betaFeedback

WorksPhysics and Astronomy

Dynamic Synchronization of Driven Self-Oscillators: Modeling and Experiment

Abstract

Synchronization of self-sustained oscillators under fixed-frequency and amplitude forcing is well understood, but how time-varying forcing mangles phase locking has been much less explored. Theory predicts that slow, deterministic modulation of the drive amplitude or frequency can lead to a peculiar synchronization regime characterized by intermittent locking of the oscillation phase beyond the Arnold-tongue boundaries associated with fixed harmonic forcing. We test these predictions in a controllable aeroacoustic self oscillator, i.e, a whistle, that exhibits a robust limit cycle and is subject to external acoustic forcing with programmable frequency and amplitude modulation. Under both slowly varying frequency or amplitude of the forcing, three regimes are observed: (i) strict synchronization (ii) intermittent synchronization, characterized by alternating phase locking and brief phase slip episodes and (iii) no synchronization, with regular phase slips. Particularly in strict synchronization regime, the phase of the oscillator will follow arbitrary slowly-varying drive phase and under amplitude modulation its amplitude fluctuations are strongly suppressed.

Introduction

Phase synchronization, or phase locking, of a self-sustained oscillator under periodic forcing refers to the adjustment of the oscillator’s phase to match that of the external periodic drive [1, 2]. It has been studied extensively in mechanics, electronics, chemistry, and biology [2, 3, 4]. A classic result in time-independent periodic forcing is the formation of Arnold tongues, tongue-shaped regions in the drive amplitude–frequency plane where n:mn:m phase locking occurs, with nn and mm the numbers of oscillation cycles of the drive and oscillator, respectively. In the setting of periodic drive, appropriate averaging of higher frequency oscillations will lead to an autonomous dynamical system which governs the stability, bifurcation, and transitions of synchronization regime [4].

Realistic forcing, however, often exhibits variations in frequency and amplitude, making the dynamics intrinsically non-autonomous. A growing number of work tackles these issues for non-autonomous oscillators [5, 6, 7, 8, 9]. Related rate-dependent phenomena have also been reported in musical-acoustics models, where slowly varying control parameters reshape basins of attraction and alter transient dynamics [10, 11]. For a one-dimensional phase oscillator with time-periodic detuning (the non-autonomous Adler equation), Gandhi et al. showed rich canard segments along the thin transition layers between them [6]. In parallel, chronotaxic theory classifies self-sustained non-autonomous oscillators with a moving point attractor on the cycle, as a deterministic, perturbation-resistant time-varying frequency motion [7]. More recently, Lucas et al. [8] showed theoretically that deterministic, slowly varying driving can stabilize oscillations beyond the boundaries of the fixed drive Arnold–tongue, by creating a finite-time intermittent regime, where system trajectories alternate between intervals with a stable locked point and intervals with no locked point and yet are stable on average. This effect persists in higher-dimensional models as well [8]. Building on these results, a finite-time dynamical-systems framework was proposed [9, 12]. It treats slow modulation explicitly (slow-fast on a bounded slow-time interval), clarifies when phases track a moving attractor, and gives qualitative criteria for finite-time stability under modulation [9, 12].

This work aims at building upon these theoretical predictions, and investigating experimentally dynamic synchronization using a controllable physical oscillator. To that end, we employ a whistle and dynamically force it around its natural self-oscillation frequency. In absence of forcing, above a critical air mass flow, the aeroacoustic system becomes linearly unstable due to a constructive interaction between one of the shear layer eigenmodes and the Helmholtz mode of the whistle, and a limit cycle emerges via a supercritical Hopf bifurcation. Recent analyses and experiments on grazing-flow cavities and compact apertures quantify the feedback mechanism between shear-layer dynamics and the acoustic field, the radiation losses, and the nonlinear saturation that sets the oscillation amplitude [13, 14, 15]. Weak external acoustic forcing is applied to drive the aeroacoustic self-oscillation, providing a clean testbed for investigating synchronization under modulation. This combination of robust self-sustained oscillations and precise programmable external driving makes the whistle a natural platform for validating non-autonomous synchronization theory in the laboratory.

Near the Hopf bifurcation, the amplitude of the aeroacoustic mode of the whistle admits a Stuart–Landau oscillator description [16]; under weak forcing from an incident sound field with programmable phase or amplitude, standard phase reduction (e.g., [17]) yields a one-dimensional phase equation with time-varying drive.

Taking advantage of a single dominant aeroacoustic mode and a straightforward implementation of slow/fast frequency and amplitude sweeps, we carry out a systematic experimental validation of non-autonomous synchronization and demonstrate intermittently stable regime beyond static Arnold–tongue boundaries, as predicted by Lucas et al. [8]. Furthermore, phase tracking of a moving attractor under sufficiently slow modulation yet arbitrary aperiodic signal [9, 12] is also tested.

The paper is organized as follows. Section II describes the aeroacoustic system, presents the derivation of the classical slow-flow equations for the oscillator’s amplitude and phase, and discusses the effects of amplitude and frequency modulation. Section III presents experimental results, including instantaneous-frequency analysis and synchronization maps (Arnold tongues) under periodic frequency/amplitude modulation, and responses to aperiodic phase modulation. Section IV concludes and provides an outlook.

Asymptotic dynamics under time modulation

We model the self-sustained aeroacoustic oscillations in the whistle subject to external acoustic forcing as a Van der Pol (vdP) oscillator weakly forced by an external drive. In Figure 1a, the vdP oscillator is sketched as a mass–spring–damper system, and Figure 1c shows the aeroacoustic analogue, in which the cavity compliance, the air mass in the whistle orifice, and the shear-layer inducing a nonlinear negative acoustic resistance correspond to the mechanical elements.

Figure 1. (a) Van der Pol type mass–spring–damper: linear stiffness kk (green), negative linear damping coefficient μ∈R+\mu\in\mathbb{R}^{+} and quadratic nonlinear damping term (orange), and external drive F(t)\mathcal{F}(t) (blue). (b) Phasor sketch of phase locking: oscillator response (black) and external forcing (blue) on the limit-cycle circle, with examples of probability density functions of the phase difference P(Δϕ(t))P(\Delta\phi(t)) in the non-synchronized (blue), quasi-synchronized (pink) and synchronized state (violet). (c) Whistle as a self-sustained oscillator: cavity compliance 1/k1/k (green), air mass mm in the whistle orifice (gray), flow-induced nonlinear negative damping (orange), and side-aperture forcing F(t)=P(t)σ\mathcal{F}(t)=P(t)\sigma (blue), where the oscillating pressure P(t)P(t) is produced by a loudspeaker and σ\sigma is the aperture area.

In dimensional form, the vdP equation reads

p¨−εμ(1−p2R2)p˙+ωn2p=εF(t)(1)\ddot{p}-\varepsilon\mu\left(1-\frac{p^{2}}{R^{2}}\right)\dot{p}+\omega_{n}^{2}p=\varepsilon\mathcal{F}(t) \tag*{(1)}

where μ>0\mu>0 defines the linear component of the negative nonlinear damping induced by the air flow, which gives birth to a stable limit cycle for flow rates that exceed the Hopf bifurcation threshold. In Eq. (1), RR is the limit-cycle amplitude of the unforced vdP oscillator, i.e when F(t)=0\mathcal{F}(t)=0, and ωn\omega_n is the natural frequency. The variable p(t)p(t) corresponds to the acoustic pressure in the whistle cavity. The small parameter 0<ε≪10<\varepsilon\ll1 makes explicit that we consider weakly damped, weakly forced self-oscillations, so that the envelope evolves on a slow time scale.

The acoustic forcing imposed by an external loudspeaker through a side aperture is expressed as

F(t)=FA(t)cos⁡(ΩF(t)t+ϕ0)(2)\mathcal{F}(t)=F A(t)\cos\left(\Omega_F(t)t+\phi_0\right) \tag*{(2)}

with F=O(1)F=\mathcal{O}(1), A(t)A(t) and ΩF(t)\Omega_F(t) the slowly varying drive amplitude and frequency, and ϕ0\phi_0 the phase at origin, which, without loss of generality, is set to zero in the following.

In the weak forcing, time independent case (A=1A=1, ΩF=Ωu=const.\Omega_F=\Omega_u=\mathrm{const.}, with Ωu\Omega_u the unmodulated drive frequency), writing the detuning as Δω=Ωu−ωn\Delta\omega=\Omega_u-\omega_n and approximating

p(t)≈rcos⁡(Ωut+θ),p(t)\approx r\cos(\Omega_u t+\theta),

where rr is the forced limit-cycle amplitude, the system exhibits phase synchronization when the phase difference θ\theta approaches a constant value [2, 18]. Equivalently, the phase difference Δθ(t)=θp(t)−θF(t)\Delta\theta(t)=\theta_p(t)-\theta_F(t) between oscillator and drive can be viewed as an angle on the unit circle. Thus, synchronization corresponds to Δθ(t)\Delta\theta(t) clustering around a fixed value so that its distribution P(Δθ)P(\Delta\theta) is sharply peaked rather than uniform, as sketched in Figure 1b. Plotting the synchronization regions in the forcing amplitude and detuning plane yields the classical Arnold tongues [2, 18].

Here we focus on nonautonomous cases where A(εt)A(\varepsilon t) and/or ΩF(εt)\Omega_F(\varepsilon t) vary slowly in time. To analyze such modulation, we introduce a standard multiple time scales reduction [19],

T0=t,T1=εt,0<ε≪1,T_0=t,\qquad T_1=\varepsilon t,\qquad0<\varepsilon\ll1,
p(t)=p0(T0,T1)+εp1(T0,T1)+O(ε2),(3)p(t)=p_0(T_0,T_1)+\varepsilon p_1(T_0,T_1)+O(\varepsilon^2), \tag*{(3)}

so that the amplitude and phase evolve on the slow time T1T_1. Subsequently, we will also introduce the explicit modulation phase

θm(T1)=εΩmt=ΩmT1,(4)\theta_m(T_1)=\varepsilon\Omega_m t=\Omega_mT_1, \tag*{(4)}

with modulation frequency Ωm\Omega_m. Thus θm\theta_m advances on the slow time scale T1T_1, and Ωm\Omega_m is chosen much smaller than the natural frequency of the self-oscillator. This classical procedure (e.g., [19, 20]) yields coupled slow-flow equations for the instantaneous amplitude and phase, capturing synchronization, phase slips, and the effects of slow modulation.

In what follows we treat amplitude modulation and frequency (phase) modulation separately. For each case we derive the slow-flow equations, identify the locking criteria, and discuss the implications for the experimentally observed synchronization maps.

Regarding the validity of the present analysis, in contrast with the case of fixed amplitude harmonic forcing, the perturbation approximation incurs a O(ε)O(\varepsilon) error on time intervals T=O(1/ε)T=O(1/\varepsilon) in the slow-modulation regime, which is however sufficient for our experimental comparisons.

It is also important to note that Adler equations are often derived for finding locking criteria without accounting for amplitude dynamics, e.g [5, 21] or [17] for more modern analysis, by assuming that the oscillator amplitude is confined to its limit cycle.

By contrast, we derive the coupled slow-flow equations for both amplitude and phase, which provides a complete picture that is essential for self-oscillators driven by frequency or amplitude modulated forcing.

Frequency Modulation

The case of a slowly varying frequency of the external drive is considered first with the time scales T0=tT_0=t and T1=εtT_1=\varepsilon t and A≡1A\equiv1. The forcing now reads

F(T0,T1)=Fcos⁡(ΩuT0+kfsin⁡θm(T1)),(5)\mathcal{F}(T_0,T_1)=F\cos\left(\Omega_uT_0+k_f\sin\theta_m(T_1)\right), \tag*{(5)}

with the mean (unmodulated) drive angular frequency Ωu\Omega_u, frequency-modulation depth kfk_f, and modulation phase θm(T1)=ΩmT1\theta_m(T_1)=\Omega_mT_1. Using complex notation, we write the forcing as

F(T0,T1)=ℜ{FeiΦ(T0,T1)},\mathcal{F}(T_0,T_1)=\Re\left\{F e^{i\Phi(T_0,T_1)}\right\},

and

Φ(T0,T1)=ΩuT0+kfsin⁡θm(T1),\Phi(T_0,T_1)=\Omega_uT_0+k_f\sin\theta_m(T_1),

so that

F(T0,T1)=ℜ{FeiΩuT0eikfsin⁡θm(T1)}.(6)\mathcal{F}(T_0,T_1)=\Re\left\{F e^{i\Omega_uT_0}e^{ik_f\sin\theta_m(T_1)}\right\}. \tag*{(6)}

The Jacobi–Anger expansion [22] is applied only to the slowly varying factor,

eikfsin⁡θm(T1)=∑n=−∞∞Jn(kf)einθm(T1),(7)e^{ik_f\sin\theta_m(T_1)}=\sum_{n=-\infty}^{\infty}J_n(k_f)e^{in\theta_m(T_1)}, \tag*{(7)}

so that the complex forcing becomes

FeiΦ(T0,T1)=F∑n=−∞∞Jn(kf)ei[ΩuT0+nΩmT1].F e^{i\Phi(T_0,T_1)}=F\sum_{n=-\infty}^{\infty}J_n(k_f)e^{i\left[\Omega_uT_0+n\Omega_mT_1\right]}.

Substituting the leading-order terms into the perturbed vdP equation and applying the multiple scales method, only those near resonance forcing terms with carrier eiωnT0e^{i\omega_nT_0} contribute to the slow evolution and all other fast oscillatory terms average out. Keeping the near-resonant terms (here n=0,±1n=0,\pm1) then yields the slow-flow system for the envelope amplitude r(T1)r(T_1) and phase θ(T1)\theta(T_1),

drdT1=μ2r(1−r2R2)−FJ0(kf)2ωnsin⁡θ−FJ1(kf)ωnsin⁡(θ−θm),(8)\frac{dr}{dT_1}=\frac{\mu}{2}r\left(1-\frac{r^2}{R^2}\right)-\frac{FJ_0(k_f)}{2\omega_n}\sin\theta-\frac{FJ_1(k_f)}{\omega_n}\sin(\theta-\theta_m), \tag*{(8)}
dθdT1=Δω−FJ0(kf)2ωnrcos⁡θ−FJ1(kf)ωnrcos⁡(θ−θm),(9)\frac{d\theta}{dT_1}=\Delta\omega-\frac{FJ_0(k_f)}{2\omega_nr}\cos\theta-\frac{FJ_1(k_f)}{\omega_nr}\cos(\theta-\theta_m), \tag*{(9)}

where Δω=Ωu−ωn\Delta\omega=\Omega_u-\omega_n is the detuning. Taking kf=0k_f=0 (so J0(kf)=1J_0(k_f)=1, J1(kf)=0J_1(k_f)=0) reduces (8)–(9) to the standard truncated slow-flow equations of the harmonically forced vdP oscillator [18].

When deriving equations (8)–(9), the truncated terms are only the n=0,±1n=0,\pm1 from the Jacobi–Anger expansion, i.e. the J0J_0 and J1J_1 terms. For small kfk_f, Jn(kf)=O((kf/2)∣n∣)J_n(k_f)=O((k_f/2)^{|n|}) so ∣J±2∣=O(kf2)|J_{\pm2}|=O(k_f^2), and the associated terms are both weak and off-resonant, thus they average out at leading order. The resulting slow flow is accurate up to O(kf2)O(k_f^2) corrections. Including J2J_2 would add terms proportional to cos⁡(θ−2θm)\cos(\theta-2\theta_m) and sin⁡(θ−2θm)\sin(\theta-2\theta_m) with coefficients proportional to J2(kf)J_2(k_f).

For a phase oscillator with slowly varying detuning, Lucas et al. [8] studied the nonautonomous Adler equation ψ˙=Δω(t)+γsin⁡ψ\dot{\psi}=\Delta\omega(t)+\gamma\sin\psi, where ψ\psi is the phase difference, Δω(t)\Delta\omega(t) a prescribed time-dependent frequency detuning, and γ>0\gamma>0 the coupling constant. Instantaneous fixed points satisfy Δω(t)+γsin⁡ψ∗(t)=0\Delta\omega(t)+\gamma\sin\psi^*(t)=0 and exist only when ∣Δω(t)∣≤γ|\Delta\omega(t)|\leq\gamma. According to how often this inequality holds over one modulation period, they distinguished three regimes (I–III in their Fig. 1 and also in Figure 2): no synchronization (∣Δω(t)∣>γ, ∀t|\Delta\omega(t)|>\gamma,\ \forall t), persistent synchronization (∣Δω(t)∣<γ, ∀t|\Delta\omega(t)|<\gamma,\ \forall t), and an intermittent regime in which ∣Δω(t)∣|\Delta\omega(t)| crosses γ\gamma twice per cycle, where fixed points are created and destroyed twice per cycle. In the intermittent case the dynamics are deterministic. The system alternates between intervals of convergence toward the instantaneous fixed point and intervals of drift, producing one slip per half-cycle of the slow modulation. Lucas et al. showed that, in the plane of mean detuning versus coupling, the union of the persistent and intermittent regimes occupies a larger region than the static tongue ∣Δω∣<γ|\Delta\omega| < \gamma. The apparent widening of the Arnold tongue under modulation is due to the intermittent regime, and parameter values with negative largest Lyapunov exponent of ψ\psi are regarded as synchronized in a broad sense [8].

Figure 2. Schematic view of synchronization diagram under slow modulation. Top: time evolution over one modulation cycle of the phase dynamics (solid black), together with horizontal lines (orange dashed) indicating when fixed points exist (stable fixed points are marked as filled circles while unstable ones marked as open circles). Bottom: corresponding schematic Arnold tongues in the plane of averaged forcing frequency (horizontal) and forcing amplitude (vertical), showing three regimes: light blue—no synchronization (region I), medium blue—intermittent synchronization (region III), and dark blue—persistent synchronization (region II). The solid white boundary is the classical (unmodulated) locking threshold. The white dashed boundaries indicate the locking thresholds evaluated at the two extrema of the modulation. (a) Frequency modulation: the forcing frequency oscillates in time while the forcing amplitude is fixed. Panel is redrawn, with notation adapted from Lucas et al. [8]. (b) Amplitude modulation: the forcing amplitude oscillates while the forcing frequency is fixed. The inner white dashed lines correspond to the persistent synchronization threshold given by Eq. (27), and the outer white dashed lines correspond to the broad synchronization threshold given by Eq. (26).

In the present work we adopt the same nonautonomous modulation, but start from a full van der Pol model: we first derive slow–flow equations for the amplitude and phase of the forced vdP oscillator and then show (Appendix A) that the slow phase difference reduces, after an explicit change of variables, to a nonautonomous Adler equation of the form studied by Lucas et al. [8].

For each frozen modulation phase θm\theta_m, instantaneous equilibria of the slow–flow phase equation (9) satisfy

0=Δω−FJ0(kf)2ωnrcos⁡θ∗−FJ1(kf)ωnrcos⁡(θ∗−θm),0 = \Delta\omega- \frac{F J_0(k_f)}{2\omega_n r}\cos\theta^* - \frac{F J_1(k_f)}{\omega_n r}\cos(\theta^*-\theta_m),

which defines a time-dependent locking condition. As detailed in Appendix A, a change of variables from θ\theta to the phase difference ψ\psi recasts the slow phase dynamics into a nonautonomous Adler equation

dψdT1=Δωeff(T1)+γeff(T1)sin⁡ψ,(10)\frac{d\psi}{dT_1} = \Delta\omega_{\mathrm{eff}}(T_1) + \gamma_{\mathrm{eff}}(T_1)\sin\psi, \tag*{(10)}

with a time-varying effective detuning Δωeff(T1)\Delta\omega_{\mathrm{eff}}(T_1) and coupling γeff(T1)\gamma_{\mathrm{eff}}(T_1) that are explicit functions of θm(T1)=ΩmT1\theta_m(T_1)=\Omega_mT_1. For frozen T1T_1, instantaneous fixed points of (10) exist if and only if ∣Δωeff(T1)∣≤γeff(T1)|\Delta\omega_{\mathrm{eff}}(T_1)|\leq\gamma_{\mathrm{eff}}(T_1), exactly as in the static Adler case.

In the small–modulation regime kf≪1k_f\ll1 one has J0(kf)≈1J_0(k_f)\approx1, J1(kf)≈kf/2J_1(k_f)\approx k_f/2, and the coupling is nearly constant while the detuning oscillates. In leading order,

γeff=F2ωnr,Δωeff(T1)=Δω−Ωmkfcos⁡θm(T1),(11)\gamma_{\mathrm{eff}}=\frac{F}{2\omega_n r},\qquad\Delta\omega_{\mathrm{eff}}(T_1)=\Delta\omega-\Omega_m k_f\cos\theta_m(T_1), \tag*{(11)}

so that (10) reduces to the form analyzed by Lucas et al. [8]. Using their classification, three regimes follow from how often the instantaneous inequality ∣Δωeff(T1)∣≤γeff|\Delta\omega_{\mathrm{eff}}(T_1)|\leq\gamma_{\mathrm{eff}} is satisfied over one modulation period:

No synchronization:γeff<∣Δω∣−Ωmkf,(12)\text{No synchronization:}\qquad\gamma_{\mathrm{eff}}<|\Delta\omega|-\Omega_m k_f, \tag*{(12)}
Persistent synchronization:γeff≥∣Δω∣+Ωmkf,(13)\text{Persistent synchronization:}\qquad\gamma_{\mathrm{eff}}\geq|\Delta\omega|+\Omega_m k_f, \tag*{(13)}
Intermittent synchronization:∣Δω∣−Ωmkf≤γeff≤∣Δω∣+Ωmkf.(14)\text{Intermittent synchronization:}\qquad|\Delta\omega|-\Omega_m k_f\leq\gamma_{\mathrm{eff}}\leq|\Delta\omega|+\Omega_m k_f. \tag*{(14)}

In the intermittent regime, ∣Δωeff(T1)∣=γ|\Delta\omega_{\mathrm{eff}}(T_1)|=\gamma is crossed twice per modulation period, two saddle–node bifurcations occur, and the trajectory alternates deterministically between synchronized and nonsynchronized subintervals. These three regions correspond to the schematic

phase diagram shown in Figure 2(a) and to regions I–III in Ref. [8]. In the experimental and schematic Arnold-tongue plots (Figure 2) we use the plane of mean drive frequency and forcing amplitude (Ωu,F)(\Omega_u,F), which is directly related to the theoretical geometry in (Δω,γ)(\Delta\omega,\gamma) as discussed in Ref. [8].

A complementary, heuristic view is treating (9) as a static forcing term plus a near-resonant sideband generated by frequency modulation. The two contributions combine into an effective forcing scale

Feff(kf)=F2ωnrJ0(kf)2+[2J1(kf)]2,(15)F_{\mathrm{eff}}(k_f)=\frac{F}{2\omega_n r}\sqrt{J_0(k_f)^2+[2J_1(k_f)]^2}, \tag*{(15)}

where the square root reflects the root–mean–square combination of the carrier and first sideband amplitudes. A rough estimate of the broadened locking interval uses the static criterion ∣Δω∣≲Feff|\Delta\omega|\lesssim F_{\mathrm{eff}} applied quasi-statically over a modulation period. For small kfk_f one has J0(kf)≈1J_0(k_f)\approx1 and J1(kf)≈kf/2J_1(k_f)\approx k_f/2, giving Feff≈F1+kf2/(2ωnr)F_{\mathrm{eff}}\approx F\sqrt{1+k_f^2}/(2\omega_n r). Thus, frequency modulation redistributes drive power between the carrier and its sidebands, slightly increasing the effective coupling and widening the locking range in the broad sense of synchronization without changing the nominal drive amplitude.

Amplitude modulation

The case of amplitude modulation of the harmonic forcing is now considered. The amplitude modulation effects the instantaneous coupling strength and therefore the locking condition in time, producing the same three regimes observed under frequency modulation (persistent, intermittent and no synchronization regimes), but with a different alternation of the Arnold tongue. The forcing is now taken as

F(T0,T1)=F[1+kasin⁡θm(T1)]cos⁡(ΩuT0),(16)F(T_0,T_1)=F[1+k_a\sin\theta_m(T_1)]\cos(\Omega_uT_0), \tag*{(16)}

with now constant drive frequency Ωu\Omega_u, amplitude modulation depth kak_a, and modulation phase θm(T1)=ΩmT1\theta_m(T_1)=\Omega_mT_1.

With the slow time T1T_1 defined above and the modulation phase θm\theta_m already introduced, the slow-flow for the coupled amplitude–phase dynamics is now

drdT1=μ2r(1−r2R2)−F2ωn(1+kasin⁡θm)sin⁡θ,(17)\frac{dr}{dT_1}=\frac{\mu}{2}r\left(1-\frac{r^2}{R^2}\right)-\frac{F}{2\omega_n}(1+k_a\sin\theta_m)\sin\theta, \tag*{(17)}
dθdT1=Δω−F2ωnr(1+kasin⁡θm)cos⁡θ,(18)\frac{d\theta}{dT_1}=\Delta\omega-\frac{F}{2\omega_nr}(1+k_a\sin\theta_m)\cos\theta, \tag*{(18)}

where 0<ka<10<k_a<1 is the input amplitude modulation strength and FF is the nominal forcing level.

We first analyze persistent synchronization and the attenuation of the imposed modulation. From the phase slow-flow (18) it is convenient to define the time-dependent effective coupling

γeff(T1)=F2ωnr(T1)[1+kasin⁡θm(T1)].(19)\gamma_{\mathrm{eff}}(T_1)=\frac{F}{2\omega_nr(T_1)}[1+k_a\sin\theta_m(T_1)]. \tag*{(19)}

For frozen T1T_1, equilibria of (18) exist if and only if

∣Δω∣≤∣γeff(T1)∣.(20)|\Delta\omega|\leq|\gamma_{\mathrm{eff}}(T_1)|. \tag*{(20)}

Thus persistent synchronization throughout the modulation requires the uniform condition

∣Δω∣≤min⁡T1∣γeff(T1)∣.(21)|\Delta\omega|\leq\min_{T_1}|\gamma_{\mathrm{eff}}(T_1)|. \tag*{(21)}

Inside the synchronization regime the phase θ(T1)\theta(T_1) remains close to a slowly varying locked value, which we denote by θs(T1)\theta_s(T_1), and the amplitude remains near a forced limit-cycle value RFR_F. Writing

r(T1)=RF+δr(T1),∣δr∣≪RF,r(T_1)=R_F+\delta r(T_1),\qquad|\delta r|\ll R_F,

and linearizing (17) near RFR_F gives a sinusoidal response of the amplitude at the modulation frequency. If we write this response as

r(T1)=RF(1+kpsin⁡θm(T1)),r(T_1)=R_F(1+k_p\sin\theta_m(T_1)),

where kpk_p is the normalized output modulation depth of pp, we obtain

kp=δrRF=Fsin⁡θs2ωnRFμka=(1−RRF)ka.(22)k_p=\frac{\delta r}{R_F}=\frac{F\sin\theta_s}{2\omega_nR_F\mu}k_a=\left(1-\frac{R}{R_F}\right)k_a. \tag*{(22)}

Since RF>RR_F>R, the factor 1−R/RF1-R/R_F is strictly between 0 and 1. This implies that slow amplitude modulation is attenuated (kp<kak_p<k_a). To leading order in kak_a, it is therefore consistent to treat the amplitude as quasi-constant and replace r(T1)r(T_1) by RFR_F in the coupling.

With this approximation the effective coupling becomes

γeff(T1)≈γ0[1+kasin⁡θm(T1)],γ0=F2ωnRF.(23)\gamma_{\mathrm{eff}}(T_1)\approx\gamma_0[1+k_a\sin\theta_m(T_1)],\qquad\gamma_0=\frac{F}{2\omega_nR_F}. \tag*{(23)}

The minimum and maximum values over a modulation period are

γmin⁡=γ0(1−ka),γmax⁡=γ0(1+ka).\gamma_{\min}=\gamma_0(1-k_a),\qquad\gamma_{\max}=\gamma_0(1+k_a).

If the detuning lies in the window

γmin⁡<∣Δω∣<γmax⁡,(24)\gamma_{\min}<|\Delta\omega|<\gamma_{\max}, \tag*{(24)}

then the similar intermittent synchronization would occur. The crossing times satisfy

sin⁡θm(T1)=∣Δω∣/γ0−1ka∈[−1,1],(25)\sin\theta_m(T_1)=\frac{|\Delta\omega|/\gamma_0-1}{k_a}\in[-1,1], \tag*{(25)}

so there are two saddle–node bifurcations per modulation cycle in the intermittent synchronization regime.

Moreover, the straight lines ∣Δω∣=γ0(1±ka)|\Delta\omega|=\gamma_0(1\pm k_a) divide the geometry of synchronization diagram in the plane of detuning and nominal coupling (Δω,γ0)(\Delta\omega,\gamma_0) into regions I–III as introduced before. The persistent synchronization region satisfies

∣Δω∣≤γ0(1−ka),|\Delta\omega|\leq\gamma_0(1-k_a),

the intermittent synchronization region has the two-sided collar

γ0(1−ka)<∣Δω∣<γ0(1+ka),\gamma_0(1-k_a)<|\Delta\omega|<\gamma_0(1+k_a),

and no synchronization occurs when

∣Δω∣≥γ0(1+ka).|\Delta\omega|\geq\gamma_0(1+k_a).

Thus persistent synchronization forms an inner triangle in the (Δω,γ0)(\Delta\omega,\gamma_0) plane, while broad synchronization (persistent plus intermittent) occupies a larger outer triangle, with the intermittent regime filling the triangular side bands between the two boundaries which are marked as white dashed lines in Figure 2.

At fixed detuning ∣Δω∣|\Delta\omega|, these conditions can be rewritten in terms of the forcing amplitude FF. Any broad synchronization requires

F≥2ωnRF1+ka∣Δω∣,(26)F\geq\frac{2\omega_nR_F}{1+k_a}|\Delta\omega|, \tag*{(26)}

while persistent synchronization requires

F≥2ωnRF1−ka∣Δω∣.(27)F \ge\frac{2\omega_n R_F}{1-k_a}|\Delta\omega|. \tag*{(27)}

Amplitude modulation therefore lowers the threshold for synchronization by the factor 1/(1+ka)1/(1+k_a) in the broad sense (persistent plus intermittent) but raises the threshold for persistence to 1/(1−ka)1/(1-k_a), producing the widened outer triangle and the shrunken persistent triangle sketched in Figure 2(b).

To compare the effects of different slow modulation, frequency modulation acts as a time-varying detuning, thus moving along the classical boundary corresponds to shifting left and right in the (Δω,γ)(\Delta\omega,\gamma) plane. The broad synchronization region therefore becomes a trapezoid symmetric about the static Arnold–tongue boundary. In contrast, amplitude modulation acts as a time-varying coupling, so moving along the classical boundary corresponds to shifting up and down in (Δω,γ)(\Delta\omega,\gamma). The intermittent band then appears as two triangular side regions around a narrower persistent triangle, and the overall broad Arnold tongue widens accordingly.

Experimental Results and Discussion

Figure 1(c) illustrates our experimental externally driven self-oscillator. It is a 3D printed whistle following the design presented in [23], that is mounted as an obstacle in the middle of a one-dimensional waveguide. Air is supplied to the whistle via a volume flow controller (Bronkhorst). The air mass flow is adjusted to induce self-oscillations, while acoustic coupling to the waveguide is achieved through 4 mm apertures on the two lateral sides of the whistle. In such ducted configurations, mean-flow and acoustic interaction can be viewed as providing an effective source of acoustic gain and loss [24]. The forcing with incident acoustic waves is generated by a compression driver (CP850Nd, Beyma) located at one end of the waveguide. The waveguide, whistle, and external forcing arrangement are essentially the same as in Ref. [23], where the focus was on acoustic wave scattering under steady constant driving, whereas here the same platform is used to study synchronization under slowly varying, nonstationary forcing. In the experiments a dimensionless drive amplitude FF is prescribed in the control script, converted to a voltage command for the power amplifier and compression driver. The pressure inside the whistle is recorded using a microphone (46BD-FV G.R.A.S.), which is flush-mounted on the internal wall of the whistle. In the following analysis, an air flow of 10 L/min is steadily supplied to the whistle, resulting in self-sustained oscillations at a natural whistling frequency of 1160 Hz.

To quantify the degree of synchronization between the whistle and the external drive, the Phase Locking Value (PLV) employed in [25, 26] is used. Alternative metrics (such as the largest Lyapunov exponent estimated from delay embeddings [27] or phase-difference entropy [28]) turned out to be less robust in the present noisy, single-observable system. The instantaneous phases, θp(tk)\theta_p(t_k) and θF(tk)\theta_F(t_k), are obtained from taking the Hilbert transform of pressure signal p(t)p(t) and force F(t)F(t) at sampling times tkt_k. The raw phase difference sequence is then

Δθk=θp(tk)−θF(tk).(28)\Delta\theta_k=\theta_p(t_k)-\theta_F(t_k). \tag*{(28)}

With slow modulation of the forcing (in frequency or amplitude), Δθk\Delta\theta_k contains a deterministic component at the modulation phase θm(t)\theta_m(t). To isolate the residual clustering, the phase difference is demodulated by projecting Δθk\Delta\theta_k onto the known drive waveform and subtracting the fitted component. Specifically, Δθk\Delta\theta_k is approximated as

Δθk=aθF(tk)+Δθk⊥,(29)\Delta\theta_k=a\theta_F(t_k)+\Delta\theta_k^\perp, \tag*{(29)}

where

a=∑k=1NθF(tk)Δθk∑k=1NθF(tk)2,a=\frac{\sum_{k=1}^{N}\theta_F(t_k)\Delta\theta_k}{\sum_{k=1}^{N}\theta_F(t_k)^2},

and Δθk⊥\Delta\theta_k^\perp is the modulation-corrected phase. This projection removes the known deterministic drift from Δθk\Delta\theta_k. The modulation-corrected PLV (mcPLV) is then defined as

mcPLV⁡=∣1N∑k=1NeiΔθk⊥∣,0≤mcPLV⁡≤1.(30)\operatorname{mcPLV}=\left|\frac{1}{N}\sum_{k=1}^{N}e^{i\Delta\theta_k^\perp}\right|,\quad0\leq\operatorname{mcPLV}\leq1. \tag*{(30)}

Values near 1 indicate persistent synchronization, intermediate values typically reflect intermittent synchronization and values near 0 indicate no synchronization. In the maps below, mcPLV is reported together with circular histograms of Δθk⊥\Delta\theta_k^\perp for interpretability.

Experimental synchronization under time‐periodic modulation of forcing frequency

Guided by the schematic in Figure 2(a), the drive frequency is modulated slowly and synchronization is quantified using the mcPLV defined above. In the frequency-modulation experiments the instantaneous drive frequency is of the form

ΩF(t)=Ωu+kfΩmcos⁡(Ωmt),\Omega_F(t)=\Omega_u+k_f\Omega_m\cos(\Omega_m t),

with mean drive frequency Ωu\Omega_u, dimensionless modulation depth kfk_f, and modulation frequency Ωm\Omega_m (here fm=Ωm/2π=1f_m=\Omega_m/2\pi=1 Hz). The modulation is therefore much slower than the natural frequency 1160 Hz of the whistle, so the oscillator experiences a slowly varying detuning on the acoustic timescale. For convenience, the modulation depth is the frequency amplitude

Δf=kfΩm2π,\Delta f=\frac{k_f\Omega_m}{2\pi},

so that Δf\Delta f is the peak deviation of ΩF(t)\Omega_F(t) expressed in Hz. Figure 3 shows representative time series from three points in the synchronization map corresponding to persistent synchronization, intermittent synchronization, and no synchronization.

In the persistent regime, the instantaneous response frequency ωp(t)\omega_p(t) (blue) closely follows the modulated drive ΩF(t)\Omega_F(t) (orange), the unwrapped phase difference Δθ(t)\Delta\theta(t) remains confined near a constant offset (see Figure 4), its circular histogram is sharply peaked, and the mcPLV is close to 1. In the intermittent regime, each epoch of entrainment is interrupted by phase slips. During synchronized subintervals ωp(t)≈ΩF(t)\omega_p(t) \approx\Omega_F(t), slips appear as stepwise jumps in Δθ(t)\Delta\theta(t), producing multiple peaks in the histogram and intermediate mcPLV values between 0 and 1. The slip timing is set by the slow modulation, with one slip per modulation cycle, consistent with two saddle-node crossings per period when the instantaneous detuning traverses the static tongue boundary in Figure 2(a) and with the nonautonomous Adler dynamics described by Lucas et al. [8]. Qualitative phase-slip patterns are summarized in Figure 4(b), where persistent traces remain bounded, intermittent traces exhibit one phase slip per modulation cycle, and no synchronized traces grow without bound.

Phase slip dynamics for the three regimes

Figure 4. Phase slip dynamics for the three regimes. (a) amplitude modulation; (b) frequency modulation. Each panel plots the absolute value of the unwrapped phase difference, ∣Δθ(t)∣|\Delta\theta(t)|, for representative points. Colors: blue—persistent synchronization (bounded, no steps), red—intermittent synchronization (plateaus separated by discrete phase slips), green—no synchronization (diffusive drift with no plateau structure).

In the no synchronization regime, ωp(t)\omega_p(t) drifts relative to ΩF(t)\Omega_F(t), Δθ(t)\Delta\theta(t) behaves as an effectively unbounded walk, the phase-difference histogram is nearly uniform on (−π,π](-\pi,\pi], and the mcPLV is low. Figure 5 assembles mcPLV-based Arnold tongue diagrams in the (Ωu,F)(\Omega_u,F) plane for increasing modulation depth (Δf=(\Delta f = 0, 10/2π10/2\pi, 30/2π30/2\pi, 50/2π50/2\pi Hz). Here FF denotes the same dimensionless drive level used in the control script. For Δf=0\Delta f = 0 the classical V-shaped static tongue is recovered. As the modulation depth increases, the broad synchronization region (persistent plus intermittent synchronization) expands such that the boundary between no synchronization and intermittent synchronization moves outward, while the boundary between intermittent and persistent synchronization moves inward. Consequently, the persistent synchronization region shrinks, the intermittent synchronization region widens, and their union grows, in qualitative agreement with the regime geometry predicted by the nonautonomous Adler analysis of Lucas et al. [8]. All together, the instantaneous traces, phase-slip statistics, and mcPLV maps provide an experimental counterpart to the theoretical and numerical picture developed in Ref. [8].

Time-periodic Modulation of forcing Amplitude

The effect of slow amplitude modulation of the drive on synchronization is now examined, guided by the schematic geometry in Figure 2(b). The drive amplitude

is modulated slowly according to

F(t)=F[1+kasin⁡(Ωmt)]cos⁡(Ωut),\mathcal{F}(t) = F[1 + k_a \sin(\Omega_m t)]\cos(\Omega_u t),

with mean drive frequency Ωu\Omega_u, amplitude-modulation depth kak_a, and modulation angular frequency Ωm\Omega_m (again fm=Ωm/2π=1f_m = \Omega_m/2\pi= 1 Hz, much slower than the natural frequency). Synchronization is quantified using the same mcPLV as in the frequency-modulation case. Figure 6 shows mcPLV-based Arnold tongue diagrams in the (Ωu,F)(\Omega_u,F) plane for increasing modulation depths ka=0.15,0.30,k_a = 0.15, 0.30, and 0.500.50. As predicted by the time-varying coupling phase equation (18) and the condition ∣Δω∣≤∣γeff(T)∣|\Delta\omega| \leq|\gamma_{\mathrm{eff}}(T)|, the persistent synchronization region collapses onto an inner triangle, the intermittent synchronization region forms two triangular collars around it, and the union (broad synchronization) occupies a larger outer triangle whose size increases with kak_a, as sketched in Figure 2(b).

mcPLV-based Arnold tongues in the $(\Omega_u,F)$ plane for increasing amplitude-modulation depth $k_a$

Figure 6. (a–c) show mcPLV-based Arnold tongues in the (Ωu,F)(\Omega_u,F) plane for increasing amplitude-modulation depth kak_a. Regions I, II, and III denote no synchronization, persistent synchronization, and intermittent synchronization, respectively (as in Figure 2). Example points E1E_1 and E2E_2 in panels (b) and (c) mark two operating conditions inside region III at the same (Ωu,F)(\Omega_u,F) but different kak_a and their time traces are shown in Figure 7.

Time traces at two representative points in the intermittent region (E1E_1: ka=0.30k_a = 0.30; E2E_2: ka=0.50k_a = 0.50, with the same detuning and mean forcing level) are shown in Figure 7. During synchronized subintervals within each modulation cycle, the instantaneous frequencies ωp(t)\omega_p(t) and Ωp(t)\Omega_p(t) coevolve and the locked phase offset varies smoothly with the envelope. During nonsynchronized gaps the phase accumulates slips, leading to stepwise jumps in the unwrapped phase difference Δθ(t)\Delta\theta(t). For the same operating point in the intermittent region, increasing kak_a lengthens the unlocked portions and increases the amount of phase slips per modulation period, consistent with the theoretical picture of two saddle-node crossings of the locking condition per cycle and with the analytical intermittent window. This trend is visible in Figure 4(a): persistent traces remain bounded, intermittent traces show periodic slips and no synchronization traces show an unbounded drift.

Instantaneous frequencies and phase differences for two intermittent cases

Figure 7. Instantaneous frequencies and phase difference for two intermittent cases E1E_1 and E2E_2 (marked in Figure 6(b,Figure 6c)). Both points lie at the same frequency and forcing amplitude but have different amplitude modulation depths kak_a. As kak_a increases from E1E_1 to E2E_2, the fraction of time spent out of synchronization within one modulation period grows and additional phase slips appear in the phase difference, in direct analogy with the intermittent behavior observed under frequency modulation in Figure 3(b).

Instantaneous frequency and phase-difference analysis for the three synchronization regimes

Figure 3. Instantaneous frequency and phase–difference analysis for the three synchronization regimes (I–III) marked in Figure 5(c), at fixed drive level F=1.85F = 1.85 and mean drive frequencies Ωu/2π=1155\Omega_u/2\pi= 1155, 1146, 1140 Hz for (a), (b), and (c), respectively. Each row corresponds to one regime: left, instantaneous frequencies of the drive (orange) and oscillator (blue); middle, wrapped modulation-corrected phase difference Δθ(t)\Delta\theta(t); right, probability density of Δθ\Delta\theta together with the corresponding mcPLV. Panel (a) illustrates persistent synchronization (regime II), (b) intermittent synchronization with periodic phase slips (regime III), and panel (c) no synchronization (regime I). Tighter clustering and thus mcPLV value closer to 1 indicates stronger long–term synchronization.

Experimentally obtained Arnold–tongue maps based on mcPLV under increasing frequency modulation depth

Figure 5. Experimentally obtained Arnold–tongue maps based on mcPLV under increasing frequency modulation depth. Panels (a–d) show synchronization in the detuning–forcing plane for peak frequency shifts Δf=0,10/2π,30/2π,50/2π\Delta f = 0, 10/2\pi, 30/2\pi, 50/2\pi Hz, respectively (color code: light blue—no synchronization, region III; medium blue—intermittent, region II; deep blue—persistent, region I). As Δf\Delta f increases, the broad synchronized region (persistent and intermittent) widens, with a shrinking persistent core and a growing intermittent band, consistent with the nonautonomous detuning picture. For reference, panel (c) labels regions I–III.

The locked fraction over one slow period can be expressed in terms of the model parameters. Within the intermittent band (24), the theory predicts a locking duty cycle

Dlock=1πarccos⁡(∣Δω∣/γ0−1ka),D_{\mathrm{lock}} = \frac{1}{\pi}\arccos\left(\frac{\lvert\Delta\omega\rvert/\gamma_0 - 1}{k_a}\right),

with Dlock=1D_{\mathrm{lock}} = 1 in the persistent region and Dlock=0D_{\mathrm{lock}} = 0 above the outer boundary. At a fixed operating point inside the intermittent band, this implies that increasing kak_a reduces DlockD_{\mathrm{lock}} (more slip time) when ∣Δω∣<γ0\lvert\Delta\omega\rvert< \gamma_0 and increases DlockD_{\mathrm{lock}} when ∣Δω∣>γ0\lvert\Delta\omega\rvert> \gamma_0, in qualitative agreement with the measured fraction of time spent in synchronized intervals.

The amplitude response verifies the predicted attenuation of slow envelopes. Figure 8 compares normalized envelopes of drive and whistle. The measured output modulation depth kpk_p scales linearly with the input depth kak_a, with slope kp/ka≈0.034k_p/k_a \approx0.034. Independent measurements of the unforced and forced limit-cycle amplitudes RR and RFR_F give (1−R/RF)=0.028±0.016(1-R/R_F) = 0.028\pm0.016. This is consistent within uncertainty with the theoretical suppression factor kp=(1−R/RF)kak_p = (1-R/R_F)k_a derived in (22).

Amplitude response under slow amplitude modulation

Figure 8. Amplitude response under slow amplitude modulation. Top: normalized amplitude envelopes of drive and oscillator for ka=0.15k_a = 0.15, showing attenuation of the imposed envelope variation in the whistle response. The output modulation depth is kp=0.0051k_p = 0.0051, obtained by fitting the normalized response envelope with a single-harmonic model. Bottom: oscillator modulation depth kpk_p versus input modulation depth kak_a. The approximately linear relation (kp=0.034 kak_p = 0.034\,k_a) with slope smaller than one confirms the predicted amplitude envelopes attenuation and is in agreement with theory.

Taken together, these measurements support the nonautonomous synchronization picture for amplitude modulation. Specifically, the modulation acts as a vertical excursion in the Arnold–tongue plane, enlarging the broad synchronization region through a widened intermittent band while shrinking the persistent core. This behavior is complementary to the broadening but simultaneously to the reduce of the persistent synchronization under frequency modulation and is in quantitative agreement with the slow-flow analysis and its nonautonomous Adler reduction.

Aperiodic Modulation

To probe the limits of phase locking under time variation, the drive frequency is subjected to non-periodic variations and the ability of the whistle to maintain frequency entrainment is examined. In Figure 9(a), the driving frequency (orange) is modulated by two incommensurate sinusoids,

Figure 9. Instantaneous frequency tracking under (a) aperiodic modulation and (b) chaotic-frequency forcing. For aperiodic forcing, the driving frequency (orange) is modulated by two slow frequencies Ωm\Omega_m and 2Ωm\sqrt{2}\Omega_m. In the chaotic case, the driving frequency is modulated by a segment of a Lorenz attractor time series, while the oscillator frequency (blue) is plotted in the strict-synchronization regime. The persistent overlap of the two traces demonstrates robust phase locking despite the arbitrary, chaotic variations in the drive.

ΩF(t)=Ωu+kf,1Ωmsin⁡(θm)+kf,2Ωmsin⁡(2θm),\Omega_F(t)=\Omega_u+k_{f,1}\Omega_m\sin(\theta_m)+k_{f,2}\Omega_m\sin(\sqrt{2}\theta_m),

while in Figure 9(b) it follows a normalized segment of a Lorenz time series s(t)s(t),

ΩF(t)=Ωu+kfΩms(t).\Omega_F(t)=\Omega_u+k_f\Omega_m s(t).

In both cases the instantaneous response frequency ωp(t)\omega_p(t) (blue) follows the aperiodically modulated drive ΩF(t)\Omega_F(t) (orange) throughout the experiment.

This behavior is consistent with the nonautonomous Adler equation (see (10)): persistent synchronization for arbitrary detuning trajectories holds whenever the instantaneous locking condition

∣Δωeff(t)∣≤γeff(t)|\Delta\omega_{\mathrm{eff}}(t)|\leq\gamma_{\mathrm{eff}}(t)

is satisfied for all tt. In the present aperiodic tests, parameters are chosen such that sup⁡t∣ΩF(t)−ωn∣<γ0\sup_t|\Omega_F(t)-\omega_n|<\gamma_0, so that the trajectory remains well inside the persistent synchronization region of the Arnold tongue and does not approach the boundaries where saddle-node bifurcations and phase slips would occur.

Figure 9 shows that, despite quasiperiodic or chaotic detuning, the response frequency overlaps the drive almost perfectly and no phase slips occur over the entire record. These observations are therefore consistent, in the persistent regime, with the theoretical predictions for nonautonomous synchronization in frequency-modulated oscillators.

Conclusion

We present a unified theory–experiment study of synchronization in a self-sustained acoustic oscillator under slowly varying forcing. Previous studies have shown that frequency modulation introduces intermittent regimes and enlarges the broad sense synchronized region, and we find that amplitude modulation produces an analogous widening via a growing intermittent band. Experimental Arnold tongue maps verify the predicted geometries and regime-specific phase slip patterns. The attenuation of slow amplitude envelopes in the amplitude-modulated case is also verified. We further demonstrate persistent tracking under aperiodic (quasi-periodic and chaotic) detuning as long as the instantaneous mismatch remains below the effective coupling. Together, these results validate slow modulation effects for both amplitude and frequency modulation types in an aeroacoustic oscillator. These findings offer practical guidance: frequency modulation enlarges the usable synchronization domain without increasing nominal forcing level, in nonlinear acoustic resonators, tunable metamaterials, synchronized oscillator networks and synchronization-based communications.

Frequency–modulated phase: transformation to a nonautonomous Adler equation with time varying detuning

Starting from the slow phase dynamics (9),

dθdT1=Δω−FJ0(kf)2ωnrcos⁡θ−FJ1(kf)ωnrcos⁡(θ−θm(T1)),(31)\frac{d\theta}{dT_1}=\Delta\omega-\frac{FJ_0(k_f)}{2\omega_n r}\cos\theta-\frac{FJ_1(k_f)}{\omega_n r}\cos(\theta-\theta_m(T_1)), \tag*{(31)}

and using the definitions

γ1=FJ0(kf)2ωnr,γ2=FJ1(kf)ωnr.(32)\gamma_1=\frac{FJ_0(k_f)}{2\omega_n r},\qquad\gamma_2=\frac{FJ_1(k_f)}{\omega_n r}. \tag*{(32)}

we can write

dθdT1=Δω−[(γ1+γ2cos⁡θm)cos⁡θ+γ2sin⁡θmsin⁡θ]=Δω−R(θm)cos⁡(θ−δ(θm)),(33)\begin{aligned} \frac{d\theta}{dT_1}&=\Delta\omega-\left[(\gamma_1+\gamma_2\cos\theta_m)\cos\theta+\gamma_2\sin\theta_m\sin\theta\right]\\ &=\Delta\omega-R(\theta_m)\cos(\theta-\delta(\theta_m)), \tag*{(33)} \end{aligned}

with

R(θm)=(γ1+γ2cos⁡θm)2+(γ2sin⁡θm)2,(34)R(\theta_m)=\sqrt{(\gamma_1+\gamma_2\cos\theta_m)^2+(\gamma_2\sin\theta_m)^2}, \tag*{(34)}
δ(θm)=arctan⁡(γ2sin⁡θmγ1+γ2cos⁡θm).(35)\delta(\theta_m)=\arctan\left(\frac{\gamma_2\sin\theta_m}{\gamma_1+\gamma_2\cos\theta_m}\right). \tag*{(35)}

Introducing the new phase

ψ=θ−δ(θm)+π2,(36)\psi=\theta-\delta(\theta_m)+\frac{\pi}{2}, \tag*{(36)}

it follows

dψdT1=dθdT1−dθmdT1δ′(θm),cos⁡(θ−δ)=sin⁡ψ,(37)\frac{d\psi}{dT_1}=\frac{d\theta}{dT_1}-\frac{d\theta_m}{dT_1}\delta'(\theta_m),\qquad\cos(\theta-\delta)=\sin\psi, \tag*{(37)}

Eq. (33) becomes the nonautonomous Adler equation with time-varying detuning,

dψdT1=Δωeff(T1)+γeff(T1)sin⁡ψ,(38)\frac{d\psi}{dT_1} = \Delta\omega_{\mathrm{eff}}(T_1) + \gamma_{\mathrm{eff}}(T_1)\sin\psi, \tag*{(38)}

where

γeff(T1)=R(θm(T1)),(39)\gamma_{\mathrm{eff}}(T_1) = R\bigl(\theta_m(T_1)\bigr), \tag*{(39)}
Δωeff(T1)=Δω−Ωmδ′(θm(T1)),(40)\Delta\omega_{\mathrm{eff}}(T_1) = \Delta\omega- \Omega_m \delta'\bigl(\theta_m(T_1)\bigr), \tag*{(40)}
δ′(θm)=γ1γ2cos⁡θm+γ22γ12+2γ1γ2cos⁡θm+γ22.(41)\delta'(\theta_m) = \frac{\gamma_1\gamma_2\cos\theta_m+\gamma_2^2}{\gamma_1^2+2\gamma_1\gamma_2\cos\theta_m+\gamma_2^2}. \tag*{(41)}

For kf≪1k_f \ll1 one has J0(kf)≈1J_0(k_f) \approx1 and J1(kf)≈kf/2J_1(k_f) \approx k_f/2, hence γ2/γ1≈kf\gamma_2/\gamma_1 \approx k_f and

γeff(T1)=γ1+O(kf),(42)\gamma_{\mathrm{eff}}(T_1) = \gamma_1 + O(k_f), \tag*{(42)}
Δωeff(T1)=Δω−Ωmkfcos⁡θm(T1)+O(kf2).(43)\Delta\omega_{\mathrm{eff}}(T_1) = \Delta\omega- \Omega_m k_f\cos\theta_m(T_1) + O(k_f^2). \tag*{(43)}

Therefore, the phase dynamics becomes nearly constant coupling with a periodically modulated detuning, giving the Adler-type phase equation used in the main text.

References

References

  1. [1]R. Adler, A study of locking phenomena in oscillators, Proceedings of the IRE 34, 351 (2006).
  2. [2]A. Pikovski, M. Rosenblum, and J. Kurths, Synchronization: A Universal Concept in Nonlinear Sciences, Cambridge Nonlinear Science Series (Cambridge University Press, 2001).
  3. [3]S. H. Strogatz, Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering (Chapman and Hall/CRC, 2024).
  4. [4]S. Strogatz, Sync: The emerging science of spontaneous order (Penguin UK, 2004).
  5. [5]R. Jensen, Synchronization of driven nonlinear oscillators, American Journal of Physics 70, 607 (2002).DOI
  6. [6]P. Gandhi, E. Knobloch, and C. Beaume, Dynamics of phase slips in systems with time-periodic modulation, Phys. Rev. E 92, 062914 (2015).DOI
  7. [7]Y. F. Suprunenko, P. T. Clemson, and A. Stefanovska, Chronotaxic systems: a new class of self-sustained nonautonomous oscillators, Phys. Rev. Lett. 111, 024101 (2013).pubmed.ncbi.nlm.nih.gov/23889404
  8. [8]M. Lucas, J. Newman, and A. Stefanovska, Stabilization of dynamics of oscillatory systems by nonautonomous perturbation, Phys. Rev. E 97, 042209 (2018).DOI
  9. [9]J. Newman, M. Lucas, and A. Stefanovska, Stabilization of cyclic processes by slowly varying forcing, Chaos 31, 10.1063/5.0066641 (2021).DOI
  10. [10]B. Bergoet, S. Terrien, and C. Vergez, Predicting transient dynamics in a model of reed musical instrument with slowly time-varying control parameter, Chaos 34, 10.1063/5.0190512 (2024).
  11. [11]S. Terrien, B. Bergoet, C. Vergez, and S. Missoum, Basins of attraction in a dynamical system with a time-varying control parameter: The case of attack transients in a simple model of reed musical instrument, J. Sound Vib. , 119241 (2025).DOI
  12. [12]J. Newman, J. P. Scott, J. R. Adams, and A. Stefanovska, Intermittent phase dynamics of non-autonomous oscillators through time-varying phase, Physica D 461, 134108 (2024).DOI
  13. [13]C. Bourquard, A. Faure-Beaulieu, and N. Noiray, Whistling of deep cavities subject to turbulent grazing flow: intermittently unstable aeroacoustic feedback, J. Fluid Mech. 909, A19 (2021).DOI
  14. [14]E. Boujo, M. Bauerheim, and N. Noiray, Saturation of a turbulent mixing layer over a cavity: response to har- monic forcing around mean flows, J. Fluid Mech. 853, 386 (2018).
  15. [15]A. K. Stoychev, T. Pedergnana, and N. Noiray, Nonlinear acoustics of an aperture under grazing flow, Proceedings of the Royal Society A 480, 20230718 (2024).
  16. [16]J. D. Crawford and E. Knobloch, Symmetry and symmetry-breaking bifurcations in fluid dynamics, Annu. Rev. Fluid Mech. 23, 341 (1991).DOI
  17. [17]H. Nakao, Phase reduction approach to synchronisation of nonlinear oscillators, Contemporary Physics 57, 188 (2016).
  18. [18]A. Balanov, N. Janson, D. Postnov, and O. Sosnovtseva, From simple to complex (Springer, 2009).
  19. [19]A. H. Nayfeh and D. T. Mook, Nonlinear oscillations (John Wiley & Sons, 2024).
  20. [20]J. Warmiński, Synchronisation effects and chaos in the van der pol-mathieu oscillator, Journal of Theoretical and Applied Mechanics 39, 861 (2001).
  21. [21]R. V. Jensen, Synchronization of randomly driven nonlinear oscillators, Phys. Rev. E 58, R6907 (1998).DOI
  22. [22]D. L. Colton, R. Kress, and R. Kress, Inverse acoustic and electromagnetic scattering theory, Vol. 93 (Springer, 1998).
  23. [23]A. K. Stoychev, X. Guo, U. Kuhl, and N. Noiray, Synchronization driven acoustics: The nonlinear scattering of a self-oscillating meta-atom, arXiv:2508.14819 [physics.app-ph] (2025).arxiv.org/abs/2508.14819
  24. [24]Y. Aurégan and P. Vagneux, PT-symmetric scattering in flow duct acoustics, Phys. Rev. Lett. 118, 174301 (2017).DOI
  25. [25]J.-P. Lachaux, E. Rodriguez, J. Martinerie, and F. J. Varela, Measuring phase synchrony in brain signals, Human brain mapping 8, 194 (1999).
  26. [26]S. Aydore, A. Pantazis, and R. M. Leahy, A note on the phase locking value and its properties, Neuroimage 74, 231 (2013).DOI
  27. [27]M. T. Rosenstein, J. J. Collins, and C. J. De Luca, A practical method for calculating largest Lyapunov exponents from small data sets, Physica D 65, 117 (1993).DOI
  28. [28]P. Tass, M. Rosenblum, J. Weule, J. Kurths, A. Pikovsky, J. Volkmann, A. Schnitzler, and H.-J. Freund, Detection of n:m phase locking from noisy data: application to magnetoencephalography, Phys. Rev. Lett. 81, 3291 (1998).

Paper details

vlnc:W_m8w9jyyz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents