Chip-scale high Q-factor glassblown

microspherical shells for magnetic sensing

Cite as: AIP Advances 8, 065214 (2018); https://doi.org/10.1063/1.5030460
Submitted: 21 March 2018 • Accepted: 04 June 2018 • Published Online: 14 June 2018

Eugene Freeman, Cheng-Yu Wang, Vedant Sumaria, et al.

ARTICLES YOU MAY BE INTERESTED IN

Invited Article: Scalable high-sensitivity optomechanical magnetometers on a chip

APL Photonics 3, 120806 (2018); https://doi.org/10.1063/1.5055029

Thermally tunable whispering-gallery mode cavities for magneto-optics

Applied Physics Letters 116, 161110 (2020); https://doi.org/10.1063/5.0006367

Fabrication of high- polydimethylsiloxane optical microspheres for thermal sensing

Applied Physics Letters 94, 231119 (2009); https://doi.org/10.1063/1.3152791

AIP Advances 8, 065214 (2018); https://doi.org/10.1063/1.5030460 8, 065214

© 2018 Author(s).
 AIP ADVANCES 8, 065214 (2018)

Chip-scale high Q-factor glassblown microspherical shells

for magnetic sensing
Eugene Freeman,1 Cheng-Yu Wang,2 Vedant Sumaria,3 Steven J. Schiff,4,5
Zhiwen Liu,2 and Srinivas Tadigadapa3,a
1 Honeywell International, Aerospace Advanced Technology, Plymouth, MN 55441, USA
2 Schoolof Electrical Engineering and Computer Science, Pennsylvania State University,
University Park, PA 16802, USA
3 Department of Electrical and Computer Engineering, Northeastern University, Boston,

MA 02115, USA
4 Department of Engineering Science and Mechanics, Center for Neural Engineering,

The Pennsylvania State University, University Park, PA 16802, USA

5 Departments of Neurosurgery and Physics, The Pennsylvania State University,

University Park, PA 16802, USA

(Received 21 March 2018; accepted 4 June 2018; published online 14 June 2018)

A whispering gallery mode resonator based magnetometer using chip-scale glass

microspherical shells is described. A neodynium micro-magnet is elastically coupled
and integrated on top of the microspherical shell structure that enables transduction
of the magnetic force experienced by the magnet in external magnetic fields into an
optical resonance frequency shift. High quality factor optical microspherical shell res-
onators with ultra-smooth surfaces have been successfully fabricated and integrated
with magnets to achieve Q-factors of greater than 1.1 × 107 and have shown a reso-
nance shift of 1.43 GHz/mT (or 4.0 pm/mT) at 760 nm wavelength. The main mode
of action is mechanical deformation of the microbubble with a minor contribution
from the photoelastic effect. An experimental limit of detection of 60 nT Hz−1/2
at 100 Hz is demonstrated. A theoretical thermorefractive limited detection limit of
52 pT Hz−1/2 at 100 Hz is calculated from the experimentally derived sensitivity.
The paper describes the mode of action, sensitivity and limit of detection is evalu-
ated for the chip-scale whispering gallery mode magnetometer. © 2018 Author(s).
All article content, except where otherwise noted, is licensed under a Creative
Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
https://doi.org/10.1063/1.5030460

I. INTRODUCTION
Room temperature, low cost magnetic sensing for bio-magnetic applications are being pursued
over a wide range of technologies including atomic magnetometers,1 nitrogen-vacancy in diamond
magnetometers,2,3 and magnetoelectric magnetometers.4,5
Recently some efforts have also been made to take advantage of optical resonance methods
of sensing to probe magnetic fields.6–10 One such strategy is to immerse a fiber based resonator
into a magnetic fluid, and then quantify the spectral response while applying an external magnetic
field. Magnetic fluids can have large changes in refractive index in response to a magnetic field.7
However, ferrofluids can have unacceptably high optical absorption and need to be diluted to balance
out refractive index sensitivity in response to magnetic fields and optical losses. The full-width at
half-maximum of resonances in most of the fiber-based interferometers, reported thus far, is relatively
large, thus hindering a high sensitivity of the magnetic field sensing. Zu et al. demonstrated a sensitivity
of 167 pm/mT with a resolution of 60 µT, at low frequencies (DC-100 Hz), with a magneto-optical

a
Electronic mail: srinivas@northeastern.edu

2158-3226/2018/8(6)/065214/8 8, 065214-1 © Author(s) 2018

065214-2 Freeman et al. AIP Advances 8, 065214 (2018)

fiber.11 Recently, Amili et al., demonstrated a silicon micro-ring with a magnetic fluid cladding with
a sensitivity of 16.8 pm/mT at DC magnetic fields; however the quality factor (Q-factor) of the device
suffered significantly after application of the magnetic fluid.8 Luo et al. demonstrated an impressive
1918 pm/mT using optical fiber and magnet fluid, but did not report the frequency regime of the
measurement nor a limit of detection.9
Whispering gallery mode (WGM) optical resonators have demonstrated exceptional sensitiv-
ity to mechanical deformation, with shot noise limited displacement sensitivities of 10−19 m Hz−1/2
experimentally measured at room temperature.12 Recently, Forstner et al.10 used a toroidal WGM res-
onator with a Terfenol-D bead to experimentally demonstrate an optomechanical method of magnetic
sensing with 400 nT Hz−1/2 limit of detection at mechanical resonance of approximately 10 MHz,
which is too fast to detect biomagnetic signals. More recent work by Forstner et al.,13 improved the
optomechanical coupling to the magnetostrictive element, resulting in low frequency detection down
to 150 nT Hz−1/2 at 2 Hz using frequency and 150 pT Hz−1/2 at mechanical resonance in the MHz
range.
In this work we leverage the emerging field of optomechanical sensing, for low frequency mag-
netometer applications. We use recently developed chip-scale glass microspherical shells as the high
quality factor WGM resonator platform14 and integrate with a micro-magnet to couple magnetic force
into mechanical deformation of the microbubble. The integration method described in this work does
not add losses and maintains the ultra high > 107 Q-factor of the resonator. The sensitivity and limit
of detection is experimentally evaluated for the device. The magnetic force is analytically estimated
and the mechanical deformation is modeled using COMSOL to correlate the radial deformation of the
microbubble and the mechanically induced stress-based photoelastic component. Finally, we estimate
the thermorefractive limited magnetic limit of detection for a fused silica microbubble.

II. FABRICATION AND TESTING

The chip-scale glassblowing process was first pioneered by Eklund and Shkel15 for mechanical
resonance applications. The process was modified and adapted for optical resonance as published
in our earlier work,14 which provides more detail on the fabrication process and WGM resonance
achievable using these chip-scale microbubbles.
First, 380 µm diameter circles are patterned using positive photoresist on a silicon substrate.
A deep reactive ion etch process is used to etch 250 µm of silicon. The photoresist is stripped away
with RemoverPG at 70 ◦ C for 15 minutes, followed by a piranha clean (H2 SO4 :(30%)H2 O2 3:1) to
remove any remaining photoresist. Next, the etched silicon is anodically bonded to a 100 µm thick
Pyrex® 7740 wafer at 133 kPa (∼1000 Torr), trapping N2 inside the cavity. The wafer is diced and
a 10 minute 49% HF etch is done to thin the borosilicate glass to 37.5 µm thickness. This etching
step allows one to control the final shell thickness. The etched die is placed in a 13.3 kPa (∼100
Torr) vacuum and heated on a silicon nitride ceramic heater to 775 ◦ C for 30 seconds and allowed
to rapidly cool to approximately 550 ◦ C, followed by a gradual reduction in temperature to prevent
thermally shocking the thin glass microbubbles. During the glassblowing process, the borosilicate
softens and begins to expand until the pressure inside the microbubble equals the pressure outside the
microbubble. The final shell thickness at the equator is approximately 1.8 µm (measured by scanning
electron microscopy) after glassblowing.
By bonding at 1000 Torr and glassblowing at 100 Torr we can create a larger pressure differ-
ential that would be possible if both steps were done at atmospheric pressure. This enables larger
near spherical microbubbles where the equatorial plane is located above the silicon substrate plane
(critical for WGM light confinement with minimal coupling loss to the substrate) without using a
complex silicon wafer bonding process, to increase the gas cavity volume, as was demonstrated by
others.15 Finally, Devcon 5 minute epoxy is used to carefully attach a N48 grade 1.5 mm diameter and
1.5 mm long neodymium magnet to the top of the microbubble, perpendicular to the sample plane,
as seen in Fig. 1(a). It is critical to keep the epoxy from contaminating the equatorial plane of the
microbubble, which is crucial for the optical resonance. Earlier proof of concept experimental and
simulation work showed that epoxying the micromagnet parallel to the sample plane resulted in poor
sensitivities. COMSOL simulations shows that the in-plane magnetic forces only caused a smearing
065214-3 Freeman et al. AIP Advances 8, 065214 (2018)

FIG. 1. (a) Microscope image of the N48 micromagnet epoxied to the top of the microbubble. (b) Test setup and
electromagnetic coil for applying a controlled magnetic field.

of the microsphere, yielding a negligible change in the radius of the microsphere and subsequently
a small change in optical path length.16 The device is therefore a vector magnetic field sensor which
is sensitive to the out-of-plane magnetic field component.
The experimental setup used to apply the magnetic field and probe the optical resonance is
illustrated in Fig. 1(b). A 25 mm diameter, 17 mm long, 23 turn coil is wound and the device is
placed 8 mm from the edge of the coil. Light from a 760 nm tunable laser (Thorlabs, TLK-L780m)
is evanescently coupled to the microbubble through a tapered fiber. A set of waveplates are used
to control the polarization to optimize the coupling efficiency. The transmission is measured using
a photodiode (Thorlabs, DET36A). The magnetic field produced by the coil is calibrated using a
commercial magnetometer (Lakeshore 475, HMMA-2504-VR). To test the magnetic sensitivity, the
magnetic field is swept from 0 µT to +630 µT and then back to 0 µT, then to -630 µT and again back
to 0 µT. The shift in resonance frequency is monitored by an oscilloscope while the laser frequency
is swept. The sample is secured to the stage to minimize motion perturbation and potentially de-
coupling from the fiber. To test the limit of detection and noise characteristics of the setup, the laser
is held to the steepest portion of a WGM resonance peak and a dynamic spectrum analyzer (Stanford
Research SR760) is used observe the signal to noise floor while applying a known signal.

III. RESULTS AND DISCUSSION

A. High Q-factor WGM resonance
The glassblowing fabrication method used in this work relies on the reflow of the borosilicate
glass enabling an extremely smooth surface. This process is similar in principle to other glass reflow
methods that have been used to demonstrate near ideal Q-factors of 8 × 109 in reflown fused silica
microspheres.17
Atomic force microscopy (AFM) is used to measure the roughness on the top of the microbubble.
Fig. 2(a) shows the topography of a 5 µm x 5 µm section near the top of the microbubble. The RMS
065214-4 Freeman et al. AIP Advances 8, 065214 (2018)

FIG. 2. (a) 5 µm x 5 µm section on the top of the microbubble measured by AFM with 0.243 nm RMS roughness and an
autocorrelation length of 108 nm. (b) WGM resonance of a glassblown microbubble with 1.1 × 107 Q-factor. The red X marks
the approximate location of the steepest portion of the resonance, which is used for testing the limit of detection.

roughness is measured to be 0.243 nm and the correlation length is 108 nm. This is very close to the
0.23 nm roughness reported for glassblown fused silica microbubbles18 used for high mechanical
Q-factor gyroscopes and 0.33 nm RMS roughness reported for ultra smooth glass surfaces.19
The Q-factor (Qtot ) of an optical WGM resonator can be summarized by Eq. 1
−1
Qtot = Qrad
−1 −1
+ Qbulk −1
+ Qs.s −1
+ Qcont (1)
where Qrad is due to the radiative (curvature) losses, Qbulk represents material attenuation losses,
Qs.s is due to light scattering from material roughness, and Qcont represents losses due to contami-
nants producing surface scattering.20 For all calculations the resonant wavelength is approximated as
760 nm, the effective index of refraction is 1.467, and the diameter of the microbubble is measured to
be 880 µm. Radiative loss exponentially decreases with increasing resonant optical cavity diameter
and can be usually ignored21 when D/λ > 15. In this work the D/λ is 1158 and can be safely ignored.
The surface scattering can be calculated17 using Eq. 2
λ2D
Qs.s = (2)
2π 2 σ 2 B
where D is the diameter of the microbubble, λ is the resonance wavelength, σ is the RMS roughness
and B is the correlation length. Using the AFM data a Qs.s of 4.04 × 109 is calculated.
Some contaminants can be modeled, for example adsorption of water on the microbubble surface
has also been shown to play a role in scattering of high Q-factor WGM resonators. The Q-factor due
to adsorbed water scattering, Q4 , can be calculated22 by Eq. 3
πD
r
Q4 = (δ β4 )−1 (3)
8n3 λ
where δ is the thickness of the water layer and β4 is the absorption coefficient of the water. δ can be
assumed to be 0.2 nm and the absorption coefficient for water at 760 nm is calculated to be 2.52 m−1
based off extinction coefficient measurements.23 Q4 is calculated to be 2.38 × 1010 .
The material losses due to the attenuation of the laser is calculated17 using Eq. 4,
2πnr
Qbulk = (4)
λα
where nr is the effective index of refraction, and α is the attenuation coefficient. Specifications
for borosilicate optical fibers24 report the attenuation as 520 dB/km (approximately 0.12 m−1 ). The
Qbulk is calculated to be 1.0 × 108 and is the main theoretical limiter of the Q-factor in this work. Qtot is
therefore 9.8 × 107 . Other non-idealities such as dust and dirt on the microbubbles can further reduce
the Q-factor. The device fabricated for this experiment has an exceptional Q-factor of 1.1 × 107 , with
a full-width at half-maximum of 35 MHz as seen in Fig. 2(b). Previous studies14 yielded borosilicate
065214-5 Freeman et al. AIP Advances 8, 065214 (2018)

FIG. 3. (a) Sensitivity of the WGM is evaluated by tracking the frequency shift. (b) Frequency shift is linear with a sensitivity
of 1.43 GHz/mT and shows little hysteresis.

chip-scale WGM devices with similar Q-factors up to 5.2 × 107 , close to the theoretical absorption
limit of 9.8 × 108 calculated in this section for borosilicate WGM resonators.

B. Magnetic sensitivity and mode of operation

Fig. 3(a) shows the magnetometer sensitivity measured by applying a calibrated magnetic field
through the electromagnetic coil and the shift in WGM resonance frequency is tracked. The magnet-
ically induced response is plotted in Fig. 3(b), where a linear fit is used to determine a sensitivity of
1.43 GHz/mT. The device shows little hysteresis and the frequency shift switches from a red-shift
(negative ∆f ) to a blue-shift (positive ∆f ) when the magnetic field direction goes from negative to
positive. This confirms that the phenomenon is due to the magnetic forces and not heat from the coil,
which would have caused only a red shift. The observed hysteresis is within the drift of the laser
system that has been normally observed over the testing time.
The change in the optical path length at the equatorial plane due to the applied magnetic field (B)
induced mechanical deformation results in a shift in the resonance frequency given by Eq. 5 where f
is the resonance frequency, nr is the effective refractive index and r is the optical cavity radius, C is
the photoelastic constant, and σ is the stress.
∆f ∆nr ∆r C σ ∆r
− = + = + (5)
f ∆B nr ∆B r∆B nr ∆B r∆B
The mechanical force induced by the magnetic field can be analytically estimated using a for-
mula developed by Janssen et al. for orthogonal cuboid magnets.25 The coil in this setup is estimated
as a permanent magnet with a pseudo magnetic field strength of Bcoil = Ic Nlcr µ0 , where I c is the cur-
rent in the coil, lc is the length of the coil, N r is the number of turns, µ0 is the permittivity of
free space. The permanent cylindrical magnet is modeled as a cuboid with a matching volume and
magnetization of the neodymium N48 micro-magnet. The resulting force16 is 226 µN/mT, mostly
in the z-axis (orthogonal to the sample plane). Using these forces as an input to a COMSOL model
the mechanical deformation and radial stresses are calculated. By comparing the original and the
deformed structure in COMSOL the ∆R is simulated to be 0.324 nm for a 1 mT magnetic field. The
radial stress is averaged over the equator and determined to be 1.70 × 10−2 MPa for a 1 mT magnetic
field.
Using Eq. 5, with a photoelastic constant 26 of 4 × 10−12 Pa−1 the resonance frequency shift can
be calculated from the ∆R and σ determined from the finite element model. The modeled sensitivity
is 0.32 GHz/mT, with 0.29 GHz/mT contribution coming from the mechanical deformation and
0.03 GHz/mT from the photoelastic effect. The modeled sensitivity is smaller than the experimentally
observed sensitivity, possible due to the fact that the COMSOL model uses a uniform shell thickness,
where as in reality the shell thickness thins out at the top of the microbubble. Therefore the finite
element simulation would under estimate the true deformation induced in the microbubble.
065214-6 Freeman et al. AIP Advances 8, 065214 (2018)

C. Limit of detection
To test for the limit of detection the laser is detuned by approximately -11.6 GHz to operate the
device at the steepest slope region of the WGM resonance as shown by the “X” label in Fig. 2(b).
While applying a known 1.6 µT signal at 25 Hz, the response is monitored on the digital spectrum
analyzer and a limit of detection of 60 nT Hz−1/2 is measured at 100 Hz as seen in Fig. 4.
The dominant noise is evaluated by measuring the noise of the laser without any coupling to
the WGM. The green line in Fig. 4 shows that the laser system or the photodetector is the dominant
source of noise, not the WGM magnetometer. Next, the photodetector noise is evaluated at maximum
gain, typical gain, and minimum gain. It becomes evident that the photodetector has a lower noise
floor than the laser. As expected, there is more dark current in the photodetector as the gain increases,
producing more noise. A photodetector operating off a DC battery could offer even lower noise,
pushing the noise floor lower. Operating at minimum gain on the photodiode and assuming a stable
laser and environmental (acoustic and magnetic) noise compensation by common mode rejection
(discussed in the next paragraph) on the WGM the limit of detection of this setup can be as low as
150 pT Hz−1/2 .
Finally we perform a theoretical prediction of the fundamental magnetic limit of detection for a
proposed fused silica based microbubble, with low material loss, with an integrated micro-magnet.
Several techniques are available to measure optical frequency shifts with high sensitivity which can
mitigate the laser noise. For example the balanced homodyne spectroscopy27 or by propagating an
orthogonal polarization and comparing to the signal beam.12 At room temperature thermorefractive
noise will likely be the dominating noise source.28 Glassblown fused silica microbubbles have been
previously demonstrated by other authors for acoustic resonance phenomenon.29 It is likely that very
high Q-factor fused silica WGM optical resonators are achievable using this process, owing to the
lower optical absorption coefficient for fused silica compared to borofloat. We therefore perform
the thermorefractive analysis assuming the parameters of a fused silica microbubble of identical
dimensions to the borosilicate microbubble fabricated in this work. Furthermore, since the Young’s
modulus of fused silica is only slightly higher than borofloat (72 GPa versus 64 GPa) we assume
the a similar magnetic sensitivity of 1.43 q GHz/mT can be achieved. The thermorefractive noise,28 in
  1/2
units of Hz−1/2 , is given by Sδf /f = αn2 ST (Ω), where α n is the thermorefractive coefficient of
the resonator host material, S T (Ω) is given by Eq. 6.
! 3/2 !2 −1
kB T 2 r 2 * r 2 ρCΩ 1 r 2 ρCΩ
ST (Ω) = 1+ + + (6)
Veff 12κ , 35/2 κ 6 8l1/3 κ -

FIG. 4. (a) Magnetic equivalent noise from the WGM magnetometer (black line), laser (green line), photodiode at max gain
(red line), typical gain (dark blue line) and minimal gain (light blue line). LOD of the device is 60 nT Hz−1/2 at 100 Hz (dashed
black line), which is limited by laser noise. In an ideal situation the ultimate limit of this setup is set by the photodetector noise
at 150 pT Hz−1/2 . (b) The thermorefractive magnetic equivalent limit of detection for 300 K and 77 K are 52 pT Hz−1/2 and
14 pT Hz−1/2 respectively.
065214-7 Freeman et al. AIP Advances 8, 065214 (2018)

where k B is Boltzmann’s constant, T is the temperature, V eff is the effective mode volume, κ is the
thermal conductivity, ρ is the density, r is the resonator radius, l is the orbital mode number, and
Ω is the noise frequency. The derivation30 and constants for fused silica28 can be found elsewhere.
Using the experimentally derived sensitivity of 1.43 GHz/mT and a approximate resonant frequency
of 394 THz (∼ 760 nm) the magnetic limit of detection is plotted in Fig. 4(b) for 300 K and 77 K.
The thermorefractive magnetic equivalent limit of detection for 300 K and 77 K are 52 pT Hz−1/2
and 14 pT Hz−1/2 respectively.

IV. CONCLUSION
In conclusion, a 1.1 × 107 Q-factor chip-scale whispering gallery mode magnetometer was
experimentally demonstrated with a sensitivity of 1.43 GHz/mT. The phenomenon is determined to
be mostly due to mechanical deformation of the radius of the borosilicate microbubble with a small
contribution from the photoelastic effect. An experimental limit of detection of 60 nT Hz−1/2 was
measured for our current setup and determined to be dominated by laser noise. A thermorefractive
analysis for a fused silica microbubble predicts an ultimate limit of detection of 52 pT Hz−1/2 at
100 Hz is possible at room temperature.

ACKNOWLEDGMENTS
We acknowledge financial support from the National Institutes of Health (NIH) under the Grant
No. 1R21EY026438-01 and National Science Foundation (NSF) Grant No. 1305663.
1 T. Sander, J. Preusser, R. Mhaskar, J. Kitching, L. Trahms, and S. Knappe, “Magnetoencephalography with a chip-scale
atomic magnetometer,” Biomedical Optics Express 3, 981–990 (2012).
2 D. Le Sage, K. Arai, D. Glenn, S. DeVience, L. Pham, L. Rahn-Lee, M. Lukin, A. Yacoby, A. Komeili, and R. Walsworth,

“Optical magnetic imaging of living cells,” Nature 496, 486 (2013).

3 L. T. Hall, C. D. Hill, J. H. Cole, B. Städler, F. Caruso, P. Mulvaney, J. Wrachtrup, and L. C. Hollenberg, “Monitoring

ion-channel function in real time through quantum decoherence,” Proceedings of the National Academy of Sciences 107,
18777–18782 (2010).
4 H. Xi, X. Qian, M.-C. Lu, L. Mei, S. Rupprecht, Q. X. Yang, and Q. Zhang, “A room temperature ultrasensitive

magnetoelectric susceptometer for quantitative tissue iron detection,” Scientific Reports 6 (2016).
5 J. Reermann, P. Durdaut, S. Salzer, T. Demming, A. Piorra, E. Quandt, N. Frey, M. Höft, and G. Schmidt, “Evaluation of

magnetoelectric sensor systems for cardiological applications,” Measurement (2017).

6 C. Yu, J. Janousek, E. Sheridan, D. L. McAuslan, H. Rubinsztein-Dunlop, P. K. Lam, Y. Zhang, and W. P. Bowen,

“Optomechanical magnetometry with a macroscopic resonator,” Physical Review Applied 5, 044007 (2016).
7 S.-Y. Yang, Y. Chen, H.-E. Horng, C.-Y. Hong, W. Tse, and H.-C. Yang, “Magnetically-modulated refractive index of

magnetic fluid films,” Applied Physics Letters 81, 4931–4933 (2002).

8 A. El Amili, M. Souza, F. Vallini, N. Frateschi, and Y. Fainman, “Magnetically controllable silicon microring with ferrofluid

cladding,” Optics Letters 41, 5576–5579 (2016).

9 L. Luo, S. Pu, J. Tang, X. Zeng, and M. Lahoubi, “Highly sensitive magnetic field sensor based on microfiber coupler with

magnetic fluid,” Applied Physics Letters 106, 193507 (2015).

10 S. Forstner, S. Prams, J. Knittel, E. D. van Ooijen, J. D. Swaim, G. I. Harris, A. Szorkovszky, W. P. Bowen, and H. Rubinsztein-

Dunlop, “Cavity optomechanical magnetometer,” Physical Review Letters 108, 120801 (2012).
11 P. Zu, C. C. Chan, W. S. Lew, Y. Jin, Y. Zhang, H. F. Liew, L. H. Chen, W. C. Wong, and X. Dong, “Magneto-optical fiber

sensor based on magnetic fluid,” Optics Letters 37, 398–400 (2012).

12 A. Schliesser, G. Anetsberger, R. Rivière, O. Arcizet, and T. J. Kippenberg, “High-sensitivity monitoring of micromechanical

vibration using optical whispering gallery mode resonators,” New Journal of Physics 10, 095015 (2008).
13 S. Forstner, E. Sheridan, J. Knittel, C. L. Humphreys, G. A. Brawley, H. Rubinsztein-Dunlop, and W. P. Bowen,

“Ultrasensitive optomechanical magnetometry,” Advanced Materials 26, 6348–6353 (2014).

14 C. Zhang, A. Cocking, E. Freeman, Z. Liu, and S. Tadigadapa, “On-chip glass microspherical shell whispering gallery mode

resonators,” Scientific Reports 7, 14965 (2017).

15 E. J. Eklund and A. M. Shkel, “Glass blowing on a wafer level,” Journal of Microelectromechanical Systems 16, 232–239

(2007).
16 E. Freeman, C.-Y. Wang, V. Sumaria, C. Zhang, A. Cocking, Z. Liu, and S. Tadigadapa, “Novel chip-scale high-q whispering

gallery mode resonator as a magnetometer,” in 2017 IEEE SENSORS (2017) pp. 1–3.
17 M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, “Ultimate q of optical microsphere resonators,” Optics Letters

21, 453–455 (1996).

18 D. Senkal, M. Ahamed, A. Trusov, and A. Shkel, “High temperature micro-glassblowing process demonstrated on fused

quartz and ULE TSG,” Sensors and Actuators A: Physical 201, 525–531 (2013).
19 K. Guenther and P. Wierer, “Surface roughness assessment of ultrasmooth laser mirrors and substrates,” in 1983 International

Techincal Conference/Europe (International Society for Optics and Photonics, 1983) pp. 266–279.
065214-8 Freeman et al. AIP Advances 8, 065214 (2018)

20 J. Buck and H. Kimble, “Optimal sizes of dielectric microspheres for cavity QED with strong coupling,” Physical Review
A 67, 033806 (2003).
21 A. Chiasera, Y. Dumeige, P. Feron, M. Ferrari, Y. Jestin, G. Nunzi Conti, S. Pelli, S. Soria, and G. C. Righini, “Spherical

whispering-gallery-mode microresonators,” Laser & Photonics Reviews 4, 457–482 (2010).

22 D. Vernooy, V. S. Ilchenko, H. Mabuchi, E. Streed, and H. Kimble, “High-q measurements of fused-silica microspheres in

the near infrared,” Optics Letters 23, 247–249 (1998).

23 G. M. Hale and M. R. Querry, “Optical constants of water in the 200-nm to 200-µm wavelength region,” Applied Optics

12, 555–563 (1973).

24 I. Fiberguide Industries, “Fiber Optic Assemblies,” (2007), form No: REF 723 DS004.
25 J. L. G. Janssen, J. J. H. Paulides, E. Lomonova, F. Bölöni, A. Tounzi, and F. Piriou, “Analytical calculation of interaction

force between orthogonally magnetized permanent magnets,” Sensor Letters 7, 442–445 (2009).
26 BOROFLOAT 33 - Optical Properties, SCHOTT.
27 H. P. Yuen and V. W. Chan, “Noise in homodyne and heterodyne detection,” Optics Letters 8, 177–179 (1983).
28 A. B. Matsko, A. A. Savchenkov, N. Yu, and L. Maleki, “Whispering-gallery-mode resonators as frequency references. I.

Fundamental limitations,” JOSA B 24, 1324–1335 (2007).

29 D. Senkal, M. J. Ahamed, M. H. A. Ardakani, S. Askari, and A. M. Shkel, “Demonstration of 1 million q-factor on

microglassblown wineglass resonators with out-of-plane electrostatic transduction,” Journal of Microelectromechanical

Systems 24, 29–37 (2015).
30 M. L. Gorodetsky and I. S. Grudinin, “Fundamental thermal fluctuations in microspheres,” JOSA B 21, 697–705 (2004).