MNL Vol27 No1 Jun2016 1478024956074

MATHEMATICS NEWSLETTER
Volume 26 MarchJune 2016

Vol. 26 #4, No. 4 2016
MarchJune
CONTENTS
An Introduction to the Kakeya Problem . . . Edward Kroc 109
Bakers Question on Dynamics of Permutable Entire Functions . . . Kuldeep Singh Charak 125
The Man Who Knew Infinity: Why Ramanujan Matters . . . Ken Ono and Robert Schneider 131
For Immediate Release IFC Films & Pressman Films THE MAN WHO KNEW INFINITY Partners With Expii,
Ken Ono, and the Templeton World Charity Foundation On The Spirit of Ramanujan Math Talent Initiative
A Global Search for Undiscovered Mathematical Talent 133
Details of Workshop/Conferences in India 135
Details of Workshop/Conferences in Abroad 136
Visit: www.ramanujanmathsociety.org
Typeset in LATEX at Krishtel eMaging Solutions Pvt. Ltd., Chennai - 600 087 Phone: 2486 13 16 and printed at
United Bind Graphics, Chennai - 600 010. Phone: 2640 1531, 42769087.
An Introduction to the Kakeya Problem
Edward Kroc
In this essays first installment, we examined some funda- 4.1 Resolution of the Kakeya conjecture in R2
mental properties of the Kakeya maximal operator problem,
notably its connection with the set conjecture version of Although Cordoba was concerned with matters other than
the problem and how partial progress on the full conjecture the Kakeya problem, his result not only gives a quantitative
can be made by proving certain intermediate Lp estimates. upper bound for the norm of the maximal operator, it in fact
In this second installment, our main concerns will be the gives the optimal one. This settles the Kakeya conjecture in
L2 -theory and Bourgains seminal bush argument [7]. two dimensions. Unfortunately, the proof offers little insight
We will spend a considerable amount of time on the latter, as into the higher dimensional cases as it relies heavily upon
the ideas we develop will prove essential in the final install- the L2 structure. The crucial geometrical lemma below will
ment of this essay when we provide a treatment of Wolffs be our most valuable item in this main section and will be
Kakeya estimate [42]. utilized repeatedly to improve our understanding of the more
complicated higher dimensional cases. Although Cordobas
4. L2 -estimates estimate is on L2 , we state and prove the lemma in Rn , as
these critical geometric facts will be needed later in this greater
The Kakeya operator in two dimensions is well understood. generality.
This is largely due to the fact that the ambient geome-
Lemma 4.2 (Cordoba). For any pair of directions ek , el
try in the plane is often transparent and always pictorially
Sn1 and any pair of points a, b Rn , we have the
tractable. The Kakeya conjecture in R2 also enjoys a uniquely
estimates
affable position in no small part because of the powerful
duality properties of the L2 Hilbert space. These niceties
diam(Tek (a) Tel (b)) ! , (4.1)
quickly break down when we elevate the dimension, but |ek el |
the methods and estimates in this main section are highly and
classical and motivate the underlying techniques and fur- n
|Tek (a) Tel (b)| ! . (4.2)
ther results in the remainder of this exposition. Much of the |ek el |
now standard analysis has its roots in the original paper of
This lemma is the first critical geometric fact we need about
Cordoba [14].
intersections of tubes in Rn and we will use it repeatedly
In his 1977 paper, Cordoba was concerned primarily with
throughout the exposition. In line with our intuition, note that
studying the Bochner-Riesz spherical summation operators1
if the tubes lie very close together in space and in orienta-
and their boundedness on Lp (Rn ) for a certain range of p.
tion (i.e. if |ek el | ), then their intersection is nearly
Building largely off Feffermans work [19], Cordoba analyzed
2
full, |Tek (a) Tel (b)| ! n1 . If on the other hand the tubes
the boundedness of the Nikodym maximal function on R
are close in space but nearly orthogonal in orientation (i.e. if
to obtain a quantitative estimate on the norm of certain
|ek el | 1), then their intersection is essentially a -box in
Bochner-Riesz operators. In our context, we have the following
Rn , |Tek (a) Tel (b)| ! n . We proceed to the proof.
theorem.
Proof. Clearly we may assume that the tubes lie near each
Theorem 4.1 (Cordoba). In R2 , the following estimate
other in space otherwise there is nothing to prove. A typical
holds:
! high-intersection scenario is illustrated in Figure 1.
1
f L2 (S 1 ) ! log f L2 (R2 ) . We begin by proving (4.1). Recall that we may assume all
1
directions lie within 10
of the vertical. We can easily reduce
1 See Section 7.5 for a definition and brief discussion. to the case where the two tubes have a common center at the
Mathematics Newsletter -109- Vol. 26 #4, MarchJune 2016

Figure 2. A point x Tek Tel represented as the point of intersection of
two line segments with orientations ek and el , contained in their respective
tubes.
Figure 1. A typical high intersection scenario of two tubes in R2 .
1
origin. To see this, pick any x Tek (a) Tel (b); we claim that x = ml + tx el for some |tx | . (4.5)
2
Tek (a) 2 Tek (x) and Tel (b) 2 Tel (x). (4.3) Now tx and tx are both fixed numbers so we may assume that
|tx | |tx |. We want to bound |x| from above, so appealing to
Indeed, for any y Tek (a) we have (4.4) we see that it suffices to bound |mk | and |tx |. We already
have |mk | , so we only need a bound for |tx |.
|(y x) ek | |(y a) ek | + |(x a) ek | 1
Consider the triangle with vertices at mk , ml , and x as in
since x Tek (a) by assumption. Similarly, for any unit vector Figure 2. By the law of sines, we have that
v ek , we have sin sin
= ;
|mk ml | |tx |
|(y x) v| |(y a) v| + |(x a) v| 2.
consequently, we see that
A completely symmetric argument verifies the second contain-
|mk ml |
ment in (4.3). Without loss of generality we may translate the |tx | !
sin |ek el |
system so that x = 0; thus we see that
since sin = |ek el |. Evidently then, we have
Tek (a) Tel (b) 2 Tek (0) 2 Tel (0),

|x| |mk | + |tx | ! + ! . (4.6)
|ek el | |ek el |
so it suffices to show (4.1) with a = b = 0.
Let x Tek (0) Tel (0). We are going to parameterize x as The point x was arbitrarily chosen to lie in the intersection
we did in Section 3, only here we will have two parameteri- Tek (0) Tel (0), so (4.6) holds for all points in this intersection.
zations, one for each tube in which x lies. Accordingly, we Thus, we have
know that x lies on some unit line segment Ik that is parallel
to ek with midpoint mk B(0, ) lying within the unique diam(Tek (0) Tel (0)) = sup 2|x| !
xTek (0)Tel (0) |ek el |
hyperplane through the origin with normal vector ek . Similarly,
we know that x lies on some other unit line segment Il that by (4.6), verifying the first part of the lemma, (4.1).
is parallel to el with midpoint ml B(0, ) lying within the For the second part of the lemma, we apply another contain-
unique hyperplane through the origin with normal vector el . ment trick. We claim that
This is to be absolutely precise, but an appeal to Figure 2 should C
Tek (0) Tel (0) T |ek el | (0)
make the situation clearer. |ek el | ek
So we have two parameterizations of x Tek (0) Tel (0):
for some constant C. But this is immediate from our work
1 above. Indeed, since x Tek (0), we know that |xv| for any
x = mk + tx ek for some |tx | (4.4) " #
2 unit vector v ek and by (4.1), we know that x B 0, |ekCel |

for some constant C. Thus,
$ $
$ C $
|Tek (0) Tel (0)| $$ Te|ek k el | (0)$$ n1
|ek el | |ek el |
n
=
|ek el |
as claimed.
Figure 3. The distance between a point x Te (a) to another tube
Cordobas L2 estimate exploits a duality argument. His Te (a).
original proof has a classical flavor as he explicitly analyzes
the adjoint of the maximal operator; we will follow Wolffs We have shown that any tube in Rn can be covered by a
treatment [44] since its methods are more geometrically trans- bounded number of equidimensional tubes with a different
parent and also more prescient of later methods we will develop orientation so long as this new orientation is not very far
in the subsequent sections. from the original. Of course, this agrees with our intuition and
we can immediately see that this implies f (e) ! fC

(e )
Lemma 4.3. Let 1 < p < and let p be the dual exponent
provided |e e | < .
of p. Suppose p has the following property: if {ek } Sn1 is
% We use the above procedure now to discretize the domain
a maximal -separated set, and if n1 k |yk |p = 1, then for
of our maximal operator and write
any choice of points ak Rn we have
& & ( *1/p
&' & ')
& &
f p p
fC (e) de
& yk Te (ak ) & A. (4.7)
& k k & k D(ek ,)
p
( *1/p
Then there is a bound '
! n1
fC (ek )p .
f Lp (Sn1 ) ! Af p . (4.8) k
Proof. Let {ek } be a maximal -separated subset of Sn1 and Using duality between lp and lp we have
'
suppose that p has the property stated in the hypothesis of the f p ! n1
yk fC (ek )
lemma. We can discretize the domain of the Kakeya maximal k
%
operator by appealing to a doubling-type argument. To see for some sequence {yk } such that n1 k |yk |p = 1. Plugging
this, suppose e, e Sn1 are such that |e e | < and fix in the definition of the maximal function, we find that
' )
any x Te (a) where a is an arbitrary point in Rn . Let E 1
f p ! n1 yk C f (x)dx
denote the principal axis of Te (a) and let b E be such that k
|Tek (ak )| TeC (ak )
k
x b = inf yE x y. This defines b to be the unique point ' )

1
! n1
yk f (x/C)dx
in E closest to x (we use the norm notation here to emphasize k
|Tek (ak )| Te (ak )
k
that the distance is measured in the standard Euclidean way and for some choice of {ak }. Notice that the last inequality follows
to avoid confusion with our notation for the metric on Sn1 ). via the general containment fact that T C C T . Now, since
We will exploit the simple fact that the Euclidean distance |Tek (ak )| n1 , it follows that
between two points is bounded by the sum of the distances +) ( * ,p
'
between these two points and any intermediate point. f p !
p
yk Te (ak ) (x) f (x/C)dx
k
Consider the tube Te (a); we similarly let E denote its axis k
& &p
and define b E to satisfy b b = inf yE b y. Now &' &
& &
since the two tubes have the same center, we know that !& yk Te (ak ) & f pp
& k k &
p
b b = a b sin(|e e |) ! |e e | <
by Holders inequality. Applying (4.7) and taking pth roots2 ,
(see Figure 3). Thus, by the triangle inequality, we arrive at the bound (4.8) as claimed.
2 This pth root is the only time our image space comes into
x b x b + b b ! ,
play. Notice that the same reasoning will hold if we were after an
and so x TeC
(a) for some constant C. Lp Lq estimate instead.

We proceed to the main result on R2 . 4.2 L2 -estimates in Rn
Proof of Theorem 1.3. Using the previous lemma, it suffices
% Although Cordobas result was originally only obtained in R2 ,
to show that for any sequence {yk } with k yk2 = 1 and any
it can be easily extended to an L2 estimate on Rn for n 3.
maximal -separated subset {ek } of S1 we have
& & ! Instead of just directly generalizing the above proof, we offer
&' & 1
& & an alternative argument due to Bourgain [7]. This proof is
& yk Te (ak ) & ! log . (4.9)
& k k & purely Fourier analytic and has an elegant appeal. As with
2
We apply the above geometrical lemma to reduce the left-hand Cordobas proof, we will need a geometrical lemma to obtain
side to the desired result. The argument will force us to again consider
& &2
&' & ' a set of tubes, but this time we will utilize a convolution opera-
& &
& yk Te (ak ) & = yk yl |Tek (ak ) Tel (al )|
& k k & tion so that they are all centered at the origin. A scaling will
k,l 2
' 2 also make these tubes quite fat instead of thin.
! yk yl ,
|ek el |
k,l Lemma 4.4 (Bourgain). Let C 1 be some constant and
which by a regrouping becomes suppose C. For any fixed Rn , one has the bound
'
! yk yl . (4.10) 1
|ek el | |{e Sn1 : C Te1/ (0) }| ! .
k,l
1 + | |
We apply Cauchy-Schwarz to estimate this quantity further;
1/
thus (4.10) becomes Notice that here, the tube C Te (0) is oriented with the
' - .1/2

- .1/2 shortest axis in the e direction.

yk yl
k,l
|ek el | |ek el | Basically, this lemma gives a quantitative estimate on the
( *1/2 ( *1/2 1/
' ' size of the set of directions that define C Te (0) tubes or

yk2 yl2 slabs containing the fixed point Rn . These fat slabs appear
k,l
|ek el | k,l
|ek el |
naturally as the dual objects to thin tubes via the Fourier trans-
'
= yk2 form and indeed this is how they will appear in our main
k,l
|ek el | result. However, we first provide a proof of this geometrical
1' 2 lemma.
! log yk
k 1 1
Proof. First, suppose | | 10C so that 1+10C
1+| |
.
1 It follows trivially then that
= log .

The second to last line follows from the fact that |ek el | ! 1 |{e Sn1 : C Te1/ (0) }| |Sn1 |
and that all directions in our collection our maximally
1 1
-separated; indeed, we see then for any fixed k that ! .
1 + 10C 1 + | |
' ' ' 1 1
! = log . 1/
|ek el | 1
l 1
l If instead | | > C

, / C Te (0) for all e Sn1
then
l 1l 1l
and so the estimate is again trivial.
Thus we have (4.9), proving the theorem. 1/
Lastly, if 10C < | | C

, then to ensure that C Te (0)
Notice that Cordobas estimate does indeed settle the
for some e, the vectors and e must be nearly orthogonal.
Kakeya maximal function conjecture in R2 ; for every > 0
By definition, we have
there exists a constant C such that log 1 C and thus
> 0 C : f 2 C f 2 . C Te1/ (0)
/ 0
By Lemma 2.3, we also know that Cordobas estimate tells us C
= R : | e| C, | v|
n
for all v e .
that Kakeya sets in R2 must have full Hausdorff dimension
(and so full Minkowski dimension as well). (4.11)

We may rotate our system so that lies on the first standard The above convolution is against a rough cut-off function,
Euclidean axis, i.e., so that = (1 , 0, . . . , 0). Now using our but we can majorize this by a smooth representation to take
above description of the slab, notice that better advantage of certain properties of the Fourier transform.
Choose any function S (R), 0 such that has compact
{e Sn1 : C Te1/ (0) } 1 2
support, say in the interval C2 , C2 , and (x) 1 when
/ 0
C |x| 1. Define : Rn R by
= e : | e| C, | v| ve

/ $ $ $ $ 0 1 3 n 4x 5
j
$ $ C $$ $ C (x) = (x1 )
= e : $$ e$$ , $ v $$ ve . n1 j =2
| | | | | | | |
(4.12) where we have written x Rn as x = (x1 , x2 , . . . , xn ).
The second inequality in (4.12) is trivial; indeed, for any Let e1 denote the first canonical basis vector in Rn . We claim
e Sn1 and any v e, that e1 . To see this, notice that if x
/ Te1 (0), then
$ $ $ $
$ $ $ $ C e1 (x) = 0 (x). If instead x Te1 (0), then (x1 ) 1 and
$ $ $ $ "x #
$ | | $ $ | | $ |v| = 1 | | ,
v
j 1 for all 2 j n and so the inequality follows by
since | | C
. So (4.12) reduces to simply definition. Thus we have

/ $ $ 0
$
n1 $
$ C f (e1 ) sup ( f )(a).
eS :$ $
e . (4.13)
| | $ | | aRn
$ $
Now $ | | e$ = |e1 | where e = (e1 , . . . , en ), so we see that the Let e denote the rotation that sends the unit vector e to e1 .
remaining restriction in (4.13) applies only to a single degree of Define e = e and notice that e e by the same
freedom. More precisely, for any direction e in the set defined reasoning as before. Consequently,
by (4.13), there is no restriction on its e2 , . . . , en components;
f (e) sup (e f )(a) = e f . (4.14)
the only restriction we need make is that |e1 | C
| |
. This defines aRn
a subset of the sphere that looks much like a ring with thickness
We proceed to the analysis. First we estimate (4.14) as
1
| |
in the e1 direction. Consequently, the size of this ring
1 1
is bounded by ! | |
! 1+| |
, since | | " 1. The lemma now e f e f1
follows accordingly. )
= |e ( )| |f( )|d (4.15)
We are now ready to prove our main theorem. Rn
Theorem 4.5 (Bourgain). Let n 3. Then we have the L2 which follows by the trivial fact that the norm of the Fourier
estimate transform defined as an operator from L1 to L is bounded
n2 by 1. Rewriting the right-hand side of (4.15) and applying
f L2 (Sn1 ) ! 2 f L2 (Rn ) .
Cauchy-Schwarz, we find that
Proof. We begin by writing our maximal operator as ) - .
9
6
a maximum of convolutions against smooth functions. | e ( )|
f (e) |f( )| |e ( )| (1 + | |) d
Naturally, this will allow us to take advantage of some nice 1 + | |
properties in the Fourier space. -) .1/2
Let e (x) = 1
Te (0) (x). Then
n1
|e ( )| |f( )|2 (1 + | |)d
)
1 () *1/2
f (e) = sup n1 f (y)dy |e ( )|
aRn Te (a) d . (4.16)
) 1 + | |
1
= sup n1 Te (0) f (y a)dy
aRn Rn We turn our attention to estimating the second integral in
(4.16). To do this, we first need to estimate |e |. In fact, since
= sup (e f )(a).
aRn e = e we need only analyze . Clearly, || is bounded

universally by some constant. By a direct computation, we also by Lemma 4.4. Thus (4.18) becomes
)
have that 2
f 2 ! (n2)
|f( )|2 d = (n2) f 22
3
n Rn
( ) = (1 ) (j ). by Plancherels theorem. Taking square roots, we obtain
j =2
Theorem 4.5.
By our assumptions on we see that is supported on a tube
This result effectively illuminates the L2 behavior of our
of size C(11/. . .1/) centered at the origin and oriented
maximal operator, but clearly we would like more. In the next
so that the shortest dimension points in the e1 -direction. It now
two sections we will build up some impressive machinery to
follows directly that |e | is bounded and that e is supported
tackle the problem of Lp estimation of the Kakeya opera-
on a tube of size C(1 1/ . . . 1/) centered at the
tor in Rn . The ideas that follow are due to Bourgain [7], Tao
origin and oriented so that the shortest dimension points in the
1/
[39] and Wolff [42] and have many wide-ranging applications
e-direction, i.e., e is supported on C Te (0). Then
than just directly to the problem of Kakeya. The techniques
) ) ) will utitlize every component we have so far presented and in
|e ( )| d d
d ! = ,
Rn 1 + | | 1/
CTe (0) 1 + | | 1/
CTe1 (0) 1 + | | their totality represent a truly sophisticated piece of analytical
craftsmanship.
where the last equality follows by making the obvious unitary
transformation in the integral to rotate our region of inte- 5. The bush argument
gration. We use polar coordinates, with = (1 , ) =
(1 , 2 , . . . , n ), to rewrite the final integral above and estimate The main result of Cordoba in Section 4 gave us the estimate
) ) K(2), in the notation of Section 2.4. In this main section, we
d d d
! present Bourgains so-called bush argument which provides
|1 |C, |j | C 1 + | | |j | C 1 + | |

" #
the improved estimate K n+1 2
, and so the lower bound of n+12
) C
r n2 dr on the Hausdorff dimension of Kakeya sets in Rn . We will give
!
0 r two proofs of this estimate. The analyses will be very similar,
)
but the preliminary reductions will differ enough to be instruc-
C

! r n3 dr
0 tive. In Section 5.1, we give Bourgains [7] original proof,
following Wolffs treatment [44], and prove the restricted weak
(n2) . (4.17) " #
type- n+1
2
, n + 1 estimate
Notice that this is the only part of the argument that would need
Theorem 5.1 (Bourgain).
to be altered if we let n = 2. Obviously, the bound would then n1
become log(1/), agreeing with the bound we acquired in the f n+1, !n n+1 f n+1 ,1 . (5.1)
2
previous section. As we have seen, this is more than we need to get the
2
The estimation done, we can now prove the desired L analogous bound on the Hausdorff dimension of Kakeya sets
bound. Combining (4.16) and (4.17), we find that in Rn , as it implies the equivalent strong type estimate on
) -) . the diagonal (see Section 2.4). We feel that this analytical
2 2
f 2 ! (n2)
|f ( )| (1 + | |) |e ( )|de d. overkill is still instructive though due to the prevalency of these
Rn Sn1
(4.18) techniques in much of the literature from the 1990s.
In Section 5.2, we will once again prove the estimate
Now we are in a position to apply the lemma. Notice that for " n+1 #
K 2 , this time exploiting a similar duality argument as was
any fixed Rn ,
used in Section 4.1. However, we will refine this approach,
)
|e ( )|de ! |supp(e ( ))| following Tao [39], and thus streamline the analysis. This will
Sn1 not only illustrate the different philosophies behind deriving
1 Kakeya estimates, but it will lay the foundation for the much
= |{e Sn1 : C Te1/ (0) }| !
1 + | | more complicated arguments of Section 6.

The main strategy of the bush argument will involve Now suppose that the second inequality does not hold for
bounding the size of a set E in R from below by con- n
some , Sn1 such that |,| > 0. Then since we can always
sidering its intersection with a -separated family of tubes. The cover Sn1 by a finite number of rotated copies of ,, we
-separation will allow us to control the intersection and thus see that we would have to contradict our first inequality for
estimate |E| from below by essentially the complement of this , = Sn1 .
intersection. If we think of many lines intersecting at a single
We can now exploit this fact to prove the lemma.
point, then expanding these lines to small tubes should not
produce a large intersection. Thus, when many tubes intersect Proof of Lemma 5.2. It suffices to prove the lemma for
E at a certain point, this technique should give a good lower M = N (DE ()). By equation (5.4) and our hypothesis (5.3),
bound by virtue of the intersection being quite small. we have
|E|2
5.1 Bourgains approach n1 |DE ()| ! 2(n1) M ! .
n+1
" n+1 # Rearranging this expression, we arrive at (5.2).
Equation (5.1) is a restricted weak type- 2
, n + 1 estimate
and as such should be interpreted as So we have reduced proving the pure restricted weak
( *n+1 type estimate of (5.2) to only having to prove (5.3) for any
2
|E| n+1 -separated subset {ej }M
|{e Sn1 : (E ) (e) }| Cn n1
n+1 , j =1 DE ().
To apply Lemma 5.2, we want a lower bound on the size
(5.2) of any set E Rn , |E| > 0, given some -separated subset
for all sets E Rn with finite measure, all 0 < 1, and j =1 DE (). Using the definition of the Kakeya maximal
{ej }M
where p = n+1
, q = n + 1. We make the following reduction. function, we see that for any 1 j M,
2
|E Tej |
Lemma 5.2. Let E be a set in Rn and for 0 < 1, denote (E ) (ej ) , (5.5)
|Tej |
by DE () = {e S n1
: (E ) (e) }. Suppose that the
j =1 DE () is -separated. If
subset {ej }M for some -tube oriented in the ej direction. In this way, we may
n+1 assign a tube Tej with a relatively large E-intersection to each
|E| " n1 2 M, (5.3)
direction ej in our -separated subset of DE (). Analyzing
then equation (5.2) holds. how these tubes interact with the set E will lead us to the lower
bound necessary to apply the lemma.
The lemma will follow from a simple combinatorial fact
Fix a positive integer to be determined later; there are two
about -separated directions. For this, we make a definition.
basic cases to consider. First, our tubes may be well dispersed
Definition 5.3. For any subset , S n1
, |,| > 0, so that no point of E belongs to more than tubes. For this
r > 0, define its r-entropy Nr (,) as the maximum possible case, we use a trivial lower bound. The other alternative is that
cardinality for an r-separated subset of ,. there may be some point in E that belongs to many different,
say more than , tubes. This case will lead us to consider how
Fact 5.4. In the notation just defined, these tubes intersect at the high multiplicity point. If many
1 |,| tubes intersect a single point of E, then since their directions
" Nr (,) " n1 . (5.4)
r n1 r are all -separated, the object formed by these tubes will look
Proof. To see this, we consider , and Sn1 to be hypersurfaces like a spiky burr or bush; thus, we will be forced to analyze the
in R . If we cover the sphere by connected pieces or caps,
n
intersection properties of this bush object. This second case
each of radius about r n1
, so that the center of any particular will require the other facts that we have established in addition
cap is not contained in any others, then we see Nr (S n1
) to two more general lemmas in order to attain a suitable lower
|Sn1 | 1
r n1
r n1
. Now, for any other subset , S n1
, we clearly bound on |E|.
have Nr (,) Nr (S n1
) and so the first inequality in (5.4) Inequality (5.5) tells us that if ej DE (), then we have
follows. the bound |E Tej | " n1 . This next lemma tells us that the

same estimate will hold if we cut out a small enough portion
of the intersection.
Lemma 5.5. Let E Rn be such that |E| > 0. Let 0 < 1

and suppose we have a tube Te such that |E Te | |Te |. 2c
Figure 4. Two 1 -separated directions u1 , u2 U lying in caps with
Then for any a E Te , there exists a constant c0 dependent 2c
centers ej1 , ej2 that are not 1 -separated.
on , and e such that
This is highly suggestive of the case when many tubes intersect
|E Te B(a, c01 )c | |T |, (5.6)
2 e E at or near a single point and indeed, this is how we will utilize
where A denotes the complement of the set A in R .
c n Lemma 5.5 in the bush argument. We have a final counting
lemma before we begin the proof of Bourgains theorem.
Proof. Pick a point a E Te . For any choice of c0 , we can
write Lemma 5.6. Suppose {ej }m
j =1 S
n1
is -separated and let
c1 be some constant to be determined later. Then for any fixed
|E Te | = |E Te B(a, c01 )|+|E Te B(a, c01 )c |. k, we have
(5.7) / 0

# j : (ej , ek ) c1 " n1 m. (5.8)
We want to bound the first term in (5.7) from above. Observe
that
Proof. Clearly we need to make a combinatorial argument to
|E Te B(a, c01 )| |Te B(a, c01 )|; estimate the cardinality of the set in (5.8); this is where our
Fact 5.4 will come into play. Let C(ej , ) denote the cap on
so choose c0 (dependent on , and e) such that Sn1 with center ej and radius ; then
|Te B(a, c01 )|
|T |.
2 e
Explicitly, this can be accom-
<
m
plished as follows. If c0 , then we trivially have {ej }m
j =1 C(ej , ),
- .n j =1

|Te B(a, c01 )| |B(a, c01 )| a disjoint union of -caps. For convenience, denote this union
c0
by U . To prove (5.8), we need a lower bound on a maximal
n1
|Te |. c1 -separation of {ej }m
j =1 . We claim that the size of such a
c0 2
separation is at least as big as the maximum cardinality of a
If instead c0 < 1, then we contain the intersection
2c1 -separation of U , which, by (5.4), is bounded by
Te B(a, c01 ) in a tube centered at a of length c01
in the e direction and width in all orthogonal directions. m n1
N2c1 (U ) " n1 m. (5.9)
Consequently, the size of the intersection is dominated by the (2c1 /)n1
size of the tube which can itself be dominated by 2 |Te | for an By construction, whatever elements of U we pick for our
appropriately large c0 . maximal 2c1 -separated subset, we cannot have any two of
Combining the above analysis with our hypothesis that these lying in the same cap. So in a sense, we are in fact
|E Te | |Te | and (5.7), we obtain 2c1 -separating our caps. Notice however that we cannot
immediately assert that the centers of these caps are separated
|E Te B(a, c01 )c | |Te | |Te B(a, c01 )|
in exactly the same way, as Figure 4 demonstrates.
What we can say with certainty is that the centers of these
|Te |,
2 caps are not separated by much less. More precisely, we have
as was to be shown.
(ej1 , ej2 ) 2c1 c2 , (5.10)
This bound on the trimmed intersection will allow us to
construct a lower bound for |E| by considering many such where c2 is the absolute constant such that diam(C(e, )) =
intersections that are disjoint by virtue of the trimming action. c2 . We can now determine that whatever the value of c1

may be, we must insist that c1 > c2 . With this choice, (5.10) Applying Lemma 5.5, we have that there exists some constant
becomes c0 such that

(ej1 , ej2 ) c1 . |E Tej B(a, c01 )| " n1 (5.12)

Thus, any 2c1 -separation of U automatically gives a for all j + 1.
c1 -separation of the discrete set {ej }m Now we apply Lemma 4.2 (Cordobas geometrical lemma)
j =1 . More precisely, let
{r }Rr=1 be a 2c1 -separated subset of U . Pick r and r , to convert (5.12) into a tighter lower bound for |E|. Fix
r = r . Then r C(ejr , ) and r C(ejr , ). So by the k +1; then if j +1, the intersection Tej Tek contains
above analysis, we know that (ejr , ejr ) c1 ; thus, {ejr }Rr=1 a and has diameter !
(ej ,ek )
by Lemma 4.2. Let c1 be a suitable
is a c1 -separated subset of {ej }m constant determined by applying Lemma 5.6 to the collection
j =1 . Since this procedure +1
works for any 2c1 -separation
of U , including a maximal {ej }j =1 such that we also have c1 > c0 . Notice if (ej , ek )
= > c1 , then we have diam(Tej Tek ) ! c01 . Consequently, the
separation, we conclude that # j : (ej , ek ) c1
N2c1 (U ) for any fixed k. Applying (5.9), it now follows that sets E Tej B(a, c01 )c and E Tek B(a, c01 )c are dis-
/ 0 joint. This is the crucial geometric fact for the bush argument
and it is really just an extension of the obvious fact that two
# j : (ej , ek ) c1 " n1
lines intersect at a single point. Of course though, we have had
for any fixed k. to adapt this bit of trivia to suit our -thickened needs. What
With all the necessary lemmas established, we proceed to this affords us is that we can now bound |E| from below by the
the proof of Bourgains theorem. sum of certain trimmed intersections as in (5.12) rather than by
just a single trimmed intersection, thus tightening the bound.
Proof of Theorem 5.1. We analyze the two multiplicity cases What we have then is that
separately, choosing a suitable upon completion. '
|E| |E Tej B(a, c01 )c |
Low multiplicity case: Suppose that no point of E belongs to j : (ej ,ek )c1
more than of our tubes Tej ; i.e., / 0

" # j : (ej , ek ) c1 n1 . (5.13)
'
M
Tej (x) ,
Notice that we need not include the trimmed intersection
j =1
corresponding to the tube Tek since this would not tighten the
for all x E. Integrating over E, we have
bound by more than an adjustment to the implicit constant.
1' The cardinality of the set in (5.13) is of the form treated in
|E| |E Tej |.
j Lemma 5.6, so combining (5.13) with the estimate (5.8), we
arrive at the bound
Applying equation (5.5) and summing over all j yields
|E| " n n1 , (5.14)
|E| " M n1 . (5.11)

completing the analysis of the high multiplicity case. It is clear
Notice that this estimate holds for any we choose; however, that for any positive
" integer
# either (5.11) or (5.14) must hold.
the estimate becomes worse the larger we make . If we set 2 n1
M, then both estimates reduce to
High multiplicity case: Now suppose that some point a E n+1
|E| " n1 2 M.
belongs to more than tubes Tej . Reordering if necessary,
we may assume that a Tej for all j + 1. As before, This is exactly the bound we need to apply Lemma 5.2;
we want to utilize (5.5), but it does not suffice to just consider hence, our restricted weak type estimate follows and we have
the intersections E Tej superficially. Instead, we will cut Bourgains theorem.
out a small portion of the intersection near the high multipli- Using Proposition 2.10, we immediately see that Kakeya
city point and construct a lower bound for the remaining set. sets in Rn have Hausdorff dimension at least n+1
2
. Of course,

we can also use the methods of Section 2.4 to see that if , Sn1 is any -separated set, and if
" #
Bourgains theorem implies the estimate K n+1 , and so that & &
2 &' &
& &
the Hausdorff dimension of Kakeya sets in Rn is bounded & & #A (5.15)
&, T &
below by n+1
2
by Proposition 2.8. Whatever route we choose, p
we are still left with the feeling that this result is clearly far for any collection of (1 )-tubes {T }, , then
from optimal, at least in light of the Kakeya set conjecture. But there is a bound
before we move onto proving a stronger bound, we will take
another look at the bush argument from a different angle. This f Lp (Sn1 ) # Af p .
new perspective will prove to be especially useful when we This will allow us to streamline the bush argument and
present Wolffs hairbrush argument in the next main section. concentrate more on the inherent geometrical nature of the
problem, rather than getting too bogged down in messy,
5.2 Taos approach
analytical details. This will prove to be very useful in the next
We introduce the notation A # B if one has A C B main section when we present the considerably more com-
" #
for every > 0. Thus A # B is weaker than A ! B, but plicated hairbrush argument to refine the estimate K n+1
2
to
" n+2 #
only by a factor of . In the context of the Kakeya conjec- K 2 . The following version of the bush argument, as well
tures, we see that a Kakeya estimate A # B is just as good as as all of Section 6, will liberally exploit the classic pigeonhole
A ! B since we can always afford to take a
loss. In fact, principle. We have already used this principle several times,
examining the maximal function conjecture (Conjecture 1.8) but now it will appear in nearly every reduction we make. It is
and Proposition 2.8 directly, we see that no finite number of worth mentioning that this principle will often be disguised
extra factors in our estimates will force us to weaken our as an application of the mean value theorem, its continuous
conclusions about the norm of the operator or the dimension analogue. This simple but powerful tool will aid us in trimming
of Kakeya sets. much of the analysis down to a more intuitively palpable level.
One of the attractive features of working with restricted To prove Proposition 5.7, we will combine two intermediate
weak type estimates as in Bourgains argument is that, by lemmas.
definition of the maximal function, we are then led to analyze
Lemma 5.8. Let 1 < p < and let {T }, be any
how dense a given -tube is in a set E; this is the sub-
collection of (1 )-tubes with orientations from
stance of (5.5). However, we can simplify matters further by
any -separated set , Sn1 . Then a bound of the form
disregarding the arbitrary set E. After all, the geometry of the
& & ( * 1
problem should depend only upon how our -tubes interact &' & ' p
& &
with each other. Thus, we would like to trim the argument down & T & # A |T | (5.16)
&, & ,
p
and simplify the objects we are working with. Tao offers the
best path to this slick treatment and we will follow his lecture implies a bound
notes [40] for the remainder of this and the following main
f Lp (Sn1 ) # Af p .
section.
Basically, we want to make use of the duality approach as Clearly, we can estimate (5.16) as
in Section 4. In its present form, this may actually compli- ( * 1
' p
cate things, so we will first strengthen Lemma 4.3 to remove A |T |
1
A ( n1 #,) p
the dependency on our set of directions being maximally ,
-separated and on the prescence of the normalizing dual

# A, (5.17)
sequence {yk }. More precisely, what we are after is the
following. so if (5.16) holds, then (5.15) holds. The amazing thing though
is that this implication is reversible.
Proposition 5.7. Let 1 < p < and let p be the
dual exponent of p. Suppose p has the following property: Lemma 5.9. If (5.15) holds, then (5.16) holds.

- .1
If we can establish these two lemmas, then we immediately n1 1 p 1n
! A p log p
have the proposition by transitivity. We begin with the simpler
of the two.
# A.
Proof of Lemma 5.8. We will prove this lemma by appealing
This calculation holds for any -separated set ,, so in
to Wolffs duality argument (Lemma 4.3) directly. Let
particular if , is maximally -separated, then the hypothesis
, Sn1 be -separated and let {y }, be such that
% of Lemma 4.3 holds. Thus, its conclusion follows accordingly,
1n = |y |p . We want to prove that
& & giving us the desired bound.
&' &
&
&
&
y T & # A We proceed to the proof of the second lemma.
& &
p
Proof of Lemma 5.9. For each integer N ! 1n , let c(N)
if (5.16) holds. By the triangle inequality, it suffices to show
denote the smallest constant such that
that & &
& & &' &
&' & & &
& & & T & ! c(N ) (5.21)
& |y |T & # A, &, &
& & p
p
so there is no harm in assuming y 0. With this in place, holds for all -separated sets , with #, N. Since we are
define assuming (5.15) holds, we know that c(N ) # 1; in light of
(5.17), we need to improve this estimate to
,1 = { , : 0 y < 1},
1
j +1
c(N ) # ( n1 N) p . (5.22)
,j = { , : 2 y < 2 j
}, j 0.
? We will use a probability trick to derive this estimate from a
Clearly then , = j ,j and
recursive inequality.
' '
'
'
1n =

yp =

yp

2jp #,j . (5.18) Suppose that N 1n . By definition, we can find a
j =1 ,j j =1 -separated set , such that #, = N and
& &
Now &' &
& & & & & &
&' & ' &' & & T & = c(N ).
& & & & &, &
& y T & & y T & , (5.19) p
& & j =1 &, &
p j p
Now let U be a rotation in Rn selected uniformly at random
but we would like to remove the dependency on the y s from from the orthogonal group O(n) and consider the set U (,).
j
this expression. As such, we let z = 2 y for all ,j so It is an easy probability exercise to show that this set will be
that z 1. Then (5.19) becomes disjoint from , almost surely. If we define TU () to be the
& & & &
' & & ' &
& ' & &' &
& rotated tube U (T ), then by symmetry we have
2j & z T & ! 2j & T & & &
j
& ,
& j
& ,
& & ' &
j p j p & &
& T & = c(N ).
( * 1 &U (,) &
' ' p p
! 2 A j
|T | q q q
j ,j The binomial theroem gives us that f + gq f q + gq
n1 ' 1 for any 1 q < and nonnegative f, g; therefore,
! A p 2j (#,j ) p (5.20) & &
j & &
& ' &
by hypothesis. We apply Holder to the sum in (5.20) and

&
& T &
& 21/p c(N ). (5.23)
&, ? U (,) &
combine with (5.18) and (5.19) to deduce that p
& & ( * 1 Even though , U (,) is a disjoint union, there is nothing

&' & ' p
& & n1 1
& y T & ! A p (#j ) p 2jp #,j to guarantee that it is -separated, something that we would
& , &
p j obviously like required. To do this, we need to ensure that we

can find an appropriate rotation U , no matter what our initial the mean value theorem). This is what we are after. Fix one
set of directions , may be, so that our larger set of directions of these rotations U now and partition the set , U (,) into
, U (,) can be -separated after we remove a very small -separated pieces. To do this, for any / Sn1 , let
number of its elements.
To this end, we let a(,, U (,)) denote the number of pairs A/ = {e / : |e | for some , U (,)}.
in ,U (,) with first and second coordinates within of each
Now , is -separated, so U (,) must be also and thus so
other; i.e.,
are the sets A, and AU (,) . So we can partition , U (,)
a(,, U (,)) = #{(, ) , U (,) : | | }. into -separated parts where #A, = O( n1 N 2 ), #AU (,) =
O( n1 N 2 ), and #(, U (,) \ (A, AU (,) )) 2N , i.e., two
We can write this more explicitly as small -separated sets and one large. Using the definition of
'' c(N ) in (5.21) and the triangle inequality, we thus have that
a(,, U (,)) = B(0,) ( U ( )).
, & &
, & ' &
& & n1 2
If we let U range over the orthogonal group O(n), then we can & &
T & c(2N ) + 2c(C N )
& ?
&, U (,) &
use the normalized Haar measure3 dO(n) to write p
) '' for some constant C. Combining this with the lower bound in
a(,, U (,))dO(n) = f (, ) (5.24)
O(n) , , (5.23), we get the inequality
where
) c(N ) 21/p c(2N) + 2c(C n1 N 2 ). (5.25)

f (, ) = B(0,) ( U ( ))dO(n).
O(n) Now we want to establish (5.22), or equivalently,

Since we are integrating over the entire group O(n), symmetry c(N )( 1n N 1 )1/p # 1. Regardless of our choice of N, we
considerations tell us that f is independent of the choice of . see that there exists some k 0 such that 1n N 1 2k ; so
Thus we can write N 2k 1n . Thus it will suffice to require
) )
1
ck 2k/p c(2k 1n ) # 1 (5.26)
f (, ) = B(0,) ( U ( ))d dO(n).
|Sn1 | O(n) Sn1
If we examine the inner integral only, then and U are fixed, for all k 0. Notice that if k " log(1/ n1 ), then 2k " 1n
so and consequently N ! 1, a contradiction. Thus it is enough
) ) to prove (5.26) for all 0 k ! log(1/). Multiplying through

B(0,) ( U ( ))d = B(0,) ( e)de
by the factor 2k/p in (5.25), we see that what we have to work
Sn1 Sn1
with is the recursive inequality
= |{e Sn1 : | e| }|

n1
ck ck1 + C2k/p c2k (5.27)
,
for some constant C. Notice that if we did not have the
the surface measure of the ball B(0, ). So since
@ second summand in (5.27) (the error term), then we could
O(n) dO(n) = 1 by definition, we see that f (, )
n1
conclude immediately that ck is a nonincreasing sequence in
uniformly; thus, (5.24) becomes
) k and thus it attains its maximum at c0 . But c0 = c( 1n ) # 1
a(,, U (,))dO(n) n1 N 2 , by hypothesis, equation (5.15), so ck # 1 for all k.
O(n)
Looking back at (5.26), we see that this would imply that
2
since #(, U (,)) = N almost surely. c(N )( 1n N 1 )1/p # 1 for all N 1n ; this is exactly
What this means is that we can always find a rotation U such (5.16), and would verify our lemma.
that a(,, U (,)) ! n1 2
N just by pigeonholing (or applying Unfortunately, we cannot just drop the error term from
3 See
Conway [13], pp. 15556 for the necessary background (5.27) altogether. Nevertheless, we would like to assert that
on the Haar measure of a topological group. ck behaves as well as a nonincreasing sequence and apply the

same subsequent reasoning to conclude our lemma. To do so, if k is large enough. Since A > C, this can again be made
we will have to introduce a new sequence dk defined by greater than 1 by choosing k appropriately large. Evidently
then, we choose a k0 so that both bounds hold for all k k0 .
dk = ck (1 + C0 2k/p ),
Thus, the recursive inequality (5.28) holds for all k k0 .
for some sufficiently large C0 to be chosen later. We claim that Let kmax maximize dk in the range [k0 , O(log(1/))]. Applying
if A > C, then (5.28), we have

dkmax < dkmax 1 +C2k/p (A(d2kmax dkmax )+(dkmax 1 dkmax )).

dk < dk1 + C2k/p (A(d2k dk ) + (dk1 dk )) (5.28)
for all k k0 , where k0 is some fixed constant. To verify this Now the second summand must be negative, so we see that

claim, multiply (5.27) through by a factor of (1 + C0 2k/p ) to dkmax < dkmax 1 which forces kmax = k0 . Therefore, ck < ck0
find that for all k > k0 . Notice that we trivially have ck ! ck0 for all

k < k0 . So since ck0 c(2k0 1n )2k0 /p ! c( 1n ) = c0 # 1
dk ck1 (1 + C0 2k/p ) + C2k/p c2k (1 + C0 2k/p )
by hypothesis, we have that ck # 1 for all 0 k ! log(1/),
( *
1 + C0 2k/p

completing the proof of Lemma 5.9.
= dk1 " #
1 + C0 2(k1)/p Thus, to prove the estimate K n+1 2
, it suffices to show
( * & &
&' &
k/p 1 + C0 2k/p & & n1
+ C2 d2k . (5.29) & T & # n+1 ,
1 + C0 22k/p &, & n+1
n1
Subtracting (5.28) from (5.29) and rearranging, we see that we

where , is any -separated subset of Sn1 , and {T }, is any
aim to show
( * collection of -tubes with the given orientations. Notice that

1 + C0 2k/p
p = n+1 and pn + 1 = n1 , so the above expression really
(A + 1)2k/p dk < dk1 1 + 2k/p n1#
" n+1 n+1
1 + C0 2(k1)/p is K 2 . Simplifying this, we find that we want to show
( *
k/p 1 + C0 2k/p

) (' * n+1
n1
+ d2k 2 A . T (x) dx # 1 . (5.30)
1 + C0 22k/p
,
We substitute in for ck and simplify to arrive at the proposed
As mentioned above, we will take great advantage of the
inequality
pigeonhole principle in the following argument, but we will
+ ,
1 + C0 (2 1/p
1) + C0 2 (k1)/p also employ dyadic decomposition to further discretize the
ck < ck1
(A + 1)(1 + C0 2k/p ) objects of interest. This is another classic analytical tool that
+
, will provide us a considerable benefit when simplifying the
k/p (A C)2k/p CC0 + AC0 2k/p
+ C2 c2k . intuition.
C(A + 1)(1 + C0 2k/p )
To this end, we proceed to discretize the domain of inte-
Since we know (5.27) holds, it is enough to show that each gration in (5.30). Notice that #, = O( 1n ); this follows
%
of the bracketed terms above are greater than 1. For the first immediately by applying Fact 5.4. Thus, , T (x)
coefficient, notice for all large k we have [0, O( 1n )] Z for any x Rn . If we let range dyadically
from 1 to O( 1n ) and define the sets
1 + C0 (21/p 1) + C0 2(k1)/p 1 + C0 (21/p 1)
> A B
(A + 1)(1 + C0 2k/p )

2(A + 1) '
E = x : T (x) < 2 ,
and that this can be made greater than 1 just by chosing C0 ,
large compared to A (recall that A was already fixed to be
then we can rewrite the left-hand side of (5.30) as
larger than C). For the second coefficient, we have
( * n+1
k/p k/p k/p ') ' n1
'
(A C)2 CC0 + AC0 2 (A C)2C C0 T (x) dx
n+1
n1 |E |.
>
C(A + 1)(1 + C0 2k/p ) 2C(A + 1) E ,

Now the number of cannot exceed log( 1n ) log 1/, and Plugging in the definition of , , we arrive at
since log 1/ # 1, we see that to prove (5.30) it suffices to '
|E | n1 #, . (5.33)
show that

n+1
n1 |E | # 1 (5.31) Now we are going to apply the pigeonhole principle again
holds for each . Notice that the set E is precisely our bush so that we can analyze just a single term of (5.33). For the
object; indeed, by definition it is the set of points that lie in sake of precision, we illustrate exactly how to go about doing
tubes. Clearly, we need some control over how much of each this.
tube is contained in the set E ; i.e., we need to analyze the Define f : [ 2n , 1/2] R by f (x) = #, if x = for
density of the tubes in this set. This is what the quantity dyadic in the domain and define f by linear interpolation for
accomplished before and we will use it again, taking advantage all other x in the domain. Let denote counting measure on
of another dyadic decomposition. Before that though, we make the set of all s. Then
' ) 1
an easy reduction. 2
|E | n1
#, = x n1 f (x)d(x).
Fix and suppose |E | n . Then 2n
n+1
1n n+1
n1 |E | ! ( ) n1 n
= , 1
Now by the mean value theorem, we know that there exists
some x ( 2n , 1/2) such that
and so (5.31) is trivially satisfied. Thus we may assume
) 1
|E | n . n1
2
|E | x f (x ) d(x)
With our fixed, we use the definition of E to write 2n
) ' '
|E | T (x)dx = |T E |. x n1 f (x ) log(1/)
E , ,
# x n1 f (x ).
Now the set |T E | can range in value from 0 to |T |, so we
can discretize this range and for each each dyadic between Clearly, there exists some such that < x 2, so we
2n 1
and 2
, define may write
, = { , : |T | < |T E | 2|T |}; |E | # n1 f ()

thus, we can write
= n1 #,
' ''
|T E | = |T E |
, , for this particular . Since we trivially have that
' '
+ |T E |. (5.32) |E | n1 #, n1 #, ,
:|T E | 2n |T |
It is easy to show that the second term can be ignored. First we find that, in fact
% %
notice that , |T E | ! |E | trivially; then notice
that since 1, |E | n , and #, ! 1n , we have n1 #, |E | (5.34)
'' '
|T E | |E | |T E | must hold. This procedure may be technical overkill, but
, :|T E | 2n |T | it serves as a nice illustration as to how the mean value
theorem can make the classic pigeonhole principle analytically
" |E | 2n
precise.
1 We see that, by (5.34), this particular acts as the dominant
" |E |.
2
density of a generic tube T in E . Notice that we have an
Consequently, we can simplify (5.32) and just write
estimate on this density, namely
''
|E | |T E |.
, $ |E | (5.35)

" n+2 #
since #, = O( 1n ). So just as before we have related establish K 2
, the estimate derived via Wolffs hairbrush
our multiplicity and our density quantities, and , and argument. As we will see, there are still many reductions to
now we can verify (5.31). The remainder of the argument make before we can actually begin to construct this hairbrush
will look very similar to our first pass through the bush argument and analyze it as effectively as we have Bourgains
argument. bush.
By (5.34), we have that
' References
|E | |T E |
,
) [1] D. Alvarez, Bounds for some Kakeya-type maximal
'
= T (x)dx. functions, Berkeley Thesis (1997).
E ,
[2] M. Bateman and N. H. Katz, Kakeya sets in
cantor directions, Math. Res. Lett., 15 no. 1 (2008)
So another application of the mean value theorem tells us that
7381.
there exists at least one high multiplicity, high density point
[3] C. Bennett and R. Sharpley, Interpolation of operators,
a E for which
Academic Press, Inc. (1988).
'
T (a) $ . [4] A. S. Besicovitch, Sur Deux Questions de lIntegrabilite
, des fonctions, J. Societe des Math. et de Phys. a lUniv.
a Perm, 2 (1919) 105123.
In other words, this a is contained in $ tubes T and each
[5] A. S. Besicovitch, On Kakeyas problem and a similar
T has density inside E .
one, Math. Zeitschrift, 27 (1928) 312320.
As before, we trim the intersection |T E | by cutting
[6] A. S. Besivovitch and R. Rado, A plane set of measure
out a ball around a of radius c01 for some sufficiently large
zero containing circumferences of every radius,
constant c0 . As before, we have
J. London Math. Soc., 43 (1968) 717719.
|T E B(a, c01 )c | " n1 . [7] J. Bourgain, Besicovitch type maximal operators and
applications to Fourier analysis, Geometric and Func-
If we then c1 -separate our directions , , we know that tional Analysis, 1 (1991) 147187.
all intersections as above are disjoint for different choices of [8] J. Bourgain, Lp Estimates for oscillatory integrals in
, , . We know that the size of this c1 -separated subset several variables, Geometric and Functional Analysis,
of , , call it , , has cardinality $ n1
, and thus we may 1 (1991) 321374.
estimate [9] J. Bourgain, Some new estimates for oscillatory inte-
' grals, in Essays on Fourier Analysis in Honor of Elias
|E | |T E B(a, c01 )c | M. Stein, ed. C. Fefferman, R. Fefferman, S. Wainger,
,
Priceton University Press (1994).
$ n1 n1 . [10] J. Bourgain, On the dimension of Kakeya sets and
related maximal inequalities, Geometric and Func-
Notice that these two inequalities are identical to (5.13) and tional Analysis, 9 (1999) 256282.
(5.14). Combining the above with (5.35) and simplifying, we [11] A. Carbery, Differentiation in lacunary directions and
arrive at (5.31) as desired. an extension of the Marcinkiewicz multiplier theorem,
Much of this version of the bush argument was similar Ann. Inst. Four., 38 no. 1 (1988) 157168.
or identical to our previous rendition, but we were able to [12] A. Carbery, Restriction implies Bochner-Riesz for
develop and employ many new techniques that will prove to be paraboloids, Math. Proc. Camb. Phil. Soc., 111 (1992)
invaluable in the next main section. Indeed, part of our moti- 525529.
" #
vation behind exhibiting a second proof of K n+1
2
was to take [13] J. B. Conway, A course in functional analysis, Springer,
some of the weight off the preliminary analysis required to Second Edition (2007).

[14] A. Cordoba, The Kakeya maximal function and the University of British Columbia,
spherical summation multipliers, American Journal of open.library.ubc.ca/cIRcle/collections/24/items/
Mathematics, 99 (1977) 122. 1.0166114 (2015).
[15] R. O. Davies, Some remarks on the Kakeya problem, [31] E. Kroc and M. Pramanik, Kakeya-type sets
Proc. Camb. Phil. Soc., 69 (1971) 417421. over cantor sets of directions in Rd+1 , J. Four.
[16] J. Duoandikoetxea and A. Vargas, Directional operators Anal. & App., dx.doi.org/10.1007/s00041-015-9426-x
and radial functions on the plane, Ark. Mat., 33 (1995) (2015).
281291. [32] I. aba, From harmonic analysis to arithmetic
[17] Z. Dvir, On the size of Kakeya sets in finite fields, combinatorics, Bull. Amer. Math. Soc., 45 (2008)
J. Amer. Math. Soc., 22 (2009) 10931097. 77115.
[18] K. Falconer, Fractal geometry mathematical foun- [33] J. M. Marstrand, Packing planes in R3 , Mathematika,
dations and applications, John Wiley & Sons, Second 26 (1979) 180183.
Edition (2003). [34] A. Nagel, E. M. Stein and S. Wainger, Differentiation
[19] C. Fefferman, A note on spherical summation multi- in lacunary directions, Proc. Natl. Acad. Sci. US, 75(3)
pliers, Israeli J. Math., 15 (1973) 4452. (1978) 10601062.
[20] G. B. Folland, Real analysis: modern techniques [35] P. Sjogren and P. Sjolin, Littlewood-Paley decom-
and their applications, John Wiley & Sons, positions and Fourier multipliers with singulari-
Second Edition (1999). ties on certain sets, Ann. Inst. Four., 31 (1981)
[21] L. Grafakos, Modern Fourier analysis, Springer, 157175.
Second Edition (2009). [36] C. Sogge, Concerning nikodym-type sets in 3-dimen-
[22] J. R. Kinney, A thin set of circles, Amer. Math. Monthly, sional curved spaces, J. Amer. Math. Soc., 12 (1999)
75 (1968) 10771081. 131.
[23] S. Kakeya, Some problems on maximum and minimum [37] E. Stein, Harmonic analysis: real-variable methods,
regarding ovals, Tohoku Science Reports, 6 (1917) orthogonality, and oscillatory integrals, Princeton
7188. University Press (1993)
[24] N. H. Katz, A counterexample for maximal operators [38] E. Stein and G. Weiss, Introduction to Fourier analysis
over a Cantor set of directions, Mat. Res. Lett., 3 (1996) on euclidean spaces, Princeton University Press
527536. (1971).
[25] N. H. Katz, Remarks on maximal operators over arbi- [39] T. Tao, Lecture Notes, available at
trary sets of directions, Bul. London Math. Soc., 31 math.ucla.edu/tao/254b.1.99s/ (1999).
(1999) 700710. [40] T. Tao, The Bochner-Riesz conjecture implies the
[26] N. H. Katz, I. aba and T. Tao, An improved bound restriction conjecture, Duke Math. J., 96 (1999)
3
on the Minkowski dimension of Besicovitch sets in R , 363375.
Ann. of Math. (2), 152 (2000) 383446. [41] A. Vargas, A remark on a maximal function over a
[27] N. H. Katz and T. Tao, New bounds for Cantor set of directions, Rend. Circ. Mat. Palermo, 44
Kakeya problems, J. Anal. Math., 87 (2002) (1995) 273282.
231263. [42] T. Wolff, An improved bound for kakeya type maximal
[28] N. H. Katz and T. Tao, Recent progress on the Kakeya functions, Rev. Mat. Iberoamericana, 11 (1995)
conjecture, Publ. Mat. (2002) 161179. 651674.
[29] L. Kolasa and T. Wolff, On some variants of [43] T. Wolff, Recent work connected with the Kakeya
the Kakeya problem, Pacific J. Math., 190 (1999) problem, Prospects in Mathematics, Princeton (1996)
111154. 129162.
[30] E. Kroc, Kakeya-type sets, lacunarity, and directional [44] T. Wolff, Lectures on harmonic analysis, University
maximal operators in Euclidean space, (Dissertation), Lecture Series, 29 (2003).

Bakers Question on Dynamics of Permutable Entire Functions
Kuldeep Singh Charak
Department of Mathematics, University of Jammu, Jammu 180 006, India
E-mail: kscharak7@rediffmail.com
Abstract. We survey up to date progress on Bakers question (raised in 1984) on the dynamics of permutable entire functions:
Let f and g be two permutable (f g = g f ) non-linear entire functions. Are the Julia sets of f and g identical? Bakers question
still remains to be answered.
Keywords. Permutable entire functions, Normal family, Fatou set, Julia set, Wandering Domain.
2010 Mathematics Subject Classification. 30D35, 30D45.
1. Introduction equicontinuity and normality are equivalent for families of

holomorphic functions.
Consider the Riemann sphere C = C {} with Chordal
Definition 1.1. Let f be a meromorphic function. For n N,
metric given as follows:
let f n denote the nth iterate of f. The set
|z w|
(z, w) = C , z, w C, F (f ) := {z : {f n } is defined and forms a normal
(1 + |z|2 )(1 + |w|2 )
family in some neighbourhood of z}
if w = , then
1 is called the Fatou set or the set of normality of f and its
(z, ) = . complement J (f ) := C\F (f ) is the Julia set of f .
(1 + |z|2 )
The chordal metric is equivalent to another metric , called By definition, F (f ) is open and completely invariant:
the spherical metric on C, defined as (z, w) = the Euclidean z F (f ) if and only if f (z) F (f ), whereas J (f ) is
length of the shortest path on C between the stereographic non-empty, closed, perfect and completely invariant. Further,

images z and w of z and w. Clearly, if the chord joining z (i) F (f n ) = F (f ). and J (f n ) = J (f ) for all n 1.
and w subtends an angle at the origin, then (z, w) = . (ii) If U is a component(maximal connected subset) of
A family F of meromorphic functions on a domain D C F (f ), then f (U ) lies in some component V of F (f ).
is said to be normal in D if every sequence in F has a If Un Um = for n = m, where Un denotes the component
subsequence that converges uniformly (with respect to the of F (f ) which contains f n (U ), then U is called a wandering
metric ) on compact subsets of D to a meromorphic func- domain, else U is called a pre-periodic domain and if Un = U
tion which may be the constant . The concept of normality for some n N, then U is called periodic domain. Rational
was introduced in 1907 by Montel [27]. The study of normal functions do not have wandering domains (see [39]) whereas
families of meromorphic functions is a subject of great interest transcendental entire functions may have wandering domains
in its own right, it too is enjoying a resurgence of interest at all (see [3]).
levels due to its central role in complex dynamics. For normal Two holomorphic functions f and g are said to be per-
families of meromorphic functions, the reader is referred to mutable or commuting if f g = gf holds. If f and g are two
Schiffs book [35], Zalcmans survey article [45], Beardons permutable rational functions of degree at least 2 then Fatou
book [6], Bakers survey article [5], Bergweiler [9], Hua and [15] and Julia [22] both proved that J (f ) = J (g). Their work
Yangs book [18], Schwick [36], Drasin [14], and Blanchard was motivated by the problem of characterizing all permutable
[11] out of the huge literature on the subject. rational functions f and g. They solved this problem when
Recall Arzela-Ascoli theorem [1]: A family F of con- both f and g are polynomials. They proved that for permutable
tinuous maps of a domain D C into C is equicontinuous nonlinear polynomials f and g, there exist natural numbers
in D if and only if it is a normal family in D. So the notions of m, n such that (up to a conjugacy of linear maps) either

(i) f m (z) = g n (z); Baker [4] characterized all nonlinear entire functions
(ii) f (z) = z and g(z) = z ;
m n
which permute with the exponential function and proved the
or following result:
(iii) f (z) = Tm (z) and g(z) = Tn (z),
Theorem 1.2. Let g be a nonlinear entire function which
where Tk is the Tchebycheff polynomial determined by
permutes with f (z) = aebz + c(ab = 0, a, b, c C), then
the equation cos kw = Tk (cos w).
g = f n.
Classes of permutable rational functions are completely
Consider Hermans example [17]:
characterized by Ritt [34]. Here, by using complex dynamics
Let
technique, we prove that permutable rational functions of
degree at least 2 have identical Julia sets (see [6]): f (z) = z 1 + ez
Since rational maps satisfy a Lipschitz condition, there is g(z) = z + 2 i 1 + ez .
an M > 0 such that for all z, w C, we have
Then
(g(z), g(w)) M (z, w).
(i) f g = g f ;
Let w F (f ). Then by equicontinuity of {f } at w, given
n (ii) J (f ) = J (g);
any positive , there is a positive such that for all n, (iii) all components of F (f ) are periodic while g has a
wandering domain.
diam [f n (D(w, ))] < .
M In this example, f g = 2 i. Baker [4] solved Problem A
Since f and g commute, we deduce that for a special class of nonlinear entire functions by proving.
diam [f n g(D(w, ))] Theorem 1.3. If f and g are two nonlinear permutable entire
= diam [g f n (D(w, ))] M[f n (D(w, ))] < . functions such that f g = A, where A is some constant,
then J (f ) = J (g).
This implies that {f n } is a normal family on the open set
g(D(w, )) and so in particular, g(w) F (f ) and therefore, Given the simplicity of the proof given by Baker, one is
g and {g n } map F (f ) onto itself showing that {g n } is normal tempted to present an outline of the proof, here.
on F (f ). Thus, F (f ) F (g). By symmetry, we conclude First, note the following:
that J (f ) = J (g). A point w C is called Fatou-exceptional for f if the
In 195859, I. N. Baker [2] and G. Iyer [21] started their backward orbit of w O (w) := n1 {f n (w)} is finite.
work on the classification of all pairs of nonlinear permutable For any z J (f ) and for any w which is not a Fatou-
entire functions. They both proved that if a nonconstant poly- exceptional point for f , there is a sequence {zn } such that
nomial f is permutable with a transcendental entire func- f n (zn ) = w, zn z.
tion g, then f (z) = e2mi/k z + b for some m, k N and If N is a neighbourhood of z J (f ), then there is an n0 such
complex number b. It follows from this result, as well as Julia that for every n > n0 , f n (N ) contains any given compact
and Fatous results that in order to answer questions about set which does not contain a Fatou-exceptional point of f.
dynamics and classification of permutable entire functions, The repulsive fixed points of f form a dense subset of J (f ).
we only need to consider permutable transcendental entire
Proof of Theorem 1.3. z J (f ) if and only if z is a limit
functions.
point of repulsive fixed points of f . But g maps a repul-
Baker [4] initiated the study of dynamics of permutable
sive fixed point w of order p to another fixed point w such
transcendental entire functions and posed the following
that (f p ) (w) = (f p ) (w ). Let f and g be transcendental
problem which still continues to be an open problem:
entire functions. Suppose that z F (f ) and there is a sub-
Problem A. Let f and g be nonlinear entire functions. If f sequence {f nk } which has a finite limit in the component of
and g are permutable, is J (f ) = J (g)? F (f ) which contains z. Let U be a neighbourhood of z such

that U F (f ). Then all f nk (U ) lie in a single compact set Results in abundance have been proved concerning
on which g is uniformly continuous. We may take U so small primeness, left-primeness and unique factorizability of
that g(f nk (U )) = f nk (g(U )) has small diameter for large nk entire and meromorphic functions (see for example,
and hence g(U ) F (f ). In particular, g(z) F (f ). That is, [12,13,16,25,29,37]).
g(F (f )) F (f ). In 2001, Tuen Wai Ng [28] settled Bakers question
Next, we consider the case when f n
in U. Take a (see Problem A) for left-prime entire functions of a certain
constant B : B > |A| + 1. There exists n0 such that |f | > B n
form and proved the following result:
holds in U for all n > n0 , and hence |f (z)| > B for z f (U ) n
Theorem 1.6. Let f be a transcendental entire function which
and n > n0 . Suppose on the contrary that g(z)
/ F (f ). Then
satisfies the following conditions:
f m takes all values, with at most one exception, for arbitrarily
large m in g(U ). Thus there exists w = g( ), with U, (A1) f is not of the form H Q where H is periodic entire
such that and Q is a polynomial.
(A2) f is left-prime in the entire sense.
|f m (w)| = |f m (g( ))| = |g(f m ( ))| < 1. (A3) f has at least two distinct zeros.
Thus = f m ( ) f m (U ) and |g()| < 1. But |f ()| > B (A4) There exists a natural number N such that for
and so |A| = |f () g()| > B + 1, which contradicts any complex number c, the simultaneous equations
the choice of B. Hence g(z) F (f ). So g(F (f )) F (f ). f (z) = c, f (z) = 0 have at most N solutions.
This implies that each g n : F (f ) F (f ) omits J (f ) and (A5) The orders of zeros of f are bounded by some M N.
since J (f ) contains at least three points, Montels normality Let g be a nonlinear entire function which permutes
criterion implies the normality of {g n } on F (f ) and so with f . Then g(z) = af n (z) + b, where a is a kth root
F (f ) F (g) and hence by symmetry, F (f ) = F (g) which of unity and b C. Hence J (f ) = J (g).
in turn implies that J (f ) = J (g). Conditions (A4) and (A2) are related. For example, Ozawa
Bergweiler and Hinkkanen [10] solved Problem A for a [31] proved that if the order of f is finite and if for any c C,
class of permutable transcendental entire functions having no the simultaneous equations f (z) = c, f (z) = 0 have only
wandering domains and proved the following result: finitely many solutions, then f is left-prime in entire sense.
For any nonconstant polynomial p(z), the functions ez + p(z)
Theorem 1.4. Let f and g be permutable transcendental and sin z + p(z) are seen to satisfy conditions (A1)(A5). It is
entire functions. If both f and g have no wandering domains, observed (see Ng [28]) that if the conjecture Let f be a tran-
then J (f ) = J (g). scendental entire function such that f has at least two distinct
A slightly weaker version of the last theorem was proved zeros. If g is a nonlinear entire function which permutes with
by Langley [23]. f , then there is a C, g (a) = 0 such that f a and f g have
infinitely many distinct common zeros is true, one can replace
Definition 1.5. A meromorphic function F := f g is said to condition (A4) in Theorem 1.6 by a much weaker condition: for
have f and g as left and right factors respectively, provided any c C, the simultaneous equations f (z) = c, f (z) = 0
that f is meromorphic and g is entire (g may be meromorphic have only finitely many solutions. Moreover, these conditions
if f is rational). F is said to be prime (pseudo-prime, left- are not restrictive, almost all entire functions satisfy these con-
prime, right-prime,) if every factorization of F of the above ditions as can be seen from the following result of Noda [29]:
form implies either f is bilinear or g is linear (either f is
Theorem 1.7. Let f be a transcendental entire function.
rational or g is a polynomial, f is bilinear whenever g is tran-
There exists a countable set Ef C such that for all
scendental, g is linear whenever f is transcendental). When
/ Ef , fa (z) = f (z) + az satisfies the following conditions:
a
the factors f and g of F are restricted to entire functions,
the factorization is said to be in the entire sense. Accordingly, (B1) fa is nonperiodic.
we may use the term F is prime (left-prime) in the entire (B2) fa is prime in the entire sense.
sense. (B3) fa has infinitely many zeros.

(B4) For any complex number c, the simultaneous equations then there exists a countable set E such that if g is a trans-
fa (z) = c, fa (z) = 0 has at most one solution. cendental entire function which permutes with fa , then g = fan
(B5) The zeros of fa are simple. / E and n N.
and hence J (g) = J (f ) for all a
Nodas proof of Theorem 1.7 shows the existence of an excep- The last theorem also holds for
tional set Ef C such taht fa satisfies conditions (B1)(B4).
H (z) = sin z(cos z)
Replacement of Ef by Ef = Ef {f (c) : f (c) = 0}
ensures that fa also satisfies (B5). Clearly (Bj) implies (Aj) such that the order of H is finite and not an integer and H has
for j = 1, 2, 3, 4 and one can check that conditions (B1) and only simple zeros, and
(B2) together imply condition (A1). Hence by combining the
last two theorems, Ng [28] obtained the following result which fa (z) = H (z)[cos z a].
says that in certain sense, the answer to Bakers question is
Question. Does the last theorem hold when
YES for almost all entire functions:
Theorem 1.8. Let f be a transcendental entire function. fa (z) = H (z) + a(z),

Define fa (z) = f (z) + az. Then there exists a countable set
where H and are entire functions such that
Ef C such that for all a
/ Ef , any nonlinear entire function
g which permutes with fa is of the form g(z) = cfan (z) + d, (a) has at least one zero,
where c is a kth root of unity and d C and hence (b) T (r, ) = T (r, H ), where T (r, .) is the Nevanlinna
J (fa ) = J (g). characteristic function (see [16]),
(c) H and have no common zeros.
Further, Ng [28] proved Note that fa is prime in entire sense for most values of a
Theorem 1.9. Let q be a nonconstant entire function and p (see [12]).
be a polynomial with at least two distinct zeros. Suppose that
f (z) = p(z)eq(z) is prime in entire sense. Then any non- 2. Some recent progress on Bakers question
linear entire function g which permutes with f is of the form
g(z) = af n (z) + b, where a is a kth root of unity and b C Wang and Yang [42] considered permutable transcendental
and hence J (f ) = J (g). entire functions satisfying a certain functional equation and
Let C(f ) denote the set of all nonconstant entire functions obtained the following conclusion:
which permute with a transcendental entire function f. If we Theorem 2.1. Let f and g be two permutable transcendental
are able to show that entire functions and p be a nonconstant polynomial. Suppose
C(f ) {af + b, n 0, a = 0},
n that p(g) = ap(f ) + b, where a(= 0), b C. Then
J (f ) = J (g).
then J (f ) = J (g) for any nonconstant entire function g that
permutes with f. Noda [30] proved that C(f ) = {f n : n 0} A value a is called a singular value of f if it is either a
holds for all most all transcendental entire function f. For such critical value or an asymptotic value of f. The set of all finite
an f , every transcendental entire function in C(f ) has the same singular values of f is denoted by sing (f 1 ). We further
Julia set as that of f. Urabe [40] discussed Bakers question for define classes B and S as f B if sing (f 1 ) is bounded,
a class of transcendental entire functions. Precisely, he proved. and f S if sing (f 1 ) is finite respectively. Poon and Yang
Theorem 1.10. Let be a nonconstant entire function and [32] proved that if sing (f 1 ) and sing (g 1 ) are isolated, then
2
J (f ) = J (g). Note that this result includes the case when
H (z) = sin ze(2 cos z1)
f, g S. If f and g are assumed to be transcendental
and entire functions of finite order, then by Denjoy-Carleman-
a Ahlfors theorem we find that both f and g have finitely
fa (z) = H (z) sin 2z,
2 many asymptotic values and hence have identical Julia sets.

Hua and Wang [19] extended the last theorem to the class B. (2) deg q1 = deg q2 = = deg qn = deg (qi qj ) for all
Liao and Yang [24] proved that if there exists a nonconstant 1 i < j n.
polynomial 3(x, y) in both x and y such that 3(f, g) 0,
or
then J (f ) = J (g). This generalizes the results of Baker,
Poon-Yang and Wang where 3(x, y) is chosen respectively (1) q1 + q2 + + qn = 0;
to be x y b, x ay b, and p(x) q(y). Further, (2) deg (q1 + q2 + + qn ) + 1 = (deg q1 + deg q2 + +
Liao and Yang proved that if f and g are two permutable deg qn )/n;
transcendental pseudo-prime entire functions having a trans- or
cendental common right factor, then J (f ) = J (g). Also, they
(1) deg q1 = deg q2 = = deg qn = deg (qi qj )
proved that if g is a nonlinear entire function permutable with
@z 1 i < j n;
f (z) = p(z)eq(z) dz, where p and q are nonconstant poly-
(2) deg (q1 + q2 + + qn ) < deg q1 1;
nomials, then J (f ) = J (g). The case when p is a constant
(3) deg (p1 p2 . . . pn ) < an+1 .
remains unsettled.
Liao and Yang [26] also proved some significant variations Then there exist two nonconstant rational functions R and S
of earlier results. In particular, they proved: If f is a trans- such that R(f (z)) = S(g(z)) and J (f ) = J (g).
cendental entire function which is prime in entire sense and
Wang, Hua and Vaillancourt [44] also raised the following
whose derivative has only one zero, then any nonlinear entire
conjecture:
function g permutable with f is of the form g(z) = af n (z)+b,
where a is a kth root of unity and b C. Conjecture 2.4. Let f and g be two permutable trans-
Wang, Hua. Yang and Yang [43] obtained the following cendental entire functions with f (z) of the form
result:
f (z) := a + p1 eq1 + p2 eq2 + + pn eqn ,
Theorem 2.2. Let f = p + p1 eq1 + p2 eq2 , where qi (i = 1, 2)
where pi (z), qi (z)(1 i n) and p(z) are polynomials. Then
and p are nonconstant polynomials, pj 0(j = 1, 2). If a
there are two nonconstant rational functions R and S such that
finite order transcendental entire function g permutes with f ,
R(f (z)) = S(g(z)) which implies that J (f ) = J (g).
then J (f ) = J (g).
Singh and Wang [38] have shown that the Julia sets of f and
Further, Hua, Vaillancourt and Wang [20] proved that if f g are identical if either f has no wandering domains or f
g permutes with a solution f of a linear differential equa- is of bounded type and hence in addition if f and g permute
tion with polynomial coefficients, then under certain restric- then J (f ) = J (g).
tions on coefficients and growths of f and g, J (f ) = J (g). Very recently, Benini, Rippon and Stallard [7] proved the
Similar investigations are shown to be valid for solutions following result (compare with Theorem 1.4):
of algebraic differential equations by Zheng, Piyapong and
Theorem 2.5. If f and g are permutable trancendental entire
Keaitsuda [46].
functions having no simply connected fast escaping wandering
In 2011, Wang, Hua and Vaillancourt [44] proved the
domains, then J (f ) = J (g).
following result:
By fast escaping set of f (see [10]) we mean a set
Theorem 2.3. Let f and g be two permutable transcendental
entire functions with (g) < and p be a constant. Let f (z) A(f ) := {z C : there is L N
be of the form with |f n (z)| > M(r, f nL for n > L}
f (z) := p + p1 eq1 + p2 eq2 + + pn eqn A(f ) is always non-empty for transcendental entire functions
(see [33]).
such that either
Theorem 2.5 is in fact contained in the more precise result
(1) q1 + q2 + + qn = 0 and p1 p2 pn = constant, (see [7]): If f and g are permutable trancendental entire

functions, then F (f ) F (g) and F (g) F (f ), where [13] C. T. Chuang and C. C. Yang, Fix-points and Factori-

F (f ) is the union of those components of F (f ) that are not zation of Meromorphic Functions, World Scientific
simply connected fast escaping wandering domains. Under the Publ. Co., Singapore (1990).

hypothesis of Theorem 2.5, it follows that F (f ) = F (f ) [14] D. Drasin, Normal families and nevanlinna theory, Acta

and F (g) = F (g) and cosequently J (f ) = J (g). Further, Math., 122 (1969) 231263.
it is proved in [7] that permutable transcendental entire func- [15] P. Fatou, Sur literation analytique et les substitutions
tions have infinitely many Fatou components in common and permutables, J. Math. (9), 2 (1923) 343384.
it is conjectured there that Julia sets of such functions must be [16] F. Gross, Factorization of meromorphic functions,
identical. Mathematics Research Center, Naval Research Lab.
Washington, D.C. (1972).
References
[17] E. Herman, Examples de fractions rationelles ayant une
[1] L. V. Alhfors, Complex Analysis (third edition), orbite dense sur la shere de Riemann, Bull. Soc. Math.
McGraw-Hill (1979). France, 112 (1984) 93142.
[2] I. N. Baker, Zusemmensetzungen ganzer funktionen, [18] X. H. Hua and C. C. Yang, Dynamics of transcendental
Math. Z., 69 (1958) 121163. functions, Gordon and Breach Science Publishers
[3] I. N. Baker, An entire function which has wandering (1998).
domain, J. Austral. Math. Soc. Ser. A, 22 (1976) [19] X. H. Hua and X. Wang, Dynamics of permutable
173176. transcendental entire functions, Acta Math. Vietnamica,
[4] I. N. Baker, Wandering domains in the iteration of 27(3) (2002) 301306.
entire functions, Proc. London Math. Soc. (3), 49 (1984) [20] X. Hua, R. Vaillancourt and X. L. Wang, Per-
563576. mutable entire functions concerning differential equa-
[5] I. N. Baker, Iterations of entire functions: an intro- tions, J. Austral. Math. Soc., 83 (2007) 369384.
ductory survey, Lectures on Complex Analysis. World [21] G. Iyer, On permutable integral functions, J. London
Scientific, Singapore, New Jersey, London, and Hong Math. Soc., 34 (1959) 141144.
Kong (1987) 117. [22] G. Julia, Memoire sur la permutabilite des frac-
[6] A. F. Beardon, Iteration of rational functions, Springer- tions rationelles, Ann. Sci. Ecole Norm. Sup., 39(3)
Verlag, 1991. (1922) 131215.
[7] A. M. Benini, P. J. Rippon and G. M. Stallard, [23] J. Langley, Permutable entire functions and Baker
Permutable entire functions and multiply connected domains, Math. Proc. Camb. Phil. Soc., 125 (1999)
wandering domains, arXiv:1503.08112v1 [math.DS], 199202.
27 March (2015). [24] L. W. Liao and C. C. Yang, Julia sets of two permutable
[8] W. Bergweiler, Iterations of meromorphic functions, entire functions, J. Math. Soc. Japan, 56(1) (2004)
Bull. Amer. Math. Soc. (N.S.), 29 (1993) 151188. 169176.
[9] W. Bergweiler, Blochs principle, Comput. Methods [25] L. W. Liao and C. C. Yang, On factorization of entire
and Function Theory, 6 no. 1 (2006) 77108. functions of bounded type, Ann. Acad. Sc. Fenn. Math.,
[10] W. Bergweiler and A. Hinkkanen, On semiconjugation 29 (2004) 345356.
of entire functions, Math. Proc. Camb. Phil. Soc., 126 [26] L. W. Liao and C. C. Yang, On the Julia sets of two
(1999) 565574. permutable entire functions, Rocky Mountan J. Math.,
[11] P. Blanchard, Complex analytic dynamics on the 36(5), (2005) 16571674.
Riemann sphere, Bull. Amer. Math. Soc. (N.S.), 11 [27] P. Montel, Sur les suites infinies fonctions, Ann. Ecole
(1984) 85141. Norm. Sup., 24 (1907) 233334.
[12] K. S. Charak and I. Laine, On a class of prime entire [28] Tuen Wai Ng, Permutable entire functions and their
functions, Acta Math. Sinica, English Series, 25 no. 10 Julia sets, Math. Proc. Camb. Phil. Soc., 131 (2001)
(2009), 16471652. 129138.

[29] Y. Noda, On factorization of entire functions, Kodai [39] D. Sullivan, Quasiconformal homeomorphisms and
Math. J., 4 (1981) 480494. dynamics I; solution of the Fatou-Julia problem on
[30] Y. Noda, Julia sets of permutable entire functions, wandering domains, Ann. Math., 122 (1985) 401418.
Preprint Series in Math. TITECH-MATH 01-04(141) [40] H. Urabe, A note on permutable entire functions and
(2004). Julia sets, Bull. Kyoto Univ. Edu. Series B, 104 (2004)
[31] M. Ozawa, On certain criteria for the left-primeness 7379.
of entire functions, Kodai Math. Sem. Rep., 26 (1975) [41] Y. Wang, Fatou sets of permutable functions satis-
304317. fying functional equations, Proc. 2nd ISAAC Congress,
[32] K. K. Poon and C. C. Yang, Dynamical behavior of two Fukuoka Inst. Tech., Fukuoka, Japan, August 1523,
permutable entire functions, Ann. Polon. Math., 168 1999 (eds. H. G. W. Begehr, R. P. Gilbert and
(1998) 129138. J. Kajiwara), Vol. 2, Kluwer Academic Publishers,
[33] P. J. Rippon and G. M. Stallard, Fast escaping points of (2000) 903908.
entire functions, Proc. London Math. Soc. (3), 105(4) [42] X. Wang and C. C. Yang, On the Fatou components
(2012) 787820. of two permutable transcendental entire functions,
[34] J. F. Ritt, Permutable rational functions, Trans. Amer. J. Math. Anal. Appl., 278 (2003) 512526.
Math. Soc., 25 (1923) 399448. [43] X. Wang, X. Hua, C. C. Yang and D. Yang, Dynamics
[35] J. L. Schiff, Normal families, Springer-Verlag, of permutable transcendental entire functions, Rocky
New York (1993). Mountan J. Math., 36(6) (2006) 20412055.
[36] W. Schwick, Sharing values and normality, Arch. Math., [44] X. Wang, X. Hua and R. Vaillancourt, Permutable entire
59, (1992) 5054. functions concerning differential equations II, Com-
[37] A. P. Singh and K. S. Charak, Factorization of plex Variables and Elliptic Equations, 56(14) (2011)
certain classes of entire and meromorphic func- 155170.
tions, Indian J. Pure Appl. Math., 28(1) (1997) [45] L. Zalcman, Normal families: New perspectives, Bull.
8193. Amer. Math. Soc. (N.S.), 35 (1998) 215230.
[38] A. P. Singh and W. Yuefei, Julia sets of permutable [46] J. H. Zheng, N. Piyapong and M. Keaitsuda, Permutable
holomorphic maps, Science in China Series A: Mathe- entire functions satisfying algebraic differential equa-
matics, 49(11) (2006) 17151721. tions, J. Math. Anal. Appl., 35(4) (2009) 177183.
The Man Who Knew Infinity: Why Ramanujan Matters

Ken Ono and Robert Schneider
E-mail: ono@mathcs.emory.edu; robert.schneider@emory.edu
spare time at my disposal to work at Mathematics . . . I have

not trodden through the conventional regular course . . . but
I am striking out a new path . . . .
What followed in the letter were astonishing mathematical
formulas, so otherworldly the letters recipient could not help
but believe they were true. Written in 1913, it has taken
mankind one century to understand their meaning; along the
This article is republished with permission from Sloan

Science & Film (scienceandfilm.org), an online publication
Dear Sir . . . I beg to introduce myself to you as a clerk . . . of reporting on all things science and film, published by Museum of
the Port Trust Office at Madras . . . I have been employing the the Moving Im (New York).

way, the pursuit has led to solutions of ancient mathematical After absorbing the surprising equations in the letter, Hardy
mysteries, breakthroughs in modern physics, and ideas which invited Ramanujan to study in England, an extraordinary offer
help power the internet. for an Indian under colonial rule. Together they innovated
vast tracts of mathematics, before Ramanujan returned to
India in fragile health. Tragically, he died at 32 from a mis-
diagnosed illness, leaving three enigmatic notebooks that drive
cutting-edge research to this day.
Just as he spent his days immersed in ancient mythology,
Ramanujan shed new light on mathematical topics rooted in
classical times. Ramanujan was fascinated with the famous
number , and found many surprising formulas for this
constantin fact, the best computer algorithm ever found for
computing the digits of is based on Ramanujans work.
Moreover, he found deep formulas relating to the golden
ratio , both constants were considered in Euclids Elements,
A major motion picture based on the life of the letters
such as this continued fraction identity (which also involves
author, The Man Who Knew Infinity starring Dev Patel and
another famous constant e):
Jeremy Irons, was released April 29th (the first author of this 6
5
e2
article is an Associate Producer of the film). The mysterious 5 = e2
1+
clerk of the Port Trust Office of Madras has been in the 1+ e4
6
1+ e1+...
air a lot recently. News stories about black hole physics and
This exotic formula, and an infinity of others like it, result
discoveries in mathematics often center on his work.
from what we now call the Rogers-Ramanujan identities.
With this letter, Srinivasa Ramanujanimpoverished
(There is another romantic story here, of an obscure 19th
Hindu college dropout, self-taught in mathematics, reaching
century mathematician whose work Ramanujan himself redis-
for worlds beyond the shores of Indiaintroduced himself
covered and popularized.) In the pursuit of the meaning
not only to G. H. Hardy (superstar British mathematician), but
of relations such as these, unexpected areas of mathe-
to the history of human thought. Ramanujan spent his youth
matics and physics blossomed in the modern era: partition
sitting on the cool stone floors in the neighborhood temple,
theory, representation theory, statistical mechanics, to name a
surrounded by deities, his mind wandering the cosmos of math
few.
as he built upon the contents of a shabby textbook that was his
Ramanujans works have been an everlasting source of
bible.
suggestion for the worlds leading mathematicians. His work
on the tau-function plays a central role in the development in
modern number theory and algebraic geometry. His observa-
tions inspired the formulation of the Weil Conjectures, whose
proof earned Pierre Deligne the Fields medal in the 1970s.
These ideas form the groundwork for an entire area of mathe-
matics whose range of applications includes the internet. Other
observations anticipated the theory of modular Galois repre-
sentations, a subject which forms the groundwork for the
celebrated proof of Fermats Last Theorem by Andrew Wiles
in the 1990s. Ramanujans work on his death bed is presently
driving cutting edge research in string theory.
To mathematicians and scientists, Ramanujan symbolizes
Photos by Pressman Films. pure inspiration, the sort of creative flights more often

associated with musicians or artists. To contemporary India, he Ken Ono is the Asa Griggs Candler Professor of Mathematics
is a national hero, a household name. Furthermore, the story of at Emory University and a Fellow of the American
Ramanujan provides a modern archetype, the rise from humble Mathematical Society. He is the author of My Search for
conditions to the worlds center-stage that modern-day India Ramanujan: How I Learned to Count and served as associate
itself exemplifies. But what makes Ramanujans story worthy producer and consultant for The Man Who Knew Infinity.
of a mass audience?
Robert Schneider is a doctoral candidate in Mathematics at
First there is human interest. My association with him
Emory University. He is also a professional composer, record
is the one romantic incident in my life, Hardy wrote after
producer, and lead singer of indie band The Apples in Stereo.
Ramanujans death. At the heart of The Man Who Knew
Infinity is the awkward friendship between the self-taught,
pious Ramanujan (played by Patel), and the rigorous, avowed For Immediate Release
atheist Hardy (played by Irons). This unlikely alliance is itself
of mythological flavor. It was a miracle that Ramanujans letter IFC Films & Pressman Films
made its way to Hardys desk from India. He had reached out THE MAN WHO KNEW INFINITY
to other academics before, to no reply. That Hardy worked Partners With
through its strange contents, recognized the authors brilliance, Expii, Ken Ono, and the Templeton
and took Ramanujan under his wing, is also hard to imagine, World Charity Foundation
for Ramanujan was an unlikely student. So enraptured was he On The
with his continual discoveries that he flunked out of college
twice. (It is the worst instance that I know of the damage that Spirit of Ramanujan Math Talent
can be done by an inelastic educational system, commented Initiative
Hardy.) A Global Search for Undiscovered
Moreover, the story has an important moral in todays Mathematical Talent
society. Ramanujan matters because he represents endless
Prof. Po-Shen Loh and Prof. Ken Ono
curiosity and untapped potential, that we all have to believe
Webpage: po@expii.com; SOR@mathcs.emory.edu
in to proceed in the sciences. Science usually advances on
the work of thousands, over generations, fine-tuning and
extending the scope of understanding. But from time to Introducing the inaugural Spirit of Ramanujan Math Talent
time, creative fireballs like Ramanujan burst onto the scene Initiative, which strives to find undiscovered mathematicians
propelling human thought forward. Yet what if Ramanu- around the world and match them with advancement oppor-
jan had not reached out to, or been taken seriously by tunities in the field.
Hardy? The loss of scientific understanding is something The Spirit of Ramanujan Math Talent Initiative is
our modern world could not absorb. He matters because presented by Ken Ono, Asa Griggs Candler Professor
science matters: curiosity and creativity drive scientific of Mathematics at Emory University; Expii, the open,
inquiry. personalized learning platform; and the Templeton World
Echoing Hardys criticism of the inelastic nature of Charity Foundation, in conjunction with IFC Films and
Ramanujans formal schooling, todays educators are flooded Pressman Film, producers of the motion picture The Man Who
with a litany of complaintsdisaffected students, teacher burn- Knew Infinity.
out, over-testing, obsolete technology, inadequate funding, to The creators of the Spirit of Ramanujan Math Talent
name a few. How would we recognize and nurture an outlier Initiative were inspired by the life of Srinivasa Ramanujan,
like Ramanujan today, as Hardy did in his time? Like SETI, a gifted mathematician and the subject of the film. In 1913,
constantly listening for signals of alien intelligence, we as a Ramanujan, a poor Hindu college dropout who was self-
global civilization must scan our cities, towns and villages for taught in mathematics, reached out to renowned British mathe-
the next Ramanujans. matician G. H. Hardy in his search for mathematical peers and

colleagues. Hardy was so astonished by Ramanujans mathe- Participants in the open round will be invited to apply
matical theories that he invited Ramanujan to Cambridge to for further enrichment. Up to 20 eligible individuals will
study and collaborate. From the start, Ramanujan was fighting be awarded Templeton-Ramanujan Scholarly Develop-
against the odds because of his poverty, his location, and the ment Prizes to use for furthering their educational pursuits
inflexibility of formal schooling. Yet with his brilliant mind, and development. Just as Ramanujans journey was trans-
his dreams, and the support of the mathematical community, formed by his introduction to the mathematical community,
he defied those odds and went on to change the future of successful applicants will be invited to participate in the
mathematics. Together with Hardy, Ramanujan innovated vast 2017 Joint Mathematics Meetings, which will be held in
tracts of mathematics before returning to India in fragile health. Atlanta from January 47. There, they will engage with
Tragically, he died at 32, leaving three enigmatic notebooks award-winning mathematicians (e.g., Fields Medalists, Cole
that drive cutting-edge research to this day. Prize winners, Steele Prize winners, etc.) Organized by the
Jeremy Irons who plays Ramanujans influential Mathematical Association of America and the American
Prof. G. H. HARDY, in The Man Who Knew Infinity, says, Mathematical Society, the JMM is the largest annual
notes, We were lucky to find Ramanujan, the world would be mathematics conference in the world. Its purpose is to
a different place if he hadnt been published My character advance mathematical achievement, encourage research,
in the film, G. H. Hardy, states very directly that Ramanujans and provide the communication necessary to progress in the
genius was not discovered because of his teachers, but because field.
of Ramanujans own imagination and intuition. In the world In the spirit of Ramanujan, up to 10 participants will
today there are without a doubt talented minds out there who be offered financial support to attend and participate in
can advance humanity Im delighted that The Man Who approved summer research in mathematics. They will
Knew Infinity has shared the story of this miraculous figure, receive a Templeton-Ramanujan Summer Fellows Prize
inspiring this initiative. Hopefully we will be able to discover to cover summer program expenses.
new minds, that in turn, will lead to continued progress in the Educational materials related to Srinivasa Ramanujan and
world of mathematics. The Man Who Knew Infinity will be offered to educational
The story of Ramanujan and the program that he inspired programs worldwide.
directly relate to Sir John Templetons interest and enthusiasm
for the nurture, education, and recognition of geniuses. This To enter the open round, please visit: expii.com/ramanujan
unique project is distinguished by widespread awareness The Spirit of Ramanujan Math Talent Initiative is headed
generated by the major film release, the use of technology by Ken Ono with an advisory board that also includes Manjul
to reach extraordinarily gifted people, and a series of oppor- Bhargava (Princeton), Olga Holtz (UC Berkeley), Po-Shen
tunities for these people to further their intellectual develop- Loh (Carnegie Mellon), and Sujatha Ramdorai (U British
ment. Columbia).
Dev Patel, who plays Ramanujan in the film says of the Corporate Sponsors of the Spirit of Ramanujan Math
initiative, Math was never my thing, but after researching Talent Initiative include Expii, Texas Instruments, Springer
Ramanujan for this film Ive realized the profound contri- Publishing, and Maplesoft.
butions that mathematicians make in our lives Im fortunate The Man Who Knew Infinity (starring Jeremy Irons and Dev
to pursue my passion for acting, its great that this competition Patel) was advised by Ken Ono, a Guggenheim Fellow and
gives those who are passionate about math to pursue theirs. cofounder of the Spirit of Ramanujan, and by Fields Medalist
The Spirit of Ramanujan Math Talent Initiative strives to Manjul Bhargava. The film offers an in-depth view of one of
find undiscovered mathematicians around the world and match the human stories behind the mathematical achievements that
them with advancement opportunities. These include: we take for granted in our everyday lives. For more information
An open round hosted on expii.com (via its Solve feature) on the movie, visit www. themanwhoknewinfinity.com.
invites people worldwide to solve creative mathematical Expii is an innovative interactive education platform that
puzzles through their smartphones. embraces the creative side of learning. Founded by Carnegie

Mellon University math Prof. Po-Shen Loh, it nurtures learners mathematical education, and foster awareness and apprecia-
of all styles. For more information on Expii and Po-Shen Loh, tion of mathematics and its connections to other disciplines
visit www.expii.com. and to everyday life. For more information on AMS, visit
The Templeton World Charity Foundation (TWCF) was www.ams.org.
established by Sir John Templeton to carry out his phil- Founded in 1915, the Mathematical Association of America
anthropic aims on a truly global basis. Their mission is (MAA) is a professional society that focuses on mathe-
to find the worlds best minds and promote open-minded, matics accessible at the undergraduate level. Mem-
forward-looking inquiry in a wide range of fields. TWCF bers include university, college, and high school teach-
works with leading academic institutions and experts from ers; graduate and undergraduate students; pure and
around the world to support projects of the highest caliber applied mathematicians; computer scientists; statisti-
and to communicate the results of these projects to a cians; and many others in academia, government, busi-
wide audience. For more information on TWCF, visit ness, and industry. For more information on MMA,
www.templetonworldcharity.org. visit www.maa.org.
Founded in 1888 to further mathematical research Ken Ono, A Guggenheim Fellow and author of My Search
and scholarship, the American Mathematical Society (AMS) for Ramanujan: How I Learned to Count, is cofounder of
fulfills its mission today through programs and services the Spirit of Ramanujan along with Fields Medalist Manjul
that promote mathematical research and its uses, strengthen Bhargava.
Details of Workshop/Conferences in India
For details regarding Advanced Training in Mathematics Schools

Visit: http://www.atmschools.org/
Name: International Conference on Number Theory

Date: August 1721, 2016
Location: Kerala School of Mathematics
Visit: http://ksom.res.in/IC-KumarMurti-Aug2016.php
Name: 24th International Conference on Finite or Infinite Dimensional Complex Analysis and Applications
Date: August 2226, 2016
Location: Anand International College of Engineering, Jaipur
Visit: http://anandice.ac.in/24icfidcaa-2016/#
Name: International Conference on Special Functions & Applications

Date: September 0911, 2016
Location: Jamia Millia Islamia Central University, New Delhi
Visit: http://www.ssfaindia.webs.com/conf.htm
Name: International Conference on Applications of Mathematics in Topological Dynamics, Physical, Biological and Chemical
Systems
Date: December 911, 2016

Location: Calcutta Mathematical Society Asutosh Bhavan AE-374, Sector-I, Salt Lake City, Kolkata-700064, West Bengal,
India
Visit: http://www.calmathsoc.org/
Name: Recent Advances in Operator Theory and Operator Algebras 2016

Date: December 13December 22, 2016
Location: Indian Statistical Institute, Bangalore, India
Visit: http://www.isibang.ac.in/jay/OTOA/OTOA16/otoa16.html
Name: International Conference of The Indian Mathematics Consortium (TIMC) in cooperation with American Mathematical
Society (AMS)
Location: DST-Centre for Interdisciplinary Mathematical Sciences, Banaras Hindu University, Varanasi 221005, India
Visit: http://www.bhu.ac.in/seminar/timcams2016/
Name: SPACE 2016 Sixth International Conference on Security, Privacy and Applied Cryptographic Engineering
Location: C. R. Rao Advanced Institute of Mathematics Statistics and Computer Science, Hyderabad, India (AIMSCS)
Visit: http://www.math.umn.edu/math-sa-sara0050/space16/
Name: Statistical Methods in Finance 2016

Location: Chennai Mathematical Institute (CMI), Chennai, India
Visit: http://www.cmi.ac.in/sourish/StatFin2016/
Name: Platinum Jubilee International Conference on Applications of Statistics

Location: Department of Statistics, University of Calcutta 35, Ballygunge Circular Road Kolkata 700019, West Bengal,
India
Visit: http://stat.caluniv.in/platinum/
Name: 10th International Conference of IMBIC on Mathematical Sciences for Advancement of Science and
Technology
Location: Indismart Hotel, Salt Lake, Kolkata
Visit: http://imbic.org/forthcoming.html
Details of Workshop/Conferences in Abroad

For details regarding SIAM conferences,
Visit: http://www.siam.org/meetings/archives.php#2017

Name: Integers Conference 2016
Date: October 69, 2016
Location: University of West Georgia, Carrollton, Georgia
Visit: http://www.westga.edu/assetsCOSM/math/INTEGERS Conference 2016.pdf
Name: Mathematical Results in Quantum Physics

Location: Georgia Institute of Technology, Atlanta, Georgia
Visit: http://qmath13.gatech.edu/
Name: AIM Workshop: Boundaries of groups

Location: American Institute of Mathematics, San Jose, CA
Visit: http://aimath.org/workshops/upcoming/groupbdy/
Name: International Conference on Statistical Distributions and Applications

Location: Crowne Plaza, Niagara Falls, Canada
Visit: http://people.cst.cmich.edu/lee1c/icosda2016/
Name: 9th International Symposium on Biomathematics and Ecology

Location: College of Charleston, Charleston, SC, USA
Visit: https://about.illinoisstate.edu/biomath/beer/Pages/default.aspx
Name: Binghamton University Graduate Conference in Algebra and Topology

Location: Binghamton University, Binghamton, NY
Visit: https://sites.google.com/a/binghamton.edu/bugcat2016/
Name: Stochastic Topology and Thermodynamic Limits (Workshop)

Location: ICERM at Brown University, Providence, Rhode Island
Visit: https://icerm.brown.edu/programs/sp-f16/w2/
Name: Applications of Ergodic Theory in Number Theory

Location: CIRM Luminy, Marseille
Visit: http://lemanczyk-ferenczi.weebly.com/doctoral-school.html
Name: International Conference on Modeling, Simulation and Control 2016

Location: UC Berkeley, San Francisco Bay Area, USA
Visit: http://www.iaeng.org/WCECS2016/ICMSC2016.html

Name: AIM Workshop: Rational subvarieties in positive characteristic
Visit: http://aimath.org/workshops/upcoming/ratsubvarpos/
Name: Geometry of Mapping Class Groups and Out (Fn)

Location: Mathematical Sciences Research Institute, Berkeley, California
Visit: http://www.msri.org/workshops/771
Name: The 4th International Conference on Complex Dynamical Systems in Life Sciences: Modeling and Analysis
Location: Ibnou Zohr University, Agadir, Morocco
Visit: http://www.cosylis.org/iccds2016/
Name: IMA Annual Program Workshop: Mathematical and Physical Models of Nonlinear Optics
Date: October 31November 4, 2016
Location: IMA, University of Minnesota, Minneapolis, MN
Visit: https://www.ima.umn.edu/2016-2017/W10.31-11.4.16/
Name: AIM Workshop: Optimization strategies for transportation

Date: October 31November 4, 2016
Visit: http://aimath.org/workshops/upcoming/transportation/
Name: 36th Southeastern-Atlantic Regional Conference on Differential Equations

Date: November 56, 2016
Location: Florida Gulf Coast University, Fort Myers, FL USA
Visit: http://lebesgue.fgcu.edu/SEARCDE2016/
Name: AIM Workshop: Polyhedral Geometry and Partition Theory

Visit: http://aimath.org/workshops/upcoming/polypartition/
Name: Homological Mirror Symmetry: Methods and Structures

Location: Institute for Advanced Study, Princeton, New Jersey
Visit: http://www.math.ias.edu/sp/mirrorsymmetry
Name: First EAI International Conference on Computer Science and Engineering

Location: Batu Feringhi, Penang Island, Malaysia
Visit: http://compse-conf.org/2016/show/cf-calls

Name: Workshop III: Collective Variables in Quantum Mechanics
Location: Institute for Pure and Applied Mathematics (IPAM), Los Angeles, California
Visit: http://www.ipam.ucla.edu/programs/workshops/workshop-iii-collective-variables-in-quantum-mechanics/
Name: Maltsev Meeting

Location: Sobolev Institute of Mathematics, Novosibirsk, Russia
Visit: http://math.nsc.ru/conference/malmeet/16/
Name: Topology and Geometry in a Discrete Setting (Workshop)

Date: November 28December 2, 2016
Location: ICERM at Brown University, Providence, Rhode Island
Visit: https://icerm.brown.edu/programs/sp-f16/w3/
Name: Synergies between Machine Learning and Physical Models

Location: Institute for Pure and Applied Mathematics (IPAM), UCLA, Los Angeles, CA
Visit: http://www.ipam.ucla.edu/programs/workshops/workshop-iv-synergies-between-machine-learning-and-
physical-models/
Name: AIM Workshop: Global Langlands correspondence
Visit: http://aimath.org/workshops/upcoming/globlanglands/
Name: Amenability, Coarse Embeddability and Fixed Point Properties

Date:December 69, 2016
Location: Mathematical Sciences Research Institute, Berkeley, California
Visit: http://www.msri.org/workshops/770
Name: 11th IMA International Conference on Mathematics in Signal Processing

Location: Austin Court, Birmingham, UK
Visit: http://www.ima.org.uk/conferences/conferences calendar/11th maths in signal processing.html
Name: IMA Annual Program Workshop: Mathematical and Numerical Modeling in Optics
Location: IMA, University of Minnesota, Minneapolis, MN
Visit: https://www.ima.umn.edu/2016-2017/W12.12-16.16/
Name: Ergodic Theory and its Connections with Arithmetic and Combinatorics
Location: CIRM Luminy, Marseille
Visit: http://lemanczyk-ferenczi.weebly.com/conference.html

Name: The 21st Asian Technology Conference in Mathematics (ATCM 2016)
Location: Pattaya, Thailand
Visit: http://atcm.mathandtech.org/
Name: International Conference Anosov systems and modern dynamics

Location: Steklov Mathematical Institute & National Research University, Moscow, Russia
Visit: http://www.mathnet.ru/eng/conf812
Department of Mathematics,
I.I.T. Roorkee is Organizing the
International Conference on Mathematical Analysis and its
to be held between
Applications (ICMAA 2016)
November 28 December 02, 2016
The focus of the conference is on topics of Mathematical Analysis and its Applications, including but not limited to
Analysis of Differential Equations (including Control theory, Fractional Calculus and Stochastic PDEs)
Complex Analysis and Harmonic Analysis (including Potential theory, Harmonic Mappings and Quasi-Conformal Mappings)
Fourier and Wavelet Analysis
Inverse Problems and Non-linear Analysis
Matrix Analysis, Operator Theory and Function Spaces
Modern Methods of Summability and Approximation
Orthogonal Polynomials and Special Functions
The SIAM Activity group on Orthogonal Polynomials and Special Functions (SIAG-OPSF) is also involved in the conference.
Further, SIAM has agreed to provide travel grants to early career researchers from US Institutions to participate in the conference.
The conference proceedings will be published in the Journal of Analysis, Springer.
Important dates:
Early Registration & Payment deadline July 31, 2016

Late Registration & Payment deadline October 31, 2016
Submission of papers for conference Proceedings December, 2016
For further details and updates of the conference kindly visit www.iitr.ac.in/icmaa/2016/ or contact us at
icmaa2016@iitr.ac.in, icmaa2k16@gmail.com
The Mathematics Newsletter may be download from the RMS website at

www.ramanujanmathsociety.org

MNL Vol27 No1 Jun2016 1478024956074

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

MNL Vol27 No1 Jun2016 1478024956074

Загружено:

Авторское право:

Доступные форматы

MATHEMATICS NEWSLETTER

Volume 26 MarchJune 2016

An Introduction to the Kakeya Problem . . . Edward Kroc 109

Details of Workshop/Conferences in India 135

Details of Workshop/Conferences in Abroad 136

Mathematics Newsletter -109- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -110- Vol. 26 #4, MarchJune 2016

x b = inf yE x y. This defines b to be the unique point ' )

Mathematics Newsletter -111- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -112- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -113- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -114- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -115- Vol. 26 #4, MarchJune 2016

Lemma 5.5. Let E Rn be such that |E| > 0. Let 0 < 1

Mathematics Newsletter -116- Vol. 26 #4, MarchJune 2016

more than of our tubes Tej ; i.e., / 0

Mathematics Newsletter -117- Vol. 26 #4, MarchJune 2016

-separated and on the prescence of the normalizing dual

Mathematics Newsletter -118- Vol. 26 #4, MarchJune 2016

by hypothesis. We apply Holder to the sum in (5.20) and

& & ( * 1 Even though , U (,) is a disjoint union, there is nothing

Mathematics Newsletter -119- Vol. 26 #4, MarchJune 2016

) c(N ) 21/p c(2N) + 2c(C n1 N 2 ). (5.25)

Mathematics Newsletter -120- Vol. 26 #4, MarchJune 2016

Subtracting (5.28) from (5.29) and rearranging, we see that we

Mathematics Newsletter -121- Vol. 26 #4, MarchJune 2016

, = { , : |T | < |T E | 2|T |}; |E | # n1 f ()

Mathematics Newsletter -122- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -123- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -124- Vol. 26 #4, MarchJune 2016

1. Introduction equicontinuity and normality are equivalent for families of

Mathematics Newsletter -125- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -126- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -127- Vol. 26 #4, MarchJune 2016

Theorem 1.8. Let f be a transcendental entire function. fa (z) = H (z) + a(z),

Mathematics Newsletter -128- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -129- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -130- Vol. 26 #4, MarchJune 2016

The Man Who Knew Infinity: Why Ramanujan Matters

spare time at my disposal to work at Mathematics . . . I have

This article is republished with permission from Sloan

Mathematics Newsletter -131- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -132- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -133- Vol. 26 #4, MarchJune 2016

Mathematics Newsletter -134- Vol. 26 #4, MarchJune 2016

Details of Workshop/Conferences in India

For details regarding Advanced Training in Mathematics Schools

Name: International Conference on Number Theory

Name: International Conference on Special Functions & Applications

Mathematics Newsletter -135- Vol. 26 #4, MarchJune 2016

Name: Recent Advances in Operator Theory and Operator Algebras 2016

Name: Statistical Methods in Finance 2016

Name: Platinum Jubilee International Conference on Applications of Statistics

Details of Workshop/Conferences in Abroad

Mathematics Newsletter -136- Vol. 26 #4, MarchJune 2016

Name: Mathematical Results in Quantum Physics

Name: AIM Workshop: Boundaries of groups

Name: International Conference on Statistical Distributions and Applications

Name: 9th International Symposium on Biomathematics and Ecology

Name: Binghamton University Graduate Conference in Algebra and Topology

Name: Stochastic Topology and Thermodynamic Limits (Workshop)

Name: Applications of Ergodic Theory in Number Theory

Name: International Conference on Modeling, Simulation and Control 2016