Abstract
Adapting tools that we introduced in Jézéquel (J Spectr Theory 10(1):185–249, 2020) to study Anosov flows, we prove that the trace formula conjectured by Dyatlov and Zworski in (Ann. Sci. Éc. Norm. Supér. (4) 49(3):543–577, 2016) holds for Anosov flows in a certain class of regularity (smaller than \({\mathcal {C}}^\infty \) but larger than the class of Gevrey functions). The main ingredient of the proof is the construction of a family of anisotropic Hilbert spaces of generalized distributions on which the generator of the flow has discrete spectrum.
Similar content being viewed by others
Notes
The top-right part of the spectrum had already been unveiled by Liverani for contact Anosov flows in [28].
Notice that when g is real-valued, or when \(\left( \phi ^t\right) _{t \in {\mathbb {R}}}\) has a periodic orbit \(\gamma \) such that no other periodic orbit has the same length, then \(d_g\) is not constant.
The Ruelle resonances for a time 1 suspension of a cat map (with \(g=0\)) are the \(2 i \pi k\)’s for \(k \in {\mathbb {Z}}\), so that(1.6) holds for all \(\rho > 1\).
In fact, the results from [21] and the present paper suggest that the discrete-time analogue of (1.4) should even hold in the class \({\mathcal {C}}^{\kappa ,\upsilon }\) defined in Sect. 2 for \(\kappa > 0\) and \(\upsilon \in ]1,2[\). We think that this could be proven easily using methods from [21] and the present paper. However, in [21, Theorem 2.12, (v)-(vi)] we proved a bound on the growth of the dynamical determinant for Gevrey hyperbolic map that we do not expect to hold for \({\mathcal {C}}^{\kappa ,\upsilon }\) dynamics. This bound is one of the reasons that make us think that the dynamical determinant of a Gevrey Anosov flow has finite order. See also [22] for a detailed discussion of dynamical determinant for expanding maps of the circle in various ultradifferentiable classes.
It follows from the fact that the condition (2.1.6) from [27] is satisfied if and only if \(\upsilon \le 2\).
There is an error in the expression for \(\xi ^\alpha \partial ^\beta {\hat{f}}\left( \xi \right) \) in the proof of [21, Proposition 5.3]. However, the proof is easily fixed by using the correct formula that we give here.
It makes easier to prove that Ruelle resonances are intrinsic in Appendix A or to define the norm \(\left\| \cdot \right\| _{{\mathcal {H}}}\) in (6.1) for instance.
This ensures that it is \({\mathcal {C}}^{\kappa ,\upsilon }\) for any \(\kappa > 0\) and \(\upsilon >1\), so that all the Fourier multipliers that appear later are automatically well-defined.
Here, \(A^{\mathrm{tr}}\) denotes the transpose of A.
Notice that the condition \(D_x {\mathcal {T}}_t^{\mathrm{tr}} \left( \xi \right) \in C_0'\) implies in particular that \(\xi \in C_0\), as a consequence of (i) and (ii).
See [17, Chapter IV.11] for the definition and basic properties of Schatten classes.
If E, F are Banach spaces, \(e \in F\) and \(l \in E'\), we denote by \(e \otimes l\) the rank 1 operator defined by \(e \otimes l (u) = l(u). e\) for \(u \in E\).
There always is a linear such map.
It could well be that \({\widetilde{{\mathcal {H}}}}_0 = {\widetilde{{\mathcal {H}}}}\), see Proposition 3.3, but we do not need this fact.
Recall that \(T_\gamma \) is the length of \(\gamma \), while \(T_\gamma ^\#\) denotes its primitive length and \({\mathcal {P}}_\gamma \) is a linearized Poincaré map. We will see during the proof of the proposition that this sum converges.
Notice that the global trace formula (1.4) may be deduced from this equality using residue’s formula.
References
Adam, A..: Opérateurs de transfert et moyenes horocycliques sur les variétés fermées. Ph.D thesis, École doctoral de sciences mathématiques de Paris centre
Adam, A.: Horocycle averages on closed manifolds and transfer operators (September 2018). arXiv:1809.04062,
Baladi, V..: Dynamical Zeta Functions and Dynamical Determinants for Hyperbolic Maps, Volume 68 of Ergebnisse. Springer (2018)
Baladi, V., Tsujii, M.: Anisotropic Hölder and Sobolev spaces for hyperbolic diffeomorphisms. Ann. Inst. Fourier (Grenoble) 57(1), 127–154 (2007)
Baladi, V., Tsujii, M..: Dynamical determinants and spectrum for hyperbolic diffemorphisms. In: Geometric and Probabilistic Structures in Dynamics. Amer. Math. Soc., Providence, RI(469), pp. 29–68 (2008)
Boas, R.: Entire Functions. Academic Press, New York (1954)
Butterley, O., Liverani, C.: Smooth Anosov flows: correlation spectra and stability. J. Mod. Dyn. 1(2), 301–322 (2007)
Butterley, O., Liverani, C.: Robustly invariant sets in fiber contracting bundle flows. J. Mod. Dyn. 7(2), 255–267 (2013)
Carmichael, R.D., Pilipović, S.: On the convolution and the Laplace transformation in the space of Beurling–Gevrey tempered ultradistributions. Math. Nachr. 158, 119–131 (1992)
Cvitanović, P., Vattay, G.: Entire Fredholm determinants for evaluation of semiclassical and thermodynamical spectra. Phys. Rev. Lett. 71(25), 4138–4141 (1993)
Dyatlov, S., Guillarmou, C.: Pollicott–Ruelle resonances for open systems. Ann. Henri Poincaré 17(11), 3089–3146 (2016)
Dyatlov, S., Guillarmou, C.: Afterword: dynamical zeta functions for Axiom A flows. Bull. Amer. Math. Soc. (N.S.) 55(3), 337–342 (2018)
Dyatlov, S., Zworski, M.: Dynamical zeta functions for Anosov flows via microlocal analysis. Ann. Sci. Éc. Norm. Supér. (4) 49(3), 543–577 (2016)
Faure, F., Sjöstrand, J.: Upper bound on the density of Ruelle resonances for Anosov flows. Commun. Math. Phys. 308(2), 325–364 (2011)
Fried, D.: Meromorphic zeta functions for analytic flows. Commun. Math. Phys. 174(1), 161–190 (1995)
Giulietti, P., Liverani, C., Pollicott, M.: Anosov flows and dynamical zeta functions. Ann. Math. (2) 178(2), 687–773 (2013)
Gohberg, I., Goldberg, S., Krupnik, N.: Traces and Determinants of Linear Operators. Operator Theory: Advances and Applications, vol. 116. Birkhaüser Verlag, Basel-Boston-Berlin (2000)
Gouëzel, S., Liverani, C.: Compact locally maximal hyperbolic sets for smooth maps: fine statistical properties. J. Differ. Geom. 79(3), 433–477 (2008)
Guillemin, V.: Lectures on spectral theory of elliptic operators. Duke Math. J. 44(3), 485–517 (1977)
Hennion, H.: Sur un théorème spectral et son application aux noyaux lipchitziens. Proc. Am. Math. Soc. 118(2), 627–634 (1993)
Jézéquel, M.: Local and global trace formulae for smooth hyperbolic diffeomorphisms. J. Spectral Theory 10(1), 185–249 (2020)
Jézéquel, M.: Transfer operator for ultradifferentiable expanding maps of the circle. In: Ergodic Theory and Dynamical Systems (first view) (2020)
Jin, L., Zworski, M.: A local trace formula for Anosov flows. Ann. Henri Poincaré 18(1), 1–35, (2017). With appendices by Frédéric Naud
Kato, T.: Perturbation Theory for Linear Operators. Springer, Berlin (1966)
Katok, A., Hasselblatt, B.: Introduction to the Modern Theory of Dynamical Systems, volume 54 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, Cambridge, (1995). With a supplementary chapter by Katok and Leonardo Mendoza
Komatsu, H.: Ultradifferentiability of solutions of ordinary differential equations. Proc. Japan Acad. Ser. A Math. Sci. 56(4), 137–142 (1980)
Kriegl, A., Michor, P.W., Rainer, A.: The convenient setting for non-quasianalytic Denjoy–Carleman differentiable mappings. J. Funct. Anal. 256(11), 3510–3544 (2009)
Liverani, C.: On contact Anosov flows. Ann. of Math. (2) 159(3), 1275–1312 (2004)
Mather, J.N.: Characterization of Anosov diffeomorphisms. Nederl. Akad. Wetensch. Proc. Ser. A 71 = Indag. Math 30, 479–483 (1968)
Muñoz, V., Pérez Marco, R.: On the genus of meromorphic functions. Proc. Amer. Math. Soc 143(1), 341–351 (2015)
Muñoz, V., Pérez Marco, R.: Unified treatment of explicit and trace formulas via Poisson-Newton formula. Comm. Math. Phys. 336(3), 1201–1230 (2015)
Nussbaum, R.D.: The radius of the essential spectrum. Duke Math. J. 37, 473–478 (1970)
Pilipović, S.: Tempered ultradistributions. Boll. Un. Mat. Ital. B (7) 2(2), 235–251 (1988)
Rugh, H.H.: The correlation spectrum for hyperbolic analytic maps. Nonlinearity 5, 1237–1263 (1992)
Rugh, H.H.: Generalized Fredholm determinants and Selberg zeta functions for Axiom A dynamical systems. Ergod. Theory Dyn. Syst. 16, 805–819 (1996)
Teofanov, N., Tomić, F.: Ultradifferentiable functions of class \(M^{\tau ,\sigma }_p\) and microlocal regularity. In: Generalized functions and Fourier analysis, Volume 260 of Oper. Theory Adv. Appl., pp. 193–213. Birkhäuser/Springer, Cham (2017)
Acknowledgements
I would like to thank Viviane Baladi for her careful reading of the different versions of this work. I would also like to thank Sébastien Gouëzel, Maciej Zworski, Semion Dyatlov and Shu Shen for discussions about the trace formula and useful suggestions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by S. Dyatlov
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No 787304). This work was started while the author was affiliated with Institut de Mathématiques de Jussieu-Paris Rive Gauche.
Appendices
Appendix
Appendix A. Ruelle Resonances are Intrinsic
As pointed out before, the Banach spaces \({\mathcal {B}}\) that appear in Theorem 1.2 are highly non-canonical. To prove that Ruelle resonances do not depend on the choice of these spaces, there is a classical argument based on the investigation of a meromorphic continuation of the resolvent of X as an operator from \({\mathcal {C}}^\infty \left( M\right) \) to \({\mathcal {D}}'\left( M\right) \). To deal with spaces that are not intermediary between \({\mathcal {C}}^\infty \left( M\right) \) and \({\mathcal {D}}'\left( M\right) \), it is easier to use an approach based on the following Lemma A.2, whose proof may be found in [3, Lemma A.3] or [5, Lemma A.1]. Recall first the following definition.
Definition A.1
(Isolated eigenvalue of finite multiplicity and essential spectral radius). If \({\mathcal {B}}\) is a Banach space, X an a priori unbounded operator on \({\mathcal {B}}\) and \(\lambda \in {\mathbb {C}}\), we say that \(\lambda \) is an isolated eigenvalue of finite multiplicity for X if \(\lambda \) is an isolated point of \(\sigma \left( X\right) \) and the rank of the spectral projector
where r is any small enough positive real number so that \(\sigma \left( X\right) \cap \overline{{\mathbb {D}}}\left( \lambda ,r\right) = \left\{ \lambda \right\} \), is finite (this rank is by definition the algebraic multiplicity of \(\lambda \)).
Now, if X is bounded we define the essential spectral radius of X as the infimum of \(\rho > 0\) such that the intersection of \(\sigma \left( X\right) \) with \(\left\{ z \in {\mathbb {C}}: \left| z \right| > \rho \right\} \) contains only isolated eigenvalues of finite multiplicity.
Lemma A.2
Let \({\mathcal {B}}\) be a Hausdorff topological vector space. Let \({\mathcal {B}}_1\) and \({\mathcal {B}}_2\) be Banach spaces continuously included in \({\mathcal {B}}\) such that \({\mathcal {B}}_1 \cap {\mathcal {B}}_2\) is dense both in \({\mathcal {B}}_1\) and in \({\mathcal {B}}_2\). Let \(L : {\mathcal {B}}\rightarrow {\mathcal {B}}\) be a continuous linear map that preserves \({\mathcal {B}}_1\) and \({\mathcal {B}}_2\). Assume that the maps induced by L on \({\mathcal {B}}_1\) and \({\mathcal {B}}_2\) are bounded operators whose essential spectral radius is smaller than some number \(\rho >0\). Then the eigenvalues in \(\left\{ z \in {\mathbb {C}}: \left| z \right| > \rho \right\} \) coincide. Furthermore, the corresponding generalized eigenspaces coincide and are contained in \({\mathcal {B}}_1 \cap {\mathcal {B}}_2\).
Applying Lemma A.2 to the resolvent of X, we may prove the two following lemmas. Lemma A.3 asserts that Ruelle resonances are well-defined, while Lemma A.4 ensures that the spectrum of X acting on the space \({\mathcal {H}}\) given by Theorem 1.7 coincides with the Ruelle spectrum (recall Definition 1.3). The proofs of Lemmas A.3 and A.4 are very similar and consequently we will only prove Lemma A.4, in order to show that there are no particular difficulties when working with unusual classes of regularity.
Lemma A.3
Let \({\mathcal {B}}\) and \({\widetilde{{\mathcal {B}}}}\) be two Banach spaces and \(A > 0\) be a positive real number. Assume that \({\mathcal {B}}\) and \({\widetilde{{\mathcal {B}}}}\) both satisfy the points (i)–(iv) from Theorem 1.2 for this particular value of A. Then the intersections of \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > - A \right\} \) with the spectrum of X acting on \({\mathcal {B}}\) and \({\widetilde{{\mathcal {B}}}}\) coincide.
Lemma A.4
Assume that M, the flow \(\left( \phi ^t\right) _{t \in {\mathbb {R}}}\), and g are \({\mathcal {C}}^{\kappa ,\upsilon }\) for some \(\kappa > 0\) and \(\upsilon >1\). Let \({\mathcal {B}}\) be a Banach space such that for some \({\tilde{\upsilon }} > \upsilon \) and \(A \in {\mathbb {R}}\) we have:
-
(i)
\({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \subseteq {\mathcal {B}}\subseteq {\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) '\), all the inclusions being continuous, the first one having dense image;
-
(ii)
for all \(t \in {\mathbb {R}}_+\), the operator \({\mathcal {L}}_t\) defined by (1.2) is bounded from \({\mathcal {B}}\) to itself;
-
(iii)
\(\left( {\mathcal {L}}_t\right) _{t \ge 0}\) forms a strongly continuous semi-group of operator acting on \({\mathcal {B}}\), whose generator is X;
-
(iv)
the intersection of the spectrum of X acting on \({\mathcal {B}}\) with \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > -A \right\} \) is made of isolated eigenvalues of finite multiplicity.
Then the intersection of \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > -A \right\} \) with the spectrum of X acting on \({\mathcal {B}}\) is the intersection of \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > -A \right\} \) with the Ruelle spectrum of X from Definition 1.3.
Proof
Apply Theorem 1.2 (with the same value of A) to get a Banach space \({\widetilde{{\mathcal {B}}}}\) such that in particular the intersection of \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > -A \right\} \) with the spectrum of X acting on \({\widetilde{{\mathcal {B}}}}\) coincides with the intersection of \(\left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) > -A \right\} \) with the Ruelle spectrum of X (by definition of the Ruelle spectrum). Now choose a positive real number \(z_0\) large enough so that the resolvent \(\left( z_0 - X\right) ^{-1}\) is well-defined both on \({\mathcal {B}}\) and \({\widetilde{{\mathcal {B}}}}\). Notice that the resolvents of X acting on \({\mathcal {B}}\) and \({\widetilde{{\mathcal {B}}}}\) coincide on the intersection \({\mathcal {B}}\cap {\widetilde{{\mathcal {B}}}}\). Indeed, from (iii) and [24, Problem 1.15 p.487], it follows that, if \(u \in {\mathcal {B}}\cap {\widetilde{{\mathcal {B}}}}\), then \(\left( z_0 - X\right) ^{-1}\) is defined as an element of \({\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) '\) by
Thus we may extend \(\left( z_0 - X\right) ^{-1}\) to \({\mathcal {B}}+ {\widetilde{{\mathcal {B}}}}\) by setting that \(\left( z_0 - X\right) ^{-1} u\) is equal to \( \left( z_0 - X\right) ^{-1} v + \left( z_0 - X\right) ^{-1} w\), if \(u = v + w\) with \(v \in {\mathcal {B}}\) and \(w \in {\widetilde{{\mathcal {B}}}}\) (it does not depend on the choice of v and w). Furthermore, this extension is continuous when \({\mathcal {B}}+ {\widetilde{{\mathcal {B}}}}\) is endowed with the norm \(\left\| \cdot \right\| _{{\mathcal {B}}+ {\widetilde{{\mathcal {B}}}}}\) defined by
Let \(A' <A\) and \(R >0\), provided that \(z_0\) is large enough we have
The map \(\lambda \mapsto \left( z_0 - \lambda \right) ^{-1}\) induces a bijection between the spectrum of X acting on \({\mathcal {B}}\) and the spectrum of \(\left( z_0 - X\right) ^{-1}\) acting on \({\mathcal {B}}\), but it also sends \( \left\{ z \in {\mathbb {C}}: \mathfrak {R}\left( z\right) \le - A \right\} \) into the disc of center 0 and radius \(\frac{1}{z_0 + A}\). Consequently, the essential spectral radius of \(\left( z_0 - X\right) ^{-1}\) acting on \({\mathcal {B}}\) is less than \(\frac{1}{z_0+A}\). The same is true for the same reason replacing \({\mathcal {B}}\) by \({\widetilde{{\mathcal {B}}}}\). Thus we may apply Lemma A.2 (with \(\rho = \frac{1}{z_0 + A}\) and \({\mathcal {B}}_1,{\mathcal {B}}_2\) and \({\mathcal {B}}\) being respectively \({\mathcal {B}},{\widetilde{{\mathcal {B}}}}\) and \({\mathcal {B}}+ {\widetilde{{\mathcal {B}}}}\)) to see that the spectrum of \(\left( z_0 -X\right) ^{-1}\) outside of the disc of center 0 and radius \(\frac{1}{z_0 +A}\) is the same on \({\mathcal {B}}\) and on \({\widetilde{{\mathcal {B}}}}\). But the map \(\lambda \mapsto \left( z_0 - \lambda \right) ^{-1}\) sends \(\left\{ z \in {\mathbb {C}}: -A' \le \mathfrak {R}(z) \le z_0 \text { and } \left| \mathfrak {I}\left( z\right) \right| \le R \right\} \) outside of the disc of center 0 and radius \(\frac{1}{z_0+A}\) (see (A.1)). Consequently, the intersection of \(\left\{ z \in {\mathbb {C}}: -A' \le \mathfrak {R}(z) \le z_0 \text { and } \left| \mathfrak {I}\left( z\right) \right| \le R \right\} \) with the spectrum of X acting on \({\mathcal {B}}\) coincides with the intersection of \(\left\{ z \in {\mathbb {C}}: -A' \le \mathfrak {R}(z) \le z_0 \text { and } \left| \mathfrak {I}\left( z\right) \right| \le R \right\} \) with the set of Ruelle resonances of X. Since \(R >0\) and \(A' < A\) are arbitrary, and \(z_0\) may be chosen arbitrarily large, the lemma is proven. \(\quad \square \)
Appendix B. Proofs of Lemmas 2.3 and 2.4
Proof of Lemma 2.3
We only need to prove the first point: the same argument with \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \) replaced by \({\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) \), and \({\mathcal {L}}_t\) and X replaced by their formal adjoints gives the second point.
We start with the case \(g = 0\). Using the group property of \(\left( {\mathcal {L}}_t\right) _{t \in {\mathbb {R}}}\), we only need to prove differentiability at \(t=0\). Then we may cover M by flow boxes, and thus we only need to show that if \(u \in {\mathcal {S}}^{{\tilde{\upsilon }}}\) is supported in a compact subset K of \({\mathbb {R}}^{d+1}\) then
where \(e_{d+1}\) denotes the last vector of the canonical basis of \({\mathbb {R}}^{d+1}\). Up to enlarging K we may assume that for all \(t \in \left[ -1,1\right] \) the function \(u\left( \cdot + t e_{d+1}\right) \) is supported in K. Then if \(x \in K\), \(\alpha \in {\mathbb {N}}^{d+1}\), and \(t \in \left[ -1,1\right] \) we have with Taylor’s formula (for any \(\kappa '' >0\)):
Thus if \(\kappa ',\kappa '' > 0\) and for \(R > 0\) depending only on K, we have for all \(x \in {\mathbb {R}}^{d+1}, \alpha \in {\mathbb {N}}^{d+1}\) and \(m \in {\mathbb {N}}\):
Thus if \(\kappa ' > 0\) and \(\kappa '' > \kappa ' \), then there is a constant \(C >0\) (that only depends on \(K, {\tilde{\upsilon }},\kappa '\), and \(\kappa ''\)) such that for all \(t \in \left[ -1,1\right] \) we have
which implies (B.1) and thus ends the proof of the lemma in the case \(g =0\).
In order to deduce the result in the case of a general g from the case \(g=0\), we only need to prove that the map
is \({\mathcal {C}}^\infty \) from \({\mathbb {R}}\) to \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \). Indeed, the multiplication is continuous from the product \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \times {\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \) to \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \). The map (B.2) is easily seen to be \({\mathcal {C}}^\infty \) from \({\mathbb {R}}\) to \({\mathcal {C}}^0\left( M\right) \), and one may notice that its derivatives are valued in \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\) (recall that the classes of regularity \({\mathcal {C}}^{\kappa ,{\tilde{\upsilon }}}\), and hence \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\), are closed by composition) with uniform bounds locally in t. Then, by successive applications of Taylor’s formula at order 1 with integral remainder, one gets that the map (B.2) is \({\mathcal {C}}^\infty \) from \({\mathbb {R}}\) to \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \), ending the proof of the lemma (we use the exact formula for the remainder in order to bound it in \({\mathcal {C}}^{\infty ,{\tilde{\upsilon }}}\left( M\right) \)). \(\quad \square \)
Proof of Lemma 2.4
Denote for now the generator of \(\left( {\mathcal {L}}_t\right) _{t \ge 0}\) by \({\widetilde{X}}\). Let \(u \in {\mathcal {B}}\) be in the domain of \({\widetilde{X}}\), then the map \({\mathbb {R}}_+ \ni t \mapsto {\mathcal {L}}_t u \in {\mathcal {B}}\) is differentiable at 0 and its derivative at 0 is \({\widetilde{X}}u\) (by definition of \({\widetilde{X}}\)). Since \({\mathcal {B}}\subseteq {\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) \) is continuous, the same is true for the map \({\mathbb {R}}_+ \ni t \mapsto {\mathcal {L}}_t u \in {\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) '\), whose derivative at 0 is Xu according to Lemma 2.3. Thus \({\widetilde{X}}u = Xu \in {\mathcal {B}}\).
Reciprocally, if \(u \in {\mathcal {B}}\) is such that \(Xu \in {\mathcal {B}}\), then we may define a \({\mathcal {C}}^1\) map \(c: {\mathbb {R}}_+ \rightarrow {\mathcal {B}}\) by \(c\left( t\right) = u + \int _0^t {\mathcal {L}}_\tau X u \mathrm {d}\tau \) for all \(t \in {\mathbb {R}}_+\). Notice that \(c'\left( 0\right) = Xu\). Since \({\mathcal {B}}\subseteq {\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) \) is continuous, the map c is still \({\mathcal {C}}^1\) when seen as a map from \({\mathbb {R}}_+\) to \({\mathcal {D}}^{{\tilde{\upsilon }}}\left( M\right) \) and we have \(c\left( 0\right) = u\) and \(c'\left( t\right) = {\mathcal {L}}_t X u\) for all \(t \in {\mathbb {R}}_+\), so that \(c\left( t\right) = {\mathcal {L}}_t u\) for all \(t \in {\mathbb {R}}_+\), using Lemma 2.3. This proves that u belongs to the domain of \({\widetilde{X}}\). \(\quad \square \)
Appendix C. Factorization of the Dynamical Determinant
We prove here, under the hypotheses of Theorem 1.7, a Hadamard-like factorization (C.3) for the dynamical determinant \(d_g\) defined by (1.3). Let \(t_0 > 0\) be shorter than any periodic orbit of \(\left( \phi ^t\right) _{t \in {\mathbb {R}}}\). Then, working as in the proof of Proposition 1.9, we see that, for \(\mathfrak {R}z \gg 1\), the essential spectral radius of
is zero. Then, applying holomorphic functional calculus in finite dimension as in the proof of Lemma 6.6, we see that the spectrum of (C.1) is made of the \(\frac{e^{\lambda t_0}}{(z-\lambda )^{d+2}}\) for \(\lambda \) in the spectrum of X. Then, for \(\mathfrak {R}z \gg 1\), Proposition 5.4 implies that the right hand side of (C.1) defines a trace class operator on \({\widetilde{{\mathcal {H}}}}_0\). From Lemma A.2, we see that the spectrum of (C.1) is the same when acting on \({\mathcal {H}}\) or on \({\tilde{h}}_0\). Then, using Lidskii’s Trace Theorem and Proposition 5.4, we see that,Footnote 16
For all \(\lambda \in {\mathbb {C}}\setminus \left\{ 0 \right\} \) notice that the meromorphic map
has a unique pole in \(\lambda \) whose order is 1 and whose residue is 1. Thus there is an entire function \(G_{\lambda ,t_0}\) such that for all \(z \in {\mathbb {C}}\)
and \(G_{\lambda ,t_0}\left( 0\right) = 1\). Choose for \(G_{0,t_0}\) any logarithmic primitive of \(z \mapsto \frac{e^{-t_0 z}}{z}\).
Now, choose \(R > 0\) and if \(\left| \lambda \right| \ge 2 R\) notice that for all \(z \in {\mathbb {D}}\left( 0,R\right) \) we have
and using the fact that \(G_{\lambda ,t_0}\) has a logarithm on \({\mathbb {D}}\left( 0,R\right) \) that vanishes in 0 (since \(G_{\lambda ,t_0}\) vanishes only at \(\lambda \)) we get that, for some constance C depending only on R and all \(z \in {\mathbb {D}}\left( 0,R\right) \)
Using Proposition 1.9, this implies that the infinite product
converges uniformly on all compact subset of \({\mathbb {C}}\). Notice that the zeros of \({\widetilde{d}}_g\) are precisely the Ruelle resonances. Now, we find that
and thus, for \(\mathfrak {R}z \gg 1\),
where \(d_g\) is the usual dynamical determinant defined by (1.3). From (C.2) we deduce that there are a polynomial P of degree at most d and \(\mu \in {\mathbb {C}}\) such that, for all \(z \in {\mathbb {C}}\), we have the Hadamard-like factorization
In order to make this factorization more explicit, let us describe the \(G_{\lambda ,t_0}\)’s. For all \(\lambda \in {\mathbb {C}}\setminus \left\{ 0 \right\} \), define the polynomial
and notice that
Thus we have for all \(\lambda \in {\mathbb {C}}\setminus \left\{ 0 \right\} \) and \(z \in {\mathbb {C}}\)
The last factor is a logarithmic primitive of \( z \mapsto \frac{e^{-\left( z-\lambda \right) t_0}-1}{z-\lambda }\).
Appendix D. Applications of the Trace Formula
As applications of the trace formula, we prove here Proposition 1.5 and Corollary 1.6.
Proof of Proposition 1.5
It is folklore to prove the implication (i) \(\Rightarrow \) (ii) from residue theorem, see also [30, Theorem 17]. Let us prove the implication (i) \(\Rightarrow \) (ii).
Choose \(x >0\) large enough so that the series
converges absolutely and \(x > \mathfrak {R}\left( \lambda \right) +\epsilon \) for all the resonances \(\lambda \) and some \(\epsilon >0\). Write \(k = \lceil \rho \rceil \). Choose \(z \in {\mathbb {C}}\) such that \(\mathfrak {R}\left( z\right) > x\). Then, we can find a sequence \(\left( \varphi _n\right) _{n \in {\mathbb {N}}}\) of \({\mathcal {C}}^\infty \) functions, compactly supported in \({\mathbb {R}}_+^*\) such that
and
Then, with (D.1) and (D.2), we have
Now, since the trace formula holds (by assumption), we know that for all \(n \in {\mathbb {N}}\) we have
However, recall that
so that, using (D.2), we have,
Now, if \(\lambda \) is non-zero, we have
Thus, (D.3), with \(x > \mathfrak {R}\left( \lambda \right) +\epsilon \), and the second hypothesis provide a domination of \(L\left( \varphi _n\right) \left( - \lambda \right) \), so that we have, using the dominated convergence theorem,
Finally we have (when \(\mathfrak {R}\left( z\right) \gg 1\))
Let P denote the canonical product of genus \(k-1\) whose zeros are the Ruelle resonances of X (well-defined by [6, (2.6.4)] thanks to (1.5)). Then we see that, if z is not a Ruelle resonance for X, we have
It follows that \(\left( \ln d_g\right) ^{(k+1)} = \left( \ln P\right) ^{(k+1)}\) and consequently there is a complex number a such that for every \(z \in {\mathbb {C}}\) that is not a Ruelle resonance for X we have
With (1.5), (D.4) and dominated convergence we see that \(\left( \ln P\right) ^{(k)}(r) \underset{\begin{array}{c} r \rightarrow + \infty \\ r \in {\mathbb {R}} \end{array}}{\rightarrow } 0\). By direct inspection, we see that \(\left( \ln d_g\right) ^{(k)}(r) \underset{\begin{array}{c} r \rightarrow + \infty \\ r \in {\mathbb {R}} \end{array}}{\rightarrow } 0\), and consequently \(a = 0\). Thus, there is a polynomial Q of degree at most \(k-1 \le \rho \) such that, for every \(z \in {\mathbb {C}}\), we have
and the result follows since P has order less than \(\rho \) by [6, Theorem 2.6.5]. \(\quad \square \)
Proof of Corollary 1.6
Proposition 1.5 implies that \(d_g\) has order less than 1. But notice that \(d_g\) is bounded on a line (choose a line parallel to the imaginary axis corresponding to a large positive real part) and thus has to be constant by the Phragmén–Lindelöf Theorem [6, Theorem 1.4.1]. Finally, it has to be constant equal to 1 since \(d_g(z) \underset{\begin{array}{c} z \rightarrow + \infty \\ z \in {\mathbb {R}} \end{array}}{\rightarrow } 1\). \(\quad \square \)
Appendix E. Expanding Maps of the Circle and the Condition
In order to discuss the condition \(\upsilon < 2\) in Theorem 1.7, we can consider a very simple example: expanding maps of the circle \({\mathbb {S}}^1 = {\mathbb {R}}/ {\mathbb {Z}}\). An analogue of the space \({\mathcal {H}}\) from Theorem 1.7 would then be an isotropic space of the type (here \(\left( {\hat{f}}(n)\right) _{n \in {\mathbb {Z}}}\) denotes the sequence of Fourier coefficient of a function f)
where \(\beta > 0\) and \(\alpha \in \left]\frac{\upsilon - 1}{\upsilon },1 \right[\) (this is the same condition as in Proposition 4.4), endowed with the norm
Then the transfer operator
associated to the doubling map may be written as
where \(e_n : x \mapsto e^{2 i \pi n x}\) (the sum converges in strong operator topology on the space of continuous endomorphisms of \({\mathcal {H}}_{\alpha ,\beta }\)). Thus, the singular values of \({\mathcal {L}}\) acting on \({\mathcal {H}}_{\alpha ,\beta }\) are the \( e^{\beta \left( \ln \left( 1 + \left| n \right| \right) ^{\frac{1}{\alpha }} - \ln \left( 1 + 2\left| n \right| \right) ^{\frac{1}{\alpha }}\right) } \) for \(n \in {\mathbb {Z}}\). Using the fact that
we see that \({\mathcal {L}}\) acting on \({\mathcal {H}}_{\alpha ,\beta }\) is trace class when \(\alpha < \frac{1}{2}\) and is not trace class when \(\alpha > \frac{1}{2}\) (in the case \(\alpha = \frac{1}{2}\) it depends on the value of \(\beta \)). Thus, we need to chose \(\alpha < \frac{1}{2}\) if we want \({\mathcal {L}}\) to be nuclear. For general maps, this choice is possible only when \(\upsilon < 2\) (see the condition in Proposition 4.4).
Consequently, using our method to prove the trace formula for \({\mathcal {C}}^{\kappa ,\upsilon }\) Anosov flows (or hyperbolic diffeomorphisms as in [21]) would require to construct Hilbert spaces in a totally different way, if \(\upsilon \ge 2\).
Rights and permissions
About this article
Cite this article
Jézéquel, M. Global Trace Formula for Ultra-Differentiable Anosov Flows. Commun. Math. Phys. 385, 1771–1834 (2021). https://doi.org/10.1007/s00220-020-03930-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-020-03930-x