Abstract
We show, using semiclassical measures and unstable derivatives, that a smooth vector field X generating a contact Anosov flow on a 3-dimensional manifold \(\mathcal {M}\) has only finitely many Ruelle resonances in the vertical strips \(\{ s\in \mathbb {C}\ |\ \mathrm{Re}(s)\in [-\nu _{\min }+\varepsilon ,-\frac{1}{2}\nu _{\max }-\varepsilon ]\cup [-\frac{1}{2}\nu _{\min }+\varepsilon ,0]\}\) for all \(\varepsilon >0\), where \(0<\nu _{\min }\le \nu _{\max }\) are the minimal and maximal expansion rates of the flow (the first strip only makes sense if \(\nu _{\min }>\nu _{\max }/2\)). We also show polynomial bounds in s for the resolvent \((-X-s)^{-1}\) as \(|\mathrm{Im}(s)|\rightarrow \infty \) in Sobolev spaces, and obtain similar results for cases with a potential. This is a short proof of a particular case of the results by Faure–Tsujii, using that \(\dim E_u=\dim E_s=1\).
Similar content being viewed by others
Notes
For instance, this is satisfied for Anosov geodesic flows on orientable Riemannian surfaces.
The limit exists by Fekete’s lemma, and the same hold as well for \(V_{\max },V_{\min }\) defined later.
\(H^N(\mathcal {M})=(1+\Delta _g)^{-N/2}L^2(\mathcal {M})\) denotes the usual Sobolev space of order N on \(\mathcal {M}\).
It can be checked, by inspecting the proof, that one can choose \(c_{X}=\nu _{\min }\) and \(c_V=V_{\max }\).
And the fact that \(a \in S^m_{h, \rho , 0}\) and \(a > 0\) implies \(\sqrt{a} \in S^{m/2}_{h, \rho , 0}\).
References
Butterley, O., Liverani, C.: Smooth Anosov flows: correlation spectra and stability. J. Mod. Dyn. 1(2), 301–322 (2007)
Dolgopyat, D.: On decay of correlations in Anosov flows. Ann. Math. 147, 357–390 (1998)
Dyatlov, S.: Resonance projectors and asymptotics for r-normally hyperbolic trapped sets. J. Am. Math. Soc. 28(2), 311–381 (2015)
Dyatlov, S.: Spectral gaps for normally hyperbolic trapping. Ann. Inst. Fourier (Grenoble) 66(1), 55–82 (2016)
Dyatlov, S., Faure, F., Guillarmou, C.: Power spectrum of the geodesic flow on hyperbolic manifolds. Anal. PDE 8(4), 923–1000 (2015)
Dyatlov, S., Zworski, M.: Mathematical theory of scattering resonances, AMS Graduate Studies in Mathematics 200
Faure, F., Guillarmou, C.: Horocyclic invariance of Ruelle resonant states for contact Anosov flows in dimension 3. Math. Res. Lett. 25(5), 1405–1427 (2018)
Faure, F., Roy, N., Sjöstrand, J.: Semi-classical approach for Anosov diffeomorphisms and Ruelle resonances. Open Math. J. 1, 35–81 (2008)
Faure, F., Sjöstrand, J.: Upper bound on the density of Ruelle resonances for Anosov flows. Commun. Math. Phys. 308(2), 325–364 (2011)
Faure, F., Tsujii, M.: Band structure of the Ruelle spectrum of contact Anosov flows. Comptes Rendus Mathématique 351, 385–391 (2013)
Faure, F., Tsujii, M.: Prequantum transfer operator for Anosov diffeomorphism, Astérisque 375, (2015), SMF
Faure, F., Tsujii, M.: The semiclassical zeta function for geodesic flows on negatively curved manifolds. Invent. Math. 208, 851–998 (2017)
Faure, F., Tsujii, M.: Micro-local Analysis and Band Structure of Contact Anosov Flows. (2020). arXiv:2102.11196
Grigis, A., Sjöstrand, J.: Microlocal Analysis for Differential Operators. An Introduction. London Mathematical Society Lecture Note Series, 196. Cambridge University Press, Cambridge (1994)
Guillarmou, C., Hilgert, J., Weich, T.: Classical and quantum resonances for hyperbolic surfaces. Math. Ann. 370(3–4), 1231–1275 (2018)
Guillarmou, C., Küster, B.: Spectral theory of the frame flow on hyperbolic 3-manifolds (with an appendix by Charles Hadfield), Preprint arXiv:2005.08387
Hurder, S., Katok, A.: Differentiability, rigidity and Godbillon–Vey classes for Anosov flows. Publ. IHES 72, 5–61 (1990)
Küster, B., Weich, T.: Quantum-classical correspondence on associated vector bundles over locally symmetric spaces. IMRN (2019) rnz068
Liverani, C.: On contact Anosov flows. Ann. Math. 159(2), 1275–1312 (2004)
Nonnenmacher, S., Zworski, M.: Decay of correlations for normally hyperbolic trapping. Invent. Math. 200(2), 345–438 (2015)
Tsujii, M.: Quasi-compactness of transfer operators for contact anosov flows. Nonlinearity 23, 1495–1545 (2010)
Tsujii, M.: Contact Anosov flows and the FBI transform. Ergod. Theory Dyn. Syst. 32, 2083–2118 (2012)
Tsujii, M.: Exponential mixing for generic volume-preserving Anosov flows in dimension three. J. Math. Soc. Jpn. 70(2), 757–821 (2018)
Tsujii, M., Zhang, Z.: Smooth mixing Anosov flows in dimension three are exponential mixing, arXiv:2006.04293
Zworski, M.: Semiclassical Analysis. Graduate Studies in Mathematics, 138. American Mathematical Society, Providence, RI (2012)
Acknowledgements
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. 725967). We thank Semyon Dyatlov for several useful discussions and for suggesting to apply the method of [Dy2] in this setting.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by C. Liverani.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: An Exotic Symbol Class
Appendix A: An Exotic Symbol Class
Let \(\mathcal {M}\) be a closed n-manifold equipped with a Riemannian metric g, let \(k\ge 0\) and \(0<\rho <1\). We use the usual notation \(\langle \xi \rangle =(1+|\xi |^2)^{1/2}\), \(h>0\) will be a small semiclassical parameter, we let \(\overline{T}^*\mathcal {M}\) be the fiber radial compactification of \(T^*\mathcal {M}\) as defined in [DyZw, Section E.1.3] and \(H_p\) will denote the Hamiltonian vector field of \(p\in C^\infty (T^*\mathcal {M})\). Given an operator P, we write \({{\,\mathrm{Im}\,}}P = \frac{P - P^*}{2i}\) for the imaginary part of P and \({{\,\mathrm{Re}\,}}P = \frac{P + P^*}{2}\) for the real part; then \(P = {{\,\mathrm{Re}\,}}P + i {{\,\mathrm{Im}\,}}P\). Recall that for \(x\in \mathbb {R}\) we write \(x_+ = \max (x, 0)\).
Most of the results we gather in this appendix are simple extensions of classical results in semiclassical analysis that can be found in the books [Zw] and [DyZw, Appendix E]. We shall only point out the main differences with our setting.
For each \(m\in \mathbb {R}\), we define the exotic pseudo-differential calculus \(\Psi _{h,\rho , k}^m(\mathcal {M})\) by saying that \(A\in \Psi _{h,\rho , k}^m(\mathcal {M})\) if its Schwartz kernel \(K_A\) is in \(\mathcal {O}_{C^N(\mathcal {M}\times \mathcal {M})}(h^N)\) for all \(N>0\) outside a neighborhood of the diagonal, and near the diagonal can be written in local coordinates as
where the local symbols are in the class \(S^{m}_{h,\rho ,k}(\mathbb {R}^{2n})\) defined by the property: \(a\in S^{m}_{h,\rho ,k}(\mathbb {R}^{2n})\) if \(a\in C^\infty (\mathbb {R}^{2n})\) is an h-dependent function and satisfies in local coordinates (for some \(C_{\alpha ,\beta }\) uniform in h)
We notice that it is important here, for the calculus, that the loss of \(h^{-\rho }\) happens only in the x-derivatives and not in the \(\xi \) derivatives. First, observe the basic properties for \(0\le k'\le k\)
We define \(S^m_{h,\rho ,k}(T^*\mathcal {M})\) to be \(C^\infty (T^*\mathcal {M})\) functions that, using local coordinates on \(\mathcal {M}\), are in \(S^m_{h,\rho ,k}(\mathbb {R}^{2n})\). First, one directly sees from the formula of symbols under a change of coordinates [Zw, Theorem 9.9] that the symbol in local coordinates being in \(S^{m}_{h,\rho ,k}(\mathbb {R}^{2n})\) is invariant by change of coordinates, and moreover there is a principal symbol map
Using local charts and a partition of unity, we fix a semi-classical quantization \(\mathrm{Op}_h:S^{m}_{h,\rho ,k}(T^*\mathcal {M})\rightarrow \Psi _{h,\rho ,k}^m(\mathcal {M})\), which satisfies
We first check that symbols in this class are closed under composition. Recall from [Zw, Theorem 4.14] that if \(A\in \Psi ^{m}_{h,\rho ,k}(\mathbb {R}^n)\) and \(B\in \Psi ^{m'}_{h,\rho ,k}(\mathbb {R}^n)\) have full symbol a, b then AB has full symbol (as an oscillatory integral)
which has expansion
Here \(|\partial _\xi ^\alpha a \partial _x^\alpha b|\le C_{\alpha }h^{-|\alpha |\rho }\langle \xi \rangle ^{m+m'-|\alpha |}\) so that higher order terms in the expansion are higher powers of h and of \(\langle \xi \rangle ^{-1}\).
Lemma A .1
Let \(a \in S^{m_1}_{h, \rho , k}(\mathbb {R}^{2n})\) and \(b \in S^{m_2}_{h, \rho , k}(\mathbb {R}^{2n})\). Then \(a \# b \in S^{m_1 + m_2}_{h, \rho , k}(\mathbb {R}^{2n})\).
Proof
This follows from (A.4) and (A.2). \(\quad \square \)
For \(A\in \Psi _{h,\rho ,k}^{m}(\mathcal {M})\), we say that \((x_0,\xi _0)\in \overline{T}^*\mathcal {M}\) is not in \(\mathrm{WF}_h(A)\) if there is a small neighborhood U of \((x_0,\xi _0)\) in \(\overline{T}^*\mathcal {M}\) so that the full local symbol of A restricted to U is in \(h^{N}S^{-N}_{h,\rho ,0}(U)\) for all \(N>0\). We also define the elliptic set \(\mathrm{ell}_h(A)\) of \(A\in \Psi _{h,\rho ,k}^m(\mathcal {M})\) to be the set of points \((x_0,\xi _0)\in \overline{T}^*\mathcal {M}\) so that for a neighborhood U of \((x_0,\xi _0)\) there is \(c_0>0\) so that \(\langle \xi \rangle ^{-m}|\sigma (A)(x,\xi )|\ge c_0\), for \((x, \xi ) \in U\). We finally list some properties of the \(\Psi _{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\) calculus.
Proposition A.1
The following properties hold:
-
1.
Let \(A \in \Psi _{h, \rho , k}^m(\mathcal {M})\). If \(B, B' \in \Psi _h^{{{\,\mathrm{comp}\,}}}(\mathcal {M})\) with \(\mathrm{WF}_h(B) \cap \mathrm{WF}_h(B') = \emptyset \), then \(BAB' \in h^\infty \Psi _h^{{{\,\mathrm{comp}\,}}}(\mathcal {M})\).
-
2.
The principal symbol is well-defined as a map
$$\begin{aligned} \sigma : \Psi ^m_{h, \rho , k}(T^*\mathcal {M}) \rightarrow S^m_{h, \rho , k}(T^*\mathcal {M}) / hS^{m-1}_{h, \rho , k}(T^*\mathcal {M}) \end{aligned}$$with kernel \(h\Psi _{h,\rho ,k}^{m-1}(\mathcal {M})\), and it satisfies for any \(A\in \Psi ^m_{h, \rho , k}(\mathcal {M})\) and \(B\in \Psi ^{m'}_{h, \rho , k}(\mathcal {M})\)
$$\begin{aligned} \begin{aligned} \sigma (AB) = \sigma (A) \sigma (B) \,\,\,\mathrm {mod}\,\,\, h^{1-\rho }S^{m+m'-1}_{h,\rho ,k}(T^*\mathcal {M}), \\ \sigma (AB) = \sigma (A) \sigma (B) \,\,\,\mathrm {mod}\,\,\, hS^{m+m'-1}_{h,\rho ,k-1}(T^*\mathcal {M}) \text { if }k\ge 1. \end{aligned} \end{aligned}$$ -
3.
If \(A \in \Psi ^{m_1}_{h, \rho , k}(\mathcal {M})\) and \(B \in \Psi ^{m_2}_{h, \rho , k}(\mathcal {M})\) for \(k\ge 1\), then
$$\begin{aligned} \begin{aligned} \, [A, B]\in h^{1 - \rho } \Psi ^{m_1 + m_2 - 1}_{h, \rho , k}(\mathcal {M}), \\ h^{-1}[A, B] \in \Psi _{h, \rho , k - 1}^{m_1 + m_2 - 1}(\mathcal {M}) \text { if }k\ge 1 , \end{aligned} \end{aligned}$$and \(\sigma (h^{-1}[A, B]) = -i\{\sigma (A), \sigma (B)\}\).
-
4.
If \(P = -ihX\) for X a vector field on \({{\,\mathrm{\mathcal {M}}\,}}\), \(\Theta \in \Psi ^m_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\) and \(H_p \sigma (\Theta ) \in S^m_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\), where \(p =\sigma (P)= \xi (X)\), then \([P, \Theta ] \in h\Psi ^m_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\).
-
5.
If \(A \in \Psi _{h, \rho , k}^m(\mathcal {M})\), then \(A^* \in \Psi _{h, \rho , k}^m(\mathcal {M})\) and
$$\begin{aligned}\begin{aligned} \sigma (A^*)=\overline{\sigma (A)} \,\,\,\mathrm {mod}\,\,\, h^{1-\rho }S^{m+m'-1}_{h,\rho ,k}(T^*\mathcal {M}),\\ \sigma (A^*)=\overline{\sigma (A)} \,\,\,\mathrm {mod}\,\,\, hS^{m-1}_{h,\rho ,k-1}(T^*\mathcal {M}) \text { if }k\ge 1.\end{aligned}\end{aligned}$$ -
6.
Each \(A \in \Psi ^0_{h, \rho , 0}(\mathcal {M})\) is bounded and for each \(\varepsilon > 0\)
$$\begin{aligned}\Vert A\Vert _{L^2 \rightarrow L^2} \le (1 + \varepsilon ) \sup _{h, x, \xi } |\sigma _h(A)(x, \xi )| + \mathcal {O}_{\varepsilon }(h^{\infty }).\end{aligned}$$Moreover, for any \(A \in \Psi _{h, \rho , k}^m(\mathcal {M})\) and any \(s \in \mathbb {R}\) we have
$$\begin{aligned}\Vert A\Vert _{H_h^s \rightarrow H_h^{s-m}} \le (1 + \varepsilon ) \sup _{h, x, \xi }|\langle {\xi }\rangle ^{-m} \sigma (A)| + \mathcal {O}_{\varepsilon }(h^\infty ).\end{aligned}$$ -
7.
Let \(P \in \Psi _{h, \rho , k}^p({{\,\mathrm{\mathcal {M}}\,}})\), \(A \in \Psi ^m_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\) and \(B_1 \in \Psi ^l_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\). Assume \(\mathrm{WF}_h(A) \subset {{\,\mathrm{ell}\,}}_h(P) \cap {{\,\mathrm{ell}\,}}_h(B_1)\). Then for all \(s\in \mathbb {R}\), \(N > 0\), and u with \(B_1Pu \in H^{s-p-l}_h({{\,\mathrm{\mathcal {M}}\,}})\)
$$\begin{aligned}\Vert Au\Vert _{H^{s-m}_h} \le C\Vert B_1Pu\Vert _{H^{s - p - l}_h} + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}.\end{aligned}$$ -
8.
Assume \(k {\ge } 1\) and let \(P \in \Psi _{h, \rho , k}^1({{\,\mathrm{\mathcal {M}}\,}})\) with \({{\,\mathrm{Re}\,}}P \in \Psi ^1_h({{\,\mathrm{\mathcal {M}}\,}})\) and \({{\,\mathrm{Im}\,}}P \in h\Psi ^0_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\). Denote \(p := \sigma (P)\) and assume that for each \((x,\xi )\in \mathrm{WF}_h(A)\subset \overline{T}^*\mathcal {M}\), there is \(T>0\) such that \(e^{-TH_p}(x,\xi )\in \mathrm{ell}_h(B)\) and \(e^{-tH_p}(x,\xi )\in \mathrm{ell}_h(B_1)\) for \(t \in [0, T]\). Then for each \(u\in L^2\) with \(Pu \in L^2({{\,\mathrm{\mathcal {M}}\,}})\), and every \(N > 0\) there is a \(C > 0\) such that
$$\begin{aligned}\Vert Au\Vert _{L^2} \le C\Vert Bu\Vert _{L^2} + Ch^{-1} \Vert B_1 P u\Vert _{L^2} + Ch^N \Vert u\Vert _{H_h^{-N}}.\end{aligned}$$
Proof
-
1.
This follows from the composition formula (A.4).
-
2.
This was discussed above.
-
3.
From the composition formula (A.4), locally we have
$$\begin{aligned} a\# b-b\# a \sim h(D_\xi a \partial _x b {-} D_\xi b \partial _x a) + \frac{1}{2}h^2(D_\xi ^2 a \partial _x^2 b - D_\xi ^2 b \partial _x^2 a) + \frac{1}{6}h^2(D_\xi ^3 a \partial _x^3 b - D_\xi ^3 b \partial _x^3 a) {+} \cdots \end{aligned}$$where, after taking the expansion to a high enough order, the remainder is in \(h^NS^{m-N}_{h,\rho ,0} (\mathbb {R}^{2n})\) for some large \(N>0\). By (A.2), all these terms are in \(h^{1-\rho }S^{m_1+m_2-1}_{h,\rho ,k}(\mathbb {R}^{2n})\cap hS^{m_1+m_2-1}_{h,\rho ,k-1}(\mathbb {R}^{2n})\) if \(k\ge 1\), and the principal symbol of \(h^{-1}[A,B]\) is \(-i(\partial _\xi a \partial _x b-\partial _x a \partial _\xi b)\).
-
4.
This follows from the composition formula (A.4), the fact that \(\partial _\xi ^{\alpha } \sigma (P) = 0\) for \(|\alpha | \ge 2\) and Item 3 above.
-
5.
This follows from the fact that, if A has full local symbol a in local coordinates, the full symbol \(a^*\) of \(A^*\) is
$$\begin{aligned} a^*(x, \xi ) \sim \sum _\alpha \frac{h^{|\alpha |}}{\alpha !} \partial _\xi ^\alpha D_x^\alpha \overline{a(x, \xi )}. \end{aligned}$$ -
6.
The argument is standard (see [GrSj, Theorem 4.5]) and we just make a brief summary. Let \(a := \sigma (A) \in S^0_{h, \rho , k}(\mathcal {M})\). For \(M=\Vert a\Vert _\infty \), we can construct \(B_0 \in \Psi ^0_{h, \rho , 0}(\mathcal {M})\) with principal symbol \(b_0:=\sqrt{(1 + \varepsilon )^2M^2-|a|^2}\ge \varepsilon M>0\), so that
$$\begin{aligned} C:=(1 + \varepsilon )^2 M^2 - A^*A = B_0^*B_0 + h^{1-\rho }R_0, \end{aligned}$$for some \(R_0=R_0^* \in \Psi _{h, \rho , 0}^{-1}(\mathcal {M})\) by part 2.Footnote 5 Next we can choose \(B_1=B_0 + \tfrac{h^{1-\rho }}{2}\mathrm {Op}_h(b_0^{-1})^*R_0\) so that \(B_1^*B_1=C - h^{2(1-\rho )}R_1\) with \(R_1\in \Psi ^{-2}_{h, \rho , 0}(\mathcal {M})\). We iterate this procedure to find \(B_N \in \Psi _{h, \rho , 0}^{0}(\mathcal {M})\) such that \((1 + \varepsilon )^2 M^2 - A^*A=B_N^*B_N+h^{(N+1) (1-\rho )}R_N\) with \(R_N\in \Psi _{h,\rho ,0}^{-N-1}(\mathcal {M})\). It is a routine argument to show that \(\Vert R_N\Vert _{L^2\rightarrow L^2}=\mathcal {O}(h^{N(1-\rho ) - n})\) thus for any \(u \in L^2\)
$$\begin{aligned} \Vert Au\Vert _{L^2}^2 = (1 + \varepsilon )^2 M^2\Vert u\Vert _{L^2}^2 - \Vert B_Nu\Vert _{L^2}^2 + \mathcal {O}(h^{2(N + 1)(1-\rho )-n}) \Vert u\Vert ^2_{L^2} \end{aligned}$$which shows the desired estimate by choosing N large. The Sobolev bound is an easy consequence of this.
-
7.
This follows from the parametrix construction in the elliptic set, the main thing to notice is that if \(B \in \Psi _{h, \rho , k}^m(\mathcal {M})\), \(A \in \Psi _{h, \rho , k}^{l}(\mathcal {M})\) and \(\mathrm{WF}_h(A) \subset {{\,\mathrm{ell}\,}}_h(B)\), then
$$\begin{aligned} \frac{\sigma (A)}{\sigma (B)} \in S_{h, \rho , k}^{l-m}(\mathcal {M}). \end{aligned}$$ -
8.
We follow the proof of [DyZw, Theorem E.47] and divide the proof into steps.
Step 0: an escape function. Fix \(\beta \ge 0\). There is a \(g \in C^\infty ({\overline{T}}^*\mathcal {M})\) with \({{\,\mathrm{supp}\,}}g \subset {{\,\mathrm{ell}\,}}_h(B_1)\), such that
$$\begin{aligned} g \ge 0, \quad g > 0 \,\,\mathrm {on}\,\, \mathrm{WF}_h(A), \quad H_p g \le - \beta g, \end{aligned}$$where the last condition holds outside \({{\,\mathrm{ell}\,}}_h(B)\). Step 1. Note that \(g \in S^0(T^*{{\,\mathrm{\mathcal {M}}\,}})\) and define
$$\begin{aligned} G:= {{\,\mathrm{Op}\,}}_h(\langle {\xi }\rangle ^s g) \in \Psi _h^s({{\,\mathrm{\mathcal {M}}\,}}), \quad \mathrm{WF}_h(G) \subset {{\,\mathrm{ell}\,}}_h(B_1). \end{aligned}$$We can that \(u \in C^\infty ({{\,\mathrm{\mathcal {M}}\,}})\). If we write \(f = Pu\),
$$\begin{aligned} {{\,\mathrm{Im}\,}}\langle {f, G^*Gu}\rangle = \underbrace{{{\,\mathrm{Im}\,}}\langle {({{\,\mathrm{Re}\,}}P) u, G^*Gu}\rangle }_{\mathrm {term\,\, T_1}} + \underbrace{{{\,\mathrm{Re}\,}}\langle {({{\,\mathrm{Im}\,}}P) u, G^*Gu}\rangle }_{\mathrm {term\,\, T_2}}. \end{aligned}$$We will bound the two terms on the right separately.
Step 2: term \(T_1\). Since \(\mathrm{Re}(P)\in \Psi _h^1(\mathcal {M})\) is non exotic, this step is exactly the same as in the proof of [DyZw, Theorem E.47] and we get
$$\begin{aligned} T_1 \le (C_1 - \beta )h \Vert Gu\Vert _{L^2}^2 + Ch\Vert Bu\Vert ^2_{H_h^s} + Ch^2 \Vert B_1 u\Vert ^2_{H^{s-\frac{1}{2}}_h} + \mathcal {O}(h^\infty )\Vert u\Vert _{H_h^{-N}}^2. \end{aligned}$$Step 3: term \(T_2\). Write
$$\begin{aligned} T_2 = {{\,\mathrm{Re}\,}}\langle {({{\,\mathrm{Im}\,}}P) u, G^*Gu}\rangle = \langle {({{\,\mathrm{Im}\,}}P)Gu, Gu}\rangle + {{\,\mathrm{Re}\,}}\langle {[G, {{\,\mathrm{Im}\,}}P]u, Gu}\rangle . \end{aligned}$$We estimate the two terms on the right hand side separately. Firstly
$$\begin{aligned} |\langle {({{\,\mathrm{Im}\,}}P)Gu, Gu}\rangle | = h |\langle {(h^{-1}{{\,\mathrm{Im}\,}}P)}Gu, Gu\rangle | \le C_2 h \Vert Gu\Vert _{L^2}^2 + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}^2, \end{aligned}$$where we used the boundedness property (item 6) for the exotic operator \(h^{-1} {{\,\mathrm{Im}\,}}P\). For the second term, we need to deal with a commutator. Note first that
$$\begin{aligned} {{\,\mathrm{Re}\,}}G^*[G, {{\,\mathrm{Im}\,}}P] = h {{\,\mathrm{Re}\,}}G^* [G, h^{-1} {{\,\mathrm{Im}\,}}P] \in h^2 \Psi ^{2s-1}_{h, \rho , k-1}({{\,\mathrm{\mathcal {M}}\,}}), \end{aligned}$$by item 3. above. As \(\mathrm{WF}_h(G^*[G, {{\,\mathrm{Im}\,}}P]) \subset {{\,\mathrm{ell}\,}}_h(B_1)\), by the elliptic estimate
$$\begin{aligned} |\langle {{{\,\mathrm{Re}\,}}(G^*[G, {{\,\mathrm{Im}\,}}P]) u, u}\rangle |&= h^2|\langle {h^{-2}{{\,\mathrm{Re}\,}}(G^*[G, {{\,\mathrm{Im}\,}}P]) u, Yu}\rangle | + \mathcal {O}(h^\infty )\Vert u\Vert _{H_h^{-N}}^2\\&\le Ch^2 \Vert B_1 u\Vert ^2_{H_h^{s - \frac{1}{2}}} + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}^2, \end{aligned}$$where \(Y \in \Psi _h^0({{\,\mathrm{\mathcal {M}}\,}})\) is such that \(Y = 1 + \mathcal {O}(h^\infty )\) microlocally on \(\mathrm{WF}_h({{\,\mathrm{Re}\,}}(G^*[G, {{\,\mathrm{Im}\,}}P]))\) and \(\mathrm{WF}_h(Y) \subset {{\,\mathrm{ell}\,}}_h(B_1)\). Here we used the boundedness in the exotic class item 6., item 1., the elliptic estimate in the exotic class (item 7.). Adding the two estimates we finally obtain
$$\begin{aligned} |{{\,\mathrm{Re}\,}}\langle {({{\,\mathrm{Im}\,}}P)u, G^*Gu}\rangle | \le C_2h \Vert Gu\Vert _{L^2}^2 + Ch^2\Vert B_1 u\Vert _{H_h^{s - \frac{1}{2}}}^2 + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}^2. \end{aligned}$$Step 4. Adding the estimates in Steps 3. and 4., we obtain
$$\begin{aligned} {{\,\mathrm{Im}\,}}\langle {f, G^*Gu}\rangle \le (C_1 + C_2 - \beta )h \Vert Gu\Vert _{L^2}^2 + Ch\Vert Bu\Vert _{H_h^s}^2 + Ch^2 \Vert B_1u\Vert _{H_h^{s - \frac{1}{2}}}^2 + \mathcal {O}(h^\infty ) \Vert u\Vert ^2_{H_h^{-N}}. \end{aligned}$$By ellipticity of B on \(\mathrm{WF}_h(G)\), there is \(Q \in \Psi _h^{s}({{\,\mathrm{\mathcal {M}}\,}})\) such that \(G = QB_1 + R\), where \(R \in h^\infty \Psi _h^{-\infty }({{\,\mathrm{\mathcal {M}}\,}})\), thus
$$\begin{aligned} |\langle {f, G^*Gu}\rangle | \le |\langle {QB_1f, Gu}\rangle | + |\langle {Rf, Gu}\rangle | \le C\Vert B_1f\Vert _{H_h^s} \Vert Gu\Vert _{L^2} + \mathcal {O}(h^\infty ) \Vert u\Vert ^2_{H_h^{-N}}. \end{aligned}$$Now choose \(\beta = C_1 + C_2 + 1\) to get
$$\begin{aligned} \Vert Gu\Vert _{L^2}^2 \le C\Vert Bu\Vert ^2_{H_h^s} + Ch^{-1} \Vert B_1f\Vert _{H_h^s} \Vert Gu\Vert _{L^2} + Ch\Vert B_1u\Vert ^2_{H_h^{s-\frac{1}{2}}} + \mathcal {O}(h^\infty )\Vert u\Vert _{H_h^{-N}}^2. \end{aligned}$$We can absorb the \(\Vert Gu\Vert _{L^2}\) term to the left hand side, at the cost of the additional term \(Ch^{-2} \Vert B_1f\Vert _{H_h^s}^2\) on the right hand side. Next, use the condition \(\mathrm{WF}_h(A) \subset {{\,\mathrm{ell}\,}}_h(G)\) and elliptic estimates to derive
$$\begin{aligned} \Vert Au\Vert _{H^s_h} \le C \Vert Bu\Vert _{H_h^s} + Ch^{-1} \Vert B_1f\Vert _{H_h^s} + Ch^{\frac{1}{2}} \Vert B_1 u\Vert _{H_h^{s - \frac{1}{2}}} + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}.\nonumber \\ \end{aligned}$$(A.5)Step 5. Here, one can use the same induction procedure as in [DyZw, Proof of Th. E.47] to show that for each \(\ell \in \mathbb {N}\)
$$\begin{aligned} \Vert Au\Vert _{H^s_h} \le C \Vert Bu\Vert _{H_h^s} + Ch^{-1} \Vert B_1f\Vert _{H_h^s} + Ch^{\frac{\ell }{2}} \Vert B_1 u\Vert _{H_h^{s - \frac{\ell }{2}}} + \mathcal {O}(h^\infty ) \Vert u\Vert _{H_h^{-N}}, \end{aligned}$$where the first step of the induction is exactly where we arrived in (A.5). \(\square \)
Next we discuss semiclassical defect measures in the setting of the exotic calculus.
Proposition A.2
Assume that \(u_h\in L^2({{\,\mathrm{\mathcal {M}}\,}})\) is a family satisfying \(\Vert u_h\Vert _{L^2} = \mathcal {O}(1)\). Then there exists a Radon measure \(\mu \), called semiclassical measure, and a sequence \(h_j \rightarrow 0\), such that for any \(A \in \Psi ^{{{\,\mathrm{comp}\,}}}_{h, \rho , 0}({{\,\mathrm{\mathcal {M}}\,}})\) with \(\lim _{h\rightarrow 0} \sigma (A)(h;x,\xi ) = a_0(x,\xi )\) in \(C_c^0(T^*{{\,\mathrm{\mathcal {M}}\,}})\),
Proof
We follow the proof of [DyZw, Theorem E.42]. By Proposition A.1, we have that \(A-\mathrm{Op}_h(a_h)\in h\Psi ^{\mathrm{comp}}_{h,\rho ,0}(\mathcal {M})\) for some symbol \(a_h\in S_{h,\rho ,0}^{\mathrm{comp}}(T^*\mathcal {M})\) so that \(a_{h}\rightarrow a_0\) in \(C_c^0(T^*{{\,\mathrm{\mathcal {M}}\,}})\); it suffices to prove the claim for \(A = {{\,\mathrm {Op}\,}}_{h_j}(a_{h_j})\). We write \(I_h(a_h) := \langle {\mathrm{Op}(a_h) u_h, u_h}\rangle _{L^2}\) and claim
Indeed, by Cauchy-Schwarz and Proposition A.1 (item 6), we have
where \(C = C(\Vert u_h\Vert _{L^2}) > 0\). Take a countable, dense subset \((a_h^{\ell })_{\ell \in \mathbb {N}} \subset S^{{{\,\mathrm{comp}\,}}}_{h, \rho , 0}(T^*\mathcal {M})\), where \(S^{{{\,\mathrm{comp}\,}}}_{h, \rho , 0}(T^*\mathcal {M})\) is equipped with the inductive limit topology from the seminorms in (A.1). By a diagonal argument and since by (A.7) \(I_h(a_h^\ell )\) is bounded for all \(\ell \), we may extract a sequence \(h_j \rightarrow 0\) such that \(I_{h_j}(a_{h_j}^\ell )\) converges for all \(\ell \). For each \(a_h \in S^{{{\,\mathrm{comp}\,}}}_{h, \rho , 0}(T^*\mathcal {M})\) and \(\ell \), we get by (A.7)
Using the density of \(a^\ell _h\), we obtain that \(I_{h_j}(a_{h_j})\) is a Cauchy sequence and we may define for \(a_h\in S^{{{\,\mathrm{comp}\,}}}_{h, \rho , 0}(T^*\mathcal {M})\)
By (A.7), the map I satisfies for each \(a_h\in S^{\mathrm{comp}}_{h,\rho ,0}(T^*\mathcal {M})\)
for any \(h_0' > 0\). In particular, I restricts to a continuous linear functional on \(C_c^0(T^*\mathcal {M})\), i.e. on h-independent functions. Given \(a_h \in S^{\mathrm{comp}}_{h,\rho ,0}(T^*\mathcal {M})\) with \(\lim _{h\rightarrow 0} a_{h}=a_0 \in C_c^\infty (T^*{{\,\mathrm{\mathcal {M}}\,}})\) in the \(C_c^0(T^*\mathcal {M})\) topology, we get by (A.8)
and thus \(I(a_h)=I(a_0)\). By [DyZw, Theorem E.42] there is a Radon measure \(\mu \) such that for each \(a_0 \in C_c^\infty (T^*\mathcal {M})\)
The main claim follows from this, by using \(I(a_h) = I(a_0)\) under the given assumptions. \(\square \)
Now we prove a version of a propagation estimate for the semiclassical measure in the exotic calculus. This is a slight extension of [DyZw, Theorem E.44]. Note that if \(\sigma (P)\) is real valued and \(k\ge 1\), then \({{\,\mathrm{Im}\,}}P \in h\Psi _{h, \rho , k-1}^{m-1}(\mathcal {M})\) if \(P \in \Psi _{h, \rho , k}^m(\mathcal {M})\) by Proposition A.1 (item 5).
Proposition A.3
Assume \(\Vert u_h\Vert _{L^2} = \mathcal {O}(1)\) and \(u_h\) converges to a semiclassical measure \(\mu \). Let \(P \in \Psi ^m_{h, \rho , k}({{\,\mathrm{\mathcal {M}}\,}})\) with \(k\ge 1\), denote \(p:= \sigma (P)\) and assume that p is real-valued for all h, and define \(b := \sigma (h^{-1}{{\,\mathrm{Im}\,}}P)\). Assume that for each \(a \in C_c^\infty (T^*{{\,\mathrm{\mathcal {M}}\,}})\),
Then there is \(C>0\) such that for all \(a \in C_c^\infty (T^*{{\,\mathrm{\mathcal {M}}\,}})\) and \(Y \in \Psi _h^{{{\,\mathrm{comp}\,}}}({{\,\mathrm{\mathcal {M}}\,}})\) with \(Y = 1 + \mathcal {O}(h^\infty )\) microlocally on \({{\,\mathrm{supp}\,}}(a)\),
Proof
Assume without loss of generality that a is real valued. Let \(A \in \Psi _h^{{{\,\mathrm{comp}\,}}}(M)\) be such that \(\sigma (A) = a\) and \(A^* = A\). We compute
Now by Proposition A.1 (item 3), we have \(h^{-1}[A, P] \in \Psi ^{{{\,\mathrm{comp}\,}}}_{h, \rho , k-1}({{\,\mathrm{\mathcal {M}}\,}})\) with \(-i\sigma (h^{- 1}[A, P]) = H_p a\) and by assumptions \(\sigma ((h^{-1}{{\,\mathrm{Im}\,}}P) A) = ba\). Thus by Proposition A.2 the right hand side converges to, after possibly extracting a subsequence \(h_j \rightarrow 0\)
Moreover, the left hand side equals
from which we easily deduce the main estimate. \(\quad \square \)
Rights and permissions
About this article
Cite this article
Cekić, M., Guillarmou, C. First Band of Ruelle Resonances for Contact Anosov Flows in Dimension 3. Commun. Math. Phys. 386, 1289–1318 (2021). https://doi.org/10.1007/s00220-021-04090-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-021-04090-2