Irrationality and quasiperiodicity in driven nonlinear systems

Irrationality and Quasiperiodicity in Driven Nonlinear Systems David Cubero,1,* Jesús Casado-Pascual,2,†and Ferruccio Renzoni3,‡ 1Departamento de Física Aplicada I, EUP, Universidad de Sevilla, Calle Virgen de África 7, 41011 Sevilla, Spain 2Física Teórica, Universidad de Sevilla, Apartado de Correos 1065, 41080 Sevilla, Spain 3Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, United Kingdom (Received 30 December 2013; published 1 May 2014) We analyze the relationship between irrationality and quasiperiodicity in nonlinear driven systems. To that purpose, we consider a nonlinear system whose steady-state response is very sensitive to the periodic or quasiperiodic character of the input signal. In the infinite time limit, an input signal consisting of two incommensurate frequencies will be recognized by the system as quasiperiodic. We show that this is, in general, not true in the case of finite interaction times. An irrational ratio of the driving frequencies of the input signal is not sufficient for it to be recognized by the nonlinear system as quasiperiodic, resulting in observations which may differ by several orders of magnitude from the expected quasiperiodic behavior. Thus, the system response depends on the nature of the irrational ratio, as well as the observation time. We derive a condition for the input signal to be identified by the system as quasiperiodic. Such a condition also takes into account the sub-Fourier response of the nonlinear system. DOI: 10.1103/PhysRevLett.112.174102 PACS numbers: 05.45.-a, 05.40.-a, 05.60.Cd Introduction.—Periodic structures, in space and in time, are ubiquitous in all branches of science, and the characteristics of many systems can be traced back to their periodicity. Periodic systems can be defined by rational numbers. In the case of spatially periodic systems, the potential is the sum of harmonics with lattice constants in rational ratio, while, in the time-periodic case, the driving is made of harmonics with frequencies in rational ratio (commensurate frequencies). Quasiperiodic systems are obtained whenever the ratio between spatial or temporal harmonics is irrational (incommensurate frequencies). Quasiperiodic order has been observed in solid state structures [1] and cold atom systems [2]. Experiments and simulations on driven nonlinear systems revealed a number of distinguishing features associated with the quasiperiodic nature of the driving [3–6]. In a real experiment, quasiperiodic structures can be represented by their best periodic approximations. This raises the issue of whether the system dynamics corresponds to the original quasiperiodic structure, or to the approximant periodic one. This is an essential issue whenever periodicity and quasiperiodicity lead to completely different dynamics. A standard approach [7] relies on the “most irrational numbers,”the golden ratio being the most popular choice, as determined by using Farey tree type classification [8,9], or equivalently, the continued fraction representation of the number. The use of these numbers guarantees that their best rational approximations are the best choice for the system response to reflect the original quasiperiodicity of the temporal or spatial structure. However, this leaves open the important question of what level of irrationality is required for the system to react following a genuine quasiperiodic behavior given the finite dimensions of the experiment. This is precisely the question addressed here. In this Letter, we consider a nonlinear driven system whose response in the infinite time limit is very sensitive to the quasiperiodic nature of the driving; i.e., the system reacts in a completely different way depending on whether the driving is periodic or quasiperiodic. We study the response of the system to a biharmonic drive in the finite interaction time limit. We first show, analytically, that the frequency resolution of the system is sub-Fourier and provide an expression for it. Then, we examine the response of the system to a drive consisting of incommensurate frequencies. We show that irrationality alone is not sufficient to guarantee quasiperiodic behavior. Instead, the response of the system depends on the nature of the irrational frequency ratio and on the interaction time. Model system, periodic driving, and sub-Fourier resolution.—As a case study of a system that is very sensitive to whether the drive is periodic or quasiperiodic, we consider a driven classical particle. The system dynamics is described, in the deterministic and overdamped regime, by γ_ xðtÞ¼−U0½xðtÞþFðtÞ;(1) where the dot and the prime denote time and spatial derivatives, respectively, γis the friction coefficient, UðxÞ¼U0cosð2kxÞ=2is a spatially periodic and symmetric potential, with period λ¼π=k,andFðtÞis a driving force. Note, however, that the conclusions reported in this Letter are based only on symmetry considerations, and, thus, do not depend on the specific details of the dynamics [10]. PRL 112, 174102 (2014) PHYSICAL REVIEW LETTERS week ending 2 MAY 2014 0031-9007=14=112(17)=174102(5) 174102-1 © 2014 American Physical Society The system response to the drive can be characterized by the average velocity v¼limTs→∞½xðTsÞ−xð0Þ=Ts.Inany spatially symmetric system, vis reversed when the driving is inverted, v½−FðtÞ ¼ −v½FðtÞ:(2) Further, it must not depend on the specific choice of the time origin, v½Fðtþt0Þ ¼ v½FðtÞ. By combining both transformations, it follows that the current vanishes if the driving possesses the following symmetry Fsh∶Fðtþt0Þ¼ −FðtÞ, for all tand a given value t0. If the system is also overdamped, like the one described by (1), another symmetry that must be broken for a finite current is given by [11,12] Fs∶Fðt0−tÞ¼−Fðt0þtÞ. A common choice used to break both symmetries is the biharmonic driving F1ðtÞ¼F0½cosðω1tÞþcosðω2tÞ, with two commensurate frequencies ω1and ω2, i.e., ω2=ω1¼p=q with pand q being two coprime positive integers. Another driving that will be used here to produce a current is F2ðtÞ¼ F0½cosðω1tÞþcosð2ω1tÞcosðω2tÞ. These driving forces are particular examples of the generic driving FðtÞ¼Φðω1tþφ1;ω2tþφ2Þ, where Φis a function that is periodic in both its arguments, Φðx1þ2π;x 2Þ¼ Φðx1;x 2þ2πÞ¼Φðx1;x 2Þ, and φ1and φ2are constants. The driving period itself is given by T¼2πq=ω1¼ 2πp=ω2. The current invariance under the choice of the time origin implies invariance under the transformation φ1→φ1þω1t0,φ2→φ2þω2t0. By choosing t0¼−φ1=ω1, it follows that the current v¼vpq depends on the driving phases only through the combination θ¼qφ2−pφ1. On the other hand, using Bézout’s lemma, it is easy to show that the periodicity of Φðx1;x 2Þon both its arguments implies vpqðθþ2πÞ¼vpqðθÞ:(3) This last symmetry property, together with (2), can be readily used to show that if Φsatisfies any of the following conditions: (i) Φðx1þπ;x 2þπÞ¼−Φðx1;x 2Þ, (ii) Φðx1þπ;x 2Þ¼−Φðx1;x 2Þ, or (iii) Φðx1;x 2þπÞ¼ −Φðx1;x 2Þ,then vpqðθþπÞ¼−vpqðθÞ:(4) From (4) it readily follows that when pand qare both odd and (i) holds—as is the case for the driving F1—the symmetry Fsh holds, and thus, vpqðθÞ¼0for all θ. Similarly, the current vanishes when pis even in case (ii), and when qis even in case (iii)—the driving F2 fulfils (iii). When both symmetries Fsh and Fsare broken in a deterministic overdamped system, a finite current of the form vpq ¼ðm=nÞv0is expected [13], where mand nare two integers, and v0¼λ=T. The response of the system to a quasiperiodic driving is completely different [4–6,14]. It can be shown, on the basis of general symmetry properties only, that there is no current when the two driving frequencies ω1and ω2are incommensurate [4,6]. The above results assume an infinite sampling time Ts. For large but finite times, the following expression for the finite-time current vTs¼½xðTsÞ−xð0Þ=Ts vTs∼ 1 ðqω2−pω1ÞTsZθþðqω2−pω1ÞTs θ dθ0vpqðθ0Þ;(5) was shown in [15] to describe the asymptotic evolution of the system, when the second driving frequency ω2is in the neighborhood of ω1p=q, independent of the system details. In overdamped systems, the leading expression obtained from a functional expansion on the driving [16,17] is given by vpq ¼BcosðθÞ, which, when inserted in (5) for F1and F2, yields vTs∼vpq sin½ðqω2−pω1ÞTs ðqω2−pω1ÞTs :(6) Our numerical simulations, presented in Fig. 1, are in good agreement with (6) for the overdamped system (1), especially for larger Ts, when the current is more independent of the initial conditions. A general expression for the width of the resonance can be derived on the basis of symmetries only [15], i.e., from Eqs. (3) and (4). The integral in (5) cancels whenever ðqω2−pω1ÞTsis a multiple of 2π. Furthermore, the current, as a function of ω2¼ω1p=q þδω2, decays as 3.141588 3.141591 3.141594 3.141597 ω2 -0.5 0 0.5 1 1.5 2 v / v0 Ts=104 Ts=105 π 355/113 FIG. 1 (color online). Current vs driving frequency ω2for the overdamped system (1) with the driving F2and F0¼5.75. Reduced units are defined in the simulations such that U0¼k¼γ¼ω1¼1. Empty and filled diamonds correspond to Ts¼104and 105, respectively. The lines are the predictions given by (6) with q¼113,p¼355, and v0¼1=ð2qÞ. The horizontal bars depict the frequency width (7), showing a resolution 113 times smaller than that expected from the Fourier width 2π=Ts. PRL 112, 174102 (2014) PHYSICAL REVIEW LETTERS week ending 2 MAY 2014 174102-2 ðqδω2TsÞ−1as we move away from ω2¼ω1p=q. The distance to the first zero, thus, gives an estimation of the width of such a resonance as Δω2¼2π qTs :(7) The frequency window given by (7) is a factor qsmaller than the Fourier width ΔωF¼2π=Ts. Numerical calculations for the specific model (1) confirm the sub-Fourier width of the resonance, as shown in Fig. 1. Periodic approximations and quasiperiodicity vs irrationality.—The infinite-time current vpq associated with a periodic driving with ω2¼ω1p=q can be accurately computed in a simulation or experiment provided the observation time Tsis much larger than the driving period Tq¼2πq=ω1. In turn, given an arbitrary frequency ω2, we may look for periodic drivings ω1p=q which provide a good approximation to the current. This will occur whenever both frequencies are close enough, that is, jω2−ω1p=qj<ϵ0Δω2;(8) where ϵ0is a small dimensionless parameter. For example, by choosing ϵ0¼0.1,(6) predicts current deviations from vpq smaller than 2%. Given the values ω2and Ts, we can define the best periodic approximation from the lowest positive integers p and qsuch that (8) is satisfied. The associated period Tqcan be used to predict the periodic or quasiperiodic behavior of the system under the given driving. If Tq≪Ts, then periodic behavior is expected. Genuine quasiperiodicity would require Tq≳Tsregardless of the specific value of Ts. As we discuss below, this does not occur for all irrational ratios ω2=ω1. Equation (8) can be rewritten as ϕ≤ðTq=TsÞϵ0, where ϕ¼jω2=ω1−p=qjq2is a function which measures how well the ratio ω2=ω1is approximated by the rational approximation p=q. For a given approximation p=q, this inequality provides a range of times Tswhere the corresponding periodic approximation is expected to hold, with the maximum value Tmax sgiven by Tq¼ϕ ϵ0 Tmax s:(9) Therefore, even if ω2=ω1is irrational, if there exists a rational approximation such that as ϕ=ϵ0≪1, then there will be a certain range of observation times where Tq≪Ts, and thus, periodic, instead of quasiperiodic, behavior is expected. Our analysis identifies the error function ϕas the central quantity to determine the relationship between irrationality and quasiperiodicity in a physical system. Standard results of number theory, stated in the following, will clarify how ϕ is related to different irrational numbers. Approximation of irrationals by rationals.—Hurwitz’s theorem [18] states that for any irrational number x, there are infinitely many rational approximations p=q such that ϕ¼jx−p=qjq2<1=ffiffiffi 5 p, providing a universal upper bound to ϕ. Liouville proved [18] that for every quadratic irrational x—i.e., the roots of a quadratic equation with integer coefficients—there is a positive constant c0such that c0<ϕholds for every positive integer pand q. Provided this constant is not very small—which is, indeed, the case for x¼ffiffiffi 2 pand ffiffiffiffiffi 10 p, as shown in Fig. 2—the quadratic irrational is difficult to approximate by rational numbers, and quasiperiodic behavior is guaranteed for arbitrary Ts. Other irrational numbers admit better rational approximations. For example, Davis proved in 1978 [19] that for the transcendental number e, for each ˆ c>1=2there are infinitely many rational approximations with q≥2that satisfy ϕ≤ˆ clog½logðqÞ=logðqÞ, which is also illustrated in Fig. 2. In our nonlinear system, Davis’s result and (9) yield periodic behavior for large enough denominators q, though the decay is very slow and the corresponding times Tsover which the effect can be observed are impractically large. One of the transcendental numbers that admits the best rational approximations is the Liouville constant α¼P∞ j¼110−j!, constructed by Liouville to this effect, establishing, in this way, the existence of a transcendental number for the first time. It will also be considered to test numerically our results. Numerical examples.—The directed current in the overdamped system (1) driven by periodic drivings was numerically computed by using a multiple of the known period T as sampling time Tsafter waiting a brief relaxation time. This procedure allows for accurate estimations of vpq even though Tsis not too large. Figure 3shows results for a driving F1and three different values of the ratio ω2=ω1: the value ω2=ω1¼110 001=105, a very good rational approximation to the Liouville number 11051010 1015 1020 1025 q 0.001 0.01 0.1 1 2 10 2 3 a 11051010 1015 1020 1025 q 0.001 0.01 0.1 1 e b FIG. 2 (color online). Relative error ϕ¼jx−p=qjq2of the best rational approximations p=q—the continued fractional approximations—to several irrational numbers: (a) x¼ffiffiffi 2 p (empty circles), ffiffiffiffiffi 10 p(empty squares), and 21=3(filled circles); (b) e(empty circles) and π(filled circles), as a function of the denominator q. The solid line for eillustrates Davis’s result with ϕ¼log½logðqÞ=logðqÞ=2. PRL 112, 174102 (2014) PHYSICAL REVIEW LETTERS week ending 2 MAY 2014 174102-3 10αwith a very low error ϕ∼10−13, the Liouville number 10αitself, and the quadratic irrational ffiffiffiffiffiffiffiffi 6=5 p. First, we consider the case of ω2=ω1¼110 001=105(triangles). For low values of F0, no current is expected [20], a fact that is shown in the bottom panel of Fig. 3as a measured current of order 10−16 due to the use of double precision variables in the computer simulations, which provides 15 to 17 significant decimal digits precision. On the other hand, finite currents are observed to be a few multiples of v0¼λ=T, which, in our units, is of order 1=T, with Titself being of order 105. Now, we turn to the case of the two irrational numbers. Numerical results have to be compared to the infinite-time-limit genuine quasiperiodic behavior: a zero current irrespective of F0. For the irrational number ω2=ω1¼10α,(8) predicts for Ts¼108that the finite-time current will deviate from the infinite time limit, and will be of the same order of magnitude as the results of the periodic approximation. The precise deviations from the periodic data are due to the fact that Tsis not an exact multiple of the period T, as shown in the inset of Fig. 3. Absence of current is shown numerically to be of order 1=Ts¼10−8. In contrast, genuine quasiperiodic behavior, i.e., a zero current of order 1=Tsirrespective of F0, can be seen in the whole range of driving amplitudes shown in Fig. 3(crosses in both panels) for the quadratic irrational ω2=ω1¼ffiffiffiffiffiffiffiffi 6=5 p. These current values are 3 orders of magnitude smaller than the ones shown for the irrational 10α. Even larger factors are expected as Tsis increased, since the current will be locked to the order of magnitude provided by this periodic approximation until the next one, which has period T∼1018, replaces it. The same effect can be observed in other irrational numbers with even poorer rational approximations. Figure 4shows the results for a driving F2with a frequency ratio ω2=ω1¼π. By choosing ϵ0¼0.1, from (8) the best periodic approximation for observation times in the range Ts¼102–104is given by ω2=ω1¼355=113. This is, indeed, what is observed in the top panel of Fig. 4for Ts¼104, where the predicted periodic behavior is clearly seen. In contrast, it can also be observed that the currents obtained with the quadratic irrational ω2=ω1¼ffiffiffiffiffi 10 pare much smaller, in fact of order 1=Ts, as expected. But quasiperiodic behavior is also recovered for ω2=ω1¼π when Tsis increased to 105, as shown in the bottom panel of Fig. 4. In this case, the rational approximation 355/133 no longer satisfies (8), and the next one is given by ω2=ω1¼312 689=99 532, with an associated period that is about six times larger than Ts. This behavior can be directly traced back to the sub-Fourier resolution: As clearly seen in Fig. 1, the frequency separation jπ−355=113j≈ 2.7×10−7is within the resolution window ϵ0Δω2for Ts¼104, but is outside for Ts¼105despite being well inside the Fourier width ΔωF¼2π=Ts≈6.3×10−5and the window ϵ0ΔωF. Conclusions.—We have shown, on the basis of symmetry properties only and, thus, not relying on the specific details of the dynamics, that the response of a nonlinear system to a bifrequency drive depends on the interaction time and the nature of the frequency ratio between the driving frequencies. In particular, we showed that an -2 0 2 4 6 8 v / v0 1071081091010 Ts 0 4 8 0.5 0.75 11.25 1.5 1.75 22.25 2.5 F0 10-16 10-14 10-12 10-10 10-8 10-6 |v| F0=1.22 FIG. 3 (color online). Current vs driving amplitude for the driving F1. Filled (black) diamonds correspond to the ratio ω2=ω1¼10α¼10 P∞ j¼110−j!¼1.100010…and Ts¼108. Triangles (red) depict the results for the driving with the rational approximation ω2=ω1¼110 001=105, and the (blue) crosses to the quadratic irrational ω2=ω1¼ffiffiffiffiffiffiffiffi 6=5 p¼1.095…for Ts¼108. Inset: current as a function of Tsfor ω2=ω1¼10αand F0¼1.22, showing convergence to the rational approximation value (horizontal solid line). The bottom panel shows the top panel data in a semilogarithmic scale. -2 -1 0 1 2 v / v 0 5 5.5 66.5 77.5 8 F 0 -2 -1 0 1 2 v / v 0 (a) T s =10 4 (b) T s =10 5 FIG. 4 (color online). Current vs F0for ω2=ω1¼π(filled diamonds) with Ts¼104(top panel) and Ts¼105(bottom panel). Up triangles (red) show the results for the driving with the rational approximation ω2=ω1¼355=113, while down triangles (orange) are the results for the same frequency ratio but with the corresponding Ts¼104and 105. Crosses (blue) correspond to ω2=ω1¼ffiffiffiffiffi 10 p. PRL 112, 174102 (2014) PHYSICAL REVIEW LETTERS week ending 2 MAY 2014 174102-4 irrational ratio is not sufficient to guarantee quasiperiodic response. The system frequency resolution, being intrinsically sub-Fourier, dictates that quasiperiodicity is only guaranteed for irrational numbers which have bad rational approximations, such as many quadratic irrationals [21]. We acknowledge financial support from the Leverhulme Trust and the Ministerio de Ciencia e Innovación of Spain Grant No. FIS2008-02873 (D. C. and J. C.-P.). *[email protected] †[email protected] ‡[email protected]k [1] D. Levine and P. J. Steinhardt, Phys.Rev.Lett.53, 2477 (1984). [2] L. Guidoni, C. Triché, P. Verkerk, and G. Grynberg, Phys. Rev. Lett. 79, 3363 (1997). [3] U. Feudel, S. Kuznetsov, and A. Pikovsky, Strange Nonchaotic Attractors, Dynamics between Order and Chaos in Quasiperiodically Forced Systems (World Sicentific Publishing, Singapore, 2006). [4] E. Neumann and A. Pikovsky, Eur. Phys. J. B 26, 219 (2002). [5] R. Gommers, S. Denisov, and F. Renzoni, Phys. Rev. Lett. 96, 240604 (2006). [6] D. Cubero and F. Renzoni, Phys. Rev. E 86, 056201 (2012). [7] J. A. Glazier and A. Libchaber, IEEE Trans. Circuits Syst. II 35, 790 (1988). [8] V. T. Sós, Ann. Univ. Sci. Budapest Eötvös Sect. Math. 1, 127 (1958). [9] N. Slater, Proc. Cambridge Philos. Soc. 63, 1115 (1967). [10] See Supplemental Material at http://link.aps.org/ supplemental/10.1103/PhysRevLett.112.174102 for the explicit validation of our approach in the case of a spatially nonperiodic system. [11] P. Reimann, Phys. Rev. Lett. 86, 4992 (2001). [12] P. Reimann, Phys. Rep. 361, 57 (2002). [13] A. Ajdari, D. Mukamel, L. Peliti, and J. Prost, J. Phys. I (France) 4, 1551 (1994). [14] S. Flach and S. Denisov, Acta Phys. Pol. B 35, 1437 (2004). [15] J. Casado-Pascual, D. Cubero, and F. Renzoni, Phys. Rev. E 88, 062919 (2013). [16] N. R. Quintero, J. A. Cuesta, and R. Alvarez-Nodarse, Phys. Rev. E 81, 030102R (2010). [17] J. A. Cuesta, N. R. Quintero, and R. Alvarez-Nodarse, Phys. Rev. X 3, 041014 (2013). [18] G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers (Oxford University Press, London, 1975). [19] C. S. Davis, J. Austral. Math. Soc., Ser. A Pure Math. 25, 407 (1978). [20] P. Hänggi and F. Marchesoni, Rev. Mod. Phys. 81, 387 (2009). [21] Not all quadratic irrationals have bad rational approximations, as illustrated by the example x¼1þffiffiffi 2 p=10n, with n an arbitrarily large integer. PRL 112, 174102 (2014) PHYSICAL REVIEW LETTERS week ending 2 MAY 2014 174102-5