Yuri Lima and I have just uploaded to ArXiv our paper “Symbolic dynamics for non-uniformly hyperbolic maps with discontinuities”. The main motivation for our paper is the question of extending the celebrated (Brin prize) work of Sarig on symbolic models/Markov partitions for smooth surface diffeomorphisms to the context of billiard maps: indeed, the main result of our paper is a partial solution to a problem appearing in page 346 of Sarig’s article.

An interesting corollary of our results is a refinement of a theorem of Chernov on the number of periodic points of certain billiard maps. More precisely, for a certain class of billiard maps {f:M\rightarrow M}, Chernov proved that

\displaystyle \liminf\limits_{n\rightarrow\infty}\frac{1}{n}\log\#\textrm{Per}_n(f)\geq h

where {\textrm{Per}_n(f)} is the set of periodic points of {f} with period {n} and {h} is the Kolmogorov-Sinai entropy of the Liouville measure. From our main results, Yuri and I can show that the billiard maps {f:M\rightarrow M} studied by Chernov actually satisfy:

\displaystyle \exists\, C>0, \, p\in\mathbb{N} \textrm{ such that } \#\textrm{Per}_{np}(f)\geq C e^{hnp} \, \forall \, n\in\mathbb{N}

Remark 1 Our improvement of Chernov’s theorem is “similar in spirit” to Sarig’s improvement of Katok’s theorem on the number of periodic points for smooth surface diffeomorphisms: see Theorem 1.1 in Sarig’s paper for more details.

Below the fold, we give a slightly simplified version of the main result in our paper and we explain some steps of its proof.

Read More…

Posted by: matheuscmss | June 6, 2016

Zorich conjecture for hyperelliptic Rauzy-Veech groups

Artur Avila, Jean-Christophe Yoccoz and I have just uploaded to ArXiv our paper Zorich conjecture for hyperelliptic Rauzy-Veech groups.

This is the first installment of a series of two articles on the Kontsevich-Zorich cocycle over certain {SL(2,\mathbb{R})}-invariant loci in moduli spaces of translation surfaces obtained from cyclic cover constructions (inspired from the works of Veech and McMullen).

More precisely, the second paper of this series (still in preparation) studies the Kontsevich-Zorich cocycle over {SL(2,\mathbb{R})}-orbits of certain cyclic covers of translation surfaces in hyperelliptic components called {\mathcal{H}(2g-2)^{hyp}} and {\mathcal{H}(g-1,g-1)^{hyp}} in the literature. (The curious reader can find more explanations about this forthcoming paper in this old blog post here [cf. Remark 7].)

Of course, before studying the cyclic covers, we need to obtain some good description of the Kontsevich-Zorich cocycle on the hyperelliptic components and this is the purpose of the first article of the series.

Since the first paper of this series is not long, this post will just give a brief “reader’s guide” rather than entering into the technical details.

Remark 1 In the sequel, we will assume some familiarity with translation surfaces.

1. Rauzy-Veech groups and Zorich conjecture

The starting point of our article is a description of certain combinatorial objects — called hyperelliptic Rauzy diagrams — coding the dynamics of the Kontsevich-Zorich cocycle on {\mathcal{H}(2g-2)^{hyp}} and {\mathcal{H}(g-1,g-1)^{hyp}}.

Remark 2 By the time that the first version of our article was written, we thought that we found a new description of these diagrams, but Pascal Hubert kindly pointed out to us that G. Rauzy was aware of it.

One important feature of this description of hyperelliptic Rauzy diagrams is the fact that we can order these diagrams by “complexity” in such a way that two consecutive diagrams can be related to each other by an inductive procedure.

The behavior of the Kontsevich-Zorich cocycle (with respect to Masur-Veech measures) is described in general by the Rauzy-Veech algorithm: roughly speaking, this algorithm is a natural way to attach matrices (acting on homology groups) to the arrows of Rauzy diagrams, and, in this language, the action of the Kontsevich-Zorich cocycle is just the multiplication of the matrices attached to concatenations of arrows of Rauzy diagrams.

In particular, the features of the Kontsevich-Zorich cocycle (with respect to Masur-Veech measures) can be derived from the study of the so-called Rauzy-Veech groups, i.e., the groups generated by the matrices attached to the arrows of a given Rauzy diagram. For example, the celebrated paper of Avila and Viana solving affirmatively a conjecture of Kontsevich and Zorich proves the simplicity of the Lyapunov exponents of the Kontsevich-Zorich cocycle by establishing (inductively) the pinching and twisting properties for Rauzy-Veech groups.

In our article, we exploit the “inductive” description of hyperelliptic Rauzy diagrams to compute the hyperelliptic Rauzy-Veech groups.

Remark 3 Our arguments for the computation of hyperelliptic Rauzy-Veech groups were inspired from the calculations of {U(p,q)}-blocks of the Kontsevich-Zorich cocycle over cyclic covers in the second paper of this series. In other words, we first developed some parts of the second paper before writing the first paper.

An interesting corollary of this computation is the fact that hyperelliptic Rauzy-Veech groups are explicit finite-index subgroups of the symplectic groups {\textrm{Sp}(2g,\mathbb{Z})}, so that they are Zariski dense in {\textrm{Sp}(2g,\mathbb{R})}.

The Zariski density of (general) Rauzy-Veech groups in symplectic groups was conjectured by Zorich (see, e.g., Remark 6.12 in Avila-Viana paper) as a step towards the Kontsevich-Zorich conjecture established by Avila-Viana. Therefore, the previous paragraph means that Zorich conjecture is true for hyperelliptic Rauzy-Veech groups (and this justifies our choice for the title of our paper).

Here, it is worth to point out that Zorich conjecture asks more than what is needed to prove the Kontsevich-Zorich conjecture. Indeed, the Zariski denseness in symplectic groups imply the pinching and twisting properties of Avila-Viana, so that Zorich conjecture implies Kontsevich-Zorich conjecture. On the other hand, we saw in this previous blog post that a pinching and twisting group of symplectic matrices might not be Zariski dense: in other words, the techniques of Avila-Viana solve the Kontsevich-Zorich without addressing Zorich conjecture.

Thus, our proof of Zorich conjecture for hyperelliptic Rauzy-Veech groups gives an alternative proof of this particular case of Avila-Viana theorem.

2. Braid groups and A’Campo theorem

After a preliminary version of our article was complete, Martin Möller noticed some similarities between our characterization of hyperelliptic Rauzy-Veech groups and a result of A’Campo on the images of certain monodromy representations associated to hyperelliptic Riemann surfaces.

As it turns out, this is not a coincidence: we showed that the elements of the hyperelliptic Rauzy-Veech group associated to certain elementary loops on hyperelliptic Rauzy diagrams are induced by Dehn twists lifting the generators of a braid group; hence, this permits to recover our description of hyperelliptic Rauzy-Veech groups from A’Campo theorem.

Remark 4 In a certain sense, the previous paragraph is a sort of sanity test: the same groups found by A’Campo were rediscovered by us using different methods.

Closing this short post, let me point out that this relationship between loops in hyperelliptic Rauzy diagrams and Dehn twists in hyperelliptic surfaces reveals an interesting fact: the fundamental groups of the combinatorial model (hyperelliptic Rauzy diagram) coincides with the orbifold fundamental groups of {\mathcal{H}(2g-2)^{hyp}} and {\mathcal{H}(g-1,g-1)^{hyp}}. In other words, the hyperelliptic Rauzy diagrams “see” the topology of objects ({\mathcal{H}(2g-2)^{hyp}} and {\mathcal{H}(g-1,g-1)^{hyp}}) coded by them.

Posted by: matheuscmss | May 29, 2016

Yoccoz proof of Jakobson theorem I

Almost twenty years ago, Jean-Christophe Yoccoz gave several lectures (at ETH and Collège de France, for instance) on a proof of Jakobson theorem (based on the so-called Yoccoz puzzles), and he distributed hard copies of his lecture notes whenever they were requested (for example, he gave me such a copy when I became his post-doctoral student in 2007).

More recently, he gave a couple of lectures at Collège de France aiming to generalize his proof of Jakobson theorem to obtain some results improving upon the Wang-Young theory of rank one attractors.

Of course, his lectures (whose videos are available at his website) started by recalling several elements of his proof of Jakobson theorem and I decided to take the opportunity to write a series of posts about Jean-Christophe’s proof of Jakobson theorem (based on his hand-written lecture notes from 1997).

For the first installment of this series, we’ll review some aspects of the dynamics of the one-dimensional quadratic maps and, after that, we’ll state Jakobson’s theorem, describe the three main steps of Yoccoz proof of Jakobson theorem and implement the first step of the strategy.

1. Introduction

The dynamics of one-dimensional affine maps {F_{a,b}(z)=az+b} is fairly easy understand. The change of variables provided by the translation {T_c(z)=z+c} transforms {F_{a,b}} into {(T_c^{-1}\circ F_{a,b}\circ T_c)(z) = az + ((a-1)c+b)}. If {a\neq 1}, the choice {c=-b/(a-1)} shows that the affine map {F_{a,b}} is conjugated to its linear part {F_{a,0}(z)=az}. If {a=1}, {F_{a,b}(z)=z+b} is a translation.

In other words, the dynamics of polynomial maps {P:\mathbb{C}\rightarrow\mathbb{C}} is not very interesting when {\textrm{deg}(P)=1}. On the other hand, we shall see below that quadratic maps are already sufficiently complicated to produce all kinds of rich dynamical behaviors.

2. Quadratic family

The quadratic family is {P_c(z)=z^2+c} where {z\in\mathbb{C}} and {c\in\mathbb{C}}.

Remark 1 This family is sometimes presented in the literature as {c-z^2} or {cz(1-z)}. As it turns out, these presentations are equivalent to each other: indeed, the affine change of variables {h(z)=Az+B} converts {P_c} into

\displaystyle (h^{-1}\circ P_c\circ h)(z) = Az^2+2Bz+\frac{B^2+c-B}{A}

The dynamics of {P_c} near infinity is easy to understand: in fact, since {P_c(z)\approx z^2} for {|z|\gg 1}, one has that {\infty} attracts the orbit {\{P_c^n(z)\}_{n\in\mathbb{N}}} of any {z\in\mathbb{C}} with {|z|\gg 1}.

This means that the interesting dynamics of {P_c} occurs in the filled-in Julia set:

\displaystyle K(c):=\{z\in\mathbb{C}: P_c^n(z)\not\rightarrow\infty \textrm{ as } n\rightarrow\infty\}

Note that {K(c)} is totally invariant, that is, {P_c(K(c)) = K(c) = P_c^{-1}(K(c))}. Also, {K(c)} is a compact set because {|z|>R} implies that {|P_c(z)|>R|z|-|c|>(R-\frac{|c|}{R})|z|} and {R-\frac{|c|}{R}>1} whenever {R>\frac{1+\sqrt{1+4|c|}}{2}:=R_c}, so that

\displaystyle K(c)=\bigcap\limits_{n\in\mathbb{N}} P_c^{-n}(\mathbb{D}_{R_c}(0))

(where {\mathbb{D}_R(0):=\{z\in\mathbb{C}: |z|\leq R\}}).

Moreover, {K(c)\neq\emptyset} because it contains all periodic points of {P_c} (i.e., all solutions of the algebraic equations {P_c^n(z)=z}, {n\in\mathbb{N}}).

Remark 2 {K(c)} is a full compact set, i.e., {\mathbb{C}-K(c)} is connected: indeed, this happens because the maximum principle implies that a bounded open set {U\subset \mathbb{C}} with boundary {\partial U\subset K(c)} must be completely contained in {K(c)} (i.e., {U\subset K(c)}).

The dynamics of {P_c} on {K(c)} is influenced by the behaviour of the orbit {\{P_c^n(0)\}_{n\in\mathbb{N}}} of the critical point {0\in\mathbb{C}}. More precisely, let us consider the Mandelbrot set

\displaystyle M:=\{c\in\mathbb{C}:P_c^n(0)\not\rightarrow\infty\}:=\{c\in\mathbb{C}: 0\in K(c)\}

Read More…

Posted by: matheuscmss | March 16, 2016

Counting torus fibrations on a K3 surface (after S. Filip)

Last week, Simion Filip gave the talk “Counting torus fibrations on a K3 surface” at Université Paris 11 (Orsay). This post is a transcription of my notes from his lecture and, of course, all typos/mistakes are my sole responsibility.

1. Introduction

A classical problem in Dynamical Systems is the investigation of closed trajectories in billiards in polygons.

In the case of rational polygons (i.e., polygons whose angles belong to {\mathbb{Q}\pi}), it is known that closed trajectories are abundant. A popular way to establish this fact passes through the procedure of unfolding a rational polygon into a flat surface: roughly speaking, instead of letting the trajectories reflect on the boundary of the polygon, we reflect (finitely many times) the boundaries of the polygon in order to obtain straight line trajectories on a flat surface. See this excellent survey of Masur-Tabachnikov for more details.

For our purposes, we recall that a flat surface is the data {(X,\omega,\Omega)} of a Riemann surface {X}, a non-trivial holomorphic {1}-form {\Omega} and the flat metric {\omega} thought as the Kähler form {\omega=\frac{i}{2}\Omega\wedge\overline{\Omega}}. Note that {\Omega} has zeros, i.e., {\omega} has (conical) singularities, whenever {X} has genus {>1} (by Riemann-Roch theorem).

Some of the key features concerning closed trajectories in flat surfaces are:

  • closed geodesics of the flat metric come in families of parallel trajectories called cylinders in the literature;
  • such closed geodesics occur in a dense set of directions in the unit circle {S^1};
  • Eskin and Masur showed the following asymptotics for the problem of counting cylinders: there exists a constant {c>0} such that the number of (cylinders of) closed trajectories of length {\leq L} is {\sim c\cdot L^2}.

The goal of this post is to generalize this picture to higher dimensions or, more precisely, to K3 surfaces.

2. K3 surfaces

Definition 1 A compact complex {2}-dimensional manifold {X} is a K3 surface if

  • (i) {X} admits a (global) nowhere vanishing holomorphic {2}-form {\Omega};
  • (ii) the first Betti number {b_1(X)} is zero (this is equivalent to {\pi_1(X)=\{0\}} in this context).

Two basic examples of K3 surfaces are:

Example 1 Quartic surfaces in {\mathbb{P}^3(\mathbb{C})}, i.e., {\{F=0\}\subset \mathbb{P}^3(\mathbb{C})}, {F} is a polynomial of degree {\textrm{deg}(F)=4}.

Example 2 (Kummer examples) Let {A=\mathbb{C}^2/\Lambda} be a complex torus (i.e., {\Lambda} is a lattice of {\mathbb{C}^2}). Then, {A_0= A/\{\pm\textrm{Id}_A\}} has a subset \textrm{sing} of {16} singular points, and the blow-up {X=Bl_{\textrm{sing}}(A_0)} is a K3 surface.

Some basic properties of K3 surfaces include:

  • all K3’s are diffeomorphic;
  • all K3’s are Kähler (Siu);
  • {H^2(X,\mathbb{Z})} has rank {22}, the Hodge intersection form has signature {(3,19)} on {H^2(X,\mathbb{Z})}, and {H^2(X,\mathbb{Z})} is an even unimodular lattice;
  • the Hodge decomposition of {H^2(X,\mathbb{C})} is {H^2(X,\mathbb{C}) = H^{2,0}\oplus H^{1,1}\oplus H^{0,2}}, where {H^{2,0}=\mathbb{C}\cdot\Omega} has rank {1} and signature {(1,0)}, {H^{1,1}} has rank {20} and signature {(1,19)}, and {H^{0,2}=\mathbb{C}\cdot\overline{\Omega}} has rank {1} and signature {(1,0)};
  • the data in the previous two items determine the K3 surface (by Torelli theorem).

See, e.g., the lecture notes of D. Huybrechts for more details.

3. Special Lagrangian submanifolds

The natural generalization of closed trajectories on flat surfaces are special Lagrangian submanifolds (SLags).

Definition 2 Let {(X, \omega, \Omega)}, {X} is a compact complex {n}-dimensional manifold, {\omega} is a Kähler form, {\Omega} is a holomorphic {n}-form. A real {n}-dimensional submanifold {L\subset X} is a special Lagrangian submanifold (SLag) if

  • {L} is Lagrangian, i.e., {\omega|_{L}=0};
  • {L} is special, i.e., {\Omega|_{L} = d \textrm{Vol}_{\textrm{Riemannian}}}.

Remark 1 Special Lagrangians are minimal submanifolds (in the sense that they minimize the volume in their cohomology class).

The next example justifies the claim that special Lagrangian submanifolds are the analog of closed trajectories on flat surfaces.

Example 3 Consider the case {n=1}, i.e., {X} is a flat surface. In this situation, all real {1}-dimensional submanifolds are Lagrangian. On the other hand, since {\Omega} is locally {dz} (away from its divisors) in this setting, we see that a special Lagrangian is a horizontal geodesic of the flat metric. In particular, if we replace {\Omega} by {e^{i\theta}\Omega}, then the SLags become the straight line trajectories at angle {\theta} on {X}.

In a similar vein, special Lagrangian fibrations are the analog of cylinders of closed horizontal trajectories on flat surfaces.

4. Special Lagrangian fibrations

Definition 3 A fibration {X\rightarrow B} of {X} over a real {n}-dimensional base {B} is SLag if its fibers are compact {n}-dimensional SLags submanifolds of {X}. The volume {V\in\mathbb{R}_+} of such a fibration is {V=\int_L\Omega} where {L} is any fiber of {X\rightarrow B}.

Remark 2 The fibers of SLag fibrations are torii. One can compare this with the Arnold-Liouville theorem saying that the fibers of a fibration {M\rightarrow B} of a symplectic manifold {M} by compact Lagrangian submanifolds are necessarily torii. In particular, the base {B} has an integral affine structure whose structural group is the semi-direct product of {GL(\mathbb{Z}^n)} by {\mathbb{Z}^n}.

Remark 3 Similarly to the case {n=1}, a typical K3 surface doesn’t admit a SLag fibration.

K3 surfaces possess a significant amount of relevant structures. For example, a particular case of Yau’s solution to Calabi’s conjecture says that:

Theorem 4 (Yau) Let {X} be a K3 surface equipped with a Kähler form {\omega}. Then, there exists an unique {\omega'} in the same cohomology class of {\omega} in {H^2(X,\mathbb{R})} such that {\omega'} induces a Ricci-flat metric.

Moreover, K3 surfaces {X} with Ricci-flat metrics {g} are hyperKähler manifolds, i.e., they admit three complex structures {I, J, K} such that {(X,g,I,J,K)} has the following properties:

  • {g} is a Ricci-flat Riemannian metric;
  • {I, J, K} are complex structures satisfying the usual quaternionic relations: {IJ=-JI=K};
  • {I, J, K} are compatible with {g}: the forms {\omega_{\ast}=g(\dot, \ast\dot)} are closed (i.e. {d\omega_{\ast}=0}) for {\ast=I, J, K}.

Equivalently, we can write the data {(X,g,I,J,K)} of a hyperKähler manifold as {(X,\omega,\Omega)} where {\omega=\omega_I} and {\Omega = \omega_J+i\omega_K}. In this way, we obtain a presentation of K3 surfaces bearing some similarities with our definition of flat surfaces.

5. Statement of the main result

In this setting, we change direction of SLag fibrations (in analogy with the case of closed trajectories in flat surfaces) through the notion of twistor families. More concretely, we consider the sphere {\mathbb{S}^2:=\{(x,y,z)\in\mathbb{R}^3: x^2+y^2+z^2=1\}} and we denote by

\displaystyle \mathfrak{X} = X\times \mathbb{S}^2

where the fiber {X\times\{t\}} is equipped with the complex structure {I_t=xI+yJ+zK} for {t=(x,y,z)}.

At this point, we are almost ready to state the main result of this post: for technical reasons, we will give an impressionistic statement before explaining the true theorem in Remark 4 below.

Theorem 5 (Filip) Fix a (generic) twistor family {\mathfrak{X}=X\times\mathbb{S}^2}. Then, there exists {c>0} and {\delta>0} such that

\displaystyle \#\{t\in\mathbb{S}^2: X\times\{t\} \textrm{ admits a SLag fibration of volume }\leq V\} =

\displaystyle c\cdot V^{20} + O(V^{20-\delta})

as {V\rightarrow\infty}.

Remark 4 This statement is not quite true in the sense that one should count “equators in the twistor sphere” rather than counting points {t\in\mathbb{S}^2}. Indeed, this is so because if a complex structure {I_t} admits a SLag fibration at some point {x} in the equator {(\mathbb{R}t)^{\perp}\cap \mathbb{S}^2}, then one also has SLag fibrations with the same “angle” {x} as {I_t} varies along the equator {(\mathbb{R}x)^{\perp}\cap \mathbb{S}^2}.

At this point, Filip started running out of time, and, for this reason, he offered the following sketch of proof of his theorem.

The first step is to reduce to counting elliptic fibrations (i.e., holomorphic fibrations {X\rightarrow\mathbb{P}^1} whose fibers are elliptic curves).

The second step is to show that if the twistor family is not too special, then the counting problem reduces to the “linear algebra level” of the rank {22} vector space {H^2(X)} equipped with a form of signature {(3,19)}.

Finally, this last counting problem can be solved through quantitative equidistribution results on {\Gamma\backslash SO(3,19)} for the action of a certain {1}-parameter subgroup {a_s} on the quotient of the stabilizer in {SO(3,19)} of a null vector in {\mathbb{R}^{3,19}} by a lattice {\Gamma_0} (that is, the quotient of the semidirect product of {SO(3,18)} and {\mathbb{R}^{18}} by{\Gamma_0}).

Sometime ago, Alex Eskin and Alex Wright told me about the following question posed to them by Peter Sarnak:

How frequent are thin groups among Kontsevich-Zorich monodromies?

Instead of explaning the meaning of Sarnak’s question in general, we shall restrict ourselves to the case of Kontsevich-Zorich (KZ) monodromies associated to square-tiled surfaces.

More concretely, let {X=(M,\omega)} be a square-tiled surface (also called origamis) of genus {g\geq 1}, i.e., {p:M\rightarrow\mathbb{R}^2/\mathbb{Z}^2} is a finite branched covering which is unramified off {0\in\mathbb{R}^2/\mathbb{Z}^2} and {\omega = p^*(dz)} is the pullback of {dz=dx+i dy} on {\mathbb{R}^2/\mathbb{Z}^2 \simeq \mathbb{C}/\mathbb{Z}\oplus i\mathbb{Z}}. We have a natural representation

\displaystyle \rho_X:\textrm{Aff}(X)\rightarrow Sp(H_1^{(0)}(X,\mathbb{Z}))\simeq Sp(2g-2,\mathbb{Z})

from the group {\textrm{Aff}(X)} of affine homeomorphisms of {X} to the group {Sp(H_1^{(0)}(X,\mathbb{Z}))} of symplectic matrices of the subspace {H_1^{(0)}(X,\mathbb{Z})} of integral homology classes of {X} projecting to zero under {p}. In this setting, the Kontsevich-Zorich monodromy {\Gamma_X} (associated to the {SL(2,\mathbb{R})}-orbit of {X} in the moduli space of translation surfaces) is the image of {\rho_X}, i.e.,

\displaystyle \Gamma_X := \rho_X(\textrm{Aff}(X))

(See e.g. these posts here for more background material on square-tiled surfaces.)

By following Sarnak’s terminology, we will say that {\Gamma_X} is a thin group if {\Gamma_X} is an infinite index subgroup of {Sp(2g-2,\mathbb{Z})} whose Zariski closure is

\displaystyle \overline{\Gamma_X}^{\textrm{Zariski}} = Sp(2g-2,\mathbb{R})

In the particular case of square-tiled surfaces, Sarnak’s question above is related to the following two problems:

  • (a) find examples of square-tiled surfaces {X} whose KZ monodromies {\Gamma_X} are thin;
  • (b) decide whether the “majority” of square-tiled surfaces in a given connected component {\mathcal{C}} of a stratum of the moduli spaces of unit area translation surfaces has thin KZ monodromy (here, “majority” could mean “all but finitely many” or “almost full probability as the number of squares/tiles grows”.)

The goal of this post is to record (below the fold) some discussions with Vincent Delecroix and certain participants of MathOverFlow around item (a).

Remark 1 While we will not give answers to items (a) and/or (b) in this post, we decided to write it down anyway with the hope that it might be of interest to some readers of this blog: in fact, by the end of this post, we will show the following conditional statement: if the group generated by the matrices

\displaystyle A=\left(\begin{array}{cccc} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & -1 & 1 \\ 0 & 0 & -1 & 0 \end{array}\right), \quad B=\left(\begin{array}{cccc} -1 & 0 & 0 & -1 \\ 0 & 0 & -1 & 0 \\ 0 & 1 & -1 & 0 \\ 1 & 0 & 0 & 0 \end{array}\right)

has infinite-index in {Sp(4,\mathbb{Z})}, then a certain square-tiled surface of genus {3} answers item (a) affirmatively.

Remark 2 Some “evidence” supporting a positive answer to item (b) is provided by this recent paper of Fuchs-Rivin where it is shown that two “randomly chosen” elements (in {SL(n,\mathbb{Z})}) “tend” to generate thin groups.

Read More…

Posted by: matheuscmss | November 18, 2015

Harmonic quasi-isometries (after Benoist and Hulin)

Last September 28, Yves Benoist gave a beautiful talk on the occasion of the workshop Geometry and Dynamics on Moduli Spaces (that is, one of the four 2015 Clay Research Workshops) about his joint work with Dominique Hulin on a generalized version of the so-called Schoen-Li-Wang conjecture (on harmonic maps within bounded distance to given quasi-isometries of symmetric spaces of rank one).

Remark 1 Still concerning beautiful talks delivered in this conference, I strongly recommend taking a look at this video of Peter Scholze’s talk on cohomology of algebraic varieties: indeed, I think that he was extremely sucessful in communicating his results to a broad audience of non-experts in Algebraic Geometry (such as myself).

This post is a transcription of my notes for Yves’ lecture and, as usual, all errors/mistakes below are my entire responsibility.

Read More…

Last August 25th, Harald Helfgott gave the talk “Soficity, short cycles and the Higman group” (based on joint work with Kate Juschenko) during the Second Workshop on Combinatorics, Number Theory and Dynamical Systems coorganized by Christian Mauduit, Carlos Gustavo Moreira, Yuri Lima and myself at IMPA (Brazil).

This post is a transcription of my notes for Harald’s talk, and, evidently, all mistakes/errors are my responsibility.

1. Some notations

We denote by {Sym(n)} the group of all permutations of {\{1,\dots, n\}}.

For {g_1, g_2\in Sym(n)}, we define their distance {d_n(g_1, g_2)} by

\displaystyle d_n(g_1, g_2):=\frac{1}{n}\#\{1\leq i\leq n: g_1(i)\neq g_2(i)\}

Definition 1 (Gromov; Weiss) Let {G} be a group. Given {\delta>0} and {S\subset G} finite, we say that {\phi:S\rightarrow Sym(n)} is a {(S,n,\delta)}-sofic representation whenever

  • (a) {\phi} is an “approximate homomorphism”: {d_n(\phi(g)\phi(h),\phi(gh))<\delta} for all {g,h\in S} with {gh\in S};
  • (b) {\phi(g)\in Sym(n)} has “few” fixed points for {g\neq e}: {d_n(\phi(g),id)>1-\delta} for all {g\in S-\{e\}}.

We say that a group {G} is sofic if it has {(S,n,\delta)}-sofic representation for all {S\subset G} finite, all {\delta>0} (and some {n\in\mathbb{N}}).

Basic examples of sofic groups are: finite groups, amenable groups, etc. In general, it is known that several families of groups are sofic, but it is an important open problem to construct (or show the existence of) non-sofic groups.

The goal of this post is to discuss a candidate for non-sofic group and its connections to Number Theory.

1.1. Higman groups

For {m\geq 2}, let

\displaystyle H_m := \langle a_1,\dots, a_m : a_i^{-1} a_{i+1} a_i = a_{i+1}^2 \textrm{ for all } i\in\mathbb{Z}/m\mathbb{Z} \rangle

The groups {H_2} and {H_3} are trivial, and the group {H_4} is the so-called Higman group.

Remark 1 Several statements in this post can be generalized for {H_m} for all {m\geq 4}, but for the sake of exposition we will stick to {H_4}.

Theorem 2 (Helfgott-Juschenko) Assume that {H_4} is sofic.Then, for every {\varepsilon>0}, there exists {n\in\mathbb{N}} and a bijection {f:\mathbb{Z}/n\mathbb{Z}\rightarrow\mathbb{Z}/n\mathbb{Z}} such that

  • (a) {f} is an “almost exponential function”: {f(x+1)=2f(x)} for all {x\in S} where {S\subset\mathbb{Z}/n\mathbb{Z}} is a subset of cardinality {|S|\geq (1-\varepsilon)n}.
  • (b) {f^4(x):=f\circ f\circ f\circ f(x)=x} for all {x\in\mathbb{Z}/n\mathbb{Z}}.

Remark 2 The existence of functions {f} as above is “unlikely” when {\varepsilon>0} is small. More precisely, it is possible to show that there are no bijections {f:\mathbb{Z}/n\mathbb{Z}\rightarrow\mathbb{Z}/n\mathbb{Z}} satisfying item (a) with {\varepsilon=0} and item (b) when {n=p^5} is the fifth power of a prime {p} (cf. Remark 4 below for a more precise statement).

In other words, if we could take {\varepsilon=0} and {n=p^5} in the statement of Theorem 2, the non-soficity of the Higman group {H_4} would follow.

Unfortunately, the techniques of Helfgott and Juschenko do not allow us to take {\varepsilon=0} in Theorem 2, but they permit to control the integer {n}. More concretely, as we are going to see in Theorem 4 below, the integer {n} can be chosen from any fixed sequence {0<n_1<n_2<\dots} which is thick in the following sense:

Definition 3 A sequence {0<n_1<n_2<\dots} of positive integers is thick if for every {\varepsilon>0} there exists {K>0} such that

\displaystyle \frac{n_{k+1}}{n_k}\leq 1+\varepsilon

for all {k>K}.

Remark 3 It does not take much to be a thick sequence: for example, the sequences {\{n^3\}_{n\in\mathbb{N}}} and {\{p^5\}_{p \textrm{ prime}}} are thick.

As we already announced, the main result of Helfgott-Juschenko is the following improvement of Theorem 2:

Theorem 4 (Helfgott-Juschenko) Assume that {H_4} is sofic and let {0<n_1<n_2<\dots} be a thick sequence.Then, for every {\varepsilon>0}, there exists {k} and a bijection {f:\mathbb{Z}/n_k\mathbb{Z}\rightarrow\mathbb{Z}/n_k\mathbb{Z}} such that

  • (a) {f(x+1)=2f(x)} for all {x\in S} where {S\subset\mathbb{Z}/n_k\mathbb{Z}} is a subset of cardinality {|S|\geq (1-\varepsilon)n_k}.
  • (b) {f^4(x):=f\circ f\circ f\circ f(x)=x} for all {x\in\mathbb{Z}/n_k\mathbb{Z}}.

Remark 4 The non-soficity of Higman group {H_4} would follow from this theorem if the bijection {f:\mathbb{Z}/n_k\mathbb{Z}\rightarrow \mathbb{Z}/n_k\mathbb{Z}}, {n_k\in\{p^5\}_{p\textrm{ prime}}}, provided by this statement could be taken so that

  • (a*) {f(x+1)=2f(x)} for all {x\in S} where {S\subset\mathbb{Z}/n_k\mathbb{Z}} is a subset of cardinality {|S|\geq n_k-n_k^{1/6}}.
  • (b) {f^4(x):=f\circ f\circ f\circ f(x)=x} for all {x\in\mathbb{Z}/n_k\mathbb{Z}}.

Indeed, this is so because Glebsky and Holden-Robinson proved that there is no {f} verifying (a*) and (b).

Remark 5 A natural question related to Theorem 4 is: what happens with fewer iterations in item (b)? In this situation, it is possible to use the fact that the group {H_3} is trivial to show that, for each {\varepsilon>0}, there exists {n_{\varepsilon}\in\mathbb{N}} such that for any {n\geq n_{\varepsilon}} there is no bijection {f:\mathbb{Z}/n\mathbb{Z}\rightarrow \mathbb{Z}/n\mathbb{Z}} such that

  • (a) {f(x+1)=2f(x)} for all {x\in S} for {|S|\geq (1-\varepsilon)n}.
  • (b) {f^3(x):=f\circ f\circ f(x)=x} for all {x\in S} for {|S|\geq (1-\varepsilon)n}.

This last remark can be generalized as follows.

Theorem 5 (Helfgott-Juschenko) Let {n, g\geq 2} be coprime. Consider the function {f:\{0,1,\dots, n-1\}\rightarrow\{0,1,\dots, n-1\}} given by

\displaystyle f(x)=g^x (\textrm{mod } n)

Then, the equation

\displaystyle f^3(x):=f\circ f\circ f(x)=x

has at most {o(n)} solutions {x\in\{0,1,\dots, n-1\}}.

Remark 6 This theorem improves on a result of Glebsky-Shparlinski.

Read More…

Posted by: matheuscmss | September 27, 2015

Third Bourbaki seminar of 2015: Sophie Morel

Three months ago, Sophie Morel gave her Bourbaki seminar talk “Construction de représentations galoisiennes [d’après Scholze]”.

As it turns out, Sophie Morel was kind towards the non-experts in this subject (like myself): indeed, a large part of the talk (see the video here) was introductory, while the more advanced material was delegated to the lecture notes (available here).

In the remainder of this post, I’ll try to summarize some of the topics discussed by Sophie Morel’s talk (using the video in the link above as the main source).

Disclaimer: Since I’m not an expert on this subject, all mistakes in this post are my responsibility.

Remark 1 If you are in the Oxford area this week, then you will have the opportunity to learn about the main results in Morel’s talk directly from the authors: indeed, Boxer and Scholze (resp.) will give talks tomorrow, resp. Wednesday on their works (see this schedule here of the corresponding 2015 Clay research conference.)

Read More…

Consider a random product of two symplectic matrices {A_0, A_1 \in Sp(V)} on a real symplectic vector space {V} of dimension {\textrm{dim}(V)=2d}, that is, the (symplectic) linear cocycle

\displaystyle F: \{0,1\}^{\mathbb{Z}}\times V\rightarrow \{0,1\}^{\mathbb{Z}}\times V

given by

\displaystyle F(x,v) = (\sigma(x), A_{x_0}v)

where {x=(x_n)_{n\in\mathbb{Z}}} and {\sigma(x):=(x_{n+1})_{n\in\mathbb{Z}}} is the shift map equipped with the Bernoulli measure {\mathbb{P}=(\frac{1}{2}\delta_0 + \frac{1}{2}\delta_1)^{\mathbb{Z}}}.

By Oseledets multiplicative ergodic theorem, the Lyapunov exponents of the random product of {A_0} and {A_1} (i.e., the linear cocycle {F}) are well-defined quantities {\lambda_1\geq\dots\geq\lambda_{2d}} (depending only on {A_0} and {A_1}) describing the exponential growth of the singular values of the random products

\displaystyle A_{x_{m}}\dots A_{x_0}, \quad m\in\mathbb{N}

for any {\mathbb{P}}-typical choice of {x=(x_n)_{n\in\mathbb{Z}}}.

Moreover, the fact that {A_0} and {A_1} are symplectic matrices implies that the Lyapunov exponents are symmetric with respect to the origin, i.e., {\lambda_{2d-k-1} = -\lambda_k} for each {k=1,\dots,d}. In other words, the Lyapunov exponents of the symplectic linear cocycle {F} have the form:

\displaystyle \lambda_1\geq\dots\geq\lambda_d\geq-\lambda_d\geq\dots\geq-\lambda_1

In fact, this structure of the Lyapunov exponents of a symplectic linear cocycle reflects the fact that if {\theta} is an eigenvalue of a symplectic matrix {B}, then {\theta^{-1}} is also an eigenvalue of {B}.

A natural qualitative question about Lyapunov exponents concerns their simplicity in the sense that there are no repeated numbers in the list above (i.e., {\lambda_j > \lambda_{j+1}} for all {k=1,\dots,d}).

The simplicity property for Lyapunov exponents is the subject of several papers in the literature: see, e.g., the works of Furstenberg, Goldsheid-Margulis, Guivarch-Raugi, and Avila-Viana (among many others).

Very roughly speaking, the basic philosophy behind these papers is that the simplicity property holds whenever the monoid {\mathcal{M}} generated by {A_0} and {A_1} is rich. Of course, there are several ways to formalize the meaning of the word “rich”, for example:

  • Goldsheid-Margulis and Guivarch-Raugi asked {\mathcal{M}} to be Zariski-dense in {Sp(V)};
  • Avila-Viana required {\mathcal{M}} to be
    • pinching: there exists {C\in\mathcal{M}} whose eigenvalues are all real with distinct moduli; such a {C} is called a pinching matrix;
    • twisting: there exists a pinching matrix {C\in\mathcal{M}} and a twisting matrix {D\in\mathcal{M}} with respect to {C} in the sense that {D(F)\cap F'=\{0\}} for all isotropic {C}-invariant subspaces {F} and all coisotropic {C}-invariant subspaces {F'} with {\textrm{dim}(F) + \textrm{dim}(F')=2d}.

Of course, these notions of “richness” of a monoid {\mathcal{M}} are “close” to each other, but they differ in a subtle detail: while the Zariski-density condition on {\mathcal{M}} is an algebraic requirement, the pinching and twisting condition on {\mathcal{M}} makes no reference to the algebraic structure of the linear group {Sp(V)}.

In particular, this leads us to the main point of this post:

How the Zariski-density and pinching and twisting conditions relate to each other?

The first half of this question has a positive answer: a Zariski-dense monoid {\mathcal{M}} is also pinching and twisting. Indeed:

  • (a) a modification of the arguments in this blog post here (in Spanish) permits to prove that any Zariski-dense monoid {\mathcal{M}} contains a pinching matrix {C}, and
  • (b) the twisting condition on a matrix {D} with respect to a pinching matrix {C} can be phrased in terms of the non-vanishing of certain (isotropic) minors of the matrix of {D} written in a basis of eigenvectors of {C}; thus, a Zariski-dense monoid {\mathcal{M}} contains a twisting matrix with respect to any given pinching matrix.

On the other hand, the second half of this question has a negative answer: we exhibit below a pinching and twisting monoid {\mathcal{M}} which is not Zariski dense.

Remark 1 The existence of such examples of monoids is certainly known among experts. Nevertheless, I’m recording it here because it partly “justifies” a forthcoming article joint with Artur Avila and Jean-Christophe Yoccoz in the following sense.

The celebrated paper of Avila-Viana quoted above (on Kontsevich-Zorich conjecture) shows that the so-called “Rauzy monoids” are pinching and twisting (and this is sufficient for their purposes of proving simplicity of Lyapunov exponents of the Kontsevich-Zorich cocycle for Masur-Veech measures).

On the other hand, since a pinching and twisting monoid is not necessarily Zariski dense (as we are going to see below), the results of Avila-Viana (per se) can not answer a question of Zorich (see also Remark 6.12 in Avila-Viana paper) about the Zariski density of Rauzy monoids.

In this direction, Artur, Jean-Christophe and I solve (in an article still in preparation) Zorich’s question about Zariski density of Rauzy monoids in the special case of hyperelliptic Rauzy diagrams, and the main example of this post (which will be included in our forthcoming article with Artur and Jean-Christophe) serves to indicate that the results obtained by Artur, Jean-Christophe and myself can not be deduced as “abstract consequences” of the arguments in Avila-Viana paper.

Remark 2 The main example of this post also shows that (a version of) Prasad-Rapinchuk’s criterion for Zariski density (cf. Theorem 9.10 of Prasad-Rapinchuk paper or Theorem 1.5 in Rivin’s paper) based on Galois-pinching (in the sense of this paper here) and twisting properties is “sharp”: indeed, an important feature of the main example of this post is the failure of the Galois-pinching property (cf. Remark 4 below for more comments).

1. A monoid of 4×4 symplectic matrices

Let {\rho} be the third symmetric power of the standard representation of {SL(2,\mathbb{R})}. In concrete terms, {\rho} is constructed as follows. Consider the basis {\mathcal{B} = \{X^3, X^2Y, XY^2, Y^3\}} of the space {V} of homogenous polynomials of degree {3} on two variables {X} and {Y}. By letting {g=\left(\begin{array}{cc} a & b \\ c & d \end{array}\right)\in SL(2,\mathbb{R})} act on {X} and {Y} as {g(X)=aX+cY} and {g(Y)=bX+dY}, we get a linear map {\rho(g)} on {V} whose matrix in the basis {\mathcal{B}} is

\displaystyle \rho\left(\begin{array}{cc} a & b \\ c & d \end{array}\right) = \left(\begin{array}{cccc} a^3 & a^2 b & a b^2 & b^3 \\ 3 a^2 c & a^2 d + 2 a b c & b^2 c + 2 a b d & 3 b^2 d \\ 3 a c^2 & b c^2 + 2 a c d & a d^2 + 2 b c d & 3 b d^2 \\ c^3 & c^2 d & c d^2 & d^3 \end{array}\right)

Remark 3 The faithful representation {\rho} is the unique irreducible four-dimensional representation of {SL(2,\mathbb{R})}.

The matrices {\rho(g)} preserve the symplectic structure on {V} associated to the matrix

\displaystyle J = \left(\begin{array}{cccc} 0 & 0 & 0 & -1 \\ 0 & 0 & 1/3 & 0 \\ 0 & -1/3 & 0 & 0 \\ 1 & 0 & 0 & 0 \end{array}\right)

Indeed, a direct calculation shows that if {g=\left(\begin{array}{cc} a & b \\ c & d \end{array}\right)}, then

\displaystyle (\rho(g))^T\cdot J\cdot \rho(g) = \left(\begin{array}{cccc} 0 & 0 & 0 & -(a d - b c)^3 \\ 0 & 0 & \frac{(a d - b c)^3}{3} & 0 \\ 0 & - \frac{(a d - b c)^3}{3} & 0 & 0 \\ (a d - b c)^3 & 0 & 0 & 0 \end{array}\right)

where {(\rho(g))^T} stands for the transpose of {\rho(g)}.

Therefore, the image {H=\rho(SL(2,\mathbb{R}))} is a linear algebraic subgroup of the symplectic group {Sp(V)}, and the Zariski closure of the monoid {\mathcal{M}} generated by the matrices

\displaystyle A = \rho\left(\begin{array}{cc} 1 & 1 \\ 0 & 1 \end{array}\right) = \left(\begin{array}{cccc} 1 & 1 & 1 & 1 \\ 0 & 1 & 2 & 3 \\ 0 & 0 & 1 & 3 \\ 0 & 0 & 0 & 1\end{array}\right)


\displaystyle B = \rho\left(\begin{array}{cc} 1 & 0 \\ 1 & 1 \end{array}\right) = \left(\begin{array}{cccc} 1 & 0 & 0 & 0 \\ 3 & 1 & 0 & 0 \\ 3 & 2 & 1 & 0 \\ 1 & 1 & 1 & 1\end{array}\right)

is precisely {\overline{\mathcal{M}}^{Zariski}=H}.

Remark 4 Coming back to Remark 2, observe that {H} does not contain Galois-pinching elements of {Sp(V)} in the sense of this paper here (i.e., pinching elements of {Sp(V)} with integral entries whose characteristic polynomial has the largest possible Galois group for a reciprocal polynomial [namely, the hyperoctahedral group]) because its rank is {1}. Alternatively, a straightforward computation reveals that the characteristic polynomial of {\rho(g)} is

\displaystyle (x^2-\textrm{tr}(g)\det(g)x+\det(g)^3)\cdot (x^2 - \textrm{tr}(g)(\textrm{tr}(g)^2 - 3\det(g))x + \det(g)^3)

and, consequently, the eigenvalues of {\rho(g)} are

\displaystyle \frac{1}{2}\det(g)\left(\textrm{tr}(g)\pm\sqrt{\textrm{tr}(g)^2 - 4 \det(g)}\right),


\displaystyle \frac{1}{2}\left(\textrm{tr}(g)(\textrm{tr}(g)^2 - 3\det(g)) \pm (\textrm{tr}(g)^2 - \det(g)) \sqrt{\textrm{tr}(g)^2 - 4 \det(g)}\right)

In particular, since the characteristic polynomial of {\rho(g)} always splits, it is never the case that {\rho(g)} is Galois-pinching.

On the other hand, the element {A.B\in\mathcal{M}} is pinching because its eigenvalues are

\displaystyle 9+4\sqrt{5} > \frac{3+\sqrt{5}}{2} > \frac{3-\sqrt{5}}{2} > \frac{1}{9+4\sqrt{5}}

Also, the matrix {A\in\mathcal{M}} is twisting with respect to {A.B}. Indeed, the columns of the matrix

\displaystyle M = \left(\begin{array}{cccc} -\frac{1}{4} + \frac{(9 + 4 \sqrt{5})}{4} & 1 - \frac{(3 + \sqrt{5})}{2} & 1 - \frac{(3 - \sqrt{5})}{2} & -\frac{1}{4} + \frac{(9 - 4 \sqrt{5})}{4} \\ \frac{9}{8} + \frac{3(9 + 4 \sqrt{5})}{8} & -2 + \frac{(3 + \sqrt{5})}{2} & -2 + \frac{(3 - \sqrt{5})}{2} & \frac{9}{8} + \frac{3(9 - 4 \sqrt{5})}{8} \\ -\frac{15}{8} + \frac{3(9 + 4 \sqrt{5})}{8} & \frac{(3 + \sqrt{5})}{2} & \frac{(3 - \sqrt{5})}{2} & -\frac{15}{8} + \frac{3(9 - 4 \sqrt{5})}{8} \\ 1 & 1 & 1 & 1 \end{array}\right)

consist of eigenvectors of {A.B}. Thus, {T=M^{-1}\cdot A\cdot M} is the matrix of {A} in the corresponding basis of eigenvectors of {A.B}. Moreover, {A} is twisting with respect to {A.B} if and only if all entries of {T} and all of its {2\times 2} minors associated to isotropic planes are non-zero (cf. Lemma 4.8 in this paper here). Finally, this last fact is a consequence of the following exact calculation (see also the numerical approximations) for {T} and its matrix of {2\times 2} minors:

\displaystyle \begin{array}{rcl} T&=& \left(\begin{array}{cccc} \frac{8(5 + 2 \sqrt{5})}{25} & \frac{2(5 + 3 \sqrt{5})}{25} & \frac{(5 + \sqrt{5})}{25} & \frac{1}{( 5 \sqrt{5})} \\ -\frac{6(5 + 3 \sqrt{5})}{25} & \frac{2(5 + \sqrt{5})}{25} & \frac{7}{5 \sqrt{5}} & -\frac{3(-5 + \sqrt{5})}{25} \\ \frac{3(5 + \sqrt{5})}{25} & -\frac{7}{5 \sqrt{5}} & -\frac{2(-5 + \sqrt{5})}{25} & \frac{6(-5 + 3 \sqrt{5})}{25} \\ -\frac{1}{5 \sqrt{5}} & \frac{(5 - \sqrt{5})}{25} & \frac{2}{5} - \frac{6}{5 \sqrt{5}} & -\frac{(8(-5 + 2 \sqrt{5})}{25} \end{array}\right) \\ &=&\left(\begin{array}{cccc} 3.03108 & 0.936656 & 0.289443 & 0.0894427 \\ -2.80997 & 0.578885 & 0.626099 & 0.331672 \\ 0.868328 & -0.626099 & 0.221115 & 0.409969 \\ -0.0894427 & 0.110557 & -0.136656 & 0.168916 \end{array}\right) \end{array}


\displaystyle \begin{array}{rcl} & & 2\times 2 \textrm{ minors of } T = \\ & & \left(\begin{array}{cccccc} \frac{56}{25} + \frac{24}{5 \sqrt{5}} & \frac{32}{25} + \frac{16}{5 \sqrt{5}} & \frac{18}{25} + \frac{6}{5 \sqrt{5}} & \frac{6}{25} + \frac{2}{5 \sqrt{5}} & \frac{2}{25} + \frac{2}{5 \sqrt{5}} & \frac{1}{25} \\  -\frac{32}{25} - \frac{16}{5 \sqrt{5}} & \frac{6}{25} + \frac{2}{5 \sqrt{5}} & \frac{9}{25} + \frac{9}{5 \sqrt{5}} & \frac{3}{25} + \frac{3}{5 \sqrt{5}} & \frac{11}{25} & -\frac{2}{25} + \frac{2}{5 \sqrt{5}} \\  \frac{6}{25} + \frac{2}{5 \sqrt{5}} & -\frac{3}{25} - \frac{3}{5 \sqrt{5}} & \frac{13}{25} & -\frac{4}{25} & -\frac{3}{25} + \frac{3}{5 \sqrt{5}} & \frac{6}{25} - \frac{2}{5 \sqrt{5}} \\ \frac{18}{25} + \frac{6}{5 \sqrt{5}} & -\frac{9}{25} - \frac{9}{5 \sqrt{5}} & -\frac{36}{25} & \frac{13}{25} & -\frac{9}{25} + \frac{9}{5 \sqrt{5}} & \frac{18}{25} - \frac{6}{5 \sqrt{5}} \\  -\frac{2}{25} - \frac{2}{5 \sqrt{5}} & \frac{11}{25} & \frac{9}{25} - \frac{9}{5 \sqrt{5}} & \frac{3}{25} - \frac{3}{5 \sqrt{5}} & \frac{6}{25} - \frac{2}{5 \sqrt{5}} & -\frac{32}{25} + \frac{16}{5 \sqrt{5}} \\   \frac{1}{25} & \frac{2}{25} - \frac{2}{5 \sqrt{5}} & \frac{18}{25} - \frac{6}{5 \sqrt{5}} & \frac{6}{25} - \frac{2}{5 \sqrt{5}} & \frac{32}{25} - \frac{16}{5 \sqrt{5}} & \frac{56}{25} - \frac{24}{5 \sqrt{5}} \end{array}\right) = \\ & & \left(\begin{array}{cccccc} 4.38663 & 2.71108 & 1.25666 & 0.418885 & 0.258885 & 0.04 \\ -2.71108 & 0.418885 & 1.16498 & 0.388328 & 0.44 & 0.0988854 \\ 0.418885 & -0.388328 & 0.52 & -0.16 & 0.148328 & 0.0611146 \\ 1.25666 & -1.16498 & -1.44 & 0.52 & 0.444984 & 0.183344 \\ -0.258885 & 0.44 & -0.444984 & -0.148328 & 0.0611146 & 0.151084 \\ 0.04 & -0.0988854 & 0.183344 & 0.0611146 & -0.151084 & 0.0933747 \end{array}\right) \end{array}

In summary, the monoid {\mathcal{M}} is pinching and twisting, but not Zariski dense in {Sp(V)}.

About two weeks ago, Weixiao Shen gave the talk “Hausdorff dimension of the graphs of the classical Weierstrass functions” during the Third Palis-Balzan International Symposium on Dynamical Systems.

In the sequel, I will transcript my notes from Shen’s talk.

1. Introduction

In Real Analysis, the classical Weierstrass function is

\displaystyle W_{\lambda,b}(x) = \sum\limits_{n=0}^{\infty} \lambda^n \cos(2\pi b^n x)

with {1/b < \lambda < 1}.

Note that the Weierstrass functions have the form

\displaystyle f^{\phi}_{\lambda,b}(x) = \sum\limits_{n=0}^{\infty} \lambda^n \phi(b^n x)

where {\phi} is a {\mathbb{Z}}-periodic {C^2}-function.

Weierstrass (1872) and Hardy (1916) were interested in {W_{\lambda,b}} because they are concrete examples of continuous but nowhere differentiable functions.

Remark 1 The graph of {f^{\phi}_{\lambda,b}} tends to be a “fractal object” because {f^{\phi}_{\lambda,b}} is self-similar in the sense that

\displaystyle f^{\phi}_{\lambda, b}(x) = \phi(x) + \lambda f^{\phi}_{\lambda,b}(bx)

We will come back to this point later.

Remark 2 {f^{\phi}_{\lambda,b}} is a {C^{\alpha}}-function for all {0\leq \alpha < \frac{-\log\lambda}{\log b}}. In fact, for all {x,y\in[0,1]}, we have

\displaystyle \frac{f^{\phi}_{\lambda, b}(x) - f^{\phi}_{\lambda,b}(y)}{|x-y|^{\alpha}} = \sum\limits_{n=0}^{\infty} \lambda^n b^{n\alpha} \left(\frac{\phi(b^n x) - \phi(b^n y)}{|b^n x - b^n y|^{\alpha}}\right),

so that

\displaystyle \frac{f^{\phi}_{\lambda, b}(x) - f^{\phi}_{\lambda,b}(y)}{|x-y|^{\alpha}} \leq \|\phi\|_{C^{\alpha}} \sum\limits_{n=0}^{\infty}(\lambda b^{\alpha})^n:=C(\phi,\alpha,\lambda,b) < \infty

whenever {\lambda b^{\alpha} < 1}, i.e., {\alpha < -\log\lambda/\log b}.

The study of the graphs of {W_{\lambda,b}} as fractal sets started with the work of Besicovitch-Ursell in 1937.

Remark 3 The Hausdorff dimension of the graph of a {C^{\alpha}}-function {f:[0,1]\rightarrow\mathbb{R}} is

\displaystyle \textrm{dim}(\textrm{graph}(f))\leq 2 - \alpha

Indeed, for each {n\in\mathbb{N}}, the Hölder continuity condition

\displaystyle |f(x)-f(y)|\leq C|x-y|^{\alpha}

leads us to the “natural cover” of {G=\textrm{graph}(f)} by the family {(R_{j,n})_{j=1}^n} of rectangles given by

\displaystyle R_{j,n}:=\left[\frac{j-1}{n}, \frac{j}{n}\right] \times \left[f(j/n)-\frac{C}{n^{\alpha}}, f(j/n)+\frac{C}{n^{\alpha}}\right]

Nevertheless, a direct calculation with the family {(R_{j,n})_{j=1}^n} does not give us an appropriate bound on {\textrm{dim}(G)}. In fact, since {\textrm{diam}(R_{j,n})\leq 4C/n^{\alpha}} for each {j=1,\dots, n}, we have

\displaystyle \sum\limits_{j=1}^n\textrm{diam}(R_{j,n})^d\leq n\left(\frac{4C}{n^{\alpha}}\right)^d = (4C)^{1/\alpha} < \infty

for {d=1/\alpha}. Because {n\in\mathbb{N}} is arbitrary, we deduce that {\textrm{dim}(G)\leq 1/\alpha}. Of course, this bound is certainly suboptimal for {\alpha<1/2} (because we know that {\textrm{dim}(G)\leq 2 < 1/\alpha} anyway).Fortunately, we can refine the covering {(R_{j,n})} by taking into account that each rectangle {R_{j,n}} tends to be more vertical than horizontal (i.e., its height {2C/n^{\alpha}} is usually larger than its width {1/n}). More precisely, we can divide each rectangle {R_{j,n}} into {\lfloor n^{1-\alpha}\rfloor} squares, say

\displaystyle R_{j,n} = \bigcup\limits_{k=1}^{\lfloor n^{1-\alpha}\rfloor}Q_{j,n,k},

such that every square {Q_{j,n,k}} has diameter {\leq 2C/n}. In this way, we obtain a covering {(Q_{j,n,k})} of {G} such that

\displaystyle \sum\limits_{j=1}^n\sum\limits_{k=1}^{\lfloor n^{1-\alpha}\rfloor} \textrm{diam}(Q_{j,n,k})^d \leq n\cdot n^{1-\alpha}\cdot\left(\frac{2}{n}\right)^d\leq (2C)^{2-\alpha}<\infty

for {d=2-\alpha}. Since {n\in\mathbb{N}} is arbitrary, we conclude the desired bound

\displaystyle \textrm{dim}(G)\leq 2-\alpha

A long-standing conjecture about the fractal geometry of {W_{\lambda,b}} is:

Conjecture (Mandelbrot 1977): The Hausdorff dimension of the graph of {W_{\lambda,b}} is

\displaystyle 1<\textrm{dim}(\textrm{graph}(W_{\lambda,b})) = 2 + \frac{\log\lambda}{\log b} < 2

Remark 4 In view of remarks 2 and 3, the whole point of Mandelbrot’s conjecture is to establish the lower bound

\displaystyle \textrm{dim}(\textrm{graph}(W_{\lambda,b})) \geq 2 + \frac{\log\lambda}{\log b}

Remark 5 The analog of Mandelbrot conjecture for the box and packing dimensions is known to be true: see, e.g., these papers here and here).

In a recent paper (see here), Shen proved the following result:

Theorem 1 (Shen) For any {b\geq 2} integer and for all {1/b < \lambda < 1}, the Mandelbrot conjecture is true, i.e.,

\displaystyle \textrm{dim}(\textrm{graph}(W_{\lambda,b})) = 2 + \frac{\log\lambda}{\log b}

Remark 6 The techniques employed by Shen also allow him to show that given {\phi:\mathbb{R}\rightarrow\mathbb{R}} a {\mathbb{Z}}-periodic, non-constant, {C^2} function, and given {b\geq 2} integer, there exists {K=K(\phi,b)>1} such that

\displaystyle \textrm{dim}(\textrm{graph}(f^{\phi}_{\lambda,b})) = 2 + \frac{\log\lambda}{\log b}

for all {1/K < \lambda < 1}.

Remark 7 A previous important result towards Mandelbrot’s conjecture was obtained by Barańsky-Barány-Romanowska (in 2014): they proved that for all {b\geq 2} integer, there exists {1/b < \lambda_b < 1} such that

\displaystyle \textrm{dim}(\textrm{graph}(W_{\lambda,b})) = 2 + \frac{\log\lambda}{\log b}

for all {\lambda_b < \lambda < 1}.

The remainder of this post is dedicated to give some ideas of Shen’s proof of Theorem 1 by discussing the particular case when {1/b<\lambda<2/b} and {b\in\mathbb{N}} is large.

Read More…

« Newer Posts - Older Posts »