betaFeedback

WorksPhysics and Astronomy

Nonlinear Mode Coupling in Silicon Nitride Membrane Resonators

Abstract

Nonlinear interactions between vibrational modes play a crucial role in understanding the dynamical response of nanomechanical resonators. Here, we report the experimental observation and theoretical modeling of nonlinear mode coupling in a high-stress square silicon nitride membrane resonator. We quantify frequency shifts of the fundamental mode arising from tension-mediated geometric nonlinearity by increasing the amplitude of the fundamental mode and higher-order flexural modes. A quantitative theoretical framework based on Kirchhoff-Love plate theory is developed, which incorporates both intrinsic Duffing nonlinearity and nonlinear intermodal coupling and shows good agreement with experimental measurements for the (1,1)-(2,1) and (1,1)-(2,2) mode pairs. We further compute the nonlinear coupling matrix across mode families, revealing the role of mode symmetry and spatial overlap in governing intermodal interactions. These results establish nonlinear mode coupling as a controllable resource for multimode frequency tuning and mechanical transduction.

Mechanical resonators[1, 2] play an important role in both fundamental research and emerging technologies. Their exceptional sensitivity to external stimuli underpins advances across a wide range of research areas, including force sensing[3, 4, 5], mass spectrometry[6] [7], magnetometry[8–11], accelerometers[12, 13], pressure sensors[14–16], magnetic resonance imaging[17–19], and scanning probe microscopy[20, 21]. Beyond sensing, mechanical resonators are widely used in communications and signal processing[22–24]. They serve as key interfaces in hybrid quantum systems[25, 26], coupling mechanical motion to optical[27, 28], electrical[29, 30], and spin[31–33] degrees of freedom. These capabilities establish mechanical resonators as a unifying platform bridging classical and quantum physics[34–37], combining technological relevance with fundamental discovery[38, 39].

Among various classes of mechanical resonators, high-stress silicon nitride (Si3N4\mathrm{Si}_3\mathrm{N}_4) membranes have emerged as a leading platform. Their resonance frequencies, quality factor, and mechanical mode shapes can be precisely engineered through device geometry and boundary conditions. Their exceptionally high quality factors[40–45] and compatibility with chip-scale integration make these resonators well suited for precision sensing and for probing mechanical dynamics at room and cryogenic temperatures[46, 47]. As a result, these resonators are playing an increasingly important role in exploring light-matter interaction[48, 49], ground state cooling[50], entanglement[51], optomechanical squeezing[52], optical-to-microwave photon transducers[53], quantum memory[54], gravitational-wave detection[55], dark matter detector[56], and advanced precision sensing[57].

Most studies of Si3N4\mathrm{Si}_3\mathrm{N}_4 membrane resonators have focused on single, isolated vibrational modes[58–60], whereas practical membranes support a dense spectrum of mechanical resonances whose interactions can strongly influence device dynamics and performance. At sufficiently large oscillation amplitudes, mechanical resonators enter into a nonlinear regime governed by tension-induced geometric nonlinearity, enabling nonlinear modal coupling that gives rise to amplitude-dependent frequency tuning, energy exchange between orthogonal modes, and dynamic control of the mechanical response[61]. These effects enable multimode functionalities including hybridized resonances[62, 63], enhanced energy transfer[64, 65], bandwidth broadening[66], and tunable dissipation pathways[67–69]. Despite their importance, nonlinear modal interactions in square Si3N4\mathrm{Si}_3\mathrm{N}_4 membranes remain largely unexplored, even though they are crucial for advanced sensing, signal processing, and hybrid classical–quantum transduction architectures.

In this work, we present an experimental and theoretical study of nonlinear intra- and intermodal coupling in a high-stress square Si3N4\mathrm{Si}_3\mathrm{N}_4 membrane resonator. By driving multiple flexural modes into the nonlinear regime, we observe amplitude-dependent frequency shifts arising from both intramodal (intrinsic Duffing nonlinearity) and intermodal coupling. We develop a simple yet quantitative theoretical framework based on Kirchhoff–Love plate theory that accurately captures these effects. Through a systematic mapping of coupling strengths across mode families, we compute the coupling matrix between the fundamental and higher-order modes, revealing the role of modal overlap in governing nonlinear interactions. Furthermore, we compute the selective frequency tuning of a target mode via the excitation of orthogonal modes, establishing nonlinear mode coupling as a powerful resource for engineering multimode functionality in highstress Si3_3N4_4 membrane resonators.

The experiments are performed on a suspended silicon nitride square membrane from Norcada with side length (LL) of 500 μ\mum and a thickness (hh) of 100 nm clamped to a 500 μ\mum thick silicon substrate. The Si3_3N4_4/Si substrate is mounted on a piezoelectric actuator that enables out-of-plane excitation via an applied AC voltage V(t)=V0cos⁡(ωt)V(t) = V_0 \cos(\omega t). The frequency-amplitude response of the membrane is measured optically using a laser Doppler vibrometer (Polytec MSA 500). All measurements were performed under high vacuum at a base pressure of 2×10−52 \times10^{-5} mbar to minimize air damping. Figure 1(a) shows a schematic of the experimental setup, in which a lock-in amplifier (Zurich Instruments HF2LI) is used to simultaneously actuate and detect the mechanical motion of the membrane. An optical micrograph of the membrane and the measured mode spectrum between 700 and 1700 kHz are presented in Figure 1(b) and Figure 1(c), respectively. Three vibrational modes are observed within this frequency range; the inset shows the measured spatial mode shapes, with modes labeled (n,m)(n,m) according to the number of anti-nodes along the two in-plane directions of the membrane.

Measurement setup and vibrational modes of the device

Figure 1. Measurement setup and vibrational modes of the device. (a) Schematic of the laser Doppler vibrometry (LDV) based measurement setup. A Polytec MSA-500 system is used to optically probe the out-of-plane motion of Si3_3N4_4 membrane, while electrical actuation and signal detection are performed using a Zurich instruments HF2LI lock-in amplifier. (b) Optical micrograph of the device, showing a 500 μ\mum ×\times 500 μ\mum Si3_3N4_4 membrane suspended over a silicon substrate. (c) Measured displacement amplitude spectrum with drive frequency showing multiple mechanical resonances. Higher frequency modes are scaled by a factor of ×5\times5 and ×25\times25 for visibility. Inset shows the corresponding spatial mode shapes of the resonant modes.

We first characterize the out-of-plane vibrational modes at low drive amplitudes, where both intramodal and intermodal coupling are negligible. Figure 2(b) shows the linear frequency response of the (1,1)(1,1) mode, while the responses of the (2,1)(2,1) and (2,2)(2,2) modes are presented in Figure S4. In this regime, all modes exhibit Lorentzian line shapes. The measured resonance frequencies of the (1,1)(1,1), (2,1)(2,1), and (2,2)(2,2) modes are 814 kHz, 1286 kHz, and 1628 kHz, with corresponding quality factors of 5.2×1045.2 \times10^{4}, 12.5×10412.5 \times10^{4}, and 11.3×10411.3 \times10^{4}, respectively. Figure 2(a) displays the nonlinear frequency response of the (1,1)(1,1) mode, measured at increasing drive amplitudes, arising from intrinsic Duffing nonlinearity.

Figure 2. Intramodal coupling and Duffing constant. (a) Nonlinear frequency response of the (1, 1) mode, measured at increasing drive strengths. (b) Linear response of the (1, 1) mode at the lowest drive amplitude; solid lines indicate Lorentzian fits used to extract the resonance frequency and quality factor. (c) Amplitude-dependent frequency shift corresponding to the maximum displacement amplitudes extracted from (a). Solid line show fit using Eq. (6) to extract the Duffing coefficient. (d) Comparison of the experimentally extracted Duffing coefficients with theoretical values calculated from Eq. (5) for the (1, 1), (2, 1), and (2, 2) modes.

Figure S4. (a,d) Nonlinear frequency responses of mode (2, 1) and mode (2, 2) at increasing driving force. (b,e) Linear responses of the (2, 1) and (2, 2) modes at the lowest drive amplitudes; solid lines indicate Lorentzian fits used to extract the resonance frequencies and quality factors. (c,f) Amplitude-dependent frequency shifts corresponding to the maximum displacement amplitudes extracted from (a,d). Solid lines show fits to extract the Duffing coefficients.

We describe the membrane dynamics within the framework of Kirchhoff–Love plate theory[70]. The transverse displacement w(x,y,t)w(x,y,t) satisfies

D∇4w(x,y,t)−(Nxx∂2w∂x2+2Nxy∂2w∂x∂y+Nyy∂2w∂y2)+ρh∂2w∂t2=0,(1)D\nabla^{4}w(x,y,t)-\left(N_{xx}\frac{\partial^{2}w}{\partial x^{2}}+2N_{xy}\frac{\partial^{2}w}{\partial x\partial y}+N_{yy}\frac{\partial^{2}w}{\partial y^{2}}\right)+\rho h\frac{\partial^{2}w}{\partial t^{2}}=0, \tag*{(1)}

where DD is the flexural rigidity, Nxx,Nxy,NyyN_{xx},N_{xy},N_{yy} are the in-plane stress resultants, ρ\rho is the mass density of the membrane and hh is the thickness. Within the high stressed thin membrane approximation[71] the in-plane stress components can be expressed as

Nxx=Eh(ϵ0+12(∂w∂x)2),Nyy=Eh(ϵ0+12(∂w∂y)2),Nxy=12Eh(∂w∂x∂w∂y),(2)\begin{aligned} N_{xx} &= Eh\left(\epsilon_{0}+\frac{1}{2}\left(\frac{\partial w}{\partial x}\right)^{2}\right),\\ N_{yy} &= Eh\left(\epsilon_{0}+\frac{1}{2}\left(\frac{\partial w}{\partial y}\right)^{2}\right),\\ N_{xy} &= \frac{1}{2}Eh\left(\frac{\partial w}{\partial x}\frac{\partial w}{\partial y}\right), \tag*{(2)} \end{aligned}

where EE is the Young’s modulus, and ϵ0\epsilon_{0} is the built-in tensile strain of the membrane. The transverse displacement is expressed as w(x,y,t)=ϕ(x,y)z(t)w(x,y,t)=\phi(x,y)z(t), where ϕ(x,y)=sin⁡(nπxL)sin⁡(mπyL)\phi(x,y)=\sin\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right) is the spatial mode shape and z(t)z(t) is the corresponding out of plane displacement. Substituting the mode expansion into Eq. (1) and applying the Galerkin discretizations procedure, followed by the inclusion of linear damping (ν\nu) and external driving force (fdcos⁡(ωdt)f_d\cos(\omega_d t)), yields the Duffing equation of motion

z¨+ωnm2z+νz˙+αnmz3=fdcos⁡(ωdt),(3)\ddot{z}+\omega_{nm}^{2}z+\nu\dot{z}+\alpha_{nm}z^{3}=f_d\cos(\omega_d t), \tag*{(3)}

The expressions for the linear resonance frequency ωnm\omega_{nm} and the Duffing coefficient αnm\alpha_{nm} for a general mode (n,m)(n,m) are given by

ωnm2=Eϵ0π2(n2+m2)ρL2,(4)\omega_{nm}^{2}=\frac{E\epsilon_{0}\pi^{2}(n^{2}+m^{2})}{\rho L^{2}}, \tag*{(4)}
αnm=3E32ρ(πL)4[n4+m4−23n2m2].(5)\alpha_{nm} = \frac{3E}{32\rho}\left(\frac{\pi}{L}\right)^4\left[n^4 + m^4 - \frac{2}{3}n^2m^2\right]. \tag*{(5)}

We use the measured resonance frequency of the fundamental (1, 1) mode to extract the built-in tensile strain ϵ0=3.7×10−3\epsilon_0 = 3.7 \times10^{-3} from Eq. (4). This value is subsequently used to estimate the theoretical resonance frequencies of higher-order modes, shown by the dashed lines in Figure 1(c), which are in good agreement with the experimentally observed higher mode frequencies. Solving Eq. (3) using the method of multiple scales yields an amplitude-dependent nonlinear pulling of the resonance frequency,

ωnmpeak=ωnm+3αnm8ωnm(znmpeak)2,(6)\omega_{nm}^{\mathrm{peak}} = \omega_{nm} + \frac{3\alpha_{nm}}{8\omega_{nm}}\left(z_{nm}^{\mathrm{peak}}\right)^2, \tag*{(6)}

where zpeakz^{\mathrm{peak}} denotes the maximum displacement amplitude of the (n,m)(n,m) mode. The measured values of zpeakz^{\mathrm{peak}} is plotted as a function of the shift in peak frequency

(ωnmpeak−ωnm)(\omega_{nm}^{\mathrm{peak}}-\omega_{nm}) for the (1, 1) mode in Figure 2(c). The data is fitted using Eq. (6), with Duffing constant as the sole free parameter. The experimental Duffing coefficients for the (1, 1), (2, 1), and (2, 2) modes are extracted and found to be 2×1022 m−2s−22 \times10^{22}\ \mathrm{m}^{-2}\mathrm{s}^{-2}, 2.5×1023 m−2s−22.5 \times10^{23}\ \mathrm{m}^{-2}\mathrm{s}^{-2}, and 4.7×1023 m−2s−24.7 \times10^{23}\ \mathrm{m}^{-2}\mathrm{s}^{-2}, respectively. Figure 2(d) shows good agreement between the experimentally obtained Duffing coefficients and the theoretical values calculated from Eq. (5). The corresponding analysis for the (2, 1) and (2, 2) modes are presented in Figure S4 of the Supplementary Material.

When a higher order mode is driven into the nonlinear regime, its large amplitude oscillation modifies the tensile stress of the membrane. This displacement-dependent tension acts as an effective parametric perturbation for other modes, leading to nonlinear intermodal coupling. As a result the resonance frequency of the fundamental mode becomes dependent on the oscillation amplitude of the driven higher-order mode. To experimentally probe this effect, we strongly drive the (2,1)(2,1) and (2,2)(2,2) modes near their resonance frequencies while simultaneously monitoring the response of the fundamental (1,1)(1,1) mode using a weak probe tone. Figure 3(a) and Figure 3(c) show color plots of the fundamental-mode response as a function of probe frequency and drive frequency applied to the higher-order modes. A clear amplitude-dependent frequency shift of the (1,1)(1,1) mode is observed, indicative of nonlinear frequency pulling due to intermodal coupling. The localized enhancement in spectral amplitude of the fundamental mode observed near a drive frequency ∼1628\sim1628 kHz in Figure 3(a) arises from a subharmonic component of the drive signal. This feature is absent in Figure 3(c), where such harmonic components are not resonant with the fundamental mode.

Figure 3. Nonlinear intermodal coupling. (a,c) Resonance frequency-shifting phenomena of the fundamental (1, 1) mode induced by strong excitation of the (2, 2) and (2, 1) modes, respectively.(b,d) Extracted resonance frequency of the (1, 1) mode as a function of the displacement amplitude of the driven (2, 2) and (2, 1) modes, respectively. Solid lines represent fits to the data using Eq. (9) to extract the nonlinear intermodal coupling constants. (e) Comparison of experimentally extracted coupling constants (stars) with theoretical values (circles) calculated from Eq. (49).

In order to theoretically evaluate the intermodal coupling constant, we modify the in-plane stress component as

Nxx=Eh[ϵ0+12{(∂w1∂x)2+(∂w2∂x)2}],N_{xx}=Eh\left[\epsilon_0+\frac{1}{2}\left\{\left(\frac{\partial w_1}{\partial x}\right)^2+\left(\frac{\partial w_2}{\partial x}\right)^2\right\}\right],
Nyy=Eh[ϵ0+12{(∂w1∂y)2+(∂w2∂y)2}],N_{yy}=Eh\left[\epsilon_0+\frac{1}{2}\left\{\left(\frac{\partial w_1}{\partial y}\right)^2+\left(\frac{\partial w_2}{\partial y}\right)^2\right\}\right],
Nxy=12Eh[(∂w1∂x∂w1∂y)+(∂w2∂x∂w2∂y)],(7)N_{xy}=\frac{1}{2}Eh\left[\left(\frac{\partial w_1}{\partial x}\frac{\partial w_1}{\partial y}\right)+\left(\frac{\partial w_2}{\partial x}\frac{\partial w_2}{\partial y}\right)\right], \tag*{(7)}

where w1=sin⁡(nπxL)sin⁡(mπyL)z1(t)w_1=\sin\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right)z_1(t) and w2=sin⁡(pπxL)sin⁡(qπyL)z2(t)w_2=\sin\left(\frac{p\pi x}{L}\right)\sin\left(\frac{q\pi y}{L}\right)z_2(t) represent the transverse displacements of the probing and driving modes, respectively. Substituting the mode shapes into the governing plate equation and applying the Galerkin discretization procedure, followed by the inclusion of linear damping (ν1)(\nu_1) and external probing force (fpcos⁡(ωpt))(f_p\cos(\omega_p t)) terms, yields the probe mode equation of motion as

z¨1+ωnm2z1+ν1z˙1+αnmz13+λnmpqz1z22=fpcos⁡(ωpt),(8)\ddot{z}_1+\omega_{nm}^2z_1+\nu_1\dot{z}_1+\alpha_{nm}z_1^3+\lambda_{nm}^{pq}z_1z_2^2=f_p\cos(\omega_p t), \tag*{(8)}

where λnmpq\lambda_{nm}^{pq} represents the coupling constant between the driving mode (p,q)(p,q) and the probing mode (n,m)(n,m). Applying the method of multiple scales to Eq. (8), we obtain the effective resonance frequency of the fundamental mode arising from intermodal coupling as

ω=ωnm+3αnm8ωnm(z1peak)2+λnmpq2ωnmz22.(9)\omega=\omega_{nm}+\frac{3\alpha_{nm}}{8\omega_{nm}}(z_1^{\mathrm{peak}})^2+\frac{\lambda_{nm}^{pq}}{2\omega_{nm}}z_2^2. \tag*{(9)}

where z2z_2 is the amplitude of the driven higher order mode. The experimentally measured frequency shifts are shown in Figure 3(b) and Figure 3(d). The data are fitted using Eq. (9), with λnmpq\lambda_{nm}^{pq} as the sole fitting parameter. From these fits, we extract the intermodal coupling constants

for the (1,1)(1,1)–(2,1)(2,1) and (1,1)(1,1)–(2,2)(2,2) mode pairs, yielding values of 5.3×1022 m−2s−25.3\times10^{22}\ m^{-2}s^{-2} and 9.2×1022 m−2s−29.2\times10^{22}\ m^{-2}s^{-2}, respectively. As shown in Figure 3(e), the extracted coupling constants are in good agreement with the theoretical predictions obtained from Eq. (49). At very high drive amplitudes, the model breaks down due to the onset of higher-order nonlinearities in the driven mode, which modify the standard Duffing response [66, 72].

Using the Kirchhoff-Love plate theory, we further compute the nonlinear coupling constants between the fundamental mode and higher-order modes using Eq. (49). We first consider the probe mode with n=mn=m. Figure 4(a–Figure 4d) shows the variation of the coupling constant as the driven mode indices (p,q)(p,q) are varied from 1 to 10, while the probe modes are fixed at (1,1)(1,1), (2,2)(2,2), (3,3)(3,3), and (4,4)(4,4), respectively. For driven modes with p,q≠np,q\ne n, the coupling constant λnmpq\lambda_{nm}^{pq} increases monotonically with increasing pp or qq when the other index is held fixed. In the special case p=np=n or q=mq=m, λnmpq\lambda_{nm}^{pq} exhibits a sharp increase with increasing pp or qq, respectively, due to enhanced mode overlap. When n=m=p=qn=m=p=q, λnmpq\lambda_{nm}^{pq} reduces to intramodal coupling and equals the Duffing constant of the mode. For probe modes with n≠mn\ne m, the symmetry is broken (see Supplementary Fig. S1). In this case, λnmpq\lambda_{nm}^{pq} exhibits a sharp increase when either p=np=n or q=mq=m. When n<mn<m, the coupling constant λnmpq\lambda_{nm}^{pq} is larger for p=np=n than for q=mq=m. Conversely, when m<nm<n, λnmpq\lambda_{nm}^{pq} is larger for q=mq=m than for p=np=n. When p=np=n and q=mq=m then λnmpq\lambda_{nm}^{pq} again reduces to the intramodal coupling constant. The calculated coupling parameters for mode indices p,q=1p,q=1–44, shown in Figure 4(a), are listed in Table I. The coupling matrix exhibits clear symmetry with respect to the mode indices and reveals a systematic increase in coupling strength for higher-order modes, as discussed above. Using the coupling parameters in Table I, we predict the frequency shift of the fundamental mode as a function of the coupled-mode amplitude, as shown in Figure 4(e). These results highlight nonlinear intermodal coupling as an effective mechanism for frequency control.

Figure 4. Theoretical estimation of the nonlinear mode coupling matrix. (a–d) Calculated coupling constants as a function of the driven mode indices (p,q)(p,q) for p,q=1p,q=1–1010, coupled to the probe modes (1,1)(1,1), (2,2)(2,2), (3,3)(3,3), and (4,4)(4,4), respectively. (e) Predicted frequency shift of the fundamental mode as a function of the coupled-mode amplitude, calculated using the theoretical coupling constants listed in Table I.

q=1q = 1q=2q = 2q=3q = 3q=4q = 4
p=1p = 10.171.122.414.22
p=2p = 21.121.382.243.44
p=3p = 32.412.243.094.30
p=4p = 44.223.444.305.51

Table I. This table shows the value of coupling constant per unit modal mass for different modes with mode number (p,q)(p,q), which is coupled to (1,1)(1,1) mode. All the values here are of the order 1023 m−2s−210^{23}\,\mathrm{m}^{-2}\mathrm{s}^{-2}. The intramode coupling constant (Duffing constant) is highlighted in blue.

In conclusion, we experimentally demonstrate and quantitatively model nonlinear intra- and intermodal coupling in a high-stress silicon nitride membrane resonator arising from tension induced geometric nonlinearity. By independently driving selected flexural modes into the nonlinear regime, we directly measure amplitude-dependent frequency shifts originating from intrinsic Duffing nonlinearity as well as from nonlinear coupling between orthogonal modes. Using a Kirchhoff–Love plate theory framework, we derived expressions for the Duffing and intermodal coupling coefficients and showed quantitative agreement with experimental measurements for multiple mode pairs, including the (1,1)(1,1)–(2,1)(2,1) and (1,1)(1,1)–(2,2)(2,2) interactions. Beyond individual mode pairs, we computed the full nonlinear coupling matrix across mode families, explicitly demonstrating how mode symmetry and spatial overlap determine the magnitude and selectivity of intermodal interactions. These results establish nonlinear mode coupling in Si3N4Si_{3}N_{4} membranes as a predictable and controllable mechanism for frequency tuning and multimode transduction. The presented framework provides a direct route to engineering mechanical responses in membrane resonators, with immediate implications for multimode sensing, mechanically mediated signal processing, and hybrid opto- and electro-mechanical systems where controlled mode interactions are essential[62, 73–75].

References

Acknowledgments

A.N. acknowledges funding support from MHRD, MeitY and DST Nano Mission through NNetRA. N.A. acknowledges fellowship support under Visvesvaraya Ph.D. Scheme, Ministry of Electronics and Information Technology (MeitY), India. C.S. acknowledges financial support from the Anusandhan National Research Foundation (ANRF), India, under Project No. ANRF/ECRG/2024/006637/PMS.

Competing Interests

The authors declare no competing interests.

Supplementary Information for Nonlinear Mode Coupling in Silicon Nitride Membrane Resonators

Soumya Kanti Das; Nishta Arora; Hridhay A S; Akshay Naik; Chandan Samanta

Theoretical description

We model nonlinear mode coupling behavior of a square silicon nitride (Si3N4\mathrm{Si}_3\mathrm{N}_4) membrane resonator using Kirchhoff–Love plate theory. We first analyze the intramode coupling (intrinsic Duffing nonlinearity) of the resonator and then introduce intermodal mode-coupling terms to develop a general coupling framework. The Duffing constants and a general expression for the coupling constant between probing mode (n,m)(n,m) and driving mode (p,q)(p,q) are derived. The theoretical results show good agreement with experimental observations.

Duffing nonlinearity

With the use of Kirchhoff-Love plate theory[1], the general equation characterizing the transverse displacement w(x,y,t)w(x,y,t) of a thin plate can be written as

D∇4w(x,y,t)−(Nxx∂2w∂x2+2Nxy∂2w∂x∂y+Nyy∂2w∂y2)+ρh∂2w∂t2=0(10)D\nabla^{4}w(x,y,t)-\left(N_{xx}\frac{\partial^{2}w}{\partial x^{2}}+2N_{xy}\frac{\partial^{2}w}{\partial x\partial y}+N_{yy}\frac{\partial^{2}w}{\partial y^{2}}\right)+\rho h\frac{\partial^{2}w}{\partial t^{2}}=0 \tag*{(10)}

where w(x,y,t)=ϕ(x,y)z(t)=sin⁡(nπxL)sin⁡(mπyL)z(t)w(x,y,t)=\phi(x,y)z(t)=\sin\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right)z(t), D is the flextural rigidity, NxxN_{xx}, NyyN_{yy}, and NxyN_{xy} are the in-plane stress resultants, which can be defined as

Nxx=∫−h/2h/2σxx dz(11)N_{xx}=\int_{-h/2}^{h/2}\sigma_{xx}\,\mathrm{d}z \tag*{(11)}
Nyy=∫−h/2h/2σyy dz(12)N_{yy}=\int_{-h/2}^{h/2}\sigma_{yy}\,\mathrm{d}z \tag*{(12)}
Nxy=∫−h/2h/2σxy dz(13)N_{xy}=\int_{-h/2}^{h/2}\sigma_{xy}\,\mathrm{d}z \tag*{(13)}

here, σxx\sigma_{xx}, σyy\sigma_{yy} and σxy\sigma_{xy} is defined as

σxx=E1−ν2(εxx+νεyy),(14)\sigma_{xx}=\frac{E}{1-\nu^{2}}\left(\varepsilon_{xx}+\nu\varepsilon_{yy}\right), \tag*{(14)}
σyy=E1−ν2(εyy+νεxx),(15)\sigma_{yy}=\frac{E}{1-\nu^{2}}\left(\varepsilon_{yy}+\nu\varepsilon_{xx}\right), \tag*{(15)}
σxy=E1+νεxy.(16)\sigma_{xy}=\frac{E}{1+\nu}\varepsilon_{xy}. \tag*{(16)}

Due to the thinness of our membrane, we assume that the Poisson’s ratio (ν\nu) is zero. In the presence of nonlinear displacement, particularly due to geometrical nonlinearity, the strain components are defined as [2]

εxx=ε0+∂u0∂x+12(∂w∂x)2−z(t)∂2w∂x2,(17)\varepsilon_{xx}=\varepsilon_{0}+\frac{\partial u_{0}}{\partial x}+\frac{1}{2}\left(\frac{\partial w}{\partial x}\right)^{2}-z(t)\frac{\partial^{2}w}{\partial x^{2}}, \tag*{(17)}
εyy=ε0+∂v0∂y+12(∂w∂y)2−z(t)∂2w∂y2,(18)\varepsilon_{yy}=\varepsilon_{0}+\frac{\partial v_{0}}{\partial y}+\frac{1}{2}\left(\frac{\partial w}{\partial y}\right)^{2}-z(t)\frac{\partial^{2}w}{\partial y^{2}}, \tag*{(18)}
εxy=12(∂u0∂y+∂v0∂x)+12∂w∂x∂w∂y−z(t)∂2w∂x∂y.(19)\varepsilon_{xy}=\frac{1}{2}\left(\frac{\partial u_{0}}{\partial y}+\frac{\partial v_{0}}{\partial x}\right)+\frac{1}{2}\frac{\partial w}{\partial x}\frac{\partial w}{\partial y}-z(t)\frac{\partial^{2}w}{\partial x\partial y}. \tag*{(19)}

For a thin membrane, we assume that in-plane mid-surface displacements are minimal, namely u0,v0=0u_{0},v_{0}=0, and that in-plane displacements resulting from transverse deformation are insignificant. With these assumptions, the stress resultants are simplified to

Nxx=Eh(ε0+12(∂w∂x)2),(20)N_{xx}=Eh\left(\varepsilon_{0}+\frac{1}{2}\left(\frac{\partial w}{\partial x}\right)^{2}\right), \tag*{(20)}
Nyy=Eh(ε0+12(∂w∂y)2),(21)N_{yy}=Eh\left(\varepsilon_{0}+\frac{1}{2}\left(\frac{\partial w}{\partial y}\right)^{2}\right), \tag*{(21)}
Nxy=12Eh(∂w∂x∂w∂y).(22)N_{xy}=\frac{1}{2}Eh\left(\frac{\partial w}{\partial x}\frac{\partial w}{\partial y}\right). \tag*{(22)}

Substituting into the governing equation (S1) and employing the Galerkin method with w(x,y,t)=sin⁡(nπxL)sin⁡(mπyL)z(t)w(x,y,t)=\sin\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right)z(t), also neglecting DD for the high-stress scenario [3], we obtain

−∫0L∫0L(Nxx∂2w∂x2+2Nxy∂2w∂x∂y+Nyy∂2w∂y2)ϕ(x,y) dx dy+ρh∫0L∫0L∂2w∂t2ϕ(x,y) dx dy=0(23)-\int_{0}^{L}\int_{0}^{L}\left(N_{xx}\frac{\partial^{2}w}{\partial x^{2}}+2N_{xy}\frac{\partial^{2}w}{\partial x\partial y}+N_{yy}\frac{\partial^{2}w}{\partial y^{2}}\right)\phi(x,y)\,\mathrm{d}x\,\mathrm{d}y+\rho h\int_{0}^{L}\int_{0}^{L}\frac{\partial^{2}w}{\partial t^{2}}\phi(x,y)\,\mathrm{d}x\,\mathrm{d}y=0 \tag*{(23)}

The expressions for NxxN_{xx}, NyyN_{yy}, and NxyN_{xy} becomes

Nxx=Eh[ϵ0+12z2(t)(nπL)2sin⁡2(mπyL)×cos⁡2(nπxL)],(24)\begin{aligned} N_{xx} &= Eh\left[\epsilon_0+\frac{1}{2}z^2(t)\left(\frac{n\pi}{L}\right)^2\sin^2\left(\frac{m\pi y}{L}\right)\right.\\ &\qquad\left.\times\cos^2\left(\frac{n\pi x}{L}\right)\right], \tag*{(24)} \end{aligned}
Nyy=Eh[ϵ0+12z2(t)(mπL)2sin⁡2(nπxL)×cos⁡2(mπyL)],(25)\begin{aligned} N_{yy} &= Eh\left[\epsilon_0+\frac{1}{2}z^2(t)\left(\frac{m\pi}{L}\right)^2\sin^2\left(\frac{n\pi x}{L}\right)\right.\\ &\qquad\left.\times\cos^2\left(\frac{m\pi y}{L}\right)\right], \tag*{(25)} \end{aligned}
Nxy=12Ehz2(t)(nπL)(mπL)×sin⁡(nπxL)cos⁡(nπxL)sin⁡(mπyL)cos⁡(mπyL).(26)\begin{aligned} N_{xy}={}&\frac{1}{2}Ehz^2(t)\left(\frac{n\pi}{L}\right)\left(\frac{m\pi}{L}\right)\\ &\times\sin\left(\frac{n\pi x}{L}\right)\cos\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right)\cos\left(\frac{m\pi y}{L}\right). \tag*{(26)} \end{aligned}

Upon substituting into equation (23), we can find the integrals

K1=∫0L∫0LNxx∂2w∂x2ϕ(x,y) dxdy(27)K_1=\int_0^L\int_0^L N_{xx}\frac{\partial^2w}{\partial x^2}\phi(x,y)\,dxdy \tag*{(27)}
K2=∫0L∫0LNyy∂2w∂y2ϕ(x,y) dxdy(28)K_2=\int_0^L\int_0^L N_{yy}\frac{\partial^2w}{\partial y^2}\phi(x,y)\,dxdy \tag*{(28)}
K3=∫0L∫0L2Nxy∂2w∂x∂yϕ(x,y) dxdy(29)K_3=\int_0^L\int_0^L 2N_{xy}\frac{\partial^2w}{\partial x\partial y}\phi(x,y)\,dxdy \tag*{(29)}

By solving the integral and substituting these values into Eq. (23), we have derived the equation

ρhL24d2zdt2+Ehϵ0L24(πL)2(n2+m2)z(t)+32EhL264(πL)4[n4+m4−23n2m2]z3(t)=0(30)\rho h\frac{L^2}{4}\frac{d^2z}{dt^2}+Eh\epsilon_0\frac{L^2}{4}\left(\frac{\pi}{L}\right)^2(n^2+m^2)z(t)+\frac{3}{2}Eh\frac{L^2}{64}\left(\frac{\pi}{L}\right)^4\left[n^4+m^4-\frac{2}{3}n^2m^2\right]z^3(t)=0 \tag*{(30)}

From there we can find,

ωnm2=Eϵ0π2(n2+m2)ρL2(31)\omega_{nm}^2=\frac{E\epsilon_0\pi^2(n^2+m^2)}{\rho L^2} \tag*{(31)}

And,

αnm=3E32ρ(πL)4[n4+m4−23n2m2](32)\alpha_{nm}=\frac{3E}{32\rho}\left(\frac{\pi}{L}\right)^4\left[n^4+m^4-\frac{2}{3}n^2m^2\right] \tag*{(32)}

To determine the resonant frequency and the Duffing nonlinearity, we employed Eq. (30); however, a damping component and a driving force term will typically be present as well. To analyze the nonlinear response, it is essential to include the driving force term, as outlined in the paper. The equation of motion, incorporating both the driving force and damping terms, can be expressed as

d2zdt2+νdzdt+ωnm2z+αnmz3=fdcos⁡(ωt)(33)\frac{d^2z}{dt^2}+\nu\frac{dz}{dt}+\omega_{nm}^2z+\alpha_{nm}z^3=f_d\cos(\omega t) \tag*{(33)}

here, ν\nu represents the damping factor, while fdcos⁡(ωt)f_d\cos(\omega t) denotes the driving force. The driving force is either high or low, as indicated by the driving force amplitude fdf_d.

Nonlinear mode coupling

We have not previously looked at the mode coupling term. The in-plane tension can be adjusted for mode coupling. Considering non-linear mode coupling, the values of ϵxx\epsilon_{xx}, ϵyy\epsilon_{yy}, ϵxy\epsilon_{xy} will be

ϵxx=∂u0∂x+12[(∂w1∂x)2+(∂w2∂x)2]−z1(t)∂2w1∂x2(34)\epsilon_{xx} = \frac{\partial u_0}{\partial x} + \frac{1}{2}\left[\left(\frac{\partial w_1}{\partial x}\right)^2 + \left(\frac{\partial w_2}{\partial x}\right)^2\right] - z_1(t)\frac{\partial^2 w_1}{\partial x^2} \tag*{(34)}
ϵyy=∂v0∂y+12[(∂w1∂y)2+(∂w2∂y)2]−z1(t)∂2w1∂y2(35)\epsilon_{yy} = \frac{\partial v_0}{\partial y} + \frac{1}{2}\left[\left(\frac{\partial w_1}{\partial y}\right)^2 + \left(\frac{\partial w_2}{\partial y}\right)^2\right] - z_1(t)\frac{\partial^2 w_1}{\partial y^2} \tag*{(35)}
ϵxy=12(∂u0∂y+∂v0∂x)+12[(∂w1∂x∂w1∂y)+(∂w2∂x∂w2∂y)]−z1(t)∂2w1∂x∂y(36)\epsilon_{xy} = \frac{1}{2}\left(\frac{\partial u_0}{\partial y} + \frac{\partial v_0}{\partial x}\right) + \frac{1}{2}\left[\left(\frac{\partial w_1}{\partial x}\frac{\partial w_1}{\partial y}\right) + \left(\frac{\partial w_2}{\partial x}\frac{\partial w_2}{\partial y}\right)\right] - z_1(t)\frac{\partial^2 w_1}{\partial x\partial y} \tag*{(36)}

Here, w1=ϕ1(x,y)z1(t)=sin⁡(nπxL)sin⁡(mπyL)z1(t)w_1 = \phi_1(x,y)z_1(t) = \sin\left(\frac{n\pi x}{L}\right)\sin\left(\frac{m\pi y}{L}\right)z_1(t), where (n,m)(n,m) denotes the mode number of the probe mode, and w2=ϕ2(x,y)z2(t)=sin⁡(pπxL)sin⁡(qπyL)z2(t)w_2 = \phi_2(x,y)z_2(t) = \sin\left(\frac{p\pi x}{L}\right)\sin\left(\frac{q\pi y}{L}\right)z_2(t), where (p,q)(p,q) signifies the mode number corresponding to the pump mode. Once more, we can apply the aforementioned thin membrane approximation. The in-plane stress resultants can now be expressed as

Nxx=Eh[ϵ0+12{(∂w1∂x)2+(∂w2∂x)2}](37)N_{xx} = Eh\left[\epsilon_0 + \frac{1}{2}\left\{\left(\frac{\partial w_1}{\partial x}\right)^2 + \left(\frac{\partial w_2}{\partial x}\right)^2\right\}\right] \tag*{(37)}
Nyy=Eh[ϵ0+12{(∂w1∂y)2+(∂w2∂y)2}](38)N_{yy} = Eh\left[\epsilon_0 + \frac{1}{2}\left\{\left(\frac{\partial w_1}{\partial y}\right)^2 + \left(\frac{\partial w_2}{\partial y}\right)^2\right\}\right] \tag*{(38)}
Nxy=12Eh[(∂w1∂x∂w1∂y)+(∂w2∂x∂w2∂y)](39)N_{xy} = \frac{1}{2}Eh\left[\left(\frac{\partial w_1}{\partial x}\frac{\partial w_1}{\partial y}\right) + \left(\frac{\partial w_2}{\partial x}\frac{\partial w_2}{\partial y}\right)\right] \tag*{(39)}

The contributions to the mode coupling factor from the terms NxxN_{xx}, NyyN_{yy}, NxyN_{xy} are as follows

Nxx,λ=Eh2(∂w2∂x)2(40)N_{xx,\lambda} = \frac{Eh}{2}\left(\frac{\partial w_2}{\partial x}\right)^2 \tag*{(40)}
Nyy,λ=Eh2(∂w2∂y)2(41)N_{yy,\lambda} = \frac{Eh}{2}\left(\frac{\partial w_2}{\partial y}\right)^2 \tag*{(41)}
Nxy,λ=Eh2(∂w2∂x∂w2∂y)(42)N_{xy,\lambda} = \frac{Eh}{2}\left(\frac{\partial w_2}{\partial x}\frac{\partial w_2}{\partial y}\right) \tag*{(42)}

Substituting the values of w2w_2 yields

Nxx,λ=Eh2[(pπL)2cos⁡2(pπxL)sin⁡2(qπyL)](43)N_{xx,\lambda} = \frac{Eh}{2}\left[\left(\frac{p\pi}{L}\right)^2 \cos^2\left(\frac{p\pi x}{L}\right)\sin^2\left(\frac{q\pi y}{L}\right)\right] \tag*{(43)}
Nyy,λ=Eh2[(qπL)2sin⁡2(pπxL)cos⁡2(qπyL)](44)N_{yy,\lambda} = \frac{Eh}{2}\left[\left(\frac{q\pi}{L}\right)^2 \sin^2\left(\frac{p\pi x}{L}\right)\cos^2\left(\frac{q\pi y}{L}\right)\right] \tag*{(44)}
Nxy,λ=12Eh[(pqπL)cos⁡(pπxL)sin⁡(qπyL)(qπL)sin⁡(pπxL)cos⁡(qπyL)](45)N_{xy,\lambda} = \frac{1}{2}Eh\left[\left(\frac{pq\pi}{L}\right)\cos\left(\frac{p\pi x}{L}\right)\sin\left(\frac{q\pi y}{L}\right)\left(\frac{q\pi}{L}\right)\sin\left(\frac{p\pi x}{L}\right)\cos\left(\frac{q\pi y}{L}\right)\right] \tag*{(45)}

To calculate the coupling constant, it is necessary to solve the integrals

∫0L∫0LNxx,λ∂2w1∂x2ϕ1(x,y) dxdy(46)\int_{0}^{L}\int_{0}^{L}N_{xx,\lambda}\frac{\partial^{2}w_{1}}{\partial x^{2}}\phi_{1}(x,y)\,dxdy \tag*{(46)}
∫0L∫0LNyy,λ∂2w1∂y2ϕ1(x,y) dxdy(47)\int_{0}^{L}\int_{0}^{L}N_{yy,\lambda}\frac{\partial^{2}w_{1}}{\partial y^{2}}\phi_{1}(x,y)\,dxdy \tag*{(47)}
∫0L∫0L2Nxy,λ∂2w1∂x∂yϕ1(x,y) dxdy(48)\int_{0}^{L}\int_{0}^{L}2N_{xy,\lambda}\frac{\partial^{2}w_{1}}{\partial x\partial y}\phi_{1}(x,y)\,dxdy \tag*{(48)}

The expression for coupling constant is

λnmpq=−∫0L∫0L(Nxx,λ∂2w1∂x2ϕ1(x,y)+Nyy,λ∂2w1∂y2ϕ1(x,y)+2Nxy,λ∂2w1∂x∂yϕ1(x,y)) dxdyρh(L24)(49)\lambda_{nm}^{pq}=-\frac{\int_{0}^{L}\int_{0}^{L}\left(N_{xx,\lambda}\frac{\partial^{2}w_{1}}{\partial x^{2}}\phi_{1}(x,y)+N_{yy,\lambda}\frac{\partial^{2}w_{1}}{\partial y^{2}}\phi_{1}(x,y)+2N_{xy,\lambda}\frac{\partial^{2}w_{1}}{\partial x\partial y}\phi_{1}(x,y)\right)\,dxdy}{\rho h\left(\frac{L^{2}}{4}\right)} \tag*{(49)}

Where the denominator comes from Eq. (30). Considering the mode number (p,q)(p,q), where p=q>1p=q>1, when coupled with mode (1,1)(1,1), the expression for the coupling constant can be defined as

λ11pq=E8ρ(πL)4[p2+q2](50)\lambda_{11}^{pq}=\frac{E}{8\rho}\left(\frac{\pi}{L}\right)^{4}[p^{2}+q^{2}] \tag*{(50)}

We have derived an equation similar to (S24) that includes the mode coupling term. The equation may be expressed as

d2z1dt2+ν1dz1dt+ωnm2z1+αnmz13+λnmpqz1z22=fpcos⁡(ωt)(51)\frac{d^{2}z_{1}}{dt^{2}}+\nu_{1}\frac{dz_{1}}{dt}+\omega_{nm}^{2}z_{1}+\alpha_{nm}z_{1}^{3}+\lambda_{nm}^{pq}z_{1}z_{2}^{2}=f_{p}\cos(\omega t) \tag*{(51)}

This equation represents the mode coupling of the probe mode, which exhibits displacement along the zz axis (z1z_{1}), with the driving mode, whose displacement is z2z_{2}. ν1\nu_{1}, ωnm\omega_{nm}, αnm\alpha_{nm}, and λnmpq\lambda_{nm}^{pq} represent the damping constant, natural frequency, Duffing constant of the probe mode, respectively and λnmpq\lambda_{nm}^{pq} is the coupling constant between probe mode (n,m)(n,m) and drive mode (p,q)(p,q). We have numerically determined the coupling constant using (49). When p=np=n and q=mq=m, the coupling constant λpqnm\lambda_{pq}^{nm} reduces to αnm\alpha_{nm}, corresponding to the intramode coupling, which is basically the Duffing constant.

q=1q=1q=2q=2q=3q=3q=4q=4q=5q=5
p=1p=11.382.416.8811.7017.89
p=2p=22.412.7510.6717.8927.19
p=3p=36.8810.6712.3917.2123.40
p=4p=411.7017.8917.2122.0328.22
p=5p=517.8927.1923.4028.2234.42

Table 1 (PDF p. 13). Coupling constant for unit modal mass for mode (p,q)(p,q), where p,qp,q runs from 1 to 5, coupled with (2,2)(2,2) mode. The values of the coupling constant are of the order 1023 m−2s−210^{23}\ \mathrm{m}^{-2}\mathrm{s}^{-2}. The intramode coupling constant (Duffing constant) is highlighted in blue.

q=1q=1q=2q=2q=3q=3q=4q=4q=5q=5
p=1p=13.097.749.2926.3340.27
p=2p=27.7412.3916.2630.9744.91
p=3p=39.2916.2613.9344.1465.05
p=4p=426.3330.9744.1453.6063.50
p=5p=540.2744.9165.0563.6077.44

Table 2. Coupling constant for unit modal mass for mode (p,q)(p,q), where p,qp,q runs from 1 to 5, coupled with (3,3)(3,3) mode. The values of the coupling constant are of the order 1023 m−2s−210^{23}\ \mathrm{m}^{-2}\mathrm{s}^{-2}. The intramode coupling constant (Duffing constant) is highlighted in blue.

Theoretical estimation of the nonlinear mode coupling matrix

Figure S1. Theoretical estimation of the nonlinear mode coupling matrix. (a–f) Calculated coupling constants as a function of the driven mode indices (p,q)(p, q) for p,q=1p, q = 1–1010, coupled to the probe modes (2,1)(2, 1), (3,1)(3, 1), (3,2)(3, 2), (4,1)(4, 1), (4,2)(4, 2) and (4,3)(4, 3), respectively.

Nonlinear frequency pulling

The frequency detuning equation for mode coupling is derived using the method of multiple scales[4]. This approach alters the time scale as follows

Tn=ϵnt,(52)T_n = \epsilon^n t, \tag*{(52)}

where T0=tT_0 = t, and ϵ\epsilon is the scaling parameter. For using method of multiple scale the Eq (51) can be modified as

z¨1+2ϵμ1z˙1+ωnm2z1+ϵαnmz13+ϵλnmpqz1z22=ϵkdcos⁡(ωnmT0+σT1).(53)\ddot{z}_1 + 2\epsilon\mu_1\dot{z}_1 + \omega_{nm}^2 z_1 + \epsilon\alpha_{nm}z_1^3 + \epsilon\lambda_{nm}^{pq}z_1z_2^2 = \epsilon k_d\cos(\omega_{nm}T_0 + \sigma T_1). \tag*{(53)}

Using the method of multiple scale method we have reached a equation as

σ=3αnm8ωnm(z1peak)2±kd24ωnm2(z1peak)2−μ12+λnmpq2ωnmz22.(54)\sigma= \frac{3\alpha_{nm}}{8\omega_{nm}}(z_1^{peak})^2 \pm\sqrt{\frac{k_d^2}{4\omega_{nm}^2(z_1^{peak})^2} - \mu_1^2} + \frac{\lambda_{nm}^{pq}}{2\omega_{nm}}z_2^2. \tag*{(54)}

For steady state the middle term in the RHS becomes zero. Then we can get a equation like

σ=3αnm8ωnm(z1peak)2+λnmpq2ωnmz22.(55)\sigma= \frac{3\alpha_{nm}}{8\omega_{nm}}(z_1^{peak})^2 + \frac{\lambda_{nm}^{pq}}{2\omega_{nm}}z_2^2. \tag*{(55)}

Here, σ=ω−ωnm\sigma= \omega- \omega_{nm} denotes the frequency detuning. This equation represents the frequency detuning generated by mode coupling, enabling the determination of the coupling constant (λnmpq\lambda_{nm}^{pq}). In the absence of mode coupling, the equation assumes the following form

σ=3αnm8ωnm(z1peak)2.(56)\sigma= \frac{3\alpha_{nm}}{8\omega_{nm}}(z_1^{peak})^2. \tag*{(56)}

Equation (56) helps us for the calculation of the experimental Duffing constant (αnm\alpha_{nm}). When the applied force is sufficiently small, the nonlinear terms in Eq. (54) can be neglected. Using the coupling constant, the resulting shift in the resonance frequency of the probe mode can then be plotted as a function of the drive-mode amplitude. The resonance-frequency shifts of the (2,2)(2,2) and (3,3)(3,3) modes as functions of the peak amplitudes of the other coupled modes are shown in Figure S2. The graph illustrates the extent of the change in resonant frequency of the modes that can occur when the coupled mode possesses the specified amplitude.

Figure S2. Predicted frequency shift of the fundamental mode (a)(2,2)(2,2), (b)(3,3)(3,3) as a function of the coupled-mode amplitude.

Experimental Section

Device parameters

The device parameters used in this work are summarized in Table 3. The strain is calculated using the resonance frequency relation given in Eq. (31). Using the extracted strain and the Young’s modulus, the corresponding tensile stress is estimated to be approximately 1.05 GPa, consistent with previously reported values [5].

Physical ParameterValue
Young’s Modulus (EE)280×109 Pa280 \times10^{9}\ \mathrm{Pa}
Density (ρ\rho)3170 kg/m33170\ \mathrm{kg}/\mathrm{m}^{3} [6]
Inbuilt Strain (ϵ0\epsilon_{0})3.751×10−33.751 \times10^{-3}
Length (LL)500×10−6 m500 \times10^{-6}\ \mathrm{m}
Thickness (hh)100×10−9 m100 \times10^{-9}\ \mathrm{m}

Table 3. Device Parameters

Displacement calibration

The displacement of the membrane was calibrated using the laser Doppler vibrometer (LDV), which provides an absolute measurement of the out-of-plane displacement. To establish a conversion between the electrical readout voltage and the physical displacement, the device was driven with a low-frequency sinusoidal excitation of known amplitude while simultaneously recording the displacement using the LDV and the corresponding voltage signal using the lock-in amplifier. Figure S3 shows the measured displacement as a function of the detected voltage. A linear fit to the data yields a calibration factor of ∂z∂V=0.2 μm/V\frac{\partial z}{\partial V} = 0.2\ \mu\mathrm{m}/\mathrm{V}, where zz is the mechanical displacement and VV is the measured voltage at the lock-in amplifier.

Measured mechanical displacement as a function of detected lock-in amplifier voltage

Figure S3. Measured mechanical displacement zz as a function of the detected lock-in amplifier voltage VV. A linear fit to the data yields a displacement calibration factor of ∂z/∂V=0.2 μm V−1\partial z/\partial V = 0.2\ \mu\mathrm{m}\ \mathrm{V}^{-1}.

This calibration factor was subsequently used to convert all voltage signals measured using the lock-in amplifier into absolute displacement units. Specifically, the displacement amplitude was obtained as z(t)=αV(t)z(t) = \alpha V(t), where α=0.2 μm/V\alpha= 0.2\ \mu\mathrm{m}/\mathrm{V} is the experimentally determined conversion factor. Since the LDV provides an absolute displacement reference traceable to optical wavelength standards, this procedure enables quantitative displacement measurements without reliance on device parameters.

Frequency responses of modes (2,1) and (2,2)

Figures S4(b,e) show the linear frequency responses of the (2,1) and (2,2) modes respectively. The solid lines represent Lorentzian fits to the data, yielding quality factors of 12.5×10412.5 \times10^{4} for the (2,1) mode and 11.5×10411.5 \times10^{4} for the mode (2,2)(2,2). As the driving force is increased, the frequency response becomes progressively nonlinear, as illustrated in Figure S4,Figure S4). Figure S4,Figure S4) show the extracted frequency shift as a function of the peak amplitudes. The solid line correspond to fits using Eq. (56), from which the experimental Duffing coefficients are obtained to be 2.5×1023 m−2s−22.5 \times10^{23}\,\mathrm{m}^{-2}\mathrm{s}^{-2} and 4.7×1023 m−2s−24.7 \times10^{23}\,\mathrm{m}^{-2}\mathrm{s}^{-2} for mode (2,1)(2,1) and (2,2)(2,2) respectively.

Throughout the measurements, the LDV laser spot was fixed at the center of the membrane. This position corresponds to the anti-node of the (1,1)(1,1) mode but lies close to the nodal regions of the higher-order (2,1)(2,1) and (2,2)(2,2) modes. As a result, the measured displacement amplitudes for these modes are small, leading to increased amplitude fluctuations, as observed in Figure S4,Figure S4). To account for this effect, scaling factors were estimated from the extrema of the spatial mode shapes shown in Figure 1 in the main text. The scaling factor is 10.2±4.010.2 \pm4.0 for the (2,1)(2,1) mode and 25.2±11.025.2 \pm11.0 for the (2,2)(2,2) mode. The uncertainty arises from the mismatch between the LDV laser spot size (∼1 μm\sim1\,\mu\mathrm{m}) and the step size used in the mode-shape measurements (∼50 μm\sim50\,\mu\mathrm{m}). In addition, the localized peak near ∼1628 kHz\sim1628\,\mathrm{kHz} is attributed to higher-harmonic contributions from the parametrically driven fundamental mode.

References

  1. [1]Arthur W. Leissa. Vibration of plates. Vol. 160. Scientific, Technical Information Division, National Aeronautics, and Space Administration, 1969.
  2. [2]E Ventsel, T Krauthammer, and E Carrera. Thin plates and shells: theory, analysis, and applications. 2002.
  3. [3]D Cattiaux et al. “Geometrical nonlinearity of circular plates and membranes: An alternative method”. In: Journal of Applied Physics 128.10 (2020).arxiv.org/abs/1910.02852
  4. [4]Inna Kozinsky. “Nonlinear nanoelectromechanical systems”. In: Ph. D. Thesis (2007).DOI
  5. [5]Bin Shu et al. “Fabrication of high compressive stress silicon nitride membrane in strained silicon technology”. In: 2009 IEEE International Conference of Electron Devices and Solid-State Circuits (EDSSC). IEEE. 2009, pp. 365–367.DOI
  6. [6]Luis Mestre et al. “Network of parametrically driven silicon nitride mechanical membranes”. In: Physical Review Applied 24.4 (2025), p. 044072.

Paper details

vlnc:W_xamdwcud
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