We investigate the non-linear mechanical dynamics of a nano-optomechanical mirror formed by a suspended membrane pierced by a photonic crystal. By applying to the mirror a periodic electrostatic force induced by interdigitated electrodes integrated below the membrane, we evidence superharmonic resonances of our nano-electro-mechanical system; the constant phase shift of the oscillator across the resonance tongues is observed on the onset of principal harmonic and subharmonic excitation regimes.
The phenomenon of superharmonic resonances of an oscillator in response to a periodic forcing function occurs in most non-linear oscillators.1 For driving frequencies Ωd in the vicinity of a rational fraction of the oscillator natural frequency (), the non-linearity adjusts the frequency of the free-oscillation to exactly n times the frequency of the forcing, i.e., . Each resonance occurs within a tongue, axes of which are the driving frequency and driving amplitude. Within this tongue, amplitude and phase, respectively, feature a resonance and a shift. Such resonances have already been implemented in NEMS, by use of harmonic driving2–4 as well as, more scarcely, subharmonic driving5 without observation of the phase evolution across the tongue in the subharmonic excitation regime. At the same time, significant advances have been achieved in Nano-Electro-Mechanical Systems (NEMS), technology by hybridizing optoelectronic/photonic devices in NEMS architectures to form hybrid nano-opto-electromechanical systems (NOEMS).6–10 In this article, resonances under subharmonic driving are observed in an individual NEMS, formed by a suspended photonic crystal mirror. The phase evolution is, in particular, investigated: the power-dependence as well as the excitation frequency-dependence of the phase trajectories for several non-linear orders up to n = 8 are measured; these phase portraits reveal the linear evolution of the phase shift between the oscillator and the initial drive while “crossing” the resonance regions at constant force amplitude.
The NEMS device under study is a deformable mirror, formed by a two-dimensional photonic-crystal suspended InP membrane,11,12 integrating interdigitated electrodes underneath (see Fig. 1(a)) to avoid mechanical quality factor degradation.13 The external driving scheme relies on dielectric transduction.4,6,14 The suspended membrane experiences a force resulting from an electric field gradient created by biasing the periodic set of electrodes, allowing for an actuation of its motion in the MHz range. InP suspended membrane is actuated by applying an AC voltage V(t) = VAC cos(2πΩdt) with a frequency Ωd (cf. Fig. 1(a)). This modulated voltage induces a force that drives the out-of-plane oscillation of the dielectric resonator around its equilibrium position. This oscillation translates into a modulation of the mutual capacitance C(x) between the IDE's array and the hovering membrane; x denotes here the out-of-plane displacement of the membrane from its equilibrium position. The resulting position-dependent capacitive force corresponds to the first derivative of the charging energy for a given bias voltage V: with .6 This static force, which is proportional to the field gradient along the x direction, leads to an energy exchange between electronic and mechanical subsystems at a rate G. For the quantitative characterization of our electrostatic actuator, finite element modeling (FEM) is carried out yielding of 4 nm/V for a membrane-electrode distance of 400 nm.
(a) Artificially colored scanning electron microscope image of the suspended InP photonic-crystal membrane driven by integrated interdigitated electrodes. (b) Schematics of the interferometric setup: The suspended InP membrane is placed in a vacuum chamber maintained below mbar at room temperature. The NEMS resonator is electrostatically actuated by applying a voltage V(t) to an electrode via the HF2LI lock-in amplifier (Lock-in). The driven membrane displacement is optically probed in an interferometric setup. The optical signal at the output of the interferometer is collected with a photodiode (PD) and sent back to a lock-in amplifier referenced at the driving frequency. (c) Measured frequency noise spectrum recorded by scanning the drive frequency Ωd from 1 to 10 MHz at 5 V AC drive voltage. The FEM simulated profiles (magnified for clarity) of the observed mechanical modes labeled by their orders are shown.
(a) Artificially colored scanning electron microscope image of the suspended InP photonic-crystal membrane driven by integrated interdigitated electrodes. (b) Schematics of the interferometric setup: The suspended InP membrane is placed in a vacuum chamber maintained below mbar at room temperature. The NEMS resonator is electrostatically actuated by applying a voltage V(t) to an electrode via the HF2LI lock-in amplifier (Lock-in). The driven membrane displacement is optically probed in an interferometric setup. The optical signal at the output of the interferometer is collected with a photodiode (PD) and sent back to a lock-in amplifier referenced at the driving frequency. (c) Measured frequency noise spectrum recorded by scanning the drive frequency Ωd from 1 to 10 MHz at 5 V AC drive voltage. The FEM simulated profiles (magnified for clarity) of the observed mechanical modes labeled by their orders are shown.
The displacement of the forced NEMS resonator is probed using a standard optical interferometry setup: The photonic crystal membrane is placed in a vacuum chamber pumped down to mbar and acts as the end mirror in one of the arms of the interferometer as depicted in Fig. 1(b). The light from a He-Ne laser of wavelength 632.8 nm and of power 1.5 mW is sent into the interferometer and focused on the membrane with a 0.4 NA microscope objective, down to a beam waist of about 0.6 μm. The phase shift of the laser light reflected by the membrane surface (with a reflectivity R of about 50%) is measured at the interferometer output, by recording its oscillatory components with a photodetector and a lock-in amplifier. All measurements are done at room temperature. The measured mechanical spectrum is shown in Fig. 1(c) for the actuated membrane at a 5 V AC drive, highlighting the excitation of several mechanical modes.
In a first series of experiments, we perform forward and downward frequency sweeps around the fundamental frequency (i.e., harmonic excitation) at different AC drive amplitudes. Amplitude and phase (denoted ) evolutions are recorded simultaneously in an homodyne detection configuration for upward (red, right triangle) and downward (blue, left triangle) frequency sweeps at 1 V and 10 V AC drive voltage (see Fig. 2). In the following, will denote the phase at the mechanical resonance frequency .
Measured amplitude (upper panels; (a) and (b)) and phase (lower panels; (c) and (d)) spectrum for upward (red, right triangle) and downward (blue, left triangle) frequency sweeps and for 1 V (left panels; (a), (c)) and 10 V (right panels; (b), (d)) AC drive amplitudes. The green solid line is obtained by fitting the experimental curve with the non-linear Eq. (1).
Measured amplitude (upper panels; (a) and (b)) and phase (lower panels; (c) and (d)) spectrum for upward (red, right triangle) and downward (blue, left triangle) frequency sweeps and for 1 V (left panels; (a), (c)) and 10 V (right panels; (b), (d)) AC drive amplitudes. The green solid line is obtained by fitting the experimental curve with the non-linear Eq. (1).
For a low actuation amplitude (VAC = 1 V, see Figs. 2(a) and 2(c)), the amplitude and phase spectra are identical for bi-directional sweeps. In this regime, the eigen-frequency of the fundamental drum mode is found at 2.753 MHz and ; its full width of kHz is obtained by measuring the ring-down decay rate. This measurement is done at low VAC to avoid extra non-linear damping.14 We calibrate a displacement of 2.08 nm at the beam's center using the method described in Ref. 15, in fair agreement with a FEM computation yielding 4 nm.
The measured spectra start to exhibit an asymmetry around 2 V. For amplitudes larger than this critical drive voltage, amplitude (Fig. 2(b)) and phase (Fig. 2(d)) response curves become bistable between upward and downward sweeps. The observed hysteresis reveals spring hardening effect (see Fig. 2(b)), which is ascribed to mechanical nonlinearities predominantly due to the tensions in the tethers induced by their transverse displacements.16 This non-linear response can be understood by use of the non-linear differential equation17
where ω0 () is the angular frequency of the membrane. h and Ωd, respectively, denote the amplitude and frequency of the external drive and βi the reduced ith-order nonlinearity coefficients. To account for our NEMS device nonlinearity, the measured hysteresis is fitted to the steady-state solution of Eq. (1), as plotted by the green line in Fig. 2(b). From the fit retaining solely the cubic-order non-linearity, we obtain a value of .
Forcing the fundamental mechanical mode at frequencies either leads to quasi-periodic oscillations, or, when the forcing amplitude is sufficiently high, to periodic oscillations at a frequency . In the latter case, the resonant driving of the oscillator by the periodic force occurs within regions with a tongue shape, the so-called resonance tongues. Such regions are shown in Fig. 3, plotting the evolution of the noise spectrum of the fundamental mechanical mode while modulating the electric load in proximity of at various AC drive amplitudes. All 2D surface plots, up to the 8th-order of subharmonics, exhibit a characteristic “tongue” shape with a widened linewidth for increased driving amplitude. The threshold bias Vth above which forced periodic oscillations occur (see Fig. 3(b)) increases for increasing values of n, due to lower non-linear mechanical coefficients. The width of the tongues for a constant AC drive amplitude decreases with n (see Fig. 3(b) for corresponding to a bias in the saturation regime for every investigated n values). This reduction of the tongue widths may arise from the emergence of higher orders of perturbation,18 dissipation due to temperature rise19,20 or weaker non-linear coefficients. These two former physical phenomena also explain the observed saturation of the tongue width at high voltage (see inset Fig. 3(b)).
(a) Resonance tongues in amplitude space under subharmonic drive at , with n ranging from 2 to 8. (b) Evolution of the tongue width (blue circle) for a fixed AC amplitude of 10 V corresponding to saturation regime for every n values and threshold voltage (brown square) as function of . Inset: Evolution of for n = 3 as a function of AC bias; the threshold bias (vertical dashed line) as well as the saturation region (shaded region) are highlighted.
(a) Resonance tongues in amplitude space under subharmonic drive at , with n ranging from 2 to 8. (b) Evolution of the tongue width (blue circle) for a fixed AC amplitude of 10 V corresponding to saturation regime for every n values and threshold voltage (brown square) as function of . Inset: Evolution of for n = 3 as a function of AC bias; the threshold bias (vertical dashed line) as well as the saturation region (shaded region) are highlighted.
The phase dynamics of the membrane can be described as slowly varying quadratures x(t) = Xcos(2πΩ1stt) + Ysin(2πΩ1stt). The solution is then pictured as a trajectory in the X-Y phase plane. The corresponding phase and quadrature components of position can be experimentally measured during the ring-down time, as shown in Fig. 4(b). The external drive of the NEMS resonator is switched off at frequencies . In each ring-down time, the in-phase and quadrature components of the position are recorded, following a single trajectory from the initial stationary state to the origin with 300 phase points.
Phase-response curves under subharmonic driving. Following the line of the constant drive strength, the initial phase is extracted from the corresponding measured phase portraits. For the same initial excitation drive, the detuning dependence of the initial phase is plotted in (a) below the saturation regimes. Red circles and green rectangles, respectively, represent the phase-response curves for driving frequencies at for 2V AC drive and at for 10 V AC drive; note the different abscissa axes identified by their colors (red bottom axis corresponding to the 2nd-order subharmonic excitation and green top axis corresponding to the 6th-order subharmonic excitation). The black line on (a) indicates the theoretical expected value for a given frequency detuning. Corresponding phase portraits are displayed in (b) for different detuning, corresponding to initial phase shifts of, respectively, , 0, , and .
Phase-response curves under subharmonic driving. Following the line of the constant drive strength, the initial phase is extracted from the corresponding measured phase portraits. For the same initial excitation drive, the detuning dependence of the initial phase is plotted in (a) below the saturation regimes. Red circles and green rectangles, respectively, represent the phase-response curves for driving frequencies at for 2V AC drive and at for 10 V AC drive; note the different abscissa axes identified by their colors (red bottom axis corresponding to the 2nd-order subharmonic excitation and green top axis corresponding to the 6th-order subharmonic excitation). The black line on (a) indicates the theoretical expected value for a given frequency detuning. Corresponding phase portraits are displayed in (b) for different detuning, corresponding to initial phase shifts of, respectively, , 0, , and .
For harmonic and subharmonic driving, the phase difference between the oscillator and the driving force is bounded between (see Figs. 2(c) and 2(d)) and depends on the excitation frequency detuning. If the resonance tongue is “crossed” along a line of constant AC drive, i.e., if the amplitude of the force is kept constant whereas its frequency is varied, the amplitude response goes across a maximum while a phase shift is expected to vary by π. To investigate this phase shift across the tongue, ring-down measurements were performed at different frequency detuning (see inset Fig. 3(b)) for different subharmonic excitation orders. Figure 4(a) plots the measured phase difference for the 2nd- and the 6th-order subharmonic excitations, respectively, at 2 V and 10 V AC drive and at various detuning around the corresponding superharmonic resonances for the same initial drive phase. Some phase portraits for n = 2 and n = 6 are shown in Figure 4(b) at some specific detuning. As expected, the phase shift is found to remain bounded within the width the resonance tongue between . It exhibits a linear behavior, varying by π while “crossing” the tongue. Near the border of the tongue, the phase-position is almost merged to the origin in the phase space diagram, yielding an uncertainty contribution in the derived phase. In the middle of the resonant tongue (), zero degrees of phase shift are measured within 4% error.
In this article, the non-linear dynamics of a fully integrated NEMS under subharmonic excitation is experimentally investigated. Wedge-like resonance domains are observed in the phase plane spanned by the forcing frequency and amplitude; the linear phase shift between the oscillator and the force across the resonance region is measured. Such resonances, in a regime of self-sustained oscillation, could be used for implementing frequency synchonization and chaotic motion in a NOEMS device. Combined to the scalability of the platform and the possible integration of other on-chip functionalities,21 they may open the way to nano-opto-electromechanical oscillator networks, enabling the investigation of many-body dynamics in clustering systems22,23 or the implementation of neurocomputing2,5 and high precision sensing.24–26
We thank S. Barbay as well as K. Makles, T. Briant, S. Delglise, P.-F. Cohadon, and A. Heidmann for fruitful discussions and G. Hwang for his help with the critical point drying step. This work is supported by the “Agence Nationale de la Recherche” programme MiNOToRe, the French RENATECH network, and the Marie Curie Innovative Training Networks (ITN) cQOM.