INSTITUTE OF PHYSICS PUBLISHING

NONLINEARITY

Nonlinearity 14 (2001) 817–827

www.iop.org/Journals/no

PII: S0951-7715(01)18100-4

On the topological entropy of an optical Hamiltonian flow C´esar J Niche Mathematics Department, University of California at Santa Cruz, Santa Cruz, CA 95064, USA E-mail: [email protected]

Received 18 October 2000, in final form 25 April 2001 Recommended by R E Goldstein Abstract In this paper, we prove two formulae for the topological entropy of an F -optical Hamiltonian flow induced by H ∈ C ∞ (M, R), where F is a Lagrangian distribution. In these formulae, we calculate the topological entropy as the exponential growth rate of the average of the determinant of the differential of the flow, restricted to the Lagrangian distribution or to a proper modification of it. Mathematics Subject Classification: 37J05

1. Introduction Bounds or formulae for calculating the topological entropy of a dynamical system are very important tools when studying diffeomorphisms or flows on compact manifolds. Many results exist relating this conjugacy invariant to the growth rate of volumes of submanifolds or to the growth rate of the expansion on the tangent bundle (see, for instance, Cogswell [5], Kozlovski [6], Newhouse [10], Przytycki [12] and Yomdin [13]). For a closed Riemannian manifold (M, g), where the metric g is C ∞ , Ma˜ne´ [7] proved the following results for the topological entropy of the geodesic flow ϕt : SM → SM, in terms of nT (x, y), the number of geodesics of length less than or equal to T between two points x and y, of the expansion ex(dϕT )θ = max |det (dϕT )θ |S | S⊂Tθ SM

and of V (θ ) = Ker(dπ)θ , where π : SM → M is the usual projection. ˜ e [7]). Theorem 1 (Theorem 1.1, Man´   1 1 htop = lim log log nT (x, y) dx dy = lim ex(dϕT )θ dθ. T →∞ T T →∞ T M×M SM 0951-7715/01/040817+11$30.00

© 2001 IOP Publishing Ltd and LMS Publishing Ltd

Printed in the UK

817

818

C J Niche

˜ e [7]). Theorem 2 (Theorem 1.4, Man´   1 1 log log |det (dϕT )θ |V (θ) | dθ = lim vol ϕT (Sx M) dx. htop = lim T →∞ T T →∞ T SM M In spite of the fact that these results are Riemannian in nature, the second equality in theorem 1 and the first in theorem 2 are proved by symplectic techniques, relying on a canonical symplectic structure on T M, which allows us to treat the geodesic flow as a Hamiltonian one. For a clear exposition of these and related results, see Paternain [11]. There are three key facts on the proof of Ma˜ne´ ’s results. They are: the twist property of the Lagrangian vertical subspace V (θ ), which says that if the intersection of V (ϕt (θ )) and (dϕt )θ (V (θ )) is non-trivial, a slight perturbation in t makes it trivial; Przytycki’s inequality [12], which gives a bound for the topological entropy of a C 2 flow in terms of the exponential growth rate of the average of the expansion on T (SM) and a very nice change of variables and some auxiliary lemmas, which lead to an inequality necessary to bound the average of the expansion from above. It is natural then, to try to prove similar results for certain Hamiltonian flows, taking into account the importance of Lagrangian distributions in symplectic geometry. To proceed with such a generalization, the vertical distribution V (θ), Przytycki’s inequality and the technical lemmas, should be properly replaced within the Hamiltonian context. In order to recover a twist property, we need to introduce the concept of an optical Hamiltonian. We closely follow Bialy and Polterovich [1, 2]. Given a symplectic vector space (E 2n , ω), let (E) be the family of its Lagrangian subspaces, which is a manifold diffeomorphic to U (n)/O(n). For λ ∈ , the tangent space Tλ  can be canonically identified with S 2 (λ), the vector space of bilinear symmetric forms on λ, in the following way: given a curve λ(t) ⊂  with λ(0) = λ, it is described by λ(t) = S(t)λ

S(0) = Id

S(t) ∈ Sp(E, ω)

(1)

˙ where Sp(E, ω) is the group of symplectic maps in E. Thus, to the vector λ(0) ∈ Tλ  we can associate the symmetric form ˙ (ζ, η) → ω(ζ, S(0)η)

ζ, η ∈ λ.

For a fixed Lagrangian space α ∈ , we note as  (k) α = α

(2)

(3)

1kn

the stratified manifold, where (k) α is the family of Lagrangian subspaces whose intersection (k) 2 with α is k dimensional. So, if λ ∈ (k) α , we can identify Tλ /Tλ α with S (α ∩λ). Then, for (k) (k) λ ∈ α , we say that a vector in Tλ /Tλ α is α-positive if the bilinear form (2) associated with it is positive definite. A vector in Tλ  is α-positive if its image under the projection to Tλ /Tλ (k) α is α-positive. Generally, to a symplectic fibre bundle π : E → M, we can associate its Lagrange– Grassman fibre bundle  : A → M, where each fibre is Ax = (Tx M). Given a Lagrangian distribution F , that is, a section F : M → A, we see that the subbundle AF of Lagrangian subspaces not transversal to F , has fibres diffeomorphic to α in (3). Then, a vector in Tλ A is F -positive for λ ∈ AF , if its image under the projection pr : Tλ A → Tλ /Tλ (k) α is

On the topological entropy of an optical Hamiltonian flow

819

F -positive. A differentiable curve in A, is F -positive, if its tangent vectors are F -positive where they intersect AF . The following proposition gives us the proper notion of twist property.

Proposition 1 (Proposition 1.6, Bialy and Polterovich [1]). Every F -positive tangent vector to A, is transversal to the stratified manifold AF . We now define optical Hamiltonians. We consider the curve of Lagrangian subspaces (ϕt )∗ (α(ϕ−t (x)) on π : A → M, where the base M is a symplectic manifold endowed with a Lagrangian distribution F , and ϕt is the flow induced by H : M → R. Then H is F -optical if for every x ∈ M, t ∈ R, d (ϕt )∗ (α(ϕ−t (x)))|t=0 dt is an α(x)-positive vector. A Hamiltonian flow is optical, if it is generated by an optical Hamiltonian. We now state the two main results of this paper. Let (M, ω) be a symplectic manifold, endowed with a Lagrangian distribution F of Lagrangian subspaces α(x) ⊂ Tx M. Let H ∈ C ∞ (M, R) be an F -optical Hamiltonian with induced flow ϕt . Let " = H −1 (e) be a compact energy level, for a regular value e. Theorem 3. For N% = H −1 (e − %, e + %), then  1 htop (ϕt ) = lim lim log |det (dϕt )x |α(x) | dx. %→0 t→∞ t N% Theorem 4. Let us assume that there exists a continuous invariant distribution of hyperplanes T (x) transversal to the Hamiltonian vector field on ". Then  1 htop (ϕt ) = lim log |det (dϕt )x |α(x) | dx ˜ t→∞ t " where α(x) ˜ = (α(x) ∩ Tx ") + XH (x). The proof of these results is based upon proposition 2, which is a slightly more general version of the bound for the average of the expansion already mentioned. We state this result now. Given a symplectic linear cocycle, that is, a symplectic vector bundle π : S → X, a C ∞ flow on the compact manifold X and a family of differentiable maps φt∗ (x) : S(x) → S(φt (x)), such that φt ◦π = π ◦φt∗ and that for all x ∈ X, t ∈ R, φt∗ (x) is a linear symplectic isomorphism, then we say the cocycle is F -optical with respect to a Lagrangian distribution F , if it is in the sense of the previous paragraphs. Proposition 2. If φt∗ is F -optical, then there exists a constant C > 0 such that for all t ∈ R   ∗ |det φt (x)|α(x) | dx  C ex φt∗ (x) dx. X

X

These theorems are interesting given the generality of the opticity condition. For example, on (T ∗ M, ω0 ) where ω0 is the canonical symplectic form, a Hamiltonian H ∈ C ∞ (T ∗ M, R), is F -optical with respect to the distribution F = {dq = 0} of spaces tangent to the fibres iff H is convex on each fibre, that is, iff Hpp > 0. As a result of this, the usual Hamiltonians of classical mechanics, are optical for this distribution. Another important case where these results can be applied, is that of Anosov magnetic flows induced by twisted symplectic structures (see, for instance, Burns and Paternain [4]). This paper is organized as follows. In section 2 we prove proposition 2 and then in sections 3 and 4, we prove theorems 3 and 4.

820

C J Niche

2. Proof of proposition 2 Before proving proposition 2, we recall some definitions and results needed. In order to measure angles and volumes on a symplectic manifold (M, ω), we introduce a compatible Riemannian metric. Given a symplectic vector bundle π : E → M, an almost complex structure is J : E → E, such that J 2 = − IdE . It is said to be compatible with the symplectic form ω, if on each fibre Eq ω(v, w) = ω(J v, J w)

∀v, w ∈ Eq

ω(v, J v) > 0

∀v ∈ Eq .

(4)

Thus, we can define a Riemannian metric on E by gJ (v, w) = ω(v, J w). For a proof of the existence and properties of such an almost complex structure, see McDuff and Salamon [8]. Given a linear map L : E → F , where E and F are finite-dimensional Hilbert spaces, we define the determinant of L as follows. For an orthonormal base of E, {v1 , v2 , . . . , vn }, we consider the matrix aij = L(vi ), L(vj ). Then, we define √ |det L| = det A. We also define the expansion of L as ex L = max |det L|S | S⊂E

where S is a subspace in E. Finally, for E1 , E2 subspaces of E, with dim Ei = them as

1 2

dim E, we define the angle between

ang(E1 , E2 ) = |det (P |E1 )| where P : E → E2⊥ is the canonical projection. From the definition it is clear that the angle is a continuous function and that ang(E1 , E2 ) = 0 iff E1 ∩ E2 = ∅. Let π : S → X be a symplectic vector bundle, where X is a compact manifold and φt : X → X a flow of class C ∞ , which preserves a measure dx. A linear symplectic cocycle is a family of differentiable maps φt∗ (x) : S(x) → S(φt (x)), such that ∀x ∈ X, t ∈ R: ∗ (a) φt+s = φt∗ (φs (x)) φs∗ (x); (b) φt ◦ π = π ◦ φt∗ ; (c) ∀x ∈ X, t ∈ R, φt∗ (x) : S(x) → S(φt (x)) is a linear symplectic isomorphism.

For a continuous Lagrangian distribution F , we say that φt∗ (x) is F -optical if it is in the sense of section 1. To prove proposition 2, we need several auxiliary lemmas. We note the Lagrange– Grassman bundle associated to π : S → " as (S). We closely follow the proof of proposition 4.18 in Paternain [11]. Lemma 1. There exists δ > 0, an integer m  1 and an upper semicontinuous function τ : (S) × R → {0, 1/m, 2/m, . . . , 1} such that for all (x, E) ∈ (S), t ∈ R, ∗ ang(φt+τ (x)(E), α(φt+τ (x))) > δ.

On the topological entropy of an optical Hamiltonian flow

821

Proof. If we find δ > 0 and an integer m  1 such that if for all (x, E, t) ∈ (S) × R, ∗ (x)(E), α(φt+i/m (x)) > δ} Q(x, E, t) = {i ∈ Z, 0  i  m : ang(φt+i/m

is not empty, then, clearly τ (x, E, t) = min{i/m, i ∈ Q(x, E, t)} has the stated properties. Let us assume that this is not true. Then, there is a sequence (xm , Em , tm ) such that 1 (5) 2m where s ∈ Am , for Am = {j/2m , j ∈ Z, 0  j  2m }. Without loss of generality, we can assume each tm = 0. Then, as a result of the compactness of (S), there is a subsequence converging to (x, E). Then, equation (5) and the continuity of the angle and the flow imply that for all s ∈ [0, 1], ∗ (xm )(E), α(φs+tm (x)))  ang(φs+t m

φs∗ (x)(E) ∩ α(φs (x)) = {0} which contradicts proposition 1.



Lemma 2. There exists γ > 0, an integer n  1 and an upper semicontinuous function ρ : (S) → {0, 1/n, 2/n, . . . , 1} such that for all (x, E) ∈ (S), ang(E, φρ∗ (x− ) (α(x− ))) > γ where x− = φ−ρ (x). Proof. The proof is similar to that of lemma 1.



Lemma 3. For every x ∈ X, t ∈ R, there exists a Lagrangian subspace Rt (x) ⊂ S(x), which depends measurably on t and x, such that: (a) |det φt∗ (x)|Rt (x) | = ex φt∗ (x); (b) if E ⊂ S(x), with dim E = 21 dim S(x), then |det φt∗ (x)|E |  ang(E, Rt⊥ (x)) ex φt∗ (x). Proof. We consider the polar decomposition φt∗ (x) = Ot (x) Lt (x) where Lt (x) : S(x) → S(x) is symmetric and positive definite and Ot (x) : S(x) → S(φt (x)) is a linear isometry. As φt∗ (x) is a symplectic map, so is Lt (x) = (φt∗ (x) φt∗ (x)t )1/2 . As a result of this, the eigenvalues of Lt (x) are real, and it is true that λi+n = λ−1 i , i = 1, . . . , n. Let us assume that λi  1, for 1  i  n. Given an almost complex structure J (x), if vi is an eigenvector with eigenvalue λi , then J (x)vi is an eigenvector with eigenvalue λ−1 i , as Lt ◦ J = J ◦ L−1 . Then t {v1 , . . . , vn , J (x)v1 , . . . , J (x)vn }

822

C J Niche

is a base of S(x) and the subspace Rt (x) generated by {v1 , . . . , vn } is a Lagrangian one, on which the expansion takes place. This proves (a). To prove (b), let us note that Rt (x) and Rt⊥ (x) are invariant under Lt (x). If E∩Rt⊥ (x) = {0} (in contrast there is nothing to prove), then as Lt (x) ◦ P = P ◦ Lt (x), with P the orthogonal projection P : S(x) → Rt (x), we have that |det Lt (x)|Rt (x) ||det P |E | = |det P |Lt (x)(E) ||det Lt (x)|E |  |det Lt (x)|E | which implies ex φt∗ (x) ang(E, Rt⊥ (x))  |det φt∗ (x)|E |. We now prove measurability of Rt (x) on t and x. Let π : F → X be the fibre bundle with fibres π −1 = {h : S(x) → S(x)}, for h a linear symmetric map. For positive integers p and li , 1  i  p, we define F(l1 , . . . , lp ) as those (x, h) with p eigenvalues λi , their multiplicities being li . This is a Borelian set, and so is P(l1 , . . . , lp ) ⊂ X × R, defined as P(l1 , . . . , lp ) = {(x, t), Lt (x) ∈ F(l1 , . . . , lp )}, which constitute a Borel measurable partition  of X × R. As Rt (x) is continuous on each P, the result follows. Lemma 4. There exists δ > 0, an integer m  1 and measurable functions τi : X → {0, 1/m, 2/m, . . . , 1} such that if τ = τ1 + τ2 and x1 = φ−τ1 (x), x2 = φτ2 (x) and α i = α(xi ), then for all x and t, (a) ang(φτ∗1 (x1 )(α 1 ), Rt⊥ (x)) > δ; ∗ (b) ang(φt+τ (x1 )(α 1 ), α 2 ) > δ. Proof. If we find that, for integers δ1 , δ2 > 0, integers m1 , m2 and measurable functions τi : X × R → {0, 1/mi , 2/mi , . . . , 1} such that (a) holds for δ = δ1 and (b) for δ = δ2 , then the lemma is true for m = m1 m2 and δ = min{δ1 , δ2 }. Taking δ1 = γ

m1 = n

τ1 (x, t) = ρ(x, Rt⊥ (x))

for γ , n, τ as in lemma 2 and applying it to E = Rt⊥ , we obtain ang(φτ∗1 (x1 )(α 1 ), Rt⊥ ) > δ1 which proves (a). The same idea works for the proof of (b), applying lemma 1 to E = ∗ φt+τ (x1 )(α 1 ).  1 +τ2 Lemma 5. Let τ1 , x1 , α 1 be as in lemma 4. Then, there exists a constant K > 0, such that for all x, t |det φt∗ (x1 )|α1 |  K ex φt∗ (x). Proof. If E = φt∗ (x1 )(α 1 ) we take a > 0 such that |det φs∗ (y)|L |  a for all s ∈ [0, 1], (y, L) ∈ (S). Then, by (a) from lemma 4 implies that ∗ ∗ |det φt∗ (x1 )|α1 | = |det φt−τ (x)|E ||det φτ∗1 (x1 )|α1 |  a δ ex φt−τ (x). 1 1

Taking a  such that ex φs∗ (x)  a  and for K = aδa  , we see that the result is true. We now prove the main result of the section.



On the topological entropy of an optical Hamiltonian flow

823

Proof of proposition 2. For a fixed t, we make the change of variables F : X → X, given by F (x) = x1 = φ−τ1 (x,t) (x). In each set A(i) = {x ∈ X : τ1 (x, t) = i/m}. F is injective and, as a consequence of this, for the measure µ = dx and each Borel set S ⊂ A(i), we have µ(S) = µ(F (S)). If ? is integrable in X, then   ? dx = (? ◦ F ) dx. F (A(i))

A(i)

For ?  0, then   (? ◦ F ) dx =  X

=

A(i)

(? ◦ F ) dx =

i

 i



F (A(i))

? dx 

 i

A(i)

X

(? ◦ F ) dx

? dx = (m + 1)

 X

? dx.

Taking ? = |det φt∗ (x)|α(x) | and using lemma 5   1 ∗ |det φt (x)|α(x) | dx  |det φt∗ (x1 )|α1 | dx m + 1 X X  K  ex φt∗ (x) dx m+1 X which proves the lemma for C =

K . m+1



3. Proof of theorem 3 We state a result by Kozlovski [6] and a proposition that are essential in the proofs of theorems 3 and 4. Theorem 5 (Kozlovski [6]). For a C ∞ diffeomorphism f : X → X, where X is a compact manifold  1 htop (f ) = lim log ex(df n )x dx. n→∞ n X Proposition 3. Let ϕt : Y → Y be a continuous flow on a compact manifold Y. Then, for Y1 , Y2 closed invariant subsets in Y htop (ϕ, Y1 ∪ Y2 ) = max htop (ϕ, Yi ). Yi ,i=1,2

Proof of theorem 3. To the double D(N% ) of N% = H −1 [e − %, e + %], we apply theorem 5 and we obtain  1 htop (ϕt |D(N% ) ) = lim log ex(dϕt )x dx. (6) t→∞ t D(N% ) As

 D(N% )

 ex(dϕt )x dx = 2

N%

ex(dϕt )x dx

(7)

824

C J Niche

taking the exponential growth rate and through (6), (7) and proposition 3 we obtain  1 htop (ϕt |N% ) = lim log ex(dϕt )x dx t→∞ t N%

(8)

where N% = N% /∂N% . Clearly, (dϕt )x is an F -optical linear symplectic cocycle, for the given distribution F , with respect to the flow ϕt , for the fibre bundle π : T M|N% → N% . Proposition 2 and the trivial inequality ex(dϕt )x  |det (dϕt )x |α(x) | turn (8) into 1 log t→∞ t



htop (ϕt |N% ) = lim

N%

|det (dϕt )x |α(x) | dx.

(9)

Let us define the function h(%) = htop (ϕt |N% ). As h is increasing and h(0) = htop (ϕt |" ), we see that lim inf h(%)  htop (ϕt |" ).

(10)

%→0

We use now the following result, due to Bowen [3]. Proposition 4 (Corollary 18, Bowen [3]). Let X and Y be compact metric spaces and ϕt : X → X a flow. If π : X → Y is a continuous map such that π ◦ ϕt = π , then htop (ϕ) = sup htop (ϕ|π −1 (y) ). y∈Y

For ϕt |N% and H |N% , then h(%) =

sup

%0 ∈[e−%,e+%]

htop (ϕt |H −1 (%0 ) ).

Given r > 0, there is an %0 (%, r), such that h(%)  htop (ϕt |H −1 (%0 ) ) + r

%0 ∈ [e − %, e + %].

Taking the upper limit when % → 0 and using the fact that topological entropy is upper semicontinuous for C ∞ flows (Newhouse [9]), then lim sup h(t)  htop (ϕt |" ).

(11)

%→0

As a result of (9)–(11) 1 htop (ϕt |" ) = lim htop (ϕt |N% ) = lim lim log %→0 %→0 t→∞ t which proves the theorem.

 N%

|det (dϕt )x |α(x) | dx 

On the topological entropy of an optical Hamiltonian flow

825

4. Proof of theorem 4 To prove theorem 4, we make the distribution F descend and the opticity of the Hamiltonian H to a symplectic vector bundle π : S → ", where the fibres are S(x) = Tx "/XH (x). Proposition 5. The fibre bundle given by π : S → ", where S(x) = Tx "/XH (x), is symplectic for the form ωS = p∗ ω, where ω is the symplectic form on M and px : Tx " → S(x) is the canonical projection. If α(x) ˜ = α(x) ∩ Tx " + XH (x), then α(x) ˜ ⊂ Tx " is a Lagrangian subspace of (Tx M, ω) and α˜ S (x) = px (α(x)) ˜ is a Lagrangian subspace of (S(x), ωS ). Proof. For v ∈ Tx ", we note px (v) = [v]. Then, for v1 , v2 ∈ Tx ", we define ωS ([v1 ], [v2 ]) = ω" (v1 , v2 ). As ω" degenerates on XH (x) it is clear that ωS is a symplectic form on S(x). The subspace ˜ then v  = v + a XH (x), with v ∈ α(x) and α(x) ˜ ⊂ Tx " annihilates ω, as for v  ∈ α(x), a ∈ R. In the case where α(x) ⊂ Tx " and XH (x) ∈ α(x), then dim α(x) ˜ = n + 1. However, ⊥ this implies that α(x) ˜ is an (n + 1)-dimensional Lagrangian subspace, as α(x) ˜ = (α(x)) ˜ . So α(x) ˜ has dimension n and this proves that it is a Lagrangian subspace.  Proposition 6. Let F˜S be the Lagrangian distribution induced in π : S → " by proposition 5. t (x) is an F˜S -optical linear symplectic If H ∈ C ∞ (M, R) is an F -optical Hamiltonian, then dϕ cocycle. Proof. By the definition of the optical property, it follows that for a curve λ(t) = S(t)α(ϕt (x))

S(0) = Id, S(t) ∈ Sp(Tϕt (x) M, ω)

˙ ∈ Tα(x) A given by in the fibre bundle  : A → M and to the vector λ(0) d ˙ (ϕt )∗ (α(ϕ−t (x))|t=0 λ(0) = dt we can associate a bilinear symmetric positive-definite form ˙ (ζ, η) → ω(ζ, S(0)η).

(12)

Each α(x) ˜ contains XH (x) and is included in Tx ", which implies that the form (12) is positive semidefinite, as ω|" degenerates on the field. Descending to the quotient S(x), ωS from proposition 5 becomes symplectic and then ˙ ([ζ ], [η]) → ω([ζ ], [S(0)η])

[ζ ], [η] ∈ α˜ S (x)

is the symmetric bilinear form which proves the proposition.



We denote by exTx " the expansion on Tx " and by exS(x) the expansion on S(x). The following lemmas relate expansions and determinants of the differential of the flow on Tx " and S(x). We only prove the second one, as the proof of the first is easy. Lemma 6. Let π : E → X be a fibre bundle on a compact manifold X. If g1 and g2 are two continuous Riemannian metrics on E, then there is a constant K > 0, such that for the linear map L : E(x) → E(y) exg1 L  K exg2 L.

826

C J Niche

Lemma 7. For a symplectic manifold (M, ω), endowed with a Lagrangian distribution F and an F -optical Hamiltonian flow ϕt that admits a continuous invariant distribution of hyperplanes XH (x) on Tx ", then there exist constants K1 , K2 > 0 such that t (x) exTx " dϕt (x)  K1 exS(x) dϕ t (x)|α˜ (x) |  K2 |det dϕt (x)|α(x) |det dϕ |. ˜ S

Proof. The symplectic form ω is non-degenerate on T (x) due to the transversality with respect to XH (x). Then, there exists an almost complex structure J that induces an inner product · , ·T in T . We extend it to Tx " through  if v, w ∈ T

v, wT    if v ∈ XH (x), w ∈ T (x)

v, wTx " = 0    1 if v = w = XH (x). Invariance under the flow allows us to choose a base {XH (x), t1 , . . . , t2n−2 } for Tx ", where {ti }1i2n−2 is a base of T (x), such that

1 0 (dϕt )x |" = . 0 (dϕt )x |T (x) Then exTx " (dϕt )x = exT (x) (dϕt )x as the expansion is the maximum of the determinant on all minors of any dimension on this matrix. We introduce an inner product on S(x), such that the projection P |T (x) is an isometry. It is clear that if g1 , g2 are the metrics given by the inner products on T (x) and S(x), then g g t )x . ex 1 (dϕt )x = ex 2 (dϕ T (x)

S(x)

Lemma 6 leads us to the first inequality we wanted to prove. The proof of the second one, relies on a similar method.  t )x is an F˜S -optical linear symplectic Proof of theorem 4. Proposition 6 implies that (dϕ cocycle, for the distribution given by proposition 5, with respect to the flow ϕt |" , on the symplectic vector bundle π : S → ". Proposition 2 and the trivial inequality t )x |α˜ (x) | t )x  |det (dϕ ex(dϕ S lead us to   1 t )x dx = lim inf 1 log |det (dϕ t )x |α˜ (x) | dx. (13) lim inf log exS(x) (dϕ S t→∞ t t→∞ t " " Then, as a result of Kozlovski’s theorem 5, the first inequality from lemma 7, equation (13) and the second inequality from lemma 7,  1 htop (ϕt |" ) = lim inf log exTx " (dϕt )x dx t→∞ t "  1 t )x dx  lim inf log exS(x) (dϕ t→∞ t "  1 t )x |α˜ (x) | dx = lim inf log |det (dϕ S t→∞ t "  1  lim inf log |det (dϕt )x |α(x) | dx. (14) ˜ t→∞ t "

On the topological entropy of an optical Hamiltonian flow

Similar methods imply that 1 log t→∞ t

htop (ϕt |" ) = lim

 "

1  lim sup log t→∞ t which through (14) and (15) say that 1 log t→∞ t

htop (ϕt |" ) = lim which proves the theorem.

827

exTx " (dϕt )x dx  "

|det (dϕt )x |α(x) | dx ˜

(15)

 "

|det (dϕt )x |α(x) | dx ˜ 

Acknowledgments We are deeply grateful to Gabriel Paternain for his advice. The author acknowledges partial financial support from PEDECIBA, Area Matem´atica, Uruguay and from Instituto de Matem´atica y Estad´ıstica Rafael Laguardia (IMERL), Facultad de Ingenier´ıa, Universidad de la Rep´ublica, Montevideo, Uruguay. An important part of this paper was written during a stay at Centro de Investigaci´on en Matem´atica (CIMAT), Guanajuato, Mexico. This stay was supported by Comisi´on Sectorial de Investigaci´on Cient´ıfica (CSIC) from Universidad de la Rep´ublica, Montevideo, Uruguay and CIMAT. To Renato Iturriaga (CIMAT) and to these institutions, we are deeply grateful. We also thank the referees for their comments. References [1] Bialy M and Polterovich L 1992 Hamiltonian diffeomorphisms and Lagrangian distributions Geom. Funct. Anal. 2 173–210 [2] Bialy M and Polterovich L 1994 Optical Hamiltonian functions Geometry in Partial Differential Equations ed A Prastaso (Singapore: World Scientific) pp 32–50 [3] Bowen R 1971 Entropy for group endomorphisms and homogeneous spaces Trans. Am. Math. Soc. 153 401–14 [4] Burns K and Paternain G P Anosov magnetic flows, critical values and topological entropy Preprint [5] Cogswell K 2000 Entropy and volume growth Ergod. Theor. Dynam. Syst. 20 77–84 [6] Kozlovski O S 1998 A formula for the topological entropy of C ∞ maps Ergod. Theor. Dynam. Syst. 18 405–24 [7] Ma˜ne´ R 1997 On the topological entropy of geodesic flows J. Diff. Geom. 45 74–93 [8] McDuff D and Salamon D 1995 Introduction to Symplectic Topology (Oxford Mathematical Monographs) (Oxford: Oxford University Press) [9] Newhouse S 1989 Continuity properties of entropy Ann. Math. 129 215–35 [10] Newhouse S 1988 Entropy and volume Ergod. Theor. Dynam. Syst. 8* 283–99 [11] Paternain G P 1999 Geodesic Flows (Progress in Mathematics vol 180) (Boston, MA: Birkh¨auser) [12] Przytycki F 1980 An upper estimation for topological entropy of diffeomorphisms Invent. Math. 59 205–13 [13] Yomdin Y 1987 Volume growth and entropy Israel J. Math. 57 285–300

On the topological entropy of an optical Hamiltonian flow

important tools when studying diffeomorphisms or flows on compact manifolds. Many results exist relating this conjugacy invariant to the growth rate of volumes of .... F-optical with respect to the distribution F = {dq = 0} of spaces tangent to the fibres iff H is convex on each fibre, that is, iff Hpp > 0. As a result of this, the usual ...

112KB Sizes 1 Downloads 156 Views

Recommend Documents

TOPOLOGICAL ENTROPY OF A MAGNETIC FLOW ...
to the growth rate of the average number of trajectories connecting two points. 1. Introduction. Topological entropy measures how complicated the global orbit.

Topological entanglement entropy of fracton stabilizer ...
Mar 1, 2018 - Entanglement entropy provides a powerful characterization of two-dimensional gapped topological phases of quantum matter, intimately tied to their description by topological quantum field theories (TQFTs). Fracton topological orders are

Optical Flow Approaches
Feb 17, 2008 - Overall, both MEEG native or imaging data may be considered as .... 2006]). We wish to define velocity vectors as the dynamical signature of ...

Optical Flow Approaches
Feb 17, 2008 - properties of brain activity as revealed by these techniques. ..... As a matter of illustration and proof of concept, we will focus on the possible ..... advanced model for global neural assemblies may be considered in future ...

Dynamically consistent optical flow estimation - Irisa
icate situations (such as the absence of data) which are not well managed with usual ... variational data assimilation [17] . ..... pean Community through the IST FET Open FLUID Project .... on Art. Int., pages 674–679, Vancouver, Canada, 1981.

Linear-optical access to topological insulator surface ...
May 25, 2017 - View online: http://dx.doi.org/10.1063/1.4984141. View Table of ... cal applications, due to their remarkable properties, with an insulating bulk ...

Performance of Optical Flow Techniques 1 Introduction
techniques require that relative errors in the optical ow be less than 10% 10, 36]. Verri and Poggio 58] have suggested that accurate estimates of the 2-d motion eld are gen- erally inaccessible due to inherent di erences between the 2-d motion eld a

Optical Flow Measurement of Human Walking
allows an automated algorithm to predict values of OF data as a function of hip motion. By tuning model initial conditions (hip velocity at heel strike) and a control ...

Optical Flow Measurement of Human Walking
cellphone. OF-based navigation does not ...... Impaired," Instrumentation and Measurement Technology Conference, 2006. IMTC 2006. Proceedings of the IEEE, ...

On the Polarization Entropy - IEEE Xplore
polarimetric SAR image. In this paper, the authors propose a new method to calculate the polarization entropy, based on the least square method. Using a ...

A Comparison of TV-L1 Optical Flow Solvers on GPU
Mar 24, 2014 - The classic TV-L1 optical flow is to minimize one of the following ..... platform for real-time motion computation with discontinuity-preserving ...

Effect of substrate temperature on the optical properties.pdf ...
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. Effect of ...

Exploiting Symmetries in Joint Optical Flow and ...
+1. Forward-backward consistency term. Data term. Occlusion-disocclusion symmetry term. Pairwise term. H. : Per-superpixel homography for forward motion ... #2 in the Clean pass. Ground truth optical flow. Forward optical flow. Backward optical flow.

A trajectory-based computational model for optical flow ...
and Dubois utilized the concepts of data conservation and spatial smoothness in ...... D. Marr, Vision. I. L. Barron, D. J. Fleet, S. S. Beauchemin, and T. A. Burkitt,.

Imaging Brain Activation Streams from Optical Flow ...
is illustrated by simulations and analysis of brain image sequences from a ball-catching paradigm. ..... and its implementation in the. BrainStorm software [2].

Optical Flow Estimation Using Learned Sparse Model
Department of Information Engineering. The Chinese University of Hong Kong [email protected] ... term that assumes image intensities (or other advanced im- age properties) do not change over time, and a ... measures, more advanced ones such as imag

An Improved Degree Based Condition for Hamiltonian ...
Lenin Mehedy1, Md. Kamrul Hasan1 and Mohammad Kaykobad2. 1Department of Computer Engineering, Kyung Hee University, South Korea. 2Department of Computer Science and Engineering, North South University, Dhaka, Bangladesh. Email: 1{lenin, kamrul}@oslab

Optical Flow-based Video Completion in Spherical ...
In addition to the distortion on each single spherical image, the motion pattern is also special in spherical image frames. It has two properties. First, pixels on spherical images can only move along the spherical surfaces. Such movements are projec

On the Secrecy of Spread-Spectrum Flow Watermarks
law enforcement agencies to detect stepping stones used by attackers [20], to determine whether a certain ... sequence, a specific PN code with good autocorrelation features, to spread the bits of a ..... recover the original signal from those freque

Stability, Optical Flow and Stochastic Resonance in ...
Figure 2.2: To transform the convolution in Eq. (2.1) to vector-matrix ...... where P is the signal contribution to the power spectral density (PSD) to the noise power ...

Entropy of random walk range on uniformly transient ... - Project Euclid
Let G be a uniformly transient graph with bounded degrees. Then lim inf ... In addition, let |∂eB(x, r)| be the number of vertices at distance r + 1 from x ∈ V with respect to the usual .... The cardinality of any set A is denoted by |A|, the lar

On Reliable Topological Structures for Message ...
Paper approved by the Editor for Computer Communication of the ... C-0609 and in part by the National Science Foundation under Grant. The author is with the ..... by (noting that in H(n - 1, k - 1) every vertex has degree ... (c) The Graph BA.