Abstract
For \(f_t(x)= t-x^2\) the quadratic family, we define the fractional susceptibility function \(\Psi ^\Omega _{\phi ,t_0}(\eta , z)\) of \(f_t\), associated to a \(C^1\) observable \(\phi \) at a stochastic parameter \({t_0}\). We also define an approximate, “frozen,” fractional susceptibility function \(\Psi ^\mathrm {fr}_{\phi ,t_0}(\eta , z)\) such that \(\lim _{\eta \rightarrow 1} \Psi ^{\mathrm {fr}}_{\phi ,t_0}(\eta , z)\) is the susceptibility function \(\Psi _{\phi ,t_0}(z)\) studied by Ruelle. If \({t_0}\) is Misiurewicz–Thurston, we show that \(\Psi ^\mathrm {fr}_{\phi ,t_0}(1/2, z)\) has a pole at \(z=1\) for generic \(\phi \) if \({\mathcal {J}}_{1/2}(t_0)\ne 0\), where \({\mathcal {J}}_\eta (t)=\sum _{k=0}^\infty \mathrm {sgn}(Df^k_{t}(c_1))|D f^k_{t}(c_1)|^{-\eta }\), with \(c_1=t\) the critical value of \(f_{t}\). We introduce “Whitney” fractional integrals \(I^{\eta ,\Omega }\) and derivatives \(M^{\eta ,\Omega }\) on suitable sets \(\Omega \). We formulate conjectures on \(\Psi ^\Omega _{\phi ,t_0}(\eta , z)\) and \({\mathcal {J}}_\eta (t)\), supported by our results on \(M^{\eta ,\Omega }\) and \(\Psi ^{\mathrm {fr}}_{\phi ,t_0}(1/2, z)\), for the former, and numerical experiments, for the latter. In particular, we expect that \(\Psi ^\Omega _{\phi ,t_0}(1/2, z)\) is singular at \(z=1\) for Collet–Eckmann \({t_0}\) and generic \(\phi \). We view this work as a step towards the resolution of the paradox that \(\Psi _{\phi ,t_0}(z)\) is holomorphic at \(z=1\) for Misiurewicz–Thurston \(f_{t_0}\) (Jiang and Ruelle in Nonlinearity 18:2447–2453, 2005, Ruelle in Commun Math Phys 258:445–453, 2005), despite lack of linear response (Baladi et al. in Invent Math 201:773–844, 2015).
Similar content being viewed by others
Notes
The map for \(t=2\) is the full parabola \(2-x^2\) on \([-2, 2]\), which can only be perturbed by taking \(t<2\). For \(t=1\), we get a half-parabola on [0, 1].
As a Radon measure, say—using distributions of higher order does not help.
Another goal is to give a probabilistic analysis (analogous to the central limit theorem of de Lima–Smania [12] in the piecewise expanding setting) of the breakdown of \(C^{1/2}\) regularity of the acim in transversal families of smooth unimodal maps with a quadratic critical point.
It is natural that the half derivative of \({\mathbf {1}}_{x>c_k}(x-c_k)^{-1/2}\) involves \((x-c_k)^{-1}\), but we found no good reference for the computation.
Recall that \(\mathrm {supp}(\rho _t)=[c_{2,t}, c_{1,t}]\).
The threshold for \(\eta \) below is 1/2; for families with criticality d the expected threshold is 1/d.
Some results of [8] require polynomial recurrence. We expect that this is an artefact of the method used there, but maybe TSR must be strengthened to polynomial recurrence.
All discs in the present work are centered at the origin.
Recalling (6), the function \(x\mapsto M^\eta _s ({\mathcal {L}}_s \rho _{t}(x))|_{s=t}\) is supported in \(I_{t,\epsilon }\subset I_t\).
Use that \(\phi \) and \(\rho _{t}\) are compactly supported while, on the one hand, we have \(M^\eta _x (\phi \circ f_{t}^k)\in L^{p}_{loc}\) for all \(p\ge 1\), while \(\phi \circ f_{t}^k\in L^s\) for all \(s\ge 1\), and, on the other hand, we have \(M^\eta (\rho _{t}) \in L^r_{loc}\) for [39] any \(1\le r <2(1+2\eta )^{-1}\), while \(\rho _{t}\in L^{{\tilde{r}}}\) for all \(1\le {\tilde{r}} <2\).
The proof shows that the proposition holds more generally, for example for mixing TSR parameters.
If g is bounded and differentiable to the left at x, the limit as \(\eta \uparrow 1\) of \(M^\eta _{+}(g)(x)\) is equal to the left-sided derivative \(g'_-(x)\), the notation is thus confusing.
This is an advantage of Marchaud derivatives over Riemann–Liouville fractional derivatives.
One could weaken this condition, up to exchanging the limit and the derivative in (35). We shall not need this more general notion.
In particular we have shown that \(M^{1/2}_+(\phi _{x_0,+})= d I^{1/2}_+(\phi _{x_0,+})\), as expected, see Remark 4.3.
In particular, \(M^{1/2}_-(\phi _{x_0,-})= -d I^{1/2}_-(\phi _{x_0,-})\), as expected, see Remark 4.3.
In particular, \(M^{1/2}_-(\phi _{x_0,+,{\mathcal {Z}}})= -d I^{1/2}_-(\phi _{x_0,+,{\mathcal {Z}}})\), as expected.
The cutoff is slightly different in Ruelle [36, Theorem 9, Remark 16A], who observes that “other choices can be useful.”
The reference to Lemma 2.4 there should be replaced by Lemma 2.5.
It would probably be possible to apply [47, Thm 2.II.b)] instead.
If \(\mathrm {supp}(\psi )\subset I_t\), the first term is \(\int \varphi ( x) {\mathcal {L}}^j_t( \psi (x))\, \mathrm {d}x\), thus the name of the lemma.
This was already established in Lemma 5.2.
If \(\mathrm {sgn}(Df^P(c_L))=-1\), then, clearly, \({\mathcal {P}}^+_t(1)={\mathcal {P}}_t^{{\tilde{\psi }}}(1)=0\).
With respect to [9] the term \({\mathcal {W}}_{\phi , 1/2}(z)\) and the presence of the Hilbert transform are new.
The improper integral is well-defined and finite, since \(\phi \) is \(C^1\) and \([-a_t, c_1]\) contains a neighbourhood of each \(c_{\ell }\).
We identify \(\chi _k=\chi (\mathbf {Y}_k)\) with \(\mathbf {Y}_k\) in (61).
The topological entropy is the logarithm of the slope and thus constant, so there are no bifurcations. It is illuminating to construct explicitly the corresponding topological conjugacy.
A unimodal map f is called pre-Chebyshev if f is exactly m times renormalisable, for some \(m \ge 0\), each renormalisation being of period two, and, in addition, if J is the restrictive interval for the mth renormalisation, \(f^{2^m}|_J : J \rightarrow J\) is smoothly conjugate on J to \(x \mapsto 1 - 2|x|\) on \((-1, 1)\).
In particular, this deepest renormalisation is topologically conjugated to \(f_2\).
See [41, Lemma 2.1] for an analogous remark.
This is an exercise left to the reader.
In this case, the eigenvalue 1 has geometric multiplicity two, i.e. there is no Jordan block.
References
Abel, N.H.: Solution de quelques problèmes à l’aide d’intégrales définies, In Gesammelte mathematische Werke. Leipzig: Teubner, 1, 11–27 (1881). (First publ. in Mag. Naturvidenkaberne, 1, Christiania, 1823)
Abel, N.H.: Auflösung einer mechanischen Aufgabe. J. reine und angew. Mathematik 1, 153–157 (1826)
Aspenberg, M., Baladi, V., Leppänen, J., Persson, T.: On the fractional susceptibility function of piecewise expanding maps. arXiv:1910.00369
Avila, A.: Infinitesimal perturbations of rational maps. Nonlinearity 15, 695–704 (2002)
Avila, A., Lyubich, M., de Melo, W.: Regular or stochastic dynamics in real analytic families of unimodal maps. Invent. Math. 154, 451–550 (2003)
Avila, A., Moreira, C.G.: Statistical properties of unimodal maps: smooth families with negative Schwarzian derivative. Astérisque 286, 81–118 (2003)
Baladi, V.: On the susceptibility function of piecewise expanding interval maps. Commun. Math. Phys. 275, 839–859 (2007)
Baladi, V., Benedicks, M., Schnellmann, D.: Whitney Hölder continuity of the SRB measure for transversal families of smooth unimodal maps. Invent. Math. 201, 773–844 (2015)
Baladi, V., Marmi, S., Sauzin, D.: Natural boundary for the susceptibility function of generic piecewise expanding unimodal maps. Ergodic Theory Dyn. Syst. 10, 1–24 (2013)
Baladi, V., Smania, D.: Linear response formula for piecewise expanding unimodal maps, Nonlinearity 21 (2008) 677–711 (Corrigendum. Nonlinearity 25, 2203–2205 (2012)
Baladi, V., Smania, D.: Linear response for smooth deformations of generic nonuniformly hyperbolic unimodal maps. Ann. Sc. ENS 45, 861–926 (2012)
de Lima, A., Smania, D.: Central limit theorem for the modulus of continuity of averages of observables on transversal families of piecewise expanding unimodal maps. J. Institut Math. Jussieu 17, 673–733 (2018)
Dobbs, N.: Visible measures of maximal entropy in dimension one. Bull. Lond. Math. Soc. 39, 366–376 (2007)
Dobbs, N., Mihalache, N.: Diabolical entropy. Commun. Math. Phys. 365, 1091–1123 (2019)
Dobbs, N., Todd, M.: Free energy and equilibrium states for families of interval maps. arXiv:1512.09245, to appear Memoirs A.M.S
Douady, A., Hubbard, J.H.: On the dynamics of polynomial-like mappings. Ann. Sci. École Norm. Sup. 18, 287–343 (1985)
Dwight, H.B.: Tables of Integrals and Other Mathematical Data, 3rd edn. McMillan, New York (1957)
Hilfer, R.: Threefold introduction to fractional derivatives. In: Klages, R., Radons, G., Sokolov, I.M. (eds.) Anomalous Transport: Foundations and Applications. Wiley, New York (2008)
Jiang, Y., Ruelle, D.: Analyticity of the susceptibility function for unimodal Markovian maps of the interval. Nonlinearity 18, 2447–2453 (2005)
Keller, G., Nowicki, T.: Spectral theory, zeta functions and the distribution of periodic points for Collet–Eckmann maps. Commun. Math. Phys. 149, 31–69 (1992)
Kilbas, A.A., Srivastava, H.M., Trujillo, J.J.: Theory and Applications of Fractional Differential Equations. Elsevier, Amsterdam (2006)
Levin, G.: On an analytic approach to the Fatou conjecture. Fund. Math. 171, 177–196 (2002)
Lieb, E.H., Loss, M.: Analysis. Graduate Studies in Mathematics, 2nd edn, vol. 14. American Mathematical Society, Providence (2001)
Lyubich, M.: Almost every real quadratic map is either regular or stochastic. Ann. Math. 156, 1–78 (2002)
Martens, M., Nowicki, T.: Invariant measures for typical quadratic maps, in: Géométrie complexe et systèmes dynamiques (Orsay, 1995). Astérisque 261 (2000), xiii, 239–252
Miller, K.S., Ross, B.: An Introduction to the Fractional Calculus and Fractional Differential Equations. Wiley, New York (1993)
Milnor, J., Thurston, W.: On iterated maps of the interval, Dynamical systems (College Park, MD, 1986–87). Lecture Notes in Math. 1342, 465–563 (2018)
Misiurewicz, M.: Absolutely continuous measures for certain maps of an interval. Inst. Hautes Études Sci. Publ. Math. 53, 17–51 (1981)
Nowicki, T., van Strien, S.: Absolutely continuous invariant measures for C2 unimodal maps satisfying the Collet-Eckmann conditions. Invent. Math. 93, 619–635 (1988)
Ognev, A.I.: Metric properties of a certain class of mappings of a segment. Mat. Zametki 30, 723–736, 797 (1981)
Ortigueira, M.D., Machado, J.A.: What is a fractional derivative? J. Comput. Phys. 293, 4–13 (2015)
Podlubny, I.: Fractional Differential Equations. Academic Press, SanDiego (1998)
Ragazzo, C.: Scalar autonomous second order ordinary differential equations. Qual. Theory Dyn. Syst. 11, 277–415 (2012)
Ruelle, D.: General linear response formula in statistical mechanics, and the fluctuation-dissipation theorem far from equilibrium. Phys. Lett. A 245, 220–224 (1998)
Ruelle, D.: Differentiating the absolutely continuous invariant measure of an interval map f with respect to f. Commun. Math. Phys. 258, 445–453 (2005)
Ruelle, D.: Structure and f-dependence of the A.C.I.M. for a unimodal map f of Misiurewicz type. Commun. Math. Phys. 287, 1039–1070 (2009)
Samko, S.G., Kilbas, A.A., Marichev, O.I.: Fractional Integrals and Derivatives. Theory and Applications. Gordon and Breach Science Publishers, London (1993)
Schnellmann, D.: Positive Lyapunov exponents for quadratic skew-products over a Misiurewicz-Thurston map. Nonlinearity 22, 2681–2695 (2009)
Sedro, J.: Pre-threshold fractional susceptibility functions at Misiurewicz parameters (2020) arXiv:2011.13648
Thunberg, H.: Unfolding of chaotic unimodal maps and the parameter dependence of natural measures. Nonlinearity 14, 323–337 (2001)
Tsujii, M.: Positive Lyapunov exponents in families of one-dimensional dynamical systems. Invent. Math. 111, 113–137 (1993)
Tsujii, M.: On continuity of Bowen–Ruelle–Sinai measures in families of one-dimensional maps. Commun. Math. Phys. 177, 1–11 (1996)
Tsujii, M.: A simple proof for monotonicity of entropy in the quadratic family. Ergodic Theory Dyn. Syst. 20, 925–933 (2000)
Wormell, C.L., Gottwald, G.A.: On the validity of linear response theory in high-dimensional deterministic dynamical systems. J. Stat. Phys. 172, 1479–1498 (2018)
Wormell, C.L., Gottwald, G.A.: Linear response for macroscopic observables in high-dimensional systems. Chaos 29, 113127 (2019)
Young, L.S.: Decay of correlations for certain quadratic maps. Commun. Math. Phys. 146, 123–138 (1992)
Young, L.S.: Recurrence times and rates of mixing. Israel J. Math. 110, 153–188 (1999)
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.
DS was partially supported by CNPq 306622/2019-0, CNPq 430351/2018-6 and FAPESP Projeto Temático 2017/06463-3. We are grateful to the Brazilian-French Network in Mathematics for supporting VB’s visit to DS in São Carlos in 2015 and DS’s visit to VB in Paris in 2017. This work was started when VB was working at IMJ-PRG. VB is grateful to the Knut and Alice Wallenberg Foundation for invitations to Lund University in 2018, 2019, and 2020. The visit of VB to São Carlos in 2019 was supported by FAPESP Projeto Teḿatico 2017/06463-3. The visits of DS to Paris in 2018 and 2019 and VB’s research are supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No 787304). We thank Magnus Aspenberg, Genadi Levin, Tomas Persson, and Julien Sedro for useful comments. We are very grateful to Clodoaldo Ragazzo for pointing out to us that Abel was the first to notice Lemma 3.1.
Appendices
Appendix A. Proof of Lemma 3.2. (Abel’s remark, two-sided)
Recall that if \(|x|<1\) then
(In particular \(b_1=1/2\) and \(b_2=-1/8\).) We shall also use the fact [17, 195.01] that for any real numbers \(a>\tau \) and \(b>\tau \)
Proof of Lemma 3.2
Set \(x_0=x-c_k\). To prove the claim on \(I^{1/2}_{+,t} (\phi _{c_k,+,{\mathcal {Z}}})(x,t)\), it is enough to show that, for any \(c_k+t-{\mathcal {Z}}<x<c_k+t+{\mathcal {Z}}\),
where, for \(y \in (-{\mathcal {Z}}/2,{\mathcal {Z}}/2)\),
Indeed, using (68), we have (the power series below are absolutely convergent)
In particular, \(\lim _{{\mathcal {Z}}\rightarrow \infty }\sup _{ y\in (-{\mathcal {Z}}/2,{\mathcal {Z}}/2)} | \partial _y G_{\mathcal {Z}}(y)|=0\), and
We proceed to show (70). From the definition (26) of \(I^{1/2}_{+,t}\), we get
Recalling that \(x_0=x-c_k\), if \(x< c_k+t\), then we find
(This term did not appear in Lemma 3.1.) Using (69), we find
If in addition \(c_k+t-{\mathcal {Z}}<x\), then, by (68) we find
and we have shown that if \(c_k+t-{\mathcal {Z}}<x<c_k+t\) then
We now consider the case \(c_k+t<x<c_k+t+{\mathcal {Z}}\). Then
Using again (69), and we find
Similarly as for (71), we find, using (68),
We have thus shown that if \(c_k+t-{\mathcal {Z}}<x<c_k+t+{\mathcal {Z}}\) then
With (72), the above identity shows the claim on the right-handed spike (\(\sigma =+\)).
For the left-handed spike, we have, recalling \(x_0=x-c_k\) and (27),
Thus, we find
Finally, note that \(-(x-c_k-t)=x-c_k-(2x_0-t)\). \(\quad \square \)
Appendix B. Vanishing of \(X_t\) at the image of endpoints
It is sometimes convenient to assume that \(X_t\) vanishes at the endpoints \(\pm 1\). This can be achieved in several ways, as we explain now. For \(t\in (1,2)\), setting
gives a family of maps \({\tilde{f}}_t\) preserving \([-1,1]\), with \(c_{0,t}=0\), and such that
so that \(\partial _t {\tilde{f}}_t\) vanishes at the endpoints \(-1\) and 1. This is a transversal family of quadratic maps in the sense of Tsujii [41], or [5, 6]. The formula for \(\partial {\tilde{f}}_t\) being unwieldy, it is convenient to work with the family \({\tilde{h}}_t:[0,1]\rightarrow [0,1]\) defined by \({\tilde{h}}_t(x)=t x(1-x)\), \(t \in (1,4]\). The critical point of each \({\tilde{h}}_t\) is 1/2, and \(X_t(x)=\partial _t {\tilde{h}}_t \circ {\tilde{h}}_t^{-1}=tx\) (there is a typo in [8, eq. (2)] where it is stated incorrectly that \(X^{{\tilde{h}}_t}_t(x)\equiv t\)). Then \({\tilde{h}}_t(0)={\tilde{h}}_t(1)=0\) for all t, so that \(\partial _t {\tilde{h}}_t\) vanishes at the endpoints 0 and 1. A variant of \({\tilde{h}}_t\) is \(h_t(x)=t(1-x^2)-1\) for \(t\in (1,2]\) on \([-1, 1]\) (there, \(c=0\) and \(h_t(-1)=h_t(1)=-1\)). However, the formulas for \(f_t\) are easier to manipulate than those of \({\tilde{f}}_t\), \(h_t\), or \({\tilde{h}}_t\), compensating for the non vanishing of the vector field at the endpoints of a common invariant interval. In addition,Footnote 38 for any fixed \(t_0 \in (1,2)\) and all t close enough to \(t_0\), we may extend \(f_t\) on \([-2, 2]\) to a \(C^4\) map, also called \(f_t\), with negative Schwarzian derivative, such that \(Df_t\) is positive on \([-2, t-t^2]\) and negative on [t, 2], with \(f_t(-2)=f_t(2)=-2\) and \(f_t - f_{t_0}= O(|t-t_0|)\). (The extended family \(f_t\) is not needed in the present paper, but we expect it should be useful to prove equality [ii] in Conjecture A in future works.) Finally, using that \(c_{2,t}=t-t^2>-|a_t|\), one can easily show that for any \(t_0\in (1,2)\) there are \(\epsilon >0\) and an interval \(I'_{t_0}\subset (-2,2)\) such that \(f^k_t (I'_{t_0})\subset I'_{t_0}\) for all \(t \in (t_0-\epsilon , t_0+\epsilon )\). In other words, \(f_t\) for \(t \in (t_0-\epsilon , t_0+\epsilon )\) is a transversal family of unimodal maps on \(I'_{t_0}\) in the sense of Tsujii [41] since, recalling (10), if \({\mathcal {J}}({t_1})\) is absolutely convergent then \({\mathcal {J}}({t_1})\ne 0\).
Appendix C. Averaging
For regular parameters t (also called “hyperbolic”), the physical measure \(\mu _t^{sink}= P^{-1} \sum _{j=1}^P \delta _{x_{t,j}}\) is atomic, supported on an attracting periodic orbit \(f^{P} (x_{1,t})=x_{1,t}\) with \(P=P(t)\), and can be obtained as
The convergence is however not uniform in any interval of regular parameters so one cannot a priori sum over k even if \(\int _{I_t}\psi \, \mathrm {d}m =0\).
Since almost every parameter is either regular or stochastic [24] it is natural to consider, for a \(C^1\) observable, say, a Collet–Eckmann parameter t, and \(\epsilon >0\), the double Lebesgue integral
where \(\mu _{t+\delta }=\mu _{t+\delta }^{sink}\) for regular parameters, and \(\mu _{t+\delta }=\rho _{t+\delta } dm\) is the SRB measure for stochastic parameters. Then it is not hard to see, using Lebesgue differentiation, that the (ordinary) t-derivative \({\mathcal {A}}_\epsilon '(t)\) exists for almost every \(\epsilon >0\) and coincides with \(\int \phi (x) \mathrm {d}\mu _{t+\epsilon }(x)\) (see also [44, §3] and [45, (16)]). This does not resolve the paradox described in the introduction, since the derivative depends on \(\epsilon \) and does not coincide with \(\Psi _\phi (1)\) in general. (Note that the “weakening of the linear response problem” in the introduction of [8]—existence and value of the limit as \(\epsilon \rightarrow 0\) of the derivative \({\mathcal {A}}'_\epsilon \)—does not explain the paradox either.)
Appendix D. Complements to the proof of Theorem C
We record here interesting facts which are not needed for our proofs.
Remark D.1
(Spectrum on a pole-extended Banach space) For \(t \in \mathrm {MT}\), let \(\Lambda =\Lambda _t:[-a_t,a_t]\rightarrow [0,1]\) be the absolutely continuous bijection defined by \(\Lambda _t(x)=\int _{-a_t}^x \rho _t(u) \mathrm {d}u\). Then (see [30, 38]) the map \(F_t:[0,1]\rightarrow [0,1]\) defined by \(F_t (\Lambda _t(x))=\Lambda _t (f_t(x))\) is Markov (for the partition \(J_\ell \) defined by the endpoints \(\Lambda _t(c_k)\), \(k=0, \ldots L+P-1\)), and \(F_t\) is \(C^1\) on the interior of each interval of monotonicity \(J_\ell \), with \(\inf | F_t'|>1\). At the endpoints, we haveFootnote 39
(taking right or left-sided limits in the left-hand side according to the dynamical orbit). In fact, \(G_t:=1/F_t'\) extends to a \(C^1\) map on the closure of each \(J_\ell \), with \(\sup G''_t<\infty \). On the Banach space \({\mathcal {B}}_{\Lambda _t}\) of bounded functions \(\phi \) on [0, 1] such that each \(\phi |_{\mathrm {int}J_\ell }\) is \(C^1\) and admits a \(C^1\) extension to the closure of \(J_\ell \), the transfer operator \({\mathcal {L}}^\Lambda _t \phi (y)= \sum _{F_t(z)=y} \phi (z)/|F'_t(z)|\) thus has spectral radius equal to one, with a simple eigenvalue at 1, for the eigenvector \(\rho ^\Lambda _t(y):=\rho _t(\Lambda _t^{-1}(y))\), and the rest of the spectrum is contained in a disc of radius \(\kappa \) strictly smaller than 1. Then \({\mathcal {B}}_t=\{ \phi \circ \Lambda _t , \, \phi \in {\mathcal {B}}_{\Lambda _t}\}\) is a Banach space for the norm induced by \({\mathcal {B}}_{t,\Lambda }\) and the operator \({\mathcal {L}}_t\) on \({\mathcal {B}}_t\) inherits the spectral properties of \({\mathcal {L}}^\Lambda _t\) on \({\mathcal {B}}_{\Lambda _t}\). Any element of \({\mathcal {B}}_t\) belongs to \(L^1(dm)\), with \(\int _{I_t} |\varphi |\, \mathrm {d}m \le \Vert \varphi \Vert _{{\mathcal {B}}_t}\). Recall the notations \({\mathcal {Y}}_t\), \(\chi _k\), \(\chi (\mathbf {Y})\), and \({\mathcal {M}}_t\) from Remark 5.7. Then we claim that we may extend \({\mathcal {L}}_t:{\mathcal {B}}_t\rightarrow {\mathcal {B}}_t\) to a bounded operator \({\mathbb {L}}_t\) on the Banach space \({\mathbb {B}}_t:={\mathcal {B}}_t\oplus {\mathcal {Y}}_t\), whose nonzero spectrum is the union of the Pth roots of \(\mathrm {sgn}( Df^P_t (c_{L}))\) with the nonzero spectrum of \({\mathcal {L}}_t\) on \({\mathcal {B}}_t\). Morever the following holds if \(\mathrm {sgn}(Df^p(c_L))=+1\): First, setting \({\mathcal {M}}^0_t(\mathbf {Y})={\mathcal {M}}_t(\mathbf {Y})-\rho _t \int _{I_t} {\mathcal {M}}_t(\mathbf {Y}) \mathrm {d}m\), and letting \(\mathbf {S}_t\) be the fixed point of \({\mathbb {S}}_t\),
and the (rank-two) spectral projector \( \Pi _1\) for the eigenvalue 1 of \({\mathbb {L}}_t\) satisfies
Second, if \(\int _{I_t} {\mathcal {M}}_t(\mathbf {S}_t) \, \mathrm {d}m=0\) then \(\psi ^*_t+\chi (\mathbf {S}_t) \in {\mathbb {B}}_t\) is a fixed pointFootnote 40 of \({\mathbb {L}}_t\), while if \(\int _{I_t} {\mathcal {M}}_t(\mathbf {S}_t) \, \mathrm {d}m\ne 0\) then there exists a non zero \(\nu ^N_t\in {\mathbb {B}}_t^*\) such that the (rank-one) nilpotent operator for the eigenvalue 1 of \({\mathbb {L}}_t\) satisfies \(N_1^2=0\) and \(\Pi _{{\mathcal {B}}_t}\circ N_1=\nu ^N_t \cdot \rho _t\).
We justify the claims above: First, there exists \(\kappa <1\) such that, on \({\mathcal {B}}_t\),
where for any \(\epsilon >0\) there exists C such that \(\Vert {\mathcal {Q}}_t^j\Vert _{{\mathcal {B}}_t}\le C (\kappa +\epsilon )^j\) for all \(j\ge 1\). Note that \({\mathcal {M}}_t(\mathbf {Y})=\Pi _{{\mathcal {B}}_t} \bigl ( {\mathbb {L}}_t(\chi (\mathbf {Y}))\bigr )\). Identifying \(\chi (\mathbf {Y})\) and \(\mathbf {Y}\), we have
In particular if 1/z does not belong to the spectrum of \({\mathcal {L}}_t\) or \({\mathbb {S}}_t\), then
If \(\int _{I_t} {\mathcal {M}}_t (\mathbf {S}_t) \, \mathrm {d}m=0\) (with \(\mathbf {S}_t\) the fixed point of \({\mathbb {S}}_t\)) then a direct computation gives that \({\mathbb {L}}_t\) inherits a (second) fixed point \(\psi ^*_t+\chi (\mathbf {S}_t) \in {\mathbb {B}}_t\) from the fixed point \(\mathbf {S}_t\) of \({\mathbb {S}}_t\). If \(\int {\mathcal {M}}_t(\mathbf {S}_t)\, \mathrm {d}m\ne 0\) then the eigenvalue 1 of \({\mathbb {L}}_t\) has algebraic multiplicity two but geometric multiplicity one, and the associated nilpotent \(N_1\) satisfies our claim. In both cases, the claim on \(\Pi _1\) follows.
Remark D.2
In the Collet–Eckmann case with an infinite postcritical orbit, the finite matrix \({\mathbb {S}}_t\) appearing in the proof of Theorem C will be replaced by a shift to the right, also denoted \({\mathbb {S}}_t\), weighted by \(s_{1,k}=\pm 1\), acting on a space of infinite sequences (for example \(\ell ^\infty ({\mathbb {Z}}_+)\)). Then \({\mathbb {S}}_t\) does not have any eigenvalues, and its spectrum is contained in the closed unit disc. Also, \({\mathbb {M}}_t(z):=(\mathrm {id}-z {\mathbb {S}}_t)^{-1}\) is the infinite matrix with \(({\mathbb {M}}_t(z))_{j,j}=1\), \(({\mathbb {M}}_t(z))_{\ell ,j}=(-1)^{1+\ell -j} z^{\ell -j} \prod _{k=\ell }^{j-1} s_{1,k} =(-1)^{1+\ell -j} z^{\ell -j} s_{j-1,\ell } \) for \(j<\ell \), and \(({\mathbb {M}}_t(z))_{\ell ,j}=0\) for other \(\ell \), j.
Rights and permissions
About this article
Cite this article
Baladi, V., Smania, D. Fractional Susceptibility Functions for the Quadratic Family: Misiurewicz–Thurston Parameters. Commun. Math. Phys. 385, 1957–2007 (2021). https://doi.org/10.1007/s00220-021-04015-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-021-04015-z