Difference between revisions of "Hopf-Tsuji-Sullivan theorem"
(Importing text file) |
Ulf Rehmann (talk | contribs) m (tex encoded by computer) |
||
(One intermediate revision by the same user not shown) | |||
Line 1: | Line 1: | ||
− | + | <!-- | |
+ | h1103101.png | ||
+ | $#A+1 = 62 n = 0 | ||
+ | $#C+1 = 62 : ~/encyclopedia/old_files/data/H110/H.1100310 Hopf\ANDTsuji\ANDSullivan theorem | ||
+ | Automatically converted into TeX, above some diagnostics. | ||
+ | Please remove this comment and the {{TEX|auto}} line below, | ||
+ | if TeX found to be correct. | ||
+ | --> | ||
− | + | {{TEX|auto}} | |
+ | {{TEX|done}} | ||
+ | |||
+ | This theorem establishes the equivalence of several characterizations of "smallness" of a [[Riemannian manifold|Riemannian manifold]] of constant negative curvature, or, more generally, of a discrete group $ G $ | ||
+ | of isometries of the $ ( d + 1 ) $- | ||
+ | dimensional hyperbolic space $ \mathbf H ^ {d + 1 } $( | ||
+ | cf. also [[Discrete group of transformations|Discrete group of transformations]]). | ||
+ | |||
+ | Denote by $ \partial \mathbf H ^ {d + 1 } = S ^ {d} $ | ||
+ | the sphere at infinity (the visibility sphere), of $ \mathbf H ^ {d + 1 } $, | ||
+ | and fix an origin $ o \in \mathbf H ^ {d + 1 } $. | ||
+ | A point $ \gamma \in \partial \mathbf H ^ {d + 1 } $ | ||
+ | is called a radial limit point of the group $ G $ | ||
+ | if there exists a number $ R > 0 $ | ||
+ | such that the $ R $- | ||
+ | neighbourhood of the geodesic ray $ [ o, \gamma ] $ | ||
+ | contains infinitely many points from the orbit $ Go = \{ {go } : {g \in G } \} $. | ||
+ | The set $ \Omega _ {r} \subset S ^ {d} $ | ||
+ | of all radial limit points is called the radial limit set of $ G $. | ||
+ | Alternatively, let the shadow $ {\mathcal S} _ {o} ( x,R ) \subset \partial \mathbf H ^ {d + 1 } $ | ||
+ | of the ball $ B ( x,R ) $ | ||
+ | of radius $ R > 0 $ | ||
+ | centred at a point $ x \in \mathbf H ^ {d + 1 } $ | ||
+ | be the set of end-points of all geodesic rays which are issued from $ o $ | ||
+ | and intersect $ B ( x,R ) $. | ||
+ | Then $ \gamma \in \Omega _ {r} $ | ||
+ | if and only if there is an $ R > 0 $ | ||
+ | such that $ \gamma $ | ||
+ | belongs to an infinite number of shadows $ {\mathcal S} _ {o} ( go,R ) $, | ||
+ | $ g \in G $. | ||
The following conditions are equivalent: | The following conditions are equivalent: | ||
− | 1) The Poincaré series | + | 1) The Poincaré series $ \sum _ {g \in G } e ^ {- d { \mathop{\rm dist} } ( o,go ) } $ |
+ | diverges, where $ { \mathop{\rm dist} } ( \cdot, \cdot ) $ | ||
+ | is the Riemannian distance on $ \mathbf H ^ {d + 1 } $. | ||
− | 2) The quotient | + | 2) The quotient $ M = \mathbf H ^ {d + 1 } /G $ |
+ | has no Green function, i.e., $ M $ | ||
+ | has no non-constant negative subharmonic functions (cf. also [[Subharmonic function|Subharmonic function]]), or, equivalently, the [[Brownian motion|Brownian motion]] on $ M $ | ||
+ | is recurrent. | ||
− | 3) The complement of the radial limit set | + | 3) The complement of the radial limit set $ \partial \mathbf H ^ {d + 1 } \setminus \Omega _ {r} $ |
+ | has [[Lebesgue measure|Lebesgue measure]] zero. | ||
− | 4) The [[Geodesic flow|geodesic flow]] on | + | 4) The [[Geodesic flow|geodesic flow]] on $ M $ |
+ | is ergodic with respect to the Liouville-invariant measure (the one determined by the Riemannian volume). | ||
− | 5) The action of | + | 5) The action of $ G $ |
+ | on the product $ \partial \mathbf H ^ {d + 1 } \times \partial \mathbf H ^ {d + 1 } $ | ||
+ | is ergodic with respect to the [[Lebesgue measure|Lebesgue measure]]. | ||
− | Usually the term "Hopf–Tsuji–Sullivan theorem" is applied to the equivalence of 1), 3) and 4). For | + | Usually the term "Hopf–Tsuji–Sullivan theorem" is applied to the equivalence of 1), 3) and 4). For $ d = 1 $ |
+ | the implication 3) $ \Rightarrow $ | ||
+ | 4) was first proved by E. Hopf [[#References|[a1]]], [[#References|[a2]]], and the implications 4) $ \Rightarrow $ | ||
+ | 1) $ \Rightarrow $ | ||
+ | 3) by M. Tsuji, see [[#References|[a3]]]. Tsuji's proof is essentially $ 2 $- | ||
+ | dimensional, as it uses complex function theory, whereas Hopf's argument easily carries over to the higher-dimensional case. D. Sullivan [[#References|[a4]]] used an entirely different way for proving the chain of implications 4) $ \Rightarrow $ | ||
+ | 3) $ \Rightarrow $ | ||
+ | 1) $ \Rightarrow $ | ||
+ | 4) for an arbitrary dimension $ d $. | ||
− | The equivalence of 1) and 2) follows from the asymptotic equivalence of the Green function on | + | The equivalence of 1) and 2) follows from the asymptotic equivalence of the Green function on $ \mathbf H ^ {d + 1 } $ |
+ | to $ e ^ {- d { \mathop{\rm dist} } ( x,y ) } $, | ||
+ | whereas the equivalence of 3), 4) and 5) is a much more general fact, see [[Hopf alternative|Hopf alternative]]. Sullivan's idea was to deduce the implication 2) $ \Rightarrow $ | ||
+ | 5) from general properties of recurrent Markov operators. On the other hand, the implication 3) $ \Rightarrow $ | ||
+ | 1) is an easy corollary of the estimate $ \nu _ {o} ( {\mathcal S} ( R,go ) ) \sim e ^ {- d { \mathop{\rm dist} } ( o,go ) } $, | ||
+ | where $ \nu _ {o} $ | ||
+ | is the image of the Lebesgue measure on the unit tangent sphere at the point $ o $ | ||
+ | under the exponential mapping (a particular case of the Sullivan shadow lemma). | ||
− | [[Ergodicity|Ergodicity]] of the action of | + | [[Ergodicity|Ergodicity]] of the action of $ G $ |
+ | on $ \partial \mathbf H ^ {d + 1 } $( | ||
+ | i.e., absence of bounded harmonic functions on $ M $) | ||
+ | is weaker than 5). For Riemannian surfaces the implication "no Green function" $ \Rightarrow $" | ||
+ | no non-constant bounded harmonic functions" is known as the Myrberg theorem, see [[#References|[a5]]]. In probabilistic terms, this implication can be reformulated as "ergodicity of the time shift in the bilateral path space" $ \Rightarrow $" | ||
+ | ergodicity of the time shift in the unilateral path space" , or just that recurrence of the [[Brownian motion|Brownian motion]] $ M $ | ||
+ | implies absence of non-constant bounded harmonic functions [[#References|[a6]]]. The latter reformulation allows one to construct examples of discrete groups of isometries of $ \mathbf H ^ {d + 1 } $ | ||
+ | whose action on $ \partial \mathbf H ^ {d + 1 } $ | ||
+ | is ergodic and on $ \partial \mathbf H ^ {d + 1 } \times \partial \mathbf H ^ {d + 1 } $ | ||
+ | is not, in a much simpler way than original Riemann surface examples, [[#References|[a7]]]. | ||
− | An analogue of the Hopf–Tsuji–Sullivan theorem for the invariant measure of the geodesic flow corresponding to the Patterson–Sullivan measure on | + | An analogue of the Hopf–Tsuji–Sullivan theorem for the invariant measure of the geodesic flow corresponding to the Patterson–Sullivan measure on $ \partial \mathbf H ^ {d + 1 } $ |
+ | was proved in [[#References|[a8]]], see also [[#References|[a9]]]. In this setup, condition 1) is replaced by divergence of the Poincaré series at the critical exponent of the group $ G $. | ||
In the non-constant curvature case, generalizations of the Hopf–Tsuji–Sullivan theorem were obtained in [[#References|[a6]]] for the harmonic invariant measure of the geodesic flow and in [[#References|[a10]]] for the Patterson–Sullivan measure. | In the non-constant curvature case, generalizations of the Hopf–Tsuji–Sullivan theorem were obtained in [[#References|[a6]]] for the harmonic invariant measure of the geodesic flow and in [[#References|[a10]]] for the Patterson–Sullivan measure. |
Latest revision as of 22:11, 5 June 2020
This theorem establishes the equivalence of several characterizations of "smallness" of a Riemannian manifold of constant negative curvature, or, more generally, of a discrete group $ G $
of isometries of the $ ( d + 1 ) $-
dimensional hyperbolic space $ \mathbf H ^ {d + 1 } $(
cf. also Discrete group of transformations).
Denote by $ \partial \mathbf H ^ {d + 1 } = S ^ {d} $ the sphere at infinity (the visibility sphere), of $ \mathbf H ^ {d + 1 } $, and fix an origin $ o \in \mathbf H ^ {d + 1 } $. A point $ \gamma \in \partial \mathbf H ^ {d + 1 } $ is called a radial limit point of the group $ G $ if there exists a number $ R > 0 $ such that the $ R $- neighbourhood of the geodesic ray $ [ o, \gamma ] $ contains infinitely many points from the orbit $ Go = \{ {go } : {g \in G } \} $. The set $ \Omega _ {r} \subset S ^ {d} $ of all radial limit points is called the radial limit set of $ G $. Alternatively, let the shadow $ {\mathcal S} _ {o} ( x,R ) \subset \partial \mathbf H ^ {d + 1 } $ of the ball $ B ( x,R ) $ of radius $ R > 0 $ centred at a point $ x \in \mathbf H ^ {d + 1 } $ be the set of end-points of all geodesic rays which are issued from $ o $ and intersect $ B ( x,R ) $. Then $ \gamma \in \Omega _ {r} $ if and only if there is an $ R > 0 $ such that $ \gamma $ belongs to an infinite number of shadows $ {\mathcal S} _ {o} ( go,R ) $, $ g \in G $.
The following conditions are equivalent:
1) The Poincaré series $ \sum _ {g \in G } e ^ {- d { \mathop{\rm dist} } ( o,go ) } $ diverges, where $ { \mathop{\rm dist} } ( \cdot, \cdot ) $ is the Riemannian distance on $ \mathbf H ^ {d + 1 } $.
2) The quotient $ M = \mathbf H ^ {d + 1 } /G $ has no Green function, i.e., $ M $ has no non-constant negative subharmonic functions (cf. also Subharmonic function), or, equivalently, the Brownian motion on $ M $ is recurrent.
3) The complement of the radial limit set $ \partial \mathbf H ^ {d + 1 } \setminus \Omega _ {r} $ has Lebesgue measure zero.
4) The geodesic flow on $ M $ is ergodic with respect to the Liouville-invariant measure (the one determined by the Riemannian volume).
5) The action of $ G $ on the product $ \partial \mathbf H ^ {d + 1 } \times \partial \mathbf H ^ {d + 1 } $ is ergodic with respect to the Lebesgue measure.
Usually the term "Hopf–Tsuji–Sullivan theorem" is applied to the equivalence of 1), 3) and 4). For $ d = 1 $ the implication 3) $ \Rightarrow $ 4) was first proved by E. Hopf [a1], [a2], and the implications 4) $ \Rightarrow $ 1) $ \Rightarrow $ 3) by M. Tsuji, see [a3]. Tsuji's proof is essentially $ 2 $- dimensional, as it uses complex function theory, whereas Hopf's argument easily carries over to the higher-dimensional case. D. Sullivan [a4] used an entirely different way for proving the chain of implications 4) $ \Rightarrow $ 3) $ \Rightarrow $ 1) $ \Rightarrow $ 4) for an arbitrary dimension $ d $.
The equivalence of 1) and 2) follows from the asymptotic equivalence of the Green function on $ \mathbf H ^ {d + 1 } $ to $ e ^ {- d { \mathop{\rm dist} } ( x,y ) } $, whereas the equivalence of 3), 4) and 5) is a much more general fact, see Hopf alternative. Sullivan's idea was to deduce the implication 2) $ \Rightarrow $ 5) from general properties of recurrent Markov operators. On the other hand, the implication 3) $ \Rightarrow $ 1) is an easy corollary of the estimate $ \nu _ {o} ( {\mathcal S} ( R,go ) ) \sim e ^ {- d { \mathop{\rm dist} } ( o,go ) } $, where $ \nu _ {o} $ is the image of the Lebesgue measure on the unit tangent sphere at the point $ o $ under the exponential mapping (a particular case of the Sullivan shadow lemma).
Ergodicity of the action of $ G $ on $ \partial \mathbf H ^ {d + 1 } $( i.e., absence of bounded harmonic functions on $ M $) is weaker than 5). For Riemannian surfaces the implication "no Green function" $ \Rightarrow $" no non-constant bounded harmonic functions" is known as the Myrberg theorem, see [a5]. In probabilistic terms, this implication can be reformulated as "ergodicity of the time shift in the bilateral path space" $ \Rightarrow $" ergodicity of the time shift in the unilateral path space" , or just that recurrence of the Brownian motion $ M $ implies absence of non-constant bounded harmonic functions [a6]. The latter reformulation allows one to construct examples of discrete groups of isometries of $ \mathbf H ^ {d + 1 } $ whose action on $ \partial \mathbf H ^ {d + 1 } $ is ergodic and on $ \partial \mathbf H ^ {d + 1 } \times \partial \mathbf H ^ {d + 1 } $ is not, in a much simpler way than original Riemann surface examples, [a7].
An analogue of the Hopf–Tsuji–Sullivan theorem for the invariant measure of the geodesic flow corresponding to the Patterson–Sullivan measure on $ \partial \mathbf H ^ {d + 1 } $ was proved in [a8], see also [a9]. In this setup, condition 1) is replaced by divergence of the Poincaré series at the critical exponent of the group $ G $.
In the non-constant curvature case, generalizations of the Hopf–Tsuji–Sullivan theorem were obtained in [a6] for the harmonic invariant measure of the geodesic flow and in [a10] for the Patterson–Sullivan measure.
References
[a1] | E. Hopf, "Statistik der geodätischen Linien in Mannigfaltigkeiten negativer Krummung" Ber. Verh. Sachs. Akad. Wiss. Leipzig , 91 (1939) pp. 261–304 |
[a2] | E. Hopf, "Ergodic theory and the geodesic flow on surfaces of constant negative curvature" Bull. Amer. Math. Soc. , 77 (1971) pp. 863–877 |
[a3] | M. Tsuji, "Potential theory in modern function theory" , Maruzen (1959) |
[a4] | D. Sullivan, "On the ergodic theory at infinity of an arbitrary discrete group of hyperbolic motions" Ann. Math. Studies , 97 (1980) pp. 465–496 |
[a5] | L.V. Ahlfors, L. Sario, "Riemann surfaces" , Princeton Univ. Press (1960) |
[a6] | V.A. Kaimanovich, "Ergodicity of harmonic invariant measures for the geodesic flow on hyperbolic spaces" J. Reine Angew. Math. , 455 (1994) pp. 57–103 |
[a7] | T. Lyons, D. Sullivan, "Function theory, random paths and covering spaces" J. Diff. Geom. , 19 (1984) pp. 299–323 |
[a8] | D. Sullivan, "The density at infinity of a discrete group of hyperbolic motions" IHES Publ. Math. , 50 (1979) pp. 171–202 |
[a9] | P.J. Nicholls, "Ergodic theory of discrete groups" , Cambridge Univ. Press (1989) |
[a10] | C.B. Yue, "The ergodic theory of discrete isometry groups on manifolds of variable negative curvature" Trans. Amer. Math. Soc. , 348 (1996) pp. 4965–5005 |
Hopf-Tsuji-Sullivan theorem. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Hopf-Tsuji-Sullivan_theorem&oldid=18178