3D microwave optomechanical cavities · 2020. 1. 17. · Kavli Institute of Nanoscience, ......
Transcript of 3D microwave optomechanical cavities · 2020. 1. 17. · Kavli Institute of Nanoscience, ......
3D microwave optomechanical cavities
Mingyun Yuan
Vibhor Singh, Yaroslav Blanter, Martijn Cohen, Shun Yanai, Gary Steele
Kavli Institute of Nanoscience, TU Delft
1
Motivation‣ Quantum
Is quantum mechanics compatible with large, massive structures?
Nano- or micro-mechanical resonator
‣ Classical
+
-
2
Interaction between cavity field and a movable mirror
Cavity optomechanics
∆x
light field
To measure mechanical motion
To create and control mechanical motion
3
Landmarks
the zero-point motion is xzp~ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiB= 2mVmð Þ
p~4:1 fm.With a ratio of
Vm/k. 50, our system is deep within the resolved-sideband regimeandwell-suited for sidebandcooling to themechanical ground state22,23.To measure the mechanical displacement, we apply a microwave
field, which is detuned below the cavity resonance frequency byD52Vm, through heavily attenuated coaxial lines to the feed lineof our device. The upper sideband, now at vc, is amplified with acustom-built Josephson parametric amplifier26,27 followed by a low-noise cryogenic amplifier, demodulated at room temperature, andfinally monitored with a spectrum analyser. The thermal motion ofthemembrane creates an easily resolvable peak in themicrowave noisespectrum. As described previously27, this measurement scheme con-stitutes a nearly shot-noise-limited microwave interferometer withwhich we can measure mechanical displacement with minimumadded noise close to fundamental limits.In order to calibrate the demodulated signal to the membrane’s
motion, wemeasure the thermal noise spectrumwhile varying the cryo-stat temperature (Fig. 1c). Here a weakmicrowave drive (,3 photons inthe cavity) is used in order to ensure that radiation pressure dampingand cooling effects are negligible. WhenVm?k?Cm andD52Vm,the displacement spectral density Sx is related to theobservedmicrowavenoise spectral density S by Sx5 2(kVm/Gkex)
2S/Po, where kex is the
coupling rate between the cavity and the feed line, and Po is the powerof the microwave drive at the output of the cavity. According to equi-partition, the area under the resonance curve of displacement spectraldensity Sx must be proportional to the effective temperature of themechanical mode. This calibration procedure allows us to convert thesideband in the microwave power spectral density to a displacementspectral density and to extract the thermal occupation of the mech-anical mode. In Fig. 1c we show the number of thermal quanta in themechanical resonator as a function of T. The linear dependence of theintegrated power spectral density with temperature shows that themechanical mode equilibrates with the cryostat even for the lowestachievable temperature of 15mK. This temperature corresponds to athermal occupancy nm5 30, where nm5 [exp(BVm/kBT)2 1]21. Thecalibration determines the electromechanical coupling strength,G/2p5 496 2MHznm21. With these device parameters, we caninvestigate both the fundamental sensitivity of our measurementand the effects of radiation pressure cooling.The total measured displacement noise results from two sources: the
membrane’s actual mean-square motion, Sthx , and its apparent motion,Simpx , which is due to imprecisionof themeasurement. Figure 2a demon-strates how the use of low-noise parametric amplification significantly
b
ωd
a
10–1
100
101
102
103
10–1 100 101 102
Ther
mal
occ
upan
cy, n
c
Γ
Γ–
mωd – Ω
Γ = Γ – Γ–
mωd + Ωωd
Cryostat temperature (mK)
c
2 μm
+
+
Figure 1 | Schematic description of the experiment. a, False-colour scanningelectron micrograph showing the aluminium (grey) electromechanical circuitfabricated on a sapphire (blue) substrate; a 15-mm-diameter membrane issuspended 50 nm above a lower electrode. The membrane’s motion modulatesthe capacitance, and hence, the resonance frequency of the superconductingmicrowave circuit. b, A coherent microwave drive (left, vd, shown green infrequency–amplitude plot below) inductively coupled to the circuit (top)acquires modulation sidebands (red and blue in plot below) owing to thethermal motion of the membrane. The upper sideband is amplified with anearly quantum-limited Josephson parametric amplifier (filled triangle, right)within the cryostat. c, The microwave power in the upper sideband provides adirect measurement of the thermal occupancy of the mechanical mode, whichmay be calibrated in situ by varying the temperature of the cryostat (mainpanel). The mechanical mode shows thermalization with the cryostat at alltemperatures, yielding aminimum thermal occupancy of 30mechanical quantawithout using sideband-cooling techniques. Error bars, s.d. Inset, illustration ofthe concept of sideband cooling. When the circuit is excited with a detunedmicrowave drive such that D52Vm, the narrow line shape of the electricalresonance ensures that the rate to scatter photons to higher energy C1 (bluedashed arrow, blue peak) exceeds the rate to scatter to lower energy C2 (reddashed arrow, red peak). Thus, the net scattering rate C (blue solid arrow)provides a cooling mechanism for the membrane.
ba
d
c
400
300
200
100
0
Sx (fm
2 H
z–1)
–400 0 400
100
101
102
103
104
105
101
102
103
n imp
nd
10–32
10–34
100 102
10–30
10–28
imp
Sx
(m
2 H
z–1)
104
100
κ/2πg/π
m/2π′Γ
102 104
100 102 104
Cou
plin
g ra
tes
(Hz)
(ω m)/2π (Hz) – Ω
Figure 2 | Displacement sensitivity in the presence of dynamical back-action. a, The displacement spectral density Sx measured with (red) andwithout (blue) the Josephson parametric amplifier. As the parametric amplifiergreatly reduces the total noise of the microwave measurement, the timerequired to resolve the thermal motion is reduced by a factor of 1,000. b, As themicrowave drive power is increased, the absolute displacement sensitivity, Simp
ximproves, reaching a minimum of 5.53 10234m2Hz21 at the highest power.c, The parametric coupling rate g between the microwave cavity and themechanicalmode increases as
ffiffiffiffiffind
p. This coupling broadens the linewidth of the
mechanical mode C ’m from its intrinsic value of Cm5 2p3 32 Hz until itexceeds the linewidth of the microwave cavity k. d, The relative measurementimprecision, in units of mechanical quanta, depends on the product of Simp
x andC ’m. Thus, once the power is large enough that dynamical back-actionoverwhelms the intrinsic mechanical linewidth, nimp asymptoticallyapproaches a constant value (nimp5 1.9), which is a direct measure of theoverall efficiency of the photon measurement.
RESEARCH LETTER
3 6 0 | N A T U R E | V O L 4 7 5 | 2 1 J U LY 2 0 1 1
Macmillan Publishers Limited. All rights reserved©2011
Teufel et al., Nature 2011
In the resolved-sideband limit, where vm/k. 1, driving the systemwith a laser (frequency, vl) tuned to the red side of the optical cavity(detuning, D;vo2vl5vm), creates an optically induced damp-ing, cOM, of the mechanical resonance25. In the weak-couplingregime (cOM=k), the optical back-action damping is given bycOM5 4g2nc/k, where nc is the average number of drive-laser photonsstored in the cavity and g is the optomechanical coupling rate betweenthe mechanical and optical modes. This coupling rate, g, is quantifiedas the shift in the optical resonance for an amplitude ofmotion equal to
the zero-point fluctuation amplitude (xzpf~(B=2mvm)1=2, wherem is
the motional mass of the localized acoustic mode and B is Planck’sconstant divided by 2p). The optomechanical damping, which is aresult of the preferential scattering of drive photons into the upper-frequency sideband, also cools the mechanical mode. For a quantum-limited drive laser, the phonon occupancy of the mechanical oscillatorcan be reduced from nb~kBTb=Bvm?1 to !n~nb= 1zCð Þznmin,where kB is Boltzmann’s constant and C; cOM/ci is the cooperativity.The residual scattering of drive photons into the lower-frequencysideband limits the cooled phonon occupancy to nmin5 (k/4vm)
2,which is determined by the level of sideband resolution25.The drive laser, in addition to providing mechanical damping and
cooling, can be used to measure the mechanical and optical propertiesof the system through a series of calibratedmeasurements. In a first setof measurements, we use the noise power spectral density (PSD) ofthe drive laser transmitted through the optomechanical cavity toperform spectroscopy of the mechanical mode. As shown in Sup-plementary Information, the noise PSD of the photocurrent generatedby the transmitted field of the drive laser with red-sideband detun-ing (D5vm) yields a Lorentzian component of the single-sidedPSD proportional to Sb vð Þ~!nc
!v{vmð Þ2z c=2ð Þ2
" #, where
c5 ci1 cOM5 ci(11C) is the total mechanical damping rate. For ablue laser detuning of D52vm, the optically induced damping isnegative (cOM524g2nc/k) and the photocurrent noise PSD is pro-portional to Sb{ vð Þ~ !nz1ð Þc
!v{vmð Þ2z c=2ð Þ2
" #. Typical mea-
sured noise power spectra under low-power laser drive (nc5 1.4,C5 0.27), for both red detuning and blue detuning, are shown inFig. 3a. Even at these small drive powers, the effects of back-actionon the measured spectra are evident, with the red-detuned drivebroadening the mechanical line and the blue-detuned drive narrowingthe line. The noise floor in Fig. 3a (shaded in grey) corresponds to thenoise generated by the EDFA used to pre-amplify the transmitteddrive-laser signal before photodetection, and is several orders of mag-nitude greater than the electronic noise of the photoreceiver and thereal-time spectrum analyser.Calibration of the EDFA gain, along with the photoreceiver and
real-time spectrum analyser photodetection gain, makes it possibleto convert the measured area under the photocurrent noise PSD intoa mechanical mode phonon occupancy. As described in detail inSupplementary Information, we perform these calibrations, alongwithmeasurements of low-drive-power (C= 1), radio-frequency spectra ofboth detunings (D56vm), to provide accurate, local thermometry ofthe optomechanical cavity. An example of this formof calibratedmodethermometry is shown in Fig. 3b, where we plot the opticallymeasured
Taper
Cryostat
Laser
EOM
RFSG
RSA
D1
D2
EDFA
WM
VOA
CIRC
FPC
FPC
LIA
Device
Figure 2 | Experimental set-up. A single, tunable, 1,550-nm diode laser isused as the cooling andmechanical transduction beam sent into the nanobeamoptomechanical resonator cavity held in a continuous-flow helium cryostat. Awavemetre (WM) is used to track and lock the laser frequency, and a variableoptical attenuator (VOA) is used to set the laser power. The transmitted signalis amplified by an erbium-doped fibre amplifier (EDFA) and detected on ahigh-speed photodetector (D2) connected to a real-time spectrum analyser(RSA), where the mechanical noise power spectrum is measured. A slowlymodulated probe signal used for optical spectroscopy and calibration isgenerated from the cooling laser beam using an amplitude electro-opticmodulator (EOM) driven by a microwave source (RFSG). The reflectedcomponent of this signal is separated from the input by an optical circulator(CIRC), sent to a photodetector (D1) and then demodulated using a lock-inamplifier (LIA). Paddle-wheel fibre polarization controllers (FPCs) are used toset the laser polarization at the input to the EOM and the input to theoptomechanical cavity. For more detail, see Supplementary Information.
5 μm
a
1 μm
1 μm
b
dc
10–1
10
e –10 0
Figure 1 | Optomechanical resonator with phononic shield. a, Scanningelectron microscope (SEM) image of the patterned silicon nanobeam and theexternal phononic bandgap shield. b, Enlarged SEM image of the central cavityregion of the nanobeam. c, Top: normalized electric field (colour scale) of thelocalized optical resonance of the nanobeam cavity, simulated using the finite-element method (FEM). Bottom: FEM simulation of the normalizeddisplacement field of the acoustic resonance (breathing mode), which iscoupled by radiation pressure to the co-localized optical resonance. The
displacement field is indicated by the exaggerated deformation of the structure,with the relative magnitude of the local displacement (strain) indicated by thecolour. d, SEM image of the interface between the nanobeam and the phononicbandgap shield. e, FEM simulation of the normalized squared displacementfield amplitude of the localized acoustic resonance at the nanobeam–shieldinterface, indicating the strong suppression of acoustic radiation provided bythe phononic bandgap shield. The colour scale represents log[x2/max(x2)],where x is the displacement field amplitude.
RESEARCH LETTER
9 0 | N A T U R E | V O L 4 7 8 | 6 O C T O B E R 2 0 1 1
Macmillan Publishers Limited. All rights reserved©2011
7.5 GHz; 10.6 MHz
Chan et al., Nature 2011195 THz; 3.68 GHz
New optomechanical system:
SiN membrane in 3d cavity
5
1 mm SiN membrane, Q>1,000,000
Al cavity, Q>100,000
22 mm
3D cavity and SiN membrane
6
sapphire substrateE-beam evaporated Al
PECVD SiN
Al coated membrane
Sample preparation
1 mm Membrane window zoom-in
7
Cavity frequency shift:
G =∂ω0
∂Cm
∂Cm
∂x
Coupling scheme
8
3D Cavity Antenna 1
Antenna 2
Membrane
Sapphire substrate
Al antenna
PECVD SiN
Al coated membrane
gap
z
x
Networkanalyzer
SpectrumanalyzerSignal
generator
-20 dB
-20 dB
-16 dB
12 mK
3 K
300 K
isolator
amplifier
700 mK-10 dB
28 mm28 mm
8 mm
SMA
directionalcoupler
xy
z
!
!p
d
z
y
(a)
(b)
(c)
(d)
1 mm
"
Measurement setup
9
High Q 3d microwave cavity
Reflection coefficient of membrane-embedded cavity
1.0
0.8
0.6
0.4
0.2
0.0
S11
5.066255.066205.066155.06610
f (GHz)
!e/2" = 21.6 kHz!/2" = 45.5 kHzQL= 1.1e5Q0= 2.1e5
cavity
directionalcoupler
10
10-20
10-19
10-18
10-17
P (W
)
123.1275123.1270
f (kHz)
!m/2" = 3.5 mHzQm= 3.5e7
Membrane resonatortime constant: 1.5 minutes!
Al coating: enabling electrical measurement while preserving ultrahigh Q.
Thermal motion of the membrane
11
Optomechanical coupling
Cooperativity vs. driving photon numbers
C =4g2
0
κΓmnd
1.0
0.5S11
-100 -50 0 50 100f-f0 (kHz)
nd=4.86e8
1.00.80.60.40.2
S11
nd=2.44e6(c)
1.00.80.60.40.2
1.24e8
1.00.80.60.40.2
200010000-1000-2000f-f0 (Hz)
4.86e8
46
103
2
46
104
2
46
105
2
C
2 4 6 8
107
2 4 6 8
108
2 4 6 8
nd
data fit, C0= 1.79e-4
(d)
g=2.36 kHz
1.0
0.5S11
-100 -50 0 50 100f-f0 (kHz)
nd=4.86e8
(b)
!0
!!d
pmeasure
!m
!0
!d measure
!m
(a)
Cooling
OMIT
coupling extracted with OMIT
10 1
10 2
10 3
10 4
10 5
( ytivitarepooCC
)
104
105
106
107
108
# of drive photons (Nd)
Cmax = 146 000
C=146000
12
Near-ground state cooling
of the mm-scale membrane
13
Sideband cooling
For T=13 mK, initial occupancy n0 = 2200
1.00.80.60.40.2
S11
nd=2.44e6(c)
1.00.80.60.40.2
1.24e8
1.00.80.60.40.2
200010000-1000-2000f-f0 (Hz)
4.86e8
46
103
2
46
104
2
46
105
2
C
2 4 6 8
107
2 4 6 8
108
2 4 6 8
nd
data fit, C0= 1.79e-4
(d)
g=2.36 kHz
1.0
0.5S11
-100 -50 0 50 100f-f0 (kHz)
nd=4.86e8
(b)
!0
!!d
pmeasure
!m
!0
!d measure
!m
(a)
Cooling
OMIT
Occupancy with cooling nm =n0
C + 1
Cmax = 105 � 103
Raman scattered
Thermal occupancy nm =1
e�ωm/kBT − 1
14
Measurement of cooling
Area under the curve is reduced.
1.00.80.60.40.2
S 11
nd=2.44e6(c)
1.00.80.60.40.2
1.24e8
1.00.80.60.40.2
200010000-1000-2000f-f0 (Hz)
4.86e8
46
103
2
46
104
2
46
105
2
C
2 4 6 8
107
2 4 6 8
108
2 4 6 8
nd
data fit, C0= 1.79e-4
(d)
g=2.36 kHz
1.0
0.5S 11
-100 -50 0 50 100f-f0 (kHz)
nd=4.86e8
(b)
!0
!!d
pmeasure
!m
!0
!d measure
!m
(a)
Cooling
OMIT
weak cooling strong cooling
10-11
10-10
S/P
in (H
z-1)
-100 -50 0 50 100f-f0 (Hz)
10-12
10-11
-100 0 100
15
Occupancy of the membrane
Minimum occupancy:
10 0
10 1
10 2
10 3
10 4ycnapucc
On m
10-1
101
103
105
C
nm = 5.2
Tm = 34 μK
nm = 5Tm = 34µK
16
Yuan et al., Nat. Commun. 2015
High mechanical Q
of SiN membranes at mK
17
Q factor and mechanical quantum state
‣ Preparation and measurement of mechanical quantum superposition state:
— Prerequisite: deep ground state cooling
Cooling ability
— Longer state lifetime
higher Q less energy loss
∝ Q
18
Overview: silicon nitride resonators
Nanostrings: 10Cornell, JAP 2006
Beams: 10Munich, Nature
2009 Trampolines:10UCSB/Leiden, OE 2011Sankey group, McGill;
Groeblacher group, Delft: 10
Membranes: 10 , 0.3 KYale, APL 2007
7
5
6
5
7
Lower temperature?
19
Measurement of
mechanical Q
20
Methods: spectral & ringdown
10-20
10-19
10-18
10-17
P (W
)
123.1275123.1270
f (kHz)
!m/2" = 3.5 mHzQm= 3.5e7∆f
1.0
0.8
0.6
0.4
0.2
0.0
Am
plitu
de (a
.u.)
3002001000
Time (s)
amp. ∝ e−γm2 t
21
Qspec =fm
∆fQring =
ωm
γm
Optomechanical ringdown
ω
ω
ωωπ
ω
ω
ω
ωω
ω
ω
ωωπ
ω
ω
ω
ωω
22
t=0
Results
23
Temperature dependence
127 million: record high
Newly revealed temperature dependence
242 kHz
24Yuan et al., APL 2015
More modes
Dev I
Dev II}
25
Dev I Dev II
size 1.5mm x 1.5mm x 50 nm 1mm x 1mm x 50nm
tensile stress 0.8 GPa 0.09 GPa
High Q vs. low Q modes
105
106
107Q
8006004002000T (mK)
mode (1,1) mode (1,2) mode (2,1)
26
Dev II
Conclusion
‣ Optomechanical system with 3d cavity and SiN membrane
‣ Large cooperativity that enables cooling close to ground state
‣ Temperature dependence of Q below 200 mK; Q exceeding 10 at 14 mK
‣ Potentials for hybrid devices
8
27
Members
Vibhor Singh
Yaroslav Blanter
Martijn Cohen
Shun Yanai
Gary Steele
This project is funded by FOM and NWO.28