)I dO —
IXD\PN(PN9)
cjN3'2 — In order to end the proof, let us assume that K = K(N) satisfies
K(N)p2 log p.
IXPN(P8)I dO Jo
Then choosing again p = PN
N
The bound in (9) came out as a very partial answer to a question raised a few years ago by A. Koldobsky. The above result has not been useful to him, but it was later one of the basic ingredients in [11]. 2.2
drcular means for planar convex bodies
Let us consider a convex planar body B and more general U' circular means:
Average Decay of the Fourier Transform
253
1,'p
22r
f
lme(PG)l"
dO
0
Arguing as in the L' case, one can prove the following best possible bound for a
polygon P when I
f
2jr
0
(see
[7] for an extension to several variables). Note that for a disc D (1) implies 2,r
J
IXD(P®)1'
dO
0
so
that the above two estimates agree if p =
2.
This is a general fact, since Podko-
rytov ([34]) has proved the following theorem.
Theorem 2. Let B be a convex planar body. Then
f
11/2
IXB(P®)12 dO
o
J
We stress that Podkorytov's theorem requires no regularity assumptions on dB (see [9] for the extension of Podkorytov's result to several variables, see also [47]). The proof of(l0) uses (3), see also [36]. We sketch a different argument (see 110]) for the essentially equivalent inequality 1/2
dO
J
0
denotes the restriction of the Lebesgue measure to a convex finite arc (i.e. the graph of a convex real function defined on a compact interval). In order to study (11) we write where
I'
J0
dO
= J0
J
R2
= 2jr
4.
R2
[
0
f Jo (2irp Ix
JR2 JR2
— yI)
where J0 is the Bessel function of order 0. By convexity, x — y is "close" to the arc length and one can reduce the two integrals to one: p
__
254
G. Travaglini
fI I Jo(ps) ds =
Jo
pp
Jo(u)du Jo
Podkorytov's result (10) is best possible. Indeed it has been proved in [28] that for every set E C R2 with finite nonzero measure we have
/
IE
(12) R
and this implies 1/1 1/1
limsup
p- -
1
> 0.
dO
(13)
Jo
In certain cases we have a better result. For example, see [71, we have I
I
I"
cp
dO
I
—v, -
Jo
for every triangle T. The inequality (14) can be extended to other polygons. but not to all. indeed, see [7), for a square Q we have 1/"
dO} for every positive integer k. We then have the following curious fact: perhaps for
most planar convex bodies (13) can be turned into an inequality as in (14); however this does not happen for the two main examples: the square and the disc (this last fact is due to the zeroes of the Bessel function). At this point we wish to describe a phenomenon occurring when p 2. Let us say that a convex planar body B has p-order (p > I) of decay equal to a if
I
Jo
J
and
/
2,r
>0.
IIa(pe)I"do
0
We have the following result for the family of convex planar bodies having piecewise smooth boundary. Let
111
\
3
a=—ora=1+--2
p
p
.
so
that the set S U T is the set in Figure 10. Then we have the following.
p
2
Average Decay of the Fourier Transform
255
2
112
1
Fig. 10
Theorem 3. Let p > i. There exists a convex body B (having piecewise smooth boundar') with p-order of decay equal to a the set SU T. The difference between the cases p
and only (f the pair
a) belongs to
< 2 and p > 2 can be explained in the
following way.
When p < 2 the above result essentially says that polygons have p-order of decay equal to 1 + I/p. while for all other bodies we have 3/2, i.e. they behave like discs. Observe that if B is not a polygon. then its (piecewise smooth) boundary must contain an arc with positive curvature. Hence we get sharp uniform order of decay 3/2 on a positive interval of 9's, and this (together with (10)) implies the average 3/2 over 10. 2ir) when p 2. The situation is different when p > 2, where (consider for a moment the extreme case p = 00) the flat points of the boundary are the relevant ones. Here one can produce a scaling between the polygons and the disc by constructing suitable convex bodies containing a piece of the graph of the function x —+ (y > 2) within their boundaries. The above dichotomy is no longer valid for arbitrary convex bodies, where the average decay describes global geometric properties of B, as we shall see in the following two sections. 2.3 InscrIbed polygons
Given a planar convex body B we consider (as in 134], 141] or [36]) the following inscribed polygon. Choose any chord at distance S from the boundary (as in Figure 6) and name it Let us move counterclockwise constructing a finite sequence of consecutive chords (each one at distance S from the boundary) until we reach again. Then, if necessary, we replace the last chord by one consecutive to £ i. In this way we get a polygon inscribed in B and we denote it by Pf (once S is given, this polygon is uniquely determined up to the choice of the starting point, and this latter turns out to be irrelevant). Let Mf be the number of sides of It is known (see whenever [41]) that Mf Observe that Mf contains a piece of a curved arc, while, on the other hand. Mf I if B is a polygon. We have the following (see [Ill).
256
G. Travaglini
B
Fig 11
Theorem 4. Let B be a convex planar body and con3ider the inscri bed polygon P..1
(see FIgure 11). LetO_
Then
cp"2 logp.
I
Jo
(15)
for any 0 < a < 1/2. there exists a convex planar body B such that cp" and, for any e > 0, Ixa(p8)1d8 > 0.
limsup p—,+oo
JO
The fist step in the proof of Theorem 4 is to prove the inequality
ç2r
I
Jo
which shows that the polygon
Jo
Ipa (p0) dO,
is a good substitute for B while studying
xB(p8). We ate therefore reduced to estimate the average decay for a polygon with no more than cpa sides. In order to get (15) we then apply the trivial estimate (8) with p" in place of N. At this point one should expect to have obtained a poor result. The second part of the statement of Theorem 4 shows that it is not so. The counterexample follows the idea which has been used to prove (9). 2.4 Measuring the Image of the Gauss map We can get similar results starting from a different point of view. We think B close to a polygon when its boundary has relatively few normals. We then recall that,
Average Decay of the Fourier Transform
257
by convexity, at every point of a B there is a left and a right tangent, therefore a left and a right outward normal. Let C [0. 2,r) be the whole set of the directions is the image of the appearing as a left or right outward normal. In other words, (generalized) Gauss map. We say that a B has "few" or "many" normals according to We measure this set in a fractal way by defining its s-neighborhood the "size" of
<sj
= fo and assuming an inequality of the form
If B is a disc, we need d = 1. On the other hand, we can choose d = 0 if and only if B is a polygon with finitely many sides. As an intermediate example, let us consider a polygon with infinitely many sides, such that the set of the normal directions to its sides is = (y > 0). Then it is not difficult to show that 5y/(y+l) We have the following essentially sharp result (see [I 1]).
Theorem 5. Let 0
1 and assume that
Then
jfO
5211
J0 We
IXB(Pe)I dG
1
ifd=0
cp2logp
sketch the first part of the proof of Theorem 5. We can assume d > 0 and
write 5
5
I
JO
IIB(pe)1d6= I
118(pe)Ide
I
JlO. 2r )\
/(d.. I j
=11 + 12. By the Schwarz inequality, (16) and Podkorytov's L2 result (10) we have 1/2
'I
1/2
2,r
IXB(PE3)12
dO}
= The bound for 12 depends on (3) and it is more technical (see [11]).
A weaker form of this theorem can be obtained directly from Theorem 4.
25K
3
G. Travaglini
Lattice points and irregularities of distribution
3.1 Lattice points Let p8 be the dilated copy of a convex planar body B. Elementary geometric considerations show that
card(pBflZ2) —p2
181
and
=—p2181+ flIEZ2
=o(p). +00. D8(p) is called the diserepaner. as p Let us replace B by a disc D. Then the problem of improving (17) is known as the eircle problem, and it has counted several milestones during the last century. Among them we point out Sierpinski's early estimate (1906)
DD(p) = 0(p213) and Hardy's
D0(p) = c(ph/2logh/4p) (due to M. Huxley 1231).
See 1291 and 1241. The best estimate so far is 0
Sierpinski's estimate (18) can be proved through a well-known Fourier analysis argument. which may be worthwhile to recall. The proof does not change if we consider the more general case of a convex planar body B having a smooth boundary with everywhere positive curvature. We start considering the convolution xpn * (here p is a smooth positive function supported in the unit disc and such that = I. f'p = 1. while =
p2/3p (p1"3:)).
Then,
by the Poisson summation formula and (6). we have
*
XpB(')
(in)
,UEZ2
=
X(p+p
+
=
((p +
rn)
In €22
= (p + =
p2
181
181
+0
+
(p2/3)
(p
+
Average Decay of the Fourier Transform
259
and since a similar bounds holds from below, the proof is complete. Here the positive curvature has been crucial (see [24. Ch. 2] or [251 for several remarks on this point). If one replaces B by a unit square Q having sides parallel to the axes, then DQ(p) can be nothing better than 0(p). The same happens when Q has a rational slope, but when the slope of Q is irrational, the problem becomes difficult. Hardy and Littlewood ([19], 1201) have proved that in this case the discrepancy DQ(p) can be 0 (log p) (see also [42]). Davenport ((17)) proved that when the slope of Q is. e.g.. a quadratic irrational, then 1/2
JV Results on bounds for -I(p)(P) for almost every a SO(2) have been obtained in [161, [6] or [26]. Combining the above argument in the proof of Sierpinski's theorem and (2), one can prove the following average result (which extends to several variables, see (371. 1461 and [71).
Theorem 6. Let P be a polygon in 1R2 Then
f
da
clog2
50(2)
p.
Averaging over rotations and translations (i.e. considering a convex body thrown at random in the plane) leads to more complete results. We describe two of them. Being Z2 periodic, it is enough to translate inside T2. Theorem 7. Let T be a triangle and I < p
I'
I.
dad:
JV J
.
(20)
Theorem S. Let B be a convex planar body different 1mm a polygon and having a
piecewise smooth boundary Let I'
Cl p112
I
I
p
2. The,,
1'
dad:
I
JT2 JS02
il/p 5 C2 J
The proof of these results can be found in [27], [7] or (121. Let us look at the proof of (21). We consider the function t
Then, if in
0.
'(R(—t(P) =
P2 IBI +
+ 1)
G.Travaglini
260
(DaI(B)_(.)(P))
=
+ t)e_2mflm.t di
T- kEZ2
—2,rim-: j
i
at
= JR2
(po'(m)),
= while (DC-I(B)_(.)(p))
= 0. Then, by (10), dadt = p4f
ff 12
IXB (pa(m))12 dci
SO(2
(p mI)3
:5 cp. As for the estimate from below, one observes that, for any th
ff T2
SO(2)
IDU_4B)_:(P)I dcrdt
? p2f
0,
IXB (pa(fii))I dci.
SO(2)
's one gets the desired lower bound (see [12]).
Choosing suitable
The arguments in Sections 2.3 and 2.4 provide intermediate results between (20) and (21). As an example, keeping the notation in Section 2.3, we have (see [111) the following.
Theorem 9. Let B be a planar convex body such that
(0
< 1/2).
Then
fJ T2
3.2
SO(2)
Irregularities of distribution
Suppose (u(j)}7_1 is a distribution of N points in the unit square U = [0, 112, treated as '11.2. The term irregularities of distribution usually means that these N points cannot, in a certain sense, be too evenly distributed. The following theorem, due to Roth ([39]), is a fundamental result in this field. Theorem 10. For every finite set parallel to the axes, such that
there exists a square Q C p112, with sides
Average Decay of the Fourier Transform
261
Roth obtained the above inequality as a consequence of the following stronger result. Let Q
set
= {sQ
—
consists of all squares in T2 with sides parallel to the axes. Then
dids _>clogN.
(22)
The above results are best possible and they extend to higher dimensions (see [17] and [40]). The RHS in (22) turns into cN112 after replacing squares by discs. More generally, one can consider dilated, translated and rotated copies of a given convex body (having diameter smaller than I) B c T2. Beck and Montgomery ([4J, [311) independently proved the following result.
Theorem 11. Let B be a convex body in T2. For every finite set
C T2 we
have
dtdads
10 ISO(2)ITZ
cN1"2.
(23)
This bound is sharp too (see [5]). The above theorem and the corresponding upper bound have been proved in several variables. More techniques involving Fourier analysis and irregularities of distribution are described e.g. in [4], [31], [7], [30], [3] or [15].
We now use arguments from [31] and from [11] and some results from Section 2.2 to show that for certain choices of B the bound (23) holds, and it is sharp even without averaging over dilations. Theorem 12. Let T be a triangle in T2. For every finite set
ISO(2) 1T2
ISO(2) f —N
dida
TI +
Moreover there exists a finite set body B we have
IBI +
C
C
we have
(24)
such that for every convex planar
dtda
cN112.
(25)
262
G. Travaglini
Proof. It is not difficult to prove that the Fourier transform of the function N r
—* —N TI + j=I
Is N
efl" XT(0'(P?l)) for in
0. while it vanishes for in = 0 . Then, by the Parseval identity. N-
L02 JT2 —N TI +
x0
I/N =
ISO(2)4
I
IN
2
I
= Ij=I
(J(7
S0c2)
I
Now we need the see 131. Ch. 5. Theorem 121. that for every unite set (u(j)}7_1 C T2 we have
IN
2 I
N2 11=1
where QN =
= (xI.x2) : lxii
IX2i
then have
I/N IIT)_s(U(J))) — N
I
I I
I I
Ij=I
I
I
eN1"2.
,
(26)
I
By (26) and (14) we
Average Decay of the Fourier Transform
263
so that (24) is proved.
We now sketch the proof of the second part of the theorem (see LI I] for the details). Suppose first that N is a square. N = f2• Then we can put the £2 points in a "standard grid" in T2 so that its periodic extension is (a translation of) the grid £—'Z2 in 1R2. We are therefore back to a lattice points problem in The difference is that now we shrink the lattice Z2 by £ instead of dilating the body by £ = = If N is not a square, we write Hence, by (21), the LHS in (25) is it as a sum of four squares, N = m2 + n2 + p2 + q2 and we dispose the N points in the union of four suitable disjoint grids (the whole argument extends to the case U d?2). The above result says that a standard grid provides the smallest possible L2 mean for the discrepancy. This is not true for L" means when p > 2. Indeed Chen (1141) has proved (in several variables) the following result.
Theorem 13. Let B be a convex body in T2 and let 2 < p <+00. Then for any N there exists a finite set {u(j )
C T2 such that
—
Ns2
dtdadsj
eN114.
This is not an easy generalization of the previous result. Indeed, if B is a polygon is a standard grid, then (20) provides only the upper bound and (u (j )
Therefore Chen's result says that (for L" means. p >
2.
and for polygons) the
standard grid is not the best way to dispose N points. Unfortunately, his proof is of a probabilistic nature and we do not see the 'better way."
4 Generalized Radon transforms The boundedness of certain operators in Fourier analysis depends on precise estimates for Fourier transforms of suitable functions or distributions. In this section we wish to show that certain estimates for the average decay of the Fourier transform can be useful in this setting. We shall consider Radon transforms and convolution operators associated to singular measures. What follows extends to higher dimensions, and it is of interest when the curves or the surfaces involved are not smooth. since in the regular case a general theory is available (see 118] or 181). We consider the classical planar Radon transform. For any Zi and t
Rf(a.t)=f
IR
let
f(x)dx 0=1
for suitable functions f. Oberlin and Stein (1331) have proved that L"(1R2) —* L" (E1 x R)
(27)
G. Travaglini
266
2 similar to the proofs of (28) and (31), and the required L
bound
(SO(2) x
L- (1R)
is
I
JSO(2)
1/2
f
which holds true since its LHS is
}I/2
ff
(JSO(2
C
I
by the L2 average decay result for curves (II). norm with a One can obtain sharper results by replacing the L2 (S 0(2) x mixed norm and applying the results in Theorem 3 to prove inequalities of the form
f
SO(2
I/p•
f (f
*
dx
do
A related argument can be used to obtain average results for the restriction phenomenon (see [10]).
References I. Bak. J.. McMichael. D.. Vance. J.. and Wainger. S.. Fourier Irans%.rms of surface area measure on convex surfaces in 1R3, Amer J. Math. 111 (1989). 633—668.
2. Beck. J.. Irregularities of distribution I, Ada Math. 159(1987). 1—49. 3. Beck. .1., Lattice point problems: Crossroads of number theory, probability theory and Fourier analysis, this volume. pp. 1—35. 4. Beck J. and Chen, W.W.L.. Irreguluritiex of Di.vtrthution. Cambridge Univ. Press. 1987.
5. Beck. J. and Chen. W.W.L.. Note on irregularities of distribution. II, Pive. London Mat/i. Soc. 61(1990). 251—272.
6. Brandolini. L.. Coizani. L., losevich. A.. Podkorytov. A.. and Travaglini. G.. Geometry of the Gauss map and lattice points in convex domains. Mathematika. 48 (2001). 107-117. 7. Brandolini, L., Colzani, L.. and Travaglini. G.. Average decay of Fourier transforms and integer points in polyhedra. Ark. Mw. 35(1997). 253—275., estimates for overdetermined — 8. Brandolini, L.. Greenleaf. A.. and Travaglini. G.. LI' Radon transforms. preprint. 9. Brandolini. L.. Hofmann. S.. and losevich, A.. Sharp rate of average decay of the Fourier transform of a bounded set, Geom. Func:. Anal. 13 (2(X)3). 671-680. 10. Brandolini. L.. losevich. A.. and Travaglini, G.. Spherical means and the restriction phenomenon. J. FourierAnal. App!. 7(2001), 359—372. II, Brandolini, L.. losevich. A., and Travaglini, G., Planar convex bodies. Fourier transform. lattice points, and irregularities of distribution. Trans Amer Math. Soc. 355 (2003). 3513— 3535.
12. Brandolini. L.. Rigoli. M., and Travaglini. G.. Average decay of Fourier transforms and geometry of convex sets. Rei: Ma:. lberna,n. 14 (1998). 519—560.
Average Decay of the Fourier Transform 13. Bruna, J.. Nagel.
14.
267
A.. and Wainger. S.. Convex hypersurfaces and Fourier transforms. Ann.
Math. 127 (1988). 333—365. Chen, W.W.L., On irregularities of distribuiton III, J.
Math. Soc. 60 (1996), 228—
244.
15. Chen, W.W.L., Fourier techniques in the theory of irregularities of point distribution. pp. 59—82.
16. Cohn de Verdiere. Y.. Nombre de points entiers dans une famille homothétique de domains de R, Ann. Sci. École Norm. Sup. 10(1977). 559—575. 17. Davenport. H.. Notes on irregularities of distribution, Mathetnatika 3(1956), 13 1—135. 18. Greenleaf. A. and Seeger. A.. Oscillatory and Fourier integral operators with degenerate canonical relations, in Proceedings of the 6th international Conference on Harmonic Analy.ris and Partial Differential Equations (El Escorial, 2000), PubI. Mat.. Vol. Extra. 2002. pp. 93—141.
19. Hardy, G.H. and Littlewood. i.E., Some problems of Diophantine approximation: the lattice points of a right-angled triangle. Proc. London Math. Soc. 20(1922). 15—36. 20. Hardy. G.H. and Littlewood. I.E.. Some problems of Diophantine approximation: the lattice points of a right-angled triangle (second memoir). Abh. Math. Sern. Unit: Hamburg 1922). 212—249. 21. Herz. C.S.. Fourier transforms related to convex sets. Ann. of Math. 75 (1962), 81—92. 22. Hlawka. E.. Uber Integrale auf konvexen Korpern. I and II. Monaish. Math. 54 (1950). 1—36 and 81—99.
23. Huxley, M.. Exponential sums and lattice points II, Proc. London Math. Soc. 66 (1993). 279—301.
24. Huxley. M.. Area. Li.utice Points, and Erponential Stuns. Oxford Science Publications. 1996.
25. losevich. A., Curvature, combinatorics, and the Fourier transform, Notices Amer. Math.
Soc. 48(2001), 577—583. 26.
losevich. A.. Lattice points and generalized Diophantine conditions. J. Number Theory
90(2001).
19—30.
27. Kendall. D.G., On the number of lattice points in a random oval, Quart. J. Math. Oxford
Series 19(1948),
1—26.
28. Kolountzakis, M. and Wolff. T.. On the Steinhaus tiling problem. Mathematika 46, no. 2 (1999), 253—280.
29. Krätzel. E.. Lattice Points, Kluwer Academic Publishers, 1988. 30. Matousek, J.. Geometric Discrepancy, Springer-Verlag, 1999. 31. Montgomery. H.L.. Ten Lectures on the inre,face Between Analytic' Number Theory and Harmonic Analysis, CBMS Regional Conf. Ser. in Math. 84. Amer. Math. Soc. Providence. RI, 1994.
32. Nazarov, F. and Podkorytov. A.N.. On the behaviour of Lebesgue constants for twodimensional Fourier sums over polygons, St. Petersburg Math. J. 7 (1996). 663—680. 33. Oberlin, D.M. and Stein. E.M., Mapping properties of the Radon transform. indiana Univ. Math. J. 31(1982). 641—650.
34. Podkorytov. A.N., The asymptotic of a Fourier transform on a convex curve. Vesin. Leningr Univ. Mat. 24 (1991). 57—65. 35. Podkorytov. A.N.. Personal communication, 2001. 36. Podkorytov. AN.. What is it possible to say about an asymptotic of the Fourier transform of the characteristic function of a two-dimensional convex body with nonsmooth boundary?, this volume. pp. 209—215. 37. Randol, B., On the Fourier transform of the indicator function of a planar set, Trans. Atner. Math. Soc. 139(1969), 27 1—278.
268
G. Travaglini
38. Ricci, F. and Travaglini. G., Convex curves, Radon transforms and convolution operators defined by singular measures, Proc.
Math. Soc. 129 (2001), 1739—I 744.
39. Roth, K.F., On irregularities of distribution, Mathematika 1(1954), 73—79. 40. Schmidt, W.M., Irregularities of distributions, in Number Theory andAlgebra. Academic
Press, 1977, pp. 311—329. 41.
Seeger, A. and Ziesler, S.. Riesz means associated with convex domains in the plane,
Math. Z. 236(2001), 643—676. 42. Skriganov, M.M., On integer points in polygons, Ann. tnst. Fourier (Grenoble) 43(1993), 313—323.
43. Stein. E.. Harmonic Anal vsis: Real Variable Methods. Orthogonality and Oscillatory integrals, Princeton University Press, 1993. 44. Stein E. and Weiss, G., introduction to Fourier Analysis in Euclidean Spaces, Princeton University Press, 1971. 45. Strichartz, R., Convolution with kernels having singularities on a sphere, Trans. Math. Soc. 148(1970), 461—471. 46. Tarnopolska-Weiss, M., On the number of lattice points in a compact n-dimensional polyhedron. Proc. Amer. Math. Soc. 74(1979), 124—I 27. 47. Varchenko, A.N.. Number of lattice points in families of homothetic domains in R". Funk An. 17(1983), 1—6. 48. Yudin, V.A., Lower bound for Lebesgue constants, Math. Zametki 25 (1979), 119—122.
Fourier Analysis and Convexity LUCA BRANDOLINI LEONARDO COLZANS,
ALEX IOSEVICH, AND GIANCARLO TRAVAGLINI, EDITORS
Over the course of the last century. the systematic exploration of the relationship between Fourier analy8is and other branches of mathematics has lead to important advances in geometry. number theory. and analysi8. stimulated in part by Hurwitz's proof of the isoperimetnc inequality using Fourier series.
This unified, self-contained volume is dedicated to Fourier analysis. convex geometry. and related topics. Specific topics covered include: • the geometric properties of convex bodies • the study of Radon transforms • the geometry of numbers • the study of translational tilings using Fourier analysis • irregularities in distributions • Lati we point problems examined in the context of number theory. probability theory. and Fourier analysis • restriction problems for the Fourier transform The book presents both a broad overview of Fourier analysis and convexity as well as a.n intricate look at applications in some specific settings: it will be useful to graduate students and researchers in harmonic analysis. convex geometry, functional analysis. number theory. computer science, and combinatorial analysis. A wide audience will benefit from the careful demonstration of how Fourier analysis is used to distill the essence of many mathematical problems in a natural and elegant way.
CONTRIBUTORS
J. Beck C. A. Berenstein W. W. L Chen B. Green H. Groemer
A. Koldobsky M. N. Kolountzakis A. Magyar A. N. Podkorytov B. Rubin
D. Ryabogin 1. Tao C. Travaglini A. Zvavitch
ISBN O-81'6-3263-8
Birkhguser ISBN 0-8176-3263-8 www.birkhauser.com