ARTICLE

https://doi.org/10.1038/s41467-019-12796-0 OPEN

Dynamic modulation of modal coupling in

microelectromechanical gyroscopic ring resonators
Xin Zhou1,2, Chun Zhao1, Dingbang Xiao2*, Jiangkun Sun1,2, Guillermo Sobreviela1, Dustin D. Gerrard3,
Yunhan Chen3, Ian Flader3, Thomas W. Kenny3, Xuezhong Wu2 & Ashwin A. Seshia 1*
1234567890():,;

Understanding and controlling modal coupling in micro/nanomechanical devices is integral to

the design of high-accuracy timing references and inertial sensors. However, insight into
specific physical mechanisms underlying modal coupling, and the ability to tune such inter-
actions is limited. Here, we demonstrate that tuneable mode coupling can be achieved in
capacitive microelectromechanical devices with dynamic electrostatic fields enabling strong
coupling between otherwise uncoupled modes. A vacuum-sealed microelectromechanical
silicon ring resonator is employed in this work, with relevance to the gyroscopic lateral modes
of vibration. It is shown that a parametric pumping scheme can be implemented through
capacitive electrodes surrounding the device that allows for the mode coupling strength to be
dynamically tuned, as well as allowing greater flexibility in the control of the coupling stiff-
ness. Electrostatic pump based sideband coupling is demonstrated, and compared to con-
ventional strain-mediated sideband operations. Electrostatic coupling is shown to be very
efficient, enabling strong, tunable dynamical coupling.

1 Nanoscience Centre, University of Cambridge, Cambridge CB3 0FF, UK. 2 Department of Intelligent Machinery and Instruments, College of Intelligence

Science, National University of Defense Technology, Changsha 410073, China. 3 Department of Mechanical Engineering, Stanford University, Stanford, CA
94305, USA. *email: dingbangxiao@nudt.edu.cn; aas41@cam.ac.uk

NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0

T
he interactions of micro- and nanomechanical resonators with interleaved spokes. The resonator is supported by a central
with various physical fields have been researched for several anchor. The diameter and the thickness of the resonator are 720
decades and have been engineered into specific device for- and 40 µm, respectively. The width of the rings and spokes is
mats for applications to timing and frequency control1,2, sensing3–6, 3 μm. The resonator is surrounded by 16 capacitive electrodes,
as well as more fundamental studies in alternative approaches to employed to actuate, sense, and tune the in-plane response. The
information processing7–9, and quantum science10–13. Several recent capacitive gap d 0 is designed to be 1.5 µm wide. The device is
studies have also focused on fundamental emergent behaviors in fabricated using highly doped P-type (111) single-crystal silicon
coupled micro/nanomechanical systems14–17 and the mutual cou- with a resistivity of 1–3 mΩ cm, and encapsulated using the “Epi-
pling between two distinct mechanical resonators or modes18–27. By seal” process46, enabling a stable <1 Pa vacuum environment. The
regarding one of the coupled mechanical modes as a phonon capacitive electrodes introduce a nonlinear electrostatic field
cavity, optomechanics-like dynamical operations, such as cooling surrounding the mechanical resonator, which produces inho-
(dynamical coupling)22,28–31, amplification22,31,32, squeezing33,34, mogeneous force terms, stiffness modification terms, and non-
coherent manipulation28,35,36, and phonon lasing30,37, have been linear mode interactions among multiple modes. Combined with
demonstrated in micro- and nanomechanical systems. Mode cou- the intrinsic properties of the bare mechanical resonators, the
pling mechanisms based on physical mechanical linkages28, dielec- capacitive features allows for abundant manipulations of
tric coupling21,35, tension-induced parametric coupling18,23,29, and the system dynamics, the most interesting aspect studied here is
internal resonance20,25 have been previously investigated. However, the parametric pump-induced dynamical operations.
physical insight is often limited due to the limited experimental The linear and nonlinear mode-coupling mechanisms are
control of mode coupling in such systems. Often, manufacturing investigated in this system. Moreover, those couplings can be
tolerances or small asymmetries can have a huge impact, and the manipulated dynamically using a parametric pumping scheme. The
ability to define modal coupling parameters by design is still limited, experimental setup is shown in Fig. 1a; all the experiments are
often even elusive. Static electrostatic tuning in such systems is conducted at room temperature. The in-plane mechanical modes
possible but provides for only weak tunability, and the ability to involved in this study are illustrated in Fig. 1b, c, involving two pairs
define and exercise good control on modal coupling in micro- and of near-degenerate order-2 and order-3 modes, II-1, II-2, III-1, and
nanomechanical resonators remains a key challenge limiting prac- III-2, with resonant angular frequencies ωII-1 ¼ 2π ´ 134; 209 Hz,
tical applications as well. ωII-2 ¼ 2π ´ 134; 253 Hz, ωIII-1 ¼ 2π ´ 166; 498 Hz, and
In this paper, we demonstrate that dynamic capacitive tuning ωIII-2 ¼ 2π ´ 166; 949 Hz. The damping rates of the modes of
can enable significant tunability of modal coupling in micro- the same order are identical, with values of γII  2π ´ 1:05 Hz and
electromechanical devices. The intrinsic modal coupling in such γIII  2π ´ 2:51 Hz. The electrodes are marked with numbers
systems can be tuned by the nonlinear fields associated with anticlockwise from 1 to 16 (Fig. 1a). The drive signals ±V d cos ðωd tÞ
parallel plate transducers. Such fields can be dynamically are applied on electrodes 3 and 7 to actuate order-2 modes. The
modulated enabling a further knob on the tunability of system pump signal V p cos ðωp tÞ is applied on electrodes 1 and 5. A direct-
response. A vacuum-sealed micromachined gyroscopic ring current (DC) voltage V 0 of 30 V is applied on the resonator body.
resonator is employed as the experimental testbed in this work. The mechanical motion is transduced by the detection electrodes
Such a system demonstrates linear hybrid coupling between the along drive and pump axes. The current signal is introduced to a
gyroscopic near-degenerate modes due to configurational/struc- lock-in amplifier following further amplification stages. A tuning
tural asymmetries arising from manufacturing tolerances or voltage V t1 can be superposed on drive electrode 7. Another tuning
material properties, and misalignment between the principal axis voltage V t2 can be applied on off-axis electrode 6.
and the detection electrode. We demonstrate that, akin to the
tension-induced stiffness hardening mechanisms18,22,23, electro-
statically induced stiffness softening can also provide a nonlinear Hybrid state coupling and dynamical manipulation. Order-2
parametric coupling term. While the static linear and nonlinear modes II-1 and II-2 are a pair of widely used degenerate
interactions generated by the electrostatic field have been pre- modes47,48. The antinodal axes (principal axes) of the order-2
viously studied38,39, capacitive nonlinear parametric coupling normal modes have an angular interval of 45 (Fig. 2a). This
between normal modes in a single resonator has not been pre- system can be equivalent to a two-degree-of-freedom lumped
sented previously. Here, it is shown that apart from the tension- parameter system in Cartesian coordinates (Fig. 2b). The angle in
induced parametric coupling, this electrostatic nonlinear para- this equivalent system is the double of that in the real system
metric coupling may reside, or even dominate, in capacitively depicted in Fig. 2a (see Supplementary Note 1). Axes x-o-y are
transduced devices30,31. Dynamical mode coupling between defined along the directions in which mechanical motions are
gyroscopic modes is implemented by applying a red-detuned probed, the drive and pump are applied along x and y directions,
pump adjusted based on the system parameters and the built-in respectively. Coordinates xω -o-yω are defined along the principal
intrinsic coupling. This capacitive device is demonstrated to be a axes of normal modes II-1 (xω direction) and II-2 (yω direction).
coupling-abundant multiple-mode system, which used to be very If the tuning voltage V t1 is applied along x direction, the
difficult to construct due to the difficulty of combining different resonant frequencies of modes II-1 and II-2 will disperse (Fig. 2d).
types of mode coupling mechanisms in one system40. Strong An avoided crossing is obviously illustrated when V t1 is around
dynamical coupling has significant practical applications to tun- 3.5 V, which indicates that the hybrid states observed from x-o-y
ing the response of micro- and nanoelectromechanical devices, coordinates H-1 (x direction) and H-2 (y direction) can be
such as mode-localized sensors41,42, Coriolis gyroscopes43,44, regarded as two coupled resonators, as depicted in Fig. 2c. This is
wireless filter resonators45, and many other devices16,32. also revealed by the nondiagonal equations of motion in x-o-y
coordinates (see Supplementary Note 2). Those hybrid states are
combinations of normal modes II-1 and II-2. The strength of the
Results avoided crossing indicates the coupling rate of the two hybrid
Capacitive symmetric electromechanical resonator. A capacitive states. As shown in Fig. 2e, the V t1 tuning process can be
symmetric microelectromechanical ring resonator is used as the simulated (see Supplementary Note 2). This coupling originates
experimental testbed in this work44. The resonator is constructed from two factors. The first one is the frequency difference between
of 45 equispaced nested rings. Adjacent rings are interconnected the normal modes II-1 and II-2 (Δω ¼ ωII-2  ωII-1 ≠ 0), and the

2 NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0 ARTICLE

a
Vt2
– +

5 Vd cos(dt)
6 4
Vt1 y
– + 7 3
x
8 2
Vp cos(pt)

+ 9 o 1
Lock-in amplifier
(reference d)
–
Charge 10 16
amplifiers V0
+ 11 15
+ –
– 12 14
13

b c
Disp. (a.u.)
II-1 III-1
0 1

10 2
Amplitude (mV)

8 1.5
II-2
6 III-2
1
4
2 0.5

0 0
134.18 134.22 134.26 166.3 166.5 166.7 166.9 167.1
Frequency (2 kHz)

Fig. 1 The experimental platform. a Basic setup for experiments. The nested-ring resonator is actuated, pumped, tuned, and sensed by equally distributed
capacitive electrodes numbered anticlockwise from 1 to 16. The drive signals are applied on electrodes 3 and 7 in the push–pull form. The pump signal is
applied on electrodes 1 and 5. A DC voltage V 0 is applied on the resonator body. An in-axis tuning voltage V t1 and an off-axis tuning voltage V t2 are applied
on electrodes 7 and 6, respectively. In the equivalent framework for order-2 degenerate modes, coordinates x and y are defined along drive and pump
directions, respectively. The set of axes x-o-y is of fourfold rotational symmetry. The response signal along x (y) axis is detected by electrodes 11 (9) and 15
(13) differentially, amplified using charge amplifiers, and measured by a lock-in amplifier. b, c Mechanical modes involved in this study. b The
amplitude–frequency responses and mode shapes of the order-2 in-plane modes, and c those of the order-3 in-plane modes. The displacements (disp.) are
normalized. Source data are provided as a Source Data file

second one is the misalignment (denoted as θ in the equivalent amplitude (normalized by mass) of mode II-j, respectively. Δp is
Cartesian coordinates, as shown in Fig. 2b) of the principal axes the pump produced by signal V p cosðωp tÞ,
with electrode axes, which are both due to manufacturing " #
tolerances. The initial θ of the tested resonator is estimated to V 2p V 2p
be about 18 . In the setup reference system, the real misalignment Δp ¼ κ 2V 0 V p cos ðωp tÞ   cos ð2ωp tÞ ; ð3Þ
2 2
is θ=2 (Fig. 2a). The mode shapes of the order-2 modes in the V t1
tuning process can be obtained by calculating θ at different values where κ ¼ Ap ϵ0 =ðd30 mII Þ, Ap is the area of pump electrodes, ϵ0 is
of V t1 , which are selectively depicted in Fig. 2f. Both Δω and θ can the permittivity of vacuum, and mII denotes the effective mass of
be changed by V t1 or V t2 (see Supplementary Note 2). A given set mode II-1 or II-2. There is a DC term, a first-order harmonic
of V t1 and V t2 can determine a specific group of Δω and θ. term, and a second-order harmonic term in the pump. The DC
By regarding mode II-2 as a phonon cavity, a dynamical term in the pump will tune the hybrid coupling condition, thus
sideband coupling operation can be realized based on the above affecting resonant frequencies (see Supplementary Fig. 6). The
structural coupling by applying a red-detuned parametric pump. resonant frequencies of order-2 modes are functions of V p and
In this implementation, V t1 and V t2 are tuned to make θ  36
and Δω  2π ´ 39:7 Hz. An alternating current signal V p cos ωp t are denoted as ωII-&;Vp (& ¼ 1; 2). When a 3-V pump is applied,
applied along y axis will periodically change the stiffness of hybrid ωII-1;3V (ωII-2;3V ) is slightly shifted from the initial ωII-1;0V
state H-2. The dynamics of the normal modes are given by the (ωII-2;0V ) by 0.33 Hz (0.60 Hz), as shown in Fig. 3f.
equations of motion The amplitude of the first-order term is 4V 0 =V p times that of
the second-order term. The first-order term is dominant in this
€xω þ γII x_ ω þ ðω2II-1 þ Δp sin2 θÞxω þ Δp cos θ sin θyω ¼ F 1 cos ðωd tÞ; ð1Þ implementation, since V 0 is much larger than V p . The second-
order term itself and its interaction with the first-order term only
€yω þ γII y_ ω þ ðω2II-2 þ Δp cos2 θÞyω þ Δp cos θ sin θxω ¼ F 2 cos ðωd tÞ; ð2Þ
slightly contribute to higher order ( 2) coupling. In the following
where xII-j and F j ðj ¼ 1; 2Þ are the displacement and drive force interpretation parts, the DC and second-order harmonic terms

NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0

a y y b c f
 ≈ –8°

Vt1: 10.5 V Vt1: 14 V

45° x y
/2 y y
5 x
6 4
7 3
H-2
II-2 x  ≈ –11°
8 2 II-1
o o H-1

9 o 1
x x
c  ≈ –18°

Vt1: 7 V
Disp. (a.u.)
10 16 1
11 15

Amplitude (a.u.)
12 13 14 0
 ≈ 45°

Vt1: 3.5 V
d Experiment e Simulation
20 20
 ≈ 18°

Vt1: 0 V
2 6 10 Min Max
15 15
Ampl. (mV) Ampl. (a.u.)
Tuning voltage Vt1 (V)

Tuning voltage Vt1 (V)

10 10
 ≈ 9°

Vt1: –7 V Vt1: –3.5 V

5 5

0 0
 ≈ 6°
–5 –5

–10 –10
134.1 134.2 134.3 134.4 134.1 134.2 134.3 134.4 134.1 134.2 134.3 134.4
Drive frequency d (2 kHz) Drive frequency d (2 kHz) Drive frequency d (2 kHz)

Fig. 2 Hybrid state coupling caused by structural asymmetry. a Normal modes II-1 (solid) and II-2 (transparent) in the setup coordinates. o-xω and o-yω
denote the principal axes of II-1 and II-2, respectively, both of which have the property of fourfold rotational symmetry. o-xω and o-yω have an intrinsic
angular interval of 45 . b The order-2 normal modes are equivalent to an order-reduced two-degree-of-freedom lumped parameter system in Cartesian
coordinates. θ is the angular offset between the equivalent probe coordinates x-o-y and principal axes xω -o-yω . The as-fabricated θ of 18 is revealed,
corresponding to the practical misalignment θ=2 of 9 in a. c The mismatched system is equivalent to two hybrid states H-1 and H-2 along x-o-y coupled by
a crossing spring. d Amplitude (Ampl.) frequency responses detected along the x axis at ωd with different values of V t1 ; an avoided crossing is illustrated.
e Simulation of electrostatic tuning in d. f Mode shape evolution when V t1 is changed. Amplitude frequency responses include orange experimental points
and blue theoretical fitting curves. The resonance peaks are marked with downward red arrows. In each mode-shape inset, black dot-dashed arrows
indicate the principal axis of the normal mode. Source data are provided as a Source Data file

are not considered. Γ1 ¼ 2κV 0 V p sin2 θ and Γ2 ¼ 2κV 0 V p cos2 θ sidebands 11, 12, 13, 14 are also generated by first- and second-
are defined as intra-modal coupling terms, and Λ ¼ order mixing processes, which would not contribute to the
2κV 0 V p cos θ sin θ is defined as an inter-modal coupling term. dynamical coupling.
When simultaneously applying drive and red-detuned pump Sidebands 9 and 10 will produce back actions on mode II-1.
with frequencies ωd  ωII-1 and ωp  Δω, the first-order side- However, the effects of sidebands 9 and 10 on mode II-1 are
opposite to each other. Based on the second-order coupling
band dynamical coupling process is illustrated in Fig. 3a. An idler
strength model in Supplementary Eq. (58), back actions caused by
spectrum line near ωII-2 is generated due to an up-conversion
sidebands p 9 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and 10 will completely cancelffi each other when
(Λþ ) process, in which the anti-Stokes sideband 3 with frequency
ωII-1 þ ωp is produced in the dynamics of mode II-2 during the θ ¼ arctan ð3ωII-1 þ ωII-2 Þ=ðωII-1 þ 3ωII-2 Þ  45 . In this con-
dition, the second-order dynamical coupling is invisible, as shown
mixing of the inter-modal coupling pump with the displacement
in Supplementary Fig. 3a.
of mode II-1. Meanwhile, the mixing of mode II-2 displacement
Here, we monitor the frequency response along the x axis at
with the inter-modal coupling pump causes a down-conversion
ωd , and the results with different pump frequencies are shown in
(Λ ) process, which produces a Stokes sideband 2 with frequency
Fig. 3c. The pump strength is maintained at V p ¼ 3 V. The
ωII-2  ωp in the dynamics of mode II-1. Sideband 2 has a phase
vertical dotted lines indicate the pump-on resonant frequencies of
delay relative to the external actuation tone. For the red-detuned
the modes II-1 and II-2. First level normal-mode splitting can be
pump, sideband 2 is in antiphase, which results in dynamic back-
observed when ωp  Δω, which indicates strong first-order
action cooling and avoided crossing. The off-resonance sideband
1 (4) is produced by the down (up)-conversion process of mode dynamical mode coupling. The avoided crossings induced by
II-1 (II-2), which will not affect the dynamical coupling. In this the very strong first-order dynamical coupling shift the lower
implementation, sidebands are generated near the mechanical branches of the resonance peaks, which makes the second-order
modes rather than the pumps in cavity optomechanics systems49, avoided crossings take place at a pump frequency higher than
because the eigen-frequencies of the mechanical modes are much Δω=2, as shown in Fig. 3c, d. However, it should be noted that the
higher than the frequency difference here. second-order dynamical coupling still takes place at the pump
If the pump frequency is changed to ωp  Δω=2, the second- frequency of exactly Δω=2. The observed second-order avoided
order sideband coupling process is illustrated in Fig. 3b. Wave crossings higher than Δω=2 are simultaneously affected by both
mixing processes of (Γ1þ ´ Λþ ) and (Λþ ´ Γ2þ ) of mode II-1 will the first- and the second-order coupling. Some slices of response
generate sidebands 7 and 8 in dynamics of mode II-2, curves near and between the first- and second-order coupling are
respectively. Similarly, wave mixing processes of (Γ2 ´ Λ ) and also provided in Fig. 3e to better illustrate the avoided crossing
(Λ ´ Γ1 ) of mode II-2 will generate sidebands 9 and 10 in evolution process. The Stückelberg interferometry pattern is also
dynamics of mode II-1. Off-resonance sidebands 5 and 6 act as observed at the bottom of the Fig. 3c where the pump frequency
the intermediary points for those processes. Other off-resonance is low50. In this system, the intra-modal coupling terms are

4 NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0 ARTICLE

a Δ II-1 II-2 b Δ/2 II-1 II-2

Δ Δ
p d p d Γ1+
Λ+ 3 10 Λ+ 6 Λ+ 8
1 4 12 Λ– 13
11 Γ1– 14
Λ– Γ2+
Frequency Λ 5 Γ
– 2–
Frequency

2 9 7

g1 g1 g2 g2
Normal-mode splitting Normal-mode splitting

c Vp = 3 V e Vp = 3 V f p = Δ
60 p = 40 Hz Expt. 10
Expt. Expt.
Pump frequency p (2 Hz)

1 3 5 10–3 Simu.
50 II-1,Vp II-2,Vp

Pump strength Vp (V)

Ampl. (mV) 8
10–4 i
Δ
40 i
ii 10–5 6
30 iii p = 35 Hz
–3
iv 10 4
Δ/2

Ampl. (mV)
20 v

3V
7
10–4 ii i
5
10 2 3
10–5 1
p = 30 Hz
Amplitude (V)

10–3
II-1,3V II-2,3V II-1,3V II-2,3V
d 10–4 iii
g
60 10
Simu. Simu.
Pump frequency p (2 Hz)

Min Max 10–5

50 p = 25 Hz

Pump strength Vp (V)

Ampl. (a.u.) 8
10–3
Δ

40 i
ii 10–4 iv 6
30 iii –5
10
iv p = 20 Hz 4 Max
Δ/2

Ampl. (a.u.)
20 v 10–3

3V
i
10 10–4 v 2
Min
–5
10
134,200 134,240 134,280 134,200 134,240 134,280 134,200 134,240 134,280
Drive frequency d (2 Hz) Drive frequency d (2 Hz) Drive frequency d (2 Hz)

Fig. 3 Dynamical coupling between modes II-1 and II-2. a Schematic spectrum of the first-order sideband coupling. The red-detuned pump is applied at Δω
(orange tone), and external force is actuating mode II-1 (black tone). The light blue (orange) region indicates the sweeping range of the pump (drive)
frequency in c. Λþ and Λ indicate inter-modal up- and down-conversion processes, respectively. Sideband 2 is in antiphase with respect to the drive tone.
Normal-mode splitting takes place if first-order pump is applied, and the frequency split indicates the first-order coupling rate g1 . b Schematic spectrum of
the second-order sideband coupling. The red-detuned pump is applied at Δω=2 (orange tone), and external force is actuating at mode II-1 (black tone). Γiþ
and Γi (i ¼ 1; 2) indicate intra-modal up- and down-conversion processes, respectively. Sideband 9 (10) is in antiphase (phase) with respect to the drive
tone. Sideband 7 is in antiphase with respect to sideband 8. Normal-mode splitting takes place if the second-order pump is applied, and the frequency split
indicates the second-order coupling rate g2 . It should be noted that the phase relations between different frequency tones are not depicted in a and b. c
Drive frequency ωd and pump frequency ωp response along x axis detected at ωd when V p ¼ 3 V. The vertical dotted lines indicate the mode frequencies
ωII-&;3V (& ¼ 1; 2) affected by the DC term in the 3-V pump. Expt. indicates experimental data. d Simulation of the experimental results in c by solving
Eqs. (1), (2), parameters of θ ¼ 36 and Δω ¼ 2π ´ 39:7 Hz are used in the simulation. Simu. indicates simulation results. e Some slices of frequency
response curves in c and d, which are denoted by numbers i–v. The orange points are experimental data and blue curves are simulation results. f The
frequency response along x axis as a function of pump amplitude V p when ωp ¼ Δω. The DC term in the pump will slightly shift resonant frequencies (black
solid curves) ωII-&;Vp to lower values. g Simulation of the experimental results in f. Source data are provided as a Source Data file

responsible for the Stückelberg interferometry phenomenon, and (2π ´ 1:05 Hz) by more than 28 times. Besides, when V p reaches
the inter-modal coupling term leads to the first-order dynamical values around 6 V, an intermediate avoided crossing is observed,
coupling. Both intra-and inter-modal coupling terms are which originates from the interaction of the two first-order idler
necessary for higher-order dynamical couplings. The experimen- sideband. If V p exceeds this value, the frequency split will decrease.
tal data can be reproduced by solving Eqs. (1), (2), as shown in The maximum coupling rate of this study is restricted by the limited
Fig. 3d (see Supplementary Note 3). value of Δω. It is noteworthy that the dynamical coupling strengths
The dynamical coupling strength is controllable. Fig. 3f shows the are both V p and θ dependent. Adjusting θ provides a new degree of
pump strength V p dependence of the first-order coupling rate g 1 ,
freedom to control dynamical coupling.
which is described by the frequency split. The first-order coupling
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi The pump strength dependence of the first-order coupling
rate is given by g 1 ¼ κ2 V 20 V 2p sin2 ð2θÞ=ð4ωII-1 ωII-2 Þ  γ2II . The strength can be reproduced by solving Eqs. (1), (2) (Fig. 3g).
theoretical values of mode splitting are depicted by dot-dashed lines There is an overall trend of dispersing to lower frequency for the
in Fig. 3f. In this study, first-order coupling rate of more than peaks in Fig. 3f, which is caused by the DC term in the pump Δp .
2π ´ 30 Hz can be obtained, which exceeds the damping rate The experiment and simulation results for another

NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0

implementation of θ  45 and Δω  2π ´ 40:7 Hz are also The displacement square of mode-III scaled by the third-order
provided in Supplementary Fig. 3. nonlinearity coefficient ν II will directly influence the effective
stiffness of mode-II. Eqs. (4), (5) are solved using multiple-
Electrostatic nonlinear coupling and dynamical manipulation. scale analysis, and the results are provided in Supplementary Eqs.
In this setup, the ring resonator is dominated by stiffness- (75, 76).
softening electrostatic nonlinearity because of the narrow capa- The dispersive frequency shifts of modes II caused by actuation
citive gap51, which is confirmed by the V 0 -dependence of the of mode III-1 can be simulated based on Eqs. (4), (5) (see
nonlinear response (see Supplementary Note 4). Doping-induced Supplementary Fig. 9). The frequency shift of mode II σ^II caused
material nonlinearity52 and tension-induced mechanical non- by displacement of mode III and that of mode III σ^III caused by
linearity51 may also reside, but they are relatively weak compared displacement of mode II are given by
to the electrostatic one when V 0 is set to be 30 V.  2 
It was previously demonstrated that the mechanical modes in 3ν d 2 γ f 2f 2 γ2
σ^II  II 0 II II2 þ 2III 2II ; ð6Þ
clamped–clamped beam resonators can be parametrically coupled 8ωII ωII ωIII γIII
to each other due to tension-induced mechanical nonlinearity,
because the displacement of one mode produces a beam  
extension thus modifying the other mode’s resonant 3ν III d20 γII f 2III γ2II 2f 2II
σ^III  þ ; ð7Þ
frequency18,22,23. Here, we demonstrate that the origin of 8ωIII ω2III γ2III ω2II
electrostatic nonlinearity can also produce intermodal parametric
coupling. The electrostatic nonlinear coupling between order-2 where f & ¼ F & =ðd0 γ2II Þ. The frequency shift direction of mode-
and order-3 modes is experimentally and theoretically described. II (the sign of σ^II ) is determined by the sign of the third-order
Modes II-1 (II-2) and III-1 are simultaneously actuated by nonlinearity coefficient ν II , which is negative for this resonator
drive and pump electrodes, respectively. Modes II are actuated in dominated by electrostatic nonlinearity. Thus, frequency of
linear region with a drive signal amplitude of 2 mV. Mode III-1 is mode-II will shift downward.
actuated in the stiffness-softening Duffing nonlinear condition As long as the associated modes can simultaneously modulate
with a drive signal amplitude of 10 mV. The frequency responses the response of a common capacitive transducer, those modes are
of modes II are recorded, and a dispersive parametric coupling is coupled. Thus, this electrostatic nonlinear mode coupling is very
observed. If the drive frequency of mode III-1 is changed from common in capacitive micro- or nanomechanical resonators. The
low to high, the dispersion of modes II is shown in Fig. 4d. If the coupling strength is significantly impacted by the characteristics
drive frequency of mode III-1 is changed from high to low, the of the shared capacitor.
dispersion of modes II is shown in Fig. 4e, and the shifts of modes It has been illustrated that the structural asymmetry will
II-1 and II-2 when mode III-1 is at resonance are 5.9 Hz and transform normal modes II-1 and II-2 into mechanically coupled
6.2 Hz, respectively. The nonlinear bifurcation of mode III-1 is hybrid states H-1 and H-2. The hybrid states are electrostatically
revealed by the frequency dispersion of modes II. When mode coupled to order-3 modes, an anti-Stokes pump can transform
III-1 is actuated, the resonant frequencies of modes II will this electrostatic coupling into tunable strong dynamical
disperse to lower values. This phenomenon is opposite to that of coupling. In this case, order-3 modes are regarded as phonon
the mechanical nonlinear parametric coupling18,22,23, which cavities. The drive signal (V d ¼ 2 mV) sweeps from below mode
should make modes II disperse to higher resonance frequencies. II-1 to above mode II-2, and the pump signal (V p ¼ 3 V) with
The observed parametric mode coupling can be explained by frequency from below ωIII-1  ωII-2 to above ωIII-2  ωII-1 are
the model that two modes sharing one polarized capacitor. It was simultaneously applied. The frequency responses of modes II-1
previously shown that a mechanical resonator can be coupled to a and II-2 are measured, as shown in Fig. 4f, in which a skewed “#”
microwave cavity using a capacitor10,53,54, and the stiffness of the configuration with four avoided crossings is illustrated. When the
mechanical resonator is modified by varying the bias voltage on pump frequency is kept ωIII-1  ωII-1 , the dynamical coupling
the capacitor. In this model, the resonant frequency of mode II is strength of modes II-1 and III-1 depicted by the frequency split of
modified by the capacitance gap variation induced by displace- mode II-1 will increase if pump voltage is increased (Fig. 4h). The
ment of mode III. When two modes are actuated simultaneously, pump voltage dependence of coupling strength of modes II-2 and
their displacements are superposed (Fig. 4a–c). Displacement of III-1 is also obtained (Fig. 4j). Moreover, avoided crossings can be
heavily actuated mode III will cause a electrostatic stiffness observed between the higher splitting branch of II-1 and mode II-
variation for mode II, thus modifying its resonant frequency. The 2 in Fig. 4h, and between the lower splitting branch of II-2 and
capacitive parametric coupling can be described by following mode II-1 in Fig. 4j, which are produced by the structural hybrid
equations of motion (see Supplementary Note 4): coupling of H-1 and H-2.
The skewed “#” coupling configuration are simulated by
€xII þ γII x_ II þ ω2II xII þ αII xIII þ βII ðxII þ xIII Þ2 separately modeling the sequential dynamical coupling of III-1 to
ð4Þ
þ ν II ðxII þ xIII Þ3 ¼ F II sinðωd-II tÞ; II-1 and II-2 and that of III-2 to II-1 and II-2, which can be
described similarly by the following equations of motion:
€xIII þ γIII x_ III þ ω2III xIII þ αIII xII þ βIII ðxII þ xIII Þ2 €xII-1 þ γII x_ II-1 þ ω2II-1 xII-1 þ α1 xII-1 þ β1 xII-2 þ λ1 xIII
3
ð5Þ
þ ν III ðxII þ xIII Þ ¼ F III sinðωd-III tÞ; þ Λ1 cosðωp tÞðxII-1 sin θ þ xII-2 cos θ þ xIII Þ ð8Þ
where xII and ωII denote the displacement and resonant ¼ g 1 cosðωd tÞ;
angular frequency of mode II-1 or II-2, and xIII and ωIII denote
the displacement and resonant angular frequency of one order-3
mode (mode III-1 in this case). F & and ωd-& (& ¼ II, III) are
the amplitudes and frequencies of the external forces acting on €xII-2 þ γII x_ II-2 þ ω2II-2 xII-2 þ α2 xII-1 þ β2 xII-2 þ λ2 xIII
the corresponding modes. Parameters α& , β& , ν & are provided þ Λ2 cosðωp tÞðxII-1 sin θ þ xII-2 cos θ þ xIII Þ ð9Þ
by Supplementary Eqs. (70–72). The key factor for the observed
mode-II frequency shift is the 3ν II x2III xII term in expanded Eq. (4). ¼ g 2 cosðωd tÞ;

6 NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0 ARTICLE

a d Sweep up e Sweep down

III-1 III-1
actuated alone
20 20

Ampl.

Ampl.
(mV)

(mV)
0
d
Mode II
10 10

II
x
c

El
ΔV
134.25 134.25

ec
II-2 II-2

actuated simultaneously

t ro
de
Modes II and III

Frequency (2 kHz)

Frequency (2 kHz)
6.2 Hz

0
d
III
+x

III
x
134.24 134.24

II
+x

II
x
0
d
b
II-1 II-1
0
d
actuated alone

134.20 134.20
III
x

5.9 Hz
Mode III

Ampl. (mV) Ampl. (mV)

1357 1357
134.19 134.19
166,476 166,480 166,484 166,476 166,480 166,484
Frequency (2 Hz) Frequency (2 Hz)

f h 10
j 10
32.76 Expt. Expt. Expt.
Ampl. (mV)

8
Pump frequency (2 kHz)

6
32.72 8 8
Pump strength (V)

Pump strength (V)

4
2
32.68 6 6

vi 4 4

Ampl. (mV)

Ampl. (mV)
32.28 vi 7 vii 7
5 5
32.24 vii 2 3 2 3
1 1
32.20

g i k
10 10
32.76 Simu. Max Simu. Simu.
Ampl. (a.u.)
Pump frequency (2 kHz)

32.72 8 8
Pump strength (V)

Pump strength (V)

32.68 Min 6 6

Max Max
Ampl. (a.u.)

Ampl. (a.u.)
vi 4 4
32.28 vi vii

32.24 vii 2 2
Min Min
32.20
134.16 134.20 134.24 134.28 134.16 134.20 134.24 134.28 134.16 134.20 134.24 134.28
Drive frequency (2 kHz) Drive frequency (2 kHz) Drive frequency (2 kHz)

Fig. 4 Electrostatic nonlinear parametric coupling and demonstration of coupling-abundant multiple-mode system. a, b Schematic transient pattern of the
independently actuated mode II (a) and mode III (b). c Schematic transient pattern of the simultaneously actuated modes. Their superposed displacements
will both affect the capacitive gap. Modal interaction occurs when a nonlinear electrostatic potential is applied. d, e The dispersive frequency shifts of
modes II induced by the actuation of mode III-1 if the drive frequency of mode III-1 is changed from low to high (d) and high to low (e). The dotted lines
indicate the resonant frequencies of the modes II without the additional actuation of mode III-1. The bifurcation-induced jump phenomenon in nonlinear
frequency response of mode III-1 (upside) can be detected by the frequency shift of modes II. f Frequency responses of modes II when pump frequency are
changed from below ωIII-1  ωII-2 to above ωIII-2  ωII-1 depict a skewed “#” configuration. The pump voltage is 3 V. g Simulation of the experimental
results in f. h The strength of the dynamical coupling between modes II-1 and III-1 as a function of pump amplitude V p , pump frequency is set to be
ωIII-1  ωII-1 . i Simulation of the experimental results in h. j The strength of the dynamical coupling between modes II-2 and III-1 as a function of pump
amplitude V p , pump frequency is set to be ωIII-1  ωII-2 . k Simulation of the experimental results in j. Source data are provided as a Source Data file

€xIII þ γIII x_ III þ ω2III xIII þ α3 xII-1 þ β3 xII-2 þ λ3 xIII electrostatic mechanical system has the potential to couple even
ð10Þ more modes. If the dynamical couplings are implemented
þ Λ3 cosðωp tÞðxII-1 sin θ þ xII-2 cos θ þ xIII Þ ¼ 0: simultaneously, a classical analog of multiple-level system could
The subscript III refers to mode III-1 or III-2. The detailed be constructed, which would enable abundant varieties of
derivation of those equations of motion and the parameters αj , βj , interesting manipulations40.
λj , Λj , and g j (j ¼ 1; 2; 3) are provided in Supplementary Note 5.
By solving Eqs. (8)–(10), we can simulate the observed avoided Discussion
crossings, as shown in Fig. 4g. The V p dependence of coupling In this study, we report significant progress in modeling and
strengths depicted in Fig. 4h, j can also be simulated by setting manipulating the structural asymmetry and misalignment
ωp ¼ ωIII-1  ωII-1 and ωIII-1  ωII-2 , as shown in Fig. 4i, k, induced mode coupling between a pair of degenerate wine-glass
respectively. modes in ring resonators. We also discover an electrostatic field-
By showing the abundant coupling phenomena between imposed nonlinear parametric coupling effect among different
order-2 and order-3 modes, we demonstrate that this modes in a single resonator. These modal coupling effects in

NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0

capacitively transduced mechanical systems can be tuned very Received: 1 February 2019; Accepted: 18 September 2019;
significantly through the parametric pumping scheme.
Electrostatic-pump-based manipulations show consequent
advantages with respect to tension-mediated manipulations.
Depending on the topology of the resonators, the effect of
tension-induced manipulations can vary. For instance, its effect References
can be significant for clamped–clamped or thin film resonators, 1. Nguyen, C. T. C. MEMS technology for timing and frequency control. IEEE
but negligible for centrally anchored or bulk-fabrication-process Trans. Ultrason. Ferroelectr. Freq. Control 54, 251–270 (2007).
resonators, such as the ring resonator in this work, if the same 2. Ng, E. et al. The long path from MEMS resonators to timing products. In
Proceedings of 28th IEEE International Conference on Micro Electro
tension is applied. In comparison, the electrostatic-medi-
Mechanical Systems. 1–2 (IEEE, 2015).
ated pump-based manipulation removes such topology restric- 3. Roy, S. K., Sauer, V. T. K., Westwood-Bachman, J. N., Venkatasubramanian,
tions for capacitively transduced resonators, hence making it A. & Hiebert, W. K. Improving mechanical sensor performance through larger
applicable for a wider range of resonator designs. damping. Science 360, eaar5220 (2018).
Recently, significant breakthroughs in enhancing quality factor 4. Moser, J. et al. Ultrasensitive force detection with a nanotube mechanical
resonator. Nat. Nanotechnol. 8, 493–496 (2013).
(Q) of mechanical resonators have been made55–58. Though some 5. Hanay, M. S. et al. Single-protein nanomechanical mass spectrometry in real
of those resonators are dielectric, the proposed dissipation miti- time. Nat. Nanotechnol. 7, 602–608 (2012).
gating techniques could inspire conductive high-Q mechanical 6. Unterreithmeier, Q. P., Weig, E. M. & Kotthaus, J. P. Universal transduction
resonators. An interesting direction for further research is to scheme for nanomechanical systems based on dielectric forces. Nature 458,
combine the electrostatic-mediated parametric coupling and 1001–1004 (2009).
7. Hafiz, M. A. A., Kosuru, L. & Younis, M. I. Microelectromechanical
dynamical manipulations illustrated here in such conductive high-
reprogrammable logic device. Nat. Commun. 7, 11137 (2016).
Q mechanical resonators, which may enable purely phonon-based 8. Hatanaka, D., Mahboob, I., Onomitsu, K. & Yamaguchi, H. Phonon
quantum sideband manipulations at the macroscopic scale10,11. waveguides for electromechanical circuits. Nat. Nanotechnol. 9, 520–524
Some interesting dynamical tunable coupling experiments have (2014).
been implemented in capacitive mechanical systems30,31. How- 9. Mahboob, I. & Yamaguchi, H. Bit storage and bit flip operations in an
electromechanical oscillator. Nat. Nanotechnol. 3, 275–279 (2008).
ever, these effects were attributed to tension-induced parametric 10. Ockeloen-Korppi, C. F. et al. Stabilized entanglement of massive mechanical
interaction. Here, we demonstrate that electrostatic field-induced oscillators. Nature 556, 478–482 (2018).
parametric coupling should also reside in these systems, and may 11. Riedinger, R. et al. Remote quantum entanglement between two
at least partly contribute to the observed results. micromechanical oscillators. Nature 556, 473–477 (2018).
Electrostatic mechanical resonators are widely used for sensing 12. Møller, C. B. et al. Quantum back-action-evading measurement of motion in a
negative mass reference frame. Nature 547, 191–195 (2017).
applications, and their susceptibility to mode coupling has often
13. Wollman, E. E. et al. Quantum squeezing of motion in a mechanical
been not fully assessed in published studies. This paper shows resonator. Science 349, 952–955 (2015).
that mode coupling should be taken into account when designing 14. Ganesan, A., Do, C. & Seshia, A. A. Phononic frequency comb via intrinsic
sensors, as the associated interactions may greatly alter the three-wave mixing. Phys. Rev. Lett. 118, 033903 (2017).
properties of the mechanical devices. More interestingly, elec- 15. Matheny, M. H. et al. Phase synchronization of two anharmonic
nanomechanical oscillators. Phys. Rev. Lett. 112, 014101 (2014).
trostatic dynamical sideband coupling may be very useful in
16. Agrawal, D. K., Woodhouse, J. & Seshia, A. A. Observation of locked phase
terms of enhancing sensor performance, such as manipulating dynamics and enhanced frequency stability in synchronized micromechanical
energy transfer between dynamically coupled modes to improve oscillators. Phys. Rev. Lett. 111, 084101 (2013).
or decrease Q factor of a specific mode59, greatly speeding up 17. Shim, S. B., Imboden, M. & Mohanty, P. Synchronized oscillation in coupled
mode switching for Coriolis gyroscopes by replacing the stop- nanomechanical oscillators. Science 316, 95–99 (2007).
18. Westra, H. J. R., Poot, M., van der Zant, H. S. J. & Venstra, W. J. Nonlinear
decay-actuate process with coherent mode manipulation60, thus modal interactions in clamped-clamped mechanical resonators. Phys. Rev.
calibrating the bias error without affecting the bandwidth, Lett. 105, 117205 (2010).
dynamically tuning the frequency split of gyroscopes, and pro- 19. Eichler, A., del Álamo Ruiz, M., Plaza, J. A. & Bachtold, A. Strong coupling
viding tunable coupling for mode-localized sensors. between mechanical modes in a nanotube resonator. Phys. Rev. Lett. 109,
025503 (2012).
20. Antonio, D., Zanette, D. H. & López, D. Frequency stabilization in nonlinear
Methods micromechanical oscillators. Nat. Commun. 3, 806 (2012).
Experiment and simulation setup. The die embedding the vacuum-sealed MEMS 21. Faust, T. et al. Nonadiabatic dynamics of two strongly coupled
resonators is electrically packaged in a ceramic leadless chip carrier. The device is nanomechanical resonator modes. Phys. Rev. Lett. 109, 037205 (2012).
operated in an ambient room temperature environment. The bias voltages are
22. Mahboob, I., Nishiguchi, K., Okamoto, H. & Yamaguchi, H. Phonon-cavity
generated by a low noise voltage source (Keysight B2961A). The drive and pump
electromechanics. Nat. Phys. 8, 387–392 (2012).
signals are provided by a two-channel lock-in amplifier (Zurich Instruments
23. Matheny, M. H., Villanueva, L. G., Karabalin, R. B., Sader, J. E. & Roukes, M.
HF2LI). The response motion of the resonator is detected by a capacitance–voltage
L. Nonlinear mode-coupling in nanomechanical systems. Nano Lett. 13,
converting scheme that is based on charge amplifier, and measured by the lock-in
1622–1626 (2013).
amplifier. The simulation codes are based on Python 3.7 with NumPy and Mat-
plotlib packages. 24. Mahboob, I. et al. Dispersive and dissipative coupling in a micromechanical
resonator embedded with a nanomechanical resonator. Nano Lett. 15,
2312–2317 (2015).
Data availability 25. Chen, C., Zanette, D. H., Czaplewski, D. A., Shaw, S. & López, D. Direct
The authors declare that all data supporting the findings of this study are included in the observation of coherent energy transfer in nonlinear micromechanical
paper and its supplementary information files, and are available on request from the oscillators. Nat. Commun. 8, 15523 (2017).
corresponding authors. The source data underlying Figs. 1b, 2d, f, 3c, e, f, and 4d, e, f, h, j 26. Seitner, M. J., Abdi, M., Ridolfo, A., Hartmann, M. J. & Weig, E. M.
and Supplementary Figs. 2c, 3a, c, e, 4c–i, 5, 6, 7, and 8 are provided as a Source Data file, Parametric oscillation, frequency mixing, and injection locking of strongly
which is also available in figshare (https://doi.org/10.6084/m9.figshare.8397908). coupled nanomechanical resonator modes. Phys. Rev. Lett. 118, 254301
(2017).
27. Luo, G. et al. Strong indirect coupling between graphene-based mechanical
Code availability resonators via a phonon cavity. Nat. Commun. 9, 383 (2018).
The authors declare that all codes supporting the findings of this study are included in
28. Okamoto, H. et al. Coherent phonon manipulation in coupled mechanical
the paper and its supplementary information files, and are available on request from the
resonators. Nat. Phys. 9, 598–598 (2013).
corresponding authors. The simulation codes underlying Figs. 2e, 3d, g, and 4g, i, k and
29. Liu, C. H., Kim, I. S. & Lauhon, L. J. Optical control of mechanical mode-
Supplementary Figs. 2d, e, 3b, d, f, 4c–i, 6, and 9 are available in figshare (https://doi.org/
coupling within a MoS2 resonator in the strong-coupling regime. Nano Lett.
10.6084/m9.figshare.8397908).
15, 6727–6731 (2015).

8 NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12796-0 ARTICLE

30. De Alba, R. et al. Tunable phonon-cavity coupling in graphene membranes. 57. Güttinger, J. et al. Energy-dependent path of dissipation in nanomechanical
Nat. Nanotechnol. 11, 741–746 (2016). resonators. Nat. Nanotechnol. 12, 631–636 (2017).
31. Mathew, J. P., Patel, R. N., Borah, A., Vijay, R. & Deshmukh, M. M. 58. Moser, J., Eichler, A., Güttinger, J., Dykman, M. I. & Bachtold, A. Nanotube
Dynamical strong coupling and parametric amplification of mechanical mechanical resonators with quality factors of up to 5 million. Nat.
modes of graphene drums. Nat. Nanotechnol. 11, 747–751 (2016). Nanotechnol. 9, 1007–1011 (2014).
32. Sun, F., Dong, X., Zou, J., Dykman, M. I. & Chan, H. B. Correlated anomalous 59. Venstra, W. J., Westra, H. J. R. & van der Zant, H. S. J. Q-factor control of a
phase diffusion of coupled phononic modes in a sideband-driven resonator. microcantilever by mechanical sideband excitation. Appl. Phys. Lett 99,
Nat. Commun. 7, 12694 (2016). 013112 (2011).
33. Mahboob, I., Okamoto, H., Onomitsu, K. & Yamaguchi, H. Two-mode 60. Okamoto, H., Mahboob, I., Onomitsu, K. & Yamaguchi, H. Rapid switching in
thermal-noise squeezing in an electromechanical resonator. Phys. Rev. Lett. high-Q mechanical resonators. Appl. Phys. Lett. 105, 083114 (2014).
113, 167203 (2014).
34. Patil, Y. S., Chakram, S., Chang, L. & Vengalattore, M. Thermomechanical
two-mode squeezing in an ultrahigh-Q membrane resonator. Phys. Rev. Lett. Acknowledgements
115, 017202 (2015). We thank Sijun Du, Milind Pandit, Malar Chellasivalingam, and Atif Aziz for helpful
35. Faust, T., Rieger, J., Seitner, M. J., Kotthaus, J. P. & Weig, E. M. Coherent control discussions and assisting with training on laboratory equipment. This work is partly
of a classical nanomechanical two-level system. Nat. Phys.9, 485–488 (2012). supported by the National Key R&D Program of China (NKPs) (2018YFB2002304) and
36. Zhu, D. et al. Coherent phonon Rabi oscillations with a high-frequency carbon the National Natural Science Foundation of China (NSFC) (51905539, 51575521 and
nanotube phonon cavity. Nano Lett. 17, 915–921 (2017). 51705527). Experimental devices are designed and fabricated in the nano@Stanford labs,
37. Mahboob, I., Nishiguchi, K., Fujiwara, A. & Yamaguchi, H. Phonon lasing in which are supported by the NSF as part of the National Nanotechnology Coordinated
an electromechanical resonator. Phys. Rev. Lett. 110, 127202 (2013). Infrastructure under award ECCS-1542152, with support from the Defense Advanced
38. Truitt, P. A., Hertzberg, J. B., Altunkaya, E. & Schwab, K. C. Linear and Research Projects Agency Precise Robust Inertial Guidance for Munitions (PRIGM)
nonlinear coupling between transverse modes of a nanomechanical resonator. Program, managed by Dr. Robert Lutwak and Dr. Ron Polcawich, and the NSF under
J. Appl. Phys. 114, 114307 (2013). grant number CMMI-1662464. This work is primarily supported by the UK Natural
39. Buks, E. & Roukes, M. L. Electrically tunable collective response in a coupled Environment Research Council under grant number NE/N012097/1.
micromechanical array. J. Microelectromech. Syst. 11, 802–807 (2002).
40. Okamoto, H. et al. A strongly coupled Λ-type micromechanical system. Appl. Author contributions
Phys. Lett. 108, 227402 (2016). X.Z. conceived the idea and designed the research under the guidance of A.A.S. and
41. Spletzer, M., Raman, A., Wu, A. Q., Xu, X. & Reifenberger, R. Ultrasensitive assistance of C.Z. The measurements and data analyses were performed by X.Z. and C.Z.
mass sensing using mode localization in coupled microcantilevers. Appl. Phys. with the assistance of J.S. and G.S. The test circuitry was developed by D.X. and X.Z. The
Lett. 88, 254899 (2006). theoretical works were done by X.Z., supervised by A.A.S., D.X. and X.W. The device was
42. Thiruvenkatanathan, P., Yan, J., Woodhouse, J., Aziz, A. & Seshia, A. A. designed and fabricated by D.D.G., Y.C., I.F. and T.W.K. The manuscript was written by
Ultrasensitive mode-localized mass sensor with electrically tunable parametric X.Z. and A.A.S. All authors contributed to manuscript preparation. The project was
sensitivity. Appl. Phys. Lett. 96, 083562 (2010). planned by A.A.S.
43. Ayazi, F. & Najafi, K. A HARPSS polysilicon vibrating ring gyroscope. J.
Microelectromech. Syst. 10, 169–179 (2001).
44. Ahn, C. H. et al. Encapsulated high frequency (235 kHz), high-Q (100 k) disk Competing interests
resonator gyroscope with electrostatic parametric pump. Appl. Phys. Lett. 105, The authors declare no competing interests.
243504 (2014).
45. Nguyen, C. T. C. Frequency-selective MEMS for miniaturized low-power
communication device. IEEE Trans. Microw. Theory Tech. 47, 1486–1503 (1999).
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41467-
46. Yang, Y., Ng, E. J., Chen, Y., Flader, I. B. & Kenny, T. W. A unified Epi-seal
019-12796-0.
process for fabrication of high-stability microelectromechanical devices. J.
Microelectromech. Syst. 25, 489–497 (2016).
Correspondence and requests for materials should be addressed to D.X. or A.A.S.
47. Zhou, X. et al. Mitigating thermoelastic dissipation of flexural
micromechanical resonators by decoupling resonant frequency from thermal
Reprints and permission information is available at http://www.nature.com/reprints
relaxation rate. Phys. Rev. Appl. 8, 064033 (2017).
48. Nitzan, S. H. et al. Self-induced parametric amplification arising from
Peer Review Information Nature Communications thanks G. P. Guo, Zenghui Wang
nonlinear elastic coupling in a micromechanical resonating disk gyroscope.
and other anonymous reviewer(s) for their contribution to the peer review of this work.
Sci. Rep. 5, 9036 (2015).
Peer reviewer reports are available.
49. Kippenberg, T. J. & Vahala, K. J. Cavity optomechanics: back-action at the
mesoscale. Science 321, 1172–1176 (2008).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
50. Ivakhnenko, O. V., Shevchenko, S. N. & Nori, F. Simulating quantum
published maps and institutional affiliations.
dynamical phenomena using classical oscillators: Landau-Zener-Stückelberg-
Majorana interferometry, latching modulation, and motional averaging. Sci.
Rep. 8, 12218 (2018).
51. Lifshitz, R., Cross, M. C. Nonlinear Dynamics of Nanomechanical and Open Access This article is licensed under a Creative Commons
Micromechanical Resonators (ed Schuster, H. G.) 1–52 (Wiley VCH Verlag Attribution 4.0 International License, which permits use, sharing,
GmbH & Co. KGaA, 2008). adaptation, distribution and reproduction in any medium or format, as long as you give
52. Yang, Y. et al. Nonlinearity of degenerately doped bulk-mode silicon MEMS appropriate credit to the original author(s) and the source, provide a link to the Creative
resonators. J. Microelectromech. Syst. 25, 859–869 (2016). Commons license, and indicate if changes were made. The images or other third party
53. Rocheleau, T. et al. Preparation and detection of a mechanical resonator near material in this article are included in the article’s Creative Commons license, unless
the ground state of motion. Nature 463, 72–75 (2010). indicated otherwise in a credit line to the material. If material is not included in the
54. Brown, K. R. et al. Passive cooling of a micromechanical oscillator with a article’s Creative Commons license and your intended use is not permitted by statutory
resonant electric circuit. Phys. Rev. Lett. 99, 137205 (2007). regulation or exceeds the permitted use, you will need to obtain permission directly from
55. Ghadimi, A. H. et al. Elastic strain engineering for ultralow mechanical the copyright holder. To view a copy of this license, visit http://creativecommons.org/
dissipation. Science 360, 764–768 (2018). licenses/by/4.0/.
56. Tsaturyan, Y., Barg, A., Polzik, E. S. & Schliesser, A. Ultracoherent
nanomechanical resonators via soft clamping and dissipation dilution. Nat.
Nanotechnol. 12, 776–783 (2017). © The Author(s) 2019

NATURE COMMUNICATIONS | (2019)10:4980 | https://doi.org/10.1038/s41467-019-12796-0 | www.nature.com/naturecommunications 9