PHYSICAL REVIEW A 71, 013817 共2005兲

Ultrahigh-Q toroidal microresonators for cavity quantum electrodynamics

S. M. Spillane, T. J. Kippenberg, and K. J. Vahala
Thomas J. Watson Laboratory of Applied Physics, California Institute of Technology, Pasadena, California 91125, USA

K. W. Goh, E. Wilcut, and H. J. Kimble

Norman Bridge Laboratory of Physics, California Institute of Technology, Pasadena, California 91125, USA
共Received 27 August 2004; published 26 January 2005兲

We investigate the suitability of toroidal microcavities for strong-coupling cavity quantum electrodynamics
共QED兲. Numerical modeling of the optical modes demonstrate a significant reduction of the modal volume
with respect to the whispering gallery modes of dielectric spheres, while retaining the high-quality factors
representative of spherical cavities. The extra degree of freedom of toroid microcavities can be used to achieve
improved cavity QED characteristics. Numerical results for atom-cavity coupling strength g, critical atom
number N0, and critical photon number n0 for cesium are calculated and shown to exceed values currently
possible using Fabry-Perot cavities. Modeling predicts coupling rates g / 2␲ exceeding 700 MHz and critical
atom numbers approaching 10−7 in optimized structures. Furthermore, preliminary experimental measurements
of toroidal cavities at a wavelength of 852 nm indicate that quality factors in excess of 108 can be obtained in
a 50-␮m principal diameter cavity, which would result in strong-coupling values of (g / 共2␲兲 , n0 , N0)
= 共86 MHz, 4.6⫻ 10−4 , 1.0⫻ 10−3兲.

DOI: 10.1103/PhysRevA.71.013817 PACS number共s兲: 42.50.Pq, 32.80.⫺t, 42.50.Ct, 42.60.Da

I. INTRODUCTION record coupling efficiencies to an optical fiber 关17兴 共the me-

The use of an optical microcavity can greatly enhance the dium of choice for low-loss transport of classical and non-
interaction of an atom with the electromagnetic field such classical states 关18兴兲 is fundamentally important in CQED
that even a single atom or photon can significantly change and bears promise for realizing quantum networks.
the dynamical evolution of the atom-cavity system 关1兴. Recently, a new type of whispering-gallery-mode optical
Achieving the regime of “strong coupling” 关2,3兴 is critically microcavity was demonstrated, which not only retains the
dependent on the characteristics of the optical cavity and high-quality factors of spherical cavities, but also has signifi-
generally requires the optical modes to be confined in a small cant advantages in fabrication reproducibility, control, and
mode volume for extended periods of time 共or equivalently mode structure. These cavities consist of a toroidally shaped
high Q factor兲. silica cavity supported by a silicon pillar on a microelec-
Recent experimental realizations of strong coupling have tronic chip 关19兴. The toroidal cavity shape allows an extra
employed high-finesse Fabry-Perot 共FP兲 optical microcavi- level of geometric control over that provided by a spherical
ties 关4–9兴. Our experiments at Caltech include the realization cavity and thus begs the question as to how these structures
of an “atom-cavity microscope” with a single atom bound in compare with silica microspheres and other microcavity de-
orbit by single photons 关4兴 and the development of a laser signs for strong-coupling cavity QED. In this paper we nu-
that operates with “one and the same” atom 关10兴. Fabry- merically investigate the suitability of toroidal microcavities
Perot cavities, while possessing ultrahigh-quality factors and for strong-coupling cavity QED experiments, and for pur-
finesse, are difficult to manufacture and control, requiring poses of comparison, we focus on the interaction with atomic
sophisticated dielectric mirror coatings as well as accurate cesium 关4,20兴. We show that toroid microcavities can
feedback for resonant wavelength control. Due in part to achieve ultrahigh-quality factors exceeding 108 while simul-
these reasons, there has been increased interest in other mi- taneously obtaining very large coupling rates between the
crocavity systems which not only can address some or all of cavity and a cesium atom. It is found that these cavities not
the limitations of Fabry-Perot cavities, but which in principle only surpass the projected limits of FP technology 关20兴, but
can have improved optical properties. also either exceed or compare favorably to other cavity de-
Based upon the pioneering work of Braginsky and col- signs such as photonic band-gap devices 关21,22兴. Last, we
leagues 关11兴, whispering-gallery-mode cavities have also present preliminary experimental measurements of quality
been investigated for cavity QED 共CQED兲 experiments for factors for toroidal cavities at a wavelength of 852 nm, suit-
many years 关12兴. Experimental studies have demonstrated Q able for strong-coupling CQED with atomic cesium. These
factors approaching 1010 in a silica microsphere whispering results show that currently attainable Q values are already
gallery cavity 关13,14兴, with values exceeding 108 readily quite promising.
achievable over a broad range of cavity diameters and wave-
lengths. The combination of their very low cavity losses, II. STRONG COUPLING IN AN ATOM-CAVITY SYSTEM
small mode volumes, and their relative ease of fabrication
makes them promising candidates for experiments in CQED The coupling rate g between an atomic system and an
关15,16兴. Furthermore, the ability to couple these cavities with electromagnetic field is related to the single-photon Rabi fre-

1050-2947/2005/71共1兲/013817共10兲/$23.00 013817-1 ©2005 The American Physical Society

SPILLANE et al. PHYSICAL REVIEW A 71, 013817 共2005兲

quency ⍀ = 2g and can be expressed in terms of the atomic

and cavity parameters by 关1兴

ជ 共r兲/Eជ 兩冑V /V ,
g共r兲 = ␥⬜兩E 共1兲
max a m

Va = 3c␭2/共4␲␥⬜兲, 共2兲

where ␥⬜ is the transverse atomic dipole transition rate,

ជ 共r兲 / Eជ 兩 denotes the normalized electric field strength at
兩E max
the atom’s location r, Va is a characteristic atomic interaction
volume 共which depends on the atomic dipole transition rate,
the transition wavelength ␭, and the speed of light c兲, and Vm
is the cavity-electromagnetic-mode volume. Assuming the
atom interacts with the electromagnetic field for a time T,
strong atom-field coupling occurs if the rate of coupling ex-
ceeds all dissipative mechanisms—i.e., g Ⰷ 共␬ , ␥⬜ , T−1兲. In FIG. 1. Scanning electron micrograph of a toroidal microcavity.
this expression ␬ denotes the cavity field decay rate, given in The principal and minor diameters are denoted by D and d,
terms of the cavity quality factor Q by ␬ ⬅ ␲c / 共␭Q兲. The respectively.
degree of strong coupling can also be related to a set of
normalized parameters 关1兴, III. TOROIDAL MICRORESONATORS

n0 ⬅ ␥⬜
2
/共2g2兲, 共3兲 Toroidal microresonators are chip-based microcavities
that possess ultrahigh-Q 共⬎108兲 whispering-gallery type
modes 关19兴. The realization of ultrahigh-Q chip-based reso-
N0 ⬅ 2␥⬜␬/共g2兲, 共4兲 nators allows improvements in fabrication and control, while
additionally allowing integration with complementary opti-
where n0 is the critical photon number, which is the number cal, mechanical, or electrical components. In brief, these
of photons required to saturate an intracavity atom, and N0 is resonators are fabricated by standard lithographic and etch-
the critical atom number, which gives the number of atoms ing techniques, followed by a laser-reflow process, as out-
required to have an appreciable effect on the cavity transmis- lined in Ref. 关19兴. The combination of thermal isolation of
sion. Note that 共N0 , n0兲 Ⰶ 1 provides a necessary but not suf- the initial preform periphery and thermal heat sinking of the
ficient condition for strong coupling. preform interior through the strong heat conduction of the
Examining these parameters, we see that only the critical silicon support pillar results in a preferential melting of the
atom number N0 ⬀ Vm / Q is dependent on the cavity-loss rate preform along the disk periphery under CO2 laser irradiation.
共or equivalently Q factor兲. It is the possibility of realizing Surface tension then induces a collapse of the silica disk
extremely low critical atom numbers with ultrahigh-Q micro- preform, resulting in a toroidally shaped boundary, with the
cavities that has fostered the investigation of silica micro- final geometry controlled by a combination of irradiation
spheres for strong-coupling CQED experiments. However, flux and exposure time. Importantly, as the optical mode re-
the geometry of a spherical dielectric dictates a definite rela- sides in the extremely uniform and smooth 共reflowed兲 pe-
tionship between cavity-mode volume Vm and the associated riphery of the structure, the quality factors of optical
quality factor Q and, hence, of the value of the coupling whispering-gallery modes can achieve ultrahigh-Q perfor-
parameter g ⬀ Vm −1/2
while still maintaining ultrahigh-quality mance, exceeding 108. Figure 1 shows a scanning electron
factors 关23兴. This is a result of the fact that to achieve large micrograph of the side view of a typical toroidal microcavity.
atom-cavity coupling rates 共comparable to or exceeding Quality factors as high as 4 ⫻ 108 at a wavelength of
those of FP cavities兲 the cavity diameter must be made small 1550 nm 共corresponding to a photon lifetime of ⬃300 ns兲
关8-␮m-diameter sphere gives g / 共2␲兲 ⬇ 740 MHz兴 in order to have been measured 关24兴.
both lower the modal volume and to increase the electric
field strength at the atomic position 共assumed to be the cavity IV. MICROTOROID NUMERICAL MODELING
surface at the point of maximum electric field strength兲.
However, at the optimum radius for atom-coupling strength, In order to investigate the properties of microtoroids for
the tunneling loss of the microcavity results in a low achiev- CQED, this paper will focus on the D2 transition of cesium
able Q factor 共Q ⬇ 4 ⫻ 104兲, thereby raising the critical atom which occurs at a wavelength of 852.359 nm 共␥⬜ / 2␲
number. While the relatively large mode volumes of silica ⬇ 2.6 MHz兲 关20兴, with scaling to other systems accom-
microsphere cavities preclude them from competing with ul- plished in the fashion of Ref. 关23兴. Fundamentally, the cou-
trasmall mode volume cavities 共such as photonic band-gap pling between an atom and a cavity field can be specified by
cavities兲 on the basis of coupling strength alone, there is the four parameters: the atomic transition moment, the cavity
possibility to access simultaneously both ultrahigh-Q and field strength at the atom’s location, the cavity mode volume
small mode volume, using toroidal microresonators. Vm, and the cavity quality factor Q. Since the optical modes

013817-2
ULTRAHIGH-Q TOROIDAL MICRORESONATORS FOR… PHYSICAL REVIEW A 71, 013817 共2005兲

FIG. 2. Electric field magnitude for the whispering-gallery FIG. 3. Electric field profiles for a toroidal cavity with a princi-
modes of a spherical 共top row兲 cavity of diameter 18 ␮m and a pal diameter D = 20 ␮m and minor diameters d = 20, 12, 6, 3, 1.5,
toroidal cavity 共bottom row兲 with principal diameter of 18 ␮m and and 0.75 ␮m. The calculations correspond to a TM-polarized mode
a minor diameter of 1 ␮m. The left 共right兲 column shows the TE- near 850 nm. The optical mode behaves as a whispering-gallery-
共TM-兲 polarized mode near 850 nm. The arrows indicate the loca- type mode until the minor diameter is below approximately 1.5 ␮m,
tion of the maximum external electric field strength, where we as- at which point the mode approaches that of a step-index optical
sume the atom is located. The dotted lines in the two-dimensional fiber 关28兴.
field distribution indicate the cross section where the electric field is
displayed.
proximate expressions for the optical behavior of these struc-
are confined to the interior dielectric in whispering-gallery- tures for both the low transverse compression 共spherelike兲
type resonators, the atom can interact only with the evanes- and high transverse compression 共step-index, fiberlike兲 re-
cent field of the cavity mode. In the following discussion, the gimes, we are mostly interested in the intermediate geometri-
atom is assumed to be located near the resonator surface at cal regimes, as these are both experimentally accessible and
the location where the electric field strength is largest, as
retain the most desirable properties of whispering-gallery-
illustrated in Fig. 2. For TM-polarized modes 共defined such
type microcavities. To accomplish this task, a two-
that the dominant electric field component is in the radial
dimensional finite-element eigenmode-eigenvalue solver was
direction兲 this occurs at the outer cavity boundary in the
equatorial plane, while for TE-polarized modes 共dominant used to characterize the optical modes of the cavity over the
electric field component in the azimuthal-vertical direction兲 complete geometrical range, after explicitly accounting for
the location of the maximum external field strength is more the rotational symmetry. The optical modes were calculated
complicated. As the toroidal geometry is compressed with in a full-vectorial model, which provides the complete elec-
respect to a sphere 共i.e., reducing the ratio of minor-to- tric field dependence. The accuracy of the numerical tech-
principal toroid diameter兲, the maximum field strength for a nique was carefully verified by comparison with results us-
TE-polarized mode changes from the equatorial outer cavity ing the analytical solution for a microsphere cavity 关27兴. The
boundary to approaching the azimuthal axis 共see Fig. 2兲. results for the mode volumes, resonance wavelengths, and
While the precise localization of the atom at the cavity eva- field profiles were in good agreement 共fractional error was
nescent field maximum has been analyzed in detail 关25,26兴, less than 10−4 and 10−2 for the resonance wavelength and
such localization has not yet been achieved experimentally. modal volume, respectively兲. Furthermore, the error in the
Nonetheless, this assumption allows a simple way to charac- radiation quality factor was less than 10% over a wide value
terize the relative merit of this cavity geometry with respect of radiation Q’s 共103 – 1014兲, demonstrating that this method
to other cavity designs. Also, in what follows we will only can give the accuracy required to investigate the fundamental
consider the fundamental radial and azimuthal modes for radiation-loss limits in the cavity geometries of interest in
both polarizations 共TE and TM兲, as they possess the smallest this work. Due to the fact that for smaller cavity geometries
modal volumes and thus the highest coupling strengths. the resonance wavelengths do not necessarily coincide with
The microtoroid geometry, which exhibits a dumbbell- the cesium transition of interest, the data in this work were
shaped cross section, can in most cases be considered a torus, evaluated by using values calculated at the closest resonance
as the presence of the supporting disk structure only affects wavelengths, both blueshifted and redshifted with respect to
the optical mode when the torus diameter becomes compa- the desired resonance, to extrapolate values at the desired
rable to the radial extent of the optical mode. As shown in wavelength 共the mode volumes were linearly extrapolated
Fig. 3, this point occurs when the toroid minor diameter 共i.e., and the radiation quality factors exponentially extrapolated
the cross-sectional diameter of the torus兲 is below approxi- as a function of wavelength兲.
mately 1.5 ␮m for a principal diameter of 16 ␮m. Further-
more, through improvements in fabrication the influence of
the toroid support can in principle be minimized. In contrast A. Mode volume
to FP and microsphere cavities, the optical modes of a toroid
do not possess analytic solutions. While one can derive ap- The optical-mode volume is determined by

013817-3
SPILLANE et al. PHYSICAL REVIEW A 71, 013817 共2005兲

dimensional harmonic oscillator model. This approach re-

sults in a reduction of modal volume which scales as
共d / D兲1/4 with respect to that of a spherical cavity. This for-
mula holds for minor diameters greater than approximately
2 ␮m for the principal diameters considered in this work.
For smaller diameters, the spatial confinement becomes
strong enough that the optical mode is additionally com-
pressed in the radial direction. This results in a faster reduc-
tion of modal volume, with the optical modes approaching
those of a step-index optical fiber 共this occurs for a minor
diameter below approximately 1 ␮m兲 关28兴. The mode vol-
ume reduces until the point where the optical mode becomes
delocalized due to the weak geometrical confinement, caus-
ing a finite minimum value. Determination of the exact point
of the minimum modal volume upon reduction of minor di-
ameter 共for a fixed principal diameter兲 can be uncertain, as
the choice of quantization volume now plays a critical role
共as discussed above兲. For this reason the results in Fig. 4
show the modal volume only for inner diameters down to
FIG. 4. Calculated mode volumes for a silica toroidal mi-
croresonator versus minor diameter for principal diameters of 20,
0.65 ␮m, where mode volume determination was unambigu-
18, and 16 ␮m. The plot shows both TM 共squares兲 and TE 共circles兲 ous.
polarizations. As the minor diameter is reduced a slow reduction of Calculation of the modal volume and the maximum elec-
modal volume due to confinement in the azimuthal direction occurs, tric field amplitude at the exterior cavity equatorial boundary
followed by a fast reduction for large confinement when the optical is straightforward, giving a simple way to calculate both the
mode is strongly compressed in both the radial and azimuthal coupling strength and the critical photon number. In order to
directions. obtain the cavity decay rate ␬ and the critical atom number
N0, however, the cavity Q factor must be determined.

冕 VQ
⑀共rជ兲兩Eជ 共rជ兲兩2d3rជ B. Quality factor
The radiation loss of the optical modes of a spherical
Vm ⬅ , 共5兲
ជ 兩2
兩E cavity is easily found by consideration of the analytic char-
max
acteristic equation 关29兴
where VQ represents a quantization volume of the electro- 关nkRjᐉ共nkR兲兴⬘ 关kRh共1兲
ᐉ 共kR兲兴⬘
magnetic field and 兩E ជ 兩 is the electric field strength 关26兴. In n1−2b = , 共6兲
nkRjᐉ共nkR兲 kRh共1兲
ᐉ 共kR兲
these calculations, we have chosen the quantization volume
cross section to consist of a square region of approximately where n is the refractive index of the spherical cavity 共the
10 ␮m width and height centered about the radial cavity external index is assumed to be unity兲, R is the cavity radius,
boundary. This choice allows the mode volume to be deter- b represents the polarization of the optical mode 共1 for TM
mined to a good accuracy while minimizing computational and 0 for TE兲, and jᐉ 共h共1兲ᐉ 兲 represents the spherical Bessel
requirements. As a further confirmation of the validity of this 共Hankel兲 function. The prime denotes differentiation with re-
approach, we note that the radiation loss is weak for the spect to the argument of the Bessel 共Hankel兲 function. This
range of geometries modeled in this work, resulting in only a equation accounts for radiation loss through the use of an
marginal difference in the numerically calculated mode vol- outgoing wave outside the cavity, as given by the complex
ume for different choices of quantization volume. Hankel function of the first kind. Solution of this equation
Figure 4 shows the calculated modal volume for the fun- results in a complex wave number k = kRe + ikIm, which deter-
damental mode of a toroidal cavity as a function of minor mines both the resonance wavelength 共␭ = 2␲ / kRe兲 and the
diameter and for principal diameters ranging from radiation quality factor 关Qrad = kRe / 共2kIm兲兴.
16 to 20 ␮m. For clarity, only data for minor diameters be- However, while the spherical solution can provide some
low 4 ␮m are shown. Both TM 共squares兲 and TE 共circles兲 insight into the scaling of the radiation quality factor for
polarizations are shown. The calculations show a reduction toroidal cavities where the minor diameter is large 共sphere
of modal volume for both polarizations as the toroid minor like兲, the radiation loss when the optical mode is strongly
diameter is decreased. This is expected when considering the confined 共as represented by small minor diameters兲 is ex-
additional confinement provided by the toroid geometry be- pected to decrease much more rapidly. Figure 5 shows nu-
yond the spherical geometry, as illustrated in the electric field merical calculations of the radiative quality factor as the mi-
plots of Fig. 3. As the minor diameter is decreased, there is nor diameter is decreased for various principal diameters of
initially a slow reduction of modal volume, which agrees 16, 18, and 20 ␮m. We observe an initially slow reduction of
very well with a simple model that accounts for transverse the radiative quality factor in the geometrical regime where
guiding 共azimuthal direction兲 using an approximate one- the minor diameter exceeds the radial extent of the optical

013817-4
ULTRAHIGH-Q TOROIDAL MICRORESONATORS FOR… PHYSICAL REVIEW A 71, 013817 共2005兲

FIG. 5. Calculated radiation loss for a toroidal microcavity as a FIG. 6. Total quality factor for a toroidal microcavity versus
function of minor diameter, for principal diameters of 20, 18, and minor diameter for principal diameters of 16, 18, and 20 ␮m. Both
16 ␮m. Both TM 共squares兲 and TE 共circles兲 polarizations are TE 共circles兲 and TM 共squares兲 polarizations are shown. The total
shown. The data show a slow reduction of Q as the minor diameter quality factor is composed of the radiative quality factor from Fig.
is reduced while the mode behaves primarily as a whispering- 5 along with the silica-absorption-limited Qmat = 2.4⫻ 1010 at a
gallery-type mode. However, as the geometrical confinement in- wavelength of 852 nm. The plots indicate that the total quality fac-
creases to such a point as the optical mode approaches that of a tor is limited by silica absorption when the principal diameter is
step-index fiber, there is a significant reduction of the quality factor. larger than 16 ␮m and the minor diameter is larger than approxi-
mately 1 ␮m. Furthermore, both polarizations have similar quality
mode 共i.e., where the optical mode exhibits whispering- factors over the range of geometries studied.
gallery behavior兲. As the minor diameter is reduced to a level
comparable to or smaller than the radial extent of the optical
mode 共step-index fiber like regime兲, the drop-off of the ra- cation the presence of water and OH can be prevented, with
diative Q is much more dramatic, with a decrease of over an surface scattering minimized, we will focus only on the con-
order of magnitude for a reduction of inner diameter of just tributions from intrinsic silica absorption and radiation loss.
50 nm. These two mechanisms put a fundamental limit on the Q
The total optical loss of a cavity has contributions not possible in these structures.
only from radiation loss, but also includes other dissipative Figure 6 shows the calculated total quality factor for vari-
mechanisms, such as intrinsic material absorption, losses re- ous principal toroid diameters in the range of 16– 20 ␮m, as
sulting from both surface and bulk scattering, and losses a function of the minor diameter. The total quality factor is
stemming from contaminates on the resonator surface 关30兴. calculated through the relation 1 / Qtotal = 1 / Qrad + 1 / Qmat,
One of the dominant contaminates which adversely affects
where only radiation loss and silica absorption are included.
the cavity Q is OH and water adsorbed onto the cavity sur-
For principal diameters less than 18 ␮m, there is a mono-
face. While prior investigations of these loss mechanisms
tonic decrease in quality factor as the minor diameter is de-
have resulted in approximate expressions for water absorp-
tion and surface scattering 关14,31兴, only very large resonators creased. This is a result of the whispering-gallery-loss in-
were studied, as opposed to the much smaller diameter cavi- crease due to the additional confinement. For larger principal
ties studied in this work. To obtain an improved estimate of diameters, the overall quality factor is clamped near the lim-
the effect of water on the small diameter cavities in this iting value resulting from silica absorption for most minor
paper, a simple model was used which determines the frac- diameters 共with only a slight decrease as minor diameter is
tion of optical energy absorbed by a monolayer of water reduced兲, until the minor diameter is small enough that the
located at the cavity surface. This method gives an estimated radiative quality factor decreases below the quality factor
quality factor for a monolayer of water to be greater than due to silica absorption. For the principal diameters studied
1010 for the case of a spherical resonator with a principal in this work, this point occurs as a minor diameter of around
diameter of 50 ␮m. While the water-limited quality factor 1 ␮m.
will be slightly lower for the smaller principal diameter cavi-
ties in this work and also slightly lower due to the increased
overlap between the optical mode and the cavity surface in a C. Cavity QED parameters
toroidal geometry, these values are comparable to the quality
factor due solely to the intrinsic absorption of silica in the The determination of the coupling strength from the
800-nm wavelength band. As in principle with proper fabri- modal volume follows from Eq. 共1兲. Figure 7 shows the

013817-5
SPILLANE et al. PHYSICAL REVIEW A 71, 013817 共2005兲

FIG. 7. Atom-cavity coupling parameter g vs minor diameter for FIG. 8. Critical photon number n0 vs minor toroid diameter for
toroidal cavities having a principal diameter of 16, 18, and 20 ␮m, a cavity with principal diameters of 16, 18, and 20 ␮m. Both TE
with g increasing for smaller principal diameters. Both TE 共circles兲 共circles兲 and TM 共squares兲 polarizations are shown. The plots show
and TM 共squares兲 polarizations are shown. The plots indicate that that as both toroid principal diameter and minor diameter are re-
the coupling strength increases dramatically as the minor diameter duced, the critical photon number decreases. This follows directly
decreases below 1.5 ␮m, which is a result of the rapid reduction of from the behavior of the atom-cavity coupling parameter g, as in-
mode volume and the increased electric field strength at the cavity dicated in Fig. 7. The calculations show that critical photon num-
surface. bers of 6 ⫻ 10−6 are possible 共with quality factors exceeding 107兲.
atom-cavity coupling rate g / 共2␲兲 for various toroid principal critical atom number. Figure 9 shows the calculated critical
diameters as the toroid minor diameter is decreased. It can be atom number versus minor diameter for toroid principal di-
seen that there is a monotonic rise in g for higher-aspect-ratio ameters of 16, 18, and 20 ␮m. The plot shows that for the
toroids 共i.e., D / d兲, as a direct result of the compression of larger principal diameters of 18 and 20 ␮m there is a mini-
modal volume. The rate of increase of g as the minor diam- mum in the critical atom number as the toroidal minor diam-
eter is reduced increases dramatically as the toroid geometry
transitions from a whispering-gallery-type mode to a
strongly confined step-index fiber-type mode. This is due not
only to the faster rate of reduction of mode volume in the
step-index fiber like regime as the minor diameter is de-
creased, but also due to the increase in electric field strength
at the cavity surface 关as g ⬀ 兩E兩共Vm兲−1/2兴. Note that the cou-
pling strengths shown do not correspond to the absolute
maximum for these structures, as this work has focused on
the simultaneous realization of high-quality factors and small
modal volume. Therefore, mode volumes were calculated
only down to where the radiation quality factor is equal to or
slightly exceeds 107. Also, as mentioned previously, by mak-
ing this restriction we prevent any uncertainty in the calcu-
lated mode volumes 共and hence g兲 through the definition of
the modal quantization volume. Under these assumptions,
the calculations indicate that coupling parameters exceeding
700 MHz are possible.
Figure 8 shows the corresponding critical photon numbers
共n0兲. The results reveal that values as low as 6 ⫻ 10−6 are FIG. 9. Critical atom number N0 vs minor diameter for a toroi-
possible, with the associated quality factors exceeding 107. dal microcavity with principal diameters of 16, 18, and 20 ␮m. For
As will be discussed in more detail in the next section, this small minor diameters the critical atom number decreases as the
value is not only comparable to the fundamental limit of FP principal diameter increases. Both TE 共circles兲 and TM 共squares兲
technology, but also vastly exceeds that possible for fused polarizations are shown. The plots indicate that there is a minimum
silica microspheres with a comparable quality factor. value of the critical atom number near 2 ⫻ 10−7 for a toroidal cavity
One of the primary reasons high-Q whispering-gallery- with a principal diameter of 20 ␮m and an inner diameter of 1 ␮m
mode cavities are promising for CQED is their very low 共TM mode兲.

013817-6
ULTRAHIGH-Q TOROIDAL MICRORESONATORS FOR… PHYSICAL REVIEW A 71, 013817 共2005兲

eter is reduced. The minimum occurs near a minor diameter capability to obtain reproducible principal diameters is a sig-
of 1 ␮m. This minimum arises from the clamping of the total nificant improvement over spherical cavities, the ability to
quality factor 共to the quality factor resulting from silica ab- accurately control the minor diameter is particularly impor-
sorption兲 for larger minor diameters when the principal di- tant to CQED. As noted previously 关19兴, the final minor di-
ameter is greater than approximately 18 ␮m. Thus, by reduc- ameter of the fabricated structures is a result of a combina-
ing the minor diameter for a fixed principal diameter, the tion of factors, which are the initial silica preform thickness,
quality factor is nearly unchanged while the coupling the supporting pillar size, and the laser irradiation intensity
strength is monotonically increasing. The critical atom num- and duration. Minor diameters as small as 3 ␮m at principal
ber decreases until the region where the minor diameter is diameters as low as 12 ␮m have been realized experimen-
such that the overall Q is determined by whispering-gallery tally.
loss. At this point the critical atom number increases ap- We have measured the quality factor of a series of fiber-
taper-coupled toroidal microcavities at a wavelength of
proximately exponentially. The plot for the 20 ␮m principal
852 nm, using an experimental apparatus similar to previous
diameter shows that in a toroidal geometry slightly larger
work 关19,32兴. The excitation laser was a New Focus Vortex
principal diameters can offer some benefit, as the minor di- laser with a tunability of 40 GHz with a center wavelength of
ameter can be compressed more strongly while maintaining 852.359 nm. The laser output was double passed through an
high radiative quality factors and, thereby, lowering the criti- acousto-optic modulator for the purpose of performing a cav-
cal atom number. A critical atom number of approximately ity ringdown measurement. The resulting beam was able to
2 ⫻ 10−7 is possible using a toroid principal diameter of be extinguished by a TTL electrical control signal, with a
20 ␮m and a minor diameter of 1 ␮m. corresponding optical decay time of 15 ns. This beam was
then coupled into a single-mode 850 nm fiber and subse-
quently interacted with the toroidal resonators through the
V. EXPERIMENTAL MEASUREMENT
tapered portion of the fiber. Due to the limited tuning range
OF MICROTOROIDS FOR STRONG-COUPLING
of the excitation laser 共which is less than the free-spectral
CAVITY QED AT 852 nm
range between fundamental modes in the cavity principal
The presented numerical results indicate that toroidal diameters of interest兲, overlap of a fundamental resonance
cavities can theoretically obtain high values of atom-cavity with the laser wavelength range was difficult. Obtaining an
coupling while simultaneously retaining an extremely low optical fundamental mode at 852.359 nm was achieved by
critical photon number and in particular an exceedingly thermally shifting the optical resonance through the use of a
small critical atom number. While in principle the critical Peltier heating element, which allowed tuning of the cavity
atom number can be more than 100 times smaller than any resonance by up to approximately 50 GHz. Upon realization
currently demonstrated cavity, the necessity of realizing of a fundamental cavity resonance at the proper wavelength,
material-limited quality factors exceeding 2 ⫻ 1010 is experi- the intrinsic quality factor was inferred two ways 共Fig. 10兲:
mentally challenging. The current record for any cavity is through cavity ringdown 关19兴 and through the threshold for
9 ⫻ 109 关14兴, in a large-diameter microsphere cavity, whereas stimulated Raman scattering 关32兴. The results of both mea-
for toroidal cavities quality factors as high as 4 ⫻ 108 at a surements were in agreement and resulted in a measured
resonance wavelength of 1550 nm have been realized 关24兴. quality factor as high as Qtotal = 1.2⫻ 108 in a cavity with a
However, for cavity quality factors much larger than 108, the principal diameter of 50 ␮m and a minor diameter of 6 ␮m.
dominant dissipative mechanism in the atom-cavity system is For this cavity geometry, the whispering-gallery loss is neg-
the radiative decay rate of the atomic medium, which is ligible 共Qrad ⬇ 1036兲 compared to the intrinsic silica absorp-
2.61 MHz for the D2 transition of cesium. For this reason tion loss, such that the overall theoretical quality factor can
more “modest” quality factors, in the range of current experi- be as high as Qtotal ⬇ 2 ⫻ 1010. We expect upon further mea-
mentally achievable values 共e.g., a few hundred million兲 are surements that this quality factor can be increased to levels
attractive. As these values are currently realizable for toroi- comparable to measurements performed at a wavelength of
dal cavities at a wavelength of 1550 nm, we have investi- 1550 nm 共4 ⫻ 108兲.
gated experimentally the quality factors and fabrication lim- While this cavity geometry is far from the optimal geom-
its for structures designed for strong coupling to the cesium etry suggested in this paper, this structure was chosen in
transition at a wavelength of 852 nm. order to increase the likelihood of finding a fundamental
As toroidal cavities are fabricated using a combination of resonance at 852 nm. Even for this relatively large structure,
lithography and a silica reflow process, the advantages of cavity QED parameters of (g / 共2␲兲 , n0 , N0) = 共86 MHz, 4.6
lithographic control and parallelism are obtained and, in fact, ⫻ 10−4 , 1.0⫻ 10−3兲 are calculated. Comparison of these val-
are a significant step forward over spherical cavities. As the ues to current FP cavities 关4,10,20兴 indicates that even with-
shape of the initial silica preform dictates the maximum pos- out additional improvements in fabrication these results are
sible principal and minor diameter and is lithographically close in coupling strength and improved with respect to the
formed, precise control of the geometry dimensions is pos- critical atom number. Additionally, if we restrict the geom-
sible. Reproducible principal diameters ranging from etry and overall quality factor to values which are currently
⬎100 to 12␮m have been fabricated. This lower value, realizable 共i.e., a quality factor of 108 at a wavelength of
while currently dictated by the available laser power in our 852 nm with a minor diameter of 3.5 ␮m, which represents a
setup, is sufficient to obtain the range of principal diameters reasonably comfortable margin from the actual current lim-
optimally suited for CQED, as indicated above. While the its兲, the optimal principal diameter is 13 ␮m 共this geometry

013817-7
SPILLANE et al. PHYSICAL REVIEW A 71, 013817 共2005兲

atom-cavity coupling solely by changing the cavity diameter,

which is easily in the realm of current fabrication capability,
but their ultra high-quality factors result in significant im-
provement in the critical atom number 共with values ap-
proaching 3 ⫻ 10−6 possible provided that silica absorption-
limited quality factors can be obtained兲. Even using quality
factors in the range of a few hundred million, which is al-
ready experimentally demonstrated, critical atom numbers
around 10−4 are possible, which is comparable to the FP
limit. From the analysis of the previous section, we see that
toroidal cavities can attain coupling strengths comparable to
or exceeding the best values possible for either FP or micro-
sphere cavities, while at the same time providing much lower
critical atom numbers. As discussed previously, this arises
from the extra level of geometrical control possible in a to-
roidally shaped cavity, which allows one to retain both the
high-coupling strength representative of small-mode volume
cavities while preserving high-quality factors. Clearly this
FIG. 10. Experimental measurement of the intrinsic quality fac- fact, along with other advantages in control and reproducibil-
tor for a toroidal microresonator at a resonance wavelength of ity over spherical cavities, suggests that these structures are
852 nm. The main figure shows the generation of stimulated Raman promising for CQED experiments.
scattering, illustrated by the secondary peak located at a wavelength Last, a comparison with photonic band-gap 共PBG兲 cavi-
of 890 nm. The threshold pump power for stimulated Raman scat-
ties is also provided in the table. Due to the realization of
tering 共53 ␮W兲 can be used to infer the intrinsic quality factor of
optical mode volumes near the fundamental limit in a dielec-
1 ⫻ 108 for this cavity. The inset shows the temporal cavity decay
resulting from a series of ringdown measurements for a different tric cavity 关21兴, combined with recent results demonstrating
toroidal microcavity. The measured photon lifetime of ␶ = 22 ns cor- reasonably high-quality factors 共⬃45 000兲 关33兴, these cavi-
responds to a loaded quality factor of QL = 0.48⫻ 108. After correct- ties are strong candidates for chip-based strong-coupling
ing for fiber-taper loading and the presence of backscattering, an CQED 关22兴. While these structures can potentially achieve
intrinsic quality factor of 1.2⫻ 108 is obtained. atom-cavity coupling strengths g ⲏ 17 GHz 关22兴, far greater
than those possible in a silica dielectric cavity, their much
lower quality factors result in greater critical atom numbers
has a radiative quality factor of 1.8⫻ 108兲. For these values than possible in toroidal microcavities. For example, the
the TM-polarized optical mode would have CQED param- work of Ref. 关22兴 projects N0 = 6.4⫻ 10−5. We also note that
eters of (g / 共2␲兲 , n0 , N0) = 共450 MHz, 1.7⫻ 10−5 , 4.5⫻ 10−5兲, the correspondingly lower quality factors also result in mod-
which are far superior to current FP cavities. est ratios of coupling to dissipation g / max共␥⬜ , ␬兲 共a figure of
merit indicative of the number of Rabi oscillations which
occur兲 of ⬃4 关22兴, much lower than predicted for toroidal
structures 共⬃165兲. Furthermore, we can consider an addi-
VI. COMPARISON OF MICROTOROIDS tional figure of merit: namely, the “rate of optical informa-
WITH OTHER RESONATORS FOR CAVITY QED tion per atom” 关1兴, given by R ⬅ g2 / ␬. The table indicates
Table I presents a comparison of CQED parameters for that toroidal cavities compare favorably with PBG cavities in
various cavity types including toroidal, FP, and photonic this figure of merit as well.
crystal. To date, most experimental work has involved the
use of Fabry-Perot cavities, with current state-of-the-art fab-
rication technology allowing the attainment of coupling VII. CONCLUSION
strengths of 110 MHz, with corresponding critical atom
numbers of 6 ⫻ 10−3 关4兴. Estimates on the theoretical perfor-
mance limits of FP cavities have also been investigated 关20兴, Our work has demonstrated that toroidal resonators are
predicting coupling rates as large as 770 MHz, with a corre- promising cavities for investigation of the coupling of an
sponding critical atom number of 2 ⫻ 10−4. While this level atomic system to the electromagnetic field in the regime of
of performance may be theoretically possible, the current strong coupling. Not only are these structures arguably sim-
necessity of expensive and sophisticated high-reflection di- pler to manufacture and control than other structures such as
electric mirror coatings does not bode well for easy improve- microspheres and FP cavities, but also allow integration on a
ments with respect to current technology. This is one of the silicon chip, paving the way for the addition of atom traps
reasons silica microspheres are of such high interest. Calcu- 关34兴 and waveguides which can enhance the capability and
lation of the limits possible with silica microspheres 关23兴 possibly reduce the experimental complexity of CQED stud-
shows that not only is it possible to obtain high values of ies. Furthermore, we note that in addition to the enhanced

013817-8
ULTRAHIGH-Q TOROIDAL MICRORESONATORS FOR… PHYSICAL REVIEW A 71, 013817 共2005兲

TABLE I. Summary of the relevant parameters for cavity QED for a variety of resonator systems. The table shows both the experimental
state of the art 关23兴 and the projected limits for a Fabry-Perot cavity 关20兴, plus current experimental results with silica microspheres 关16兴.
Furthermore, a theoretical comparison between silica microspheres 关23兴, photonic band-gap cavities 关22兴, and toroidal microresonators 共this
work兲 is also given. The results indicate that toroidal cavities can uniformly exceed the performance on these parameters for both FP cavities
and silica microspheres. Comparison with PBG cavities indicates that toroids possess much lower atom-cavity coupling strengths 共as a result
of their much larger mode volumes兲, but still result in greatly improved critical atom numbers due to their very large quality factors.

Critical photon Critical atom Coupling to Rate of

Coupling coefficient number number dissipation ratio optical information
Resonator system g / 共2␲兲 共MHz兲 n0 N0 g / max共␥⬜ , ␬兲 R ⬅ g2 / ␬ 共Mbits/sec兲

Fabry-Perot 110 2.8⫻ 10−4 6.1⫻ 10−3 7.8 5.4⫻ 103

experimental state of the art

Fabry-Perot projected limits 770 5.7⫻ 10−6 1.9⫻ 10−4 36 1.7⫻ 105

Microsphere experimental 24 5.5⫻ 10−3 3.0⫻ 10−2 7.2 1.1⫻ 103

共D = 120 ␮m兲

Microsphere theory
Maximum g 共D = 7.25 ␮m兲 750 6.1⫻ 10−6 7.3⫻ 10−1 0.01 4.5⫻ 101
Minimum N0 共D = 18 ␮m兲 280 4.3⫻ 10−5 3.1⫻ 10−6 107 1.1⫻ 107

Photonic band-gap cavity 17000 7.6⫻ 10−9 6.4⫻ 10−5 3.9 5.1⫻ 105

Toroidal microcavity theory

Maximum g ⬎700 6.0⫻ 10−6 2.0⫻ 10−4 40 1.6⫻ 105
Minimum N0 430 2.0⫻ 10−5 2.0⫻ 10−7 165 1.6⫻ 108

performance benefit of having a toroidal geometry, the capa- As a further note, the use of higher-index contrast dielec-
bility to retain a relatively large resonator diameter over tric material can allow additional improvements in the per-
other structures results in a smaller free-spectral range formance of these structures. The use of silica as the dielec-
共FSR兲. This allows not only easier tuning of the cavity reso- tric of choice in both the spherical geometry and in the
nance location to correspond precisely to the atomic transi- toroidal microcavities studied in this work was convenient,
tion wavelength, but also may allow integration of a supple- as these structures not only possess record high-quality fac-
mental far-off-resonance trap by exciting the cavity at a tors but are currently producible. However, as the radiative
multiple of the free-spectral range. The realization of a cavity quality factor of a whispering-gallery-type cavity is strongly
with a smaller FSR may allow a closer matching of a sec- dependent on the refractive index difference between the
ondary resonance location to the pump wavelength which structure and external environment, much smaller modal vol-
corresponds to state-insensitive trapping of atomic cesium umes are possible for a given quality factor with the use of a
关7兴, which can simplify the atom-cavity dynamics. The use
higher-index resonator material. In fact, this is one of the
of a silica dielectric whispering-gallery cavity also allows
reasons PBG cavities fabricated from silicon or other high-
operation over a broad range of wavelengths, with very-high-
index dielectrics can obtain ultrasmall mode volumes. A
quality factors possible for nearly all resonances. This is in
strong contrast to the mirror reflectivity limits of coated FP simple comparison of the mode volume possible in a silicon
cavities. toroid shows that a mode volume on the order of only about
The ability to connect distant quantum nodes with high 10 times larger than PBG cavities is possible, with much
efficiency, preferably over optical fiber, is very desirable for higher-quality factors. While this work has focused on silica
quantum networks. Using FP cavities, optical fiber coupling microcavities, the reflow process is a relatively flexible
is possible; however, the overall coupling efficiency is mod- method, thus suggesting that it may be possible to also create
est 共⬃70% 兲. Fiber-taper-coupled microtoroids allow cou- high-index ultrahigh-Q quality factor cavities which come
pling efficiencies in excess of 99% 关17兴, above both FP and closer to the large coupling strengths of PBG cavities while
PBG cavities 共97%兲 关35兴. This capability to obtain near- further improving the critical atom number.
complete input and output coupling efficiencies strongly sug- Last, the current experimental ability to obtain large cou-
gests the use of fiber-coupled silica whispering-gallery cavi- pling strengths with quality factors exceeding 108 is promis-
ties, such as microtoroids, as building blocks to enable high- ing for the immediate use of these structures in strong-
performance quantum networks. coupling studies. We are currently pressing forward

013817-9
SPILLANE et al. PHYSICAL REVIEW A 71, 013817 共2005兲

on improving the fabrication capabilities and losses of these ACKNOWLEDGMENTS

structures. Coupled with the intrinsic fiber-optic compatibil- The work of K.J.V. was supported by DARPA, the
ity of these structures and the demonstration of near lossless Caltech Lee Center, and the National Science Foundation.
excitation and extraction of optical energy from these struc- The work of H.J.K. was supported by the National Science
tures using tapered optical fibers 关17兴, toroidal microcavities Foundation, by the Caltech MURI Center for Quantum Net-
can provide a highly advantageous experimental system for works, by the Advanced Research and Development Activity,
the investigation of strong-coupling cavity QED. and by the California Institute of Technology.

关1兴 H. J. Kimble, Phys. Scr. T76, 127 共1998兲. 关19兴 D. K. Armani, T. J. Kippenberg, S. M. Spillane, and K. J.
关2兴 Cavity Quantum Electrodynamics, edited by P. R. Berman Vahala, Nature 共London兲 421, 925 共2003兲.
共Academic Press, San Diego, 1994兲. 关20兴 C. J. Hood, H. J. Kimble, and J. Ye, Phys. Rev. A 64, 033804
关3兴 H. Mabuchi and A. C. Doherty, Science 298, 1372 共2002兲. 共2001兲.
关4兴 C. J. Hood, T. W. Lynn, A. C. Doherty, A. S. Parkins, and H. 关21兴 J. Vuckovic, M. Loncar, H. Mabuchi, and A. Scherer, Phys.
J. Kimble, Science 287, 1447 共2000兲. Rev. E 65, 016608 共2001兲.
关5兴 P. W. H. Pinkse, T. Fischer, P. Maunz, and G. Rempe, Nature 关22兴 B. Lev, K. Srinivasan, P. Barclay, O. Painter, and H. Mabuchi,
共London兲 404, 365 共2000兲. Nanotechnology 15, S556 共2004兲.
关6兴 Y. Shimizu, N. Shiokawa, N. Yamamoto, M. Kozuma, T. 关23兴 J. R. Buck and H. J. Kimble, Phys. Rev. A 67, 033806 共2003兲.
Kuga, L. Deng, and E. W. Hagley, Phys. Rev. Lett. 89, 233001 关24兴 T. J. Kippenberg, S. M. Spillane, and K. J. Vahala, Appl. Phys.
共2002兲. Lett. 85, 6113 共2004兲.
关7兴 J. McKeever, J. R. Buck, A. D. Boozer, A. Kuzmich, H.-C. 关25兴 H. Mabuchi and H. J. Kimble, Opt. Lett. 19, 749 共1994兲.
Naegerl, D. M. Stamper-Kurn, and H. J. Kimble, Phys. Rev. 关26兴 D. W. Vernooy and H. J. Kimble, Phys. Rev. A 55, 1239
Lett. 90, 133602 共2003兲. 共1997兲.
关8兴 P. Maunz, T. Puppe, I. Schuster, N. Syassen, P. W. H. Pinkse, 关27兴 J. A. Stratton, Electromagnetic Theory 共McGraw Hill, New
and G. Rempe, Nature 共London兲 428, 50 共2004兲. York, 1997兲.
关9兴 J. A. Sauer, K. M. Fortier, M. S. Chang, C. D. Hamley, and M. 关28兴 A. W. Snyder and J. D. Love, Optical Waveguide Theory
S. Chapman, Phys. Rev. A 69, 051804共R兲 共2004兲. 共Chapman and Hall, London, 1983兲.
关10兴 J. McKeever, A. Boca, A. D. Boozer, J. R. Buck, and H. J. 关29兴 L. A. Weinstein, Open Resonators and Open Waveguides 共Go-
Kimble, Nature 共London兲 425, 268 共2003兲. lem, Boulder, CO, 1969兲.
关11兴 V. B. Braginsky, M. L. Gorodetsky, and V. S. Ilchenko, Phys. 关30兴 M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, Opt.
Lett. A 137, 393 共1989兲. Lett. 21, 453 共1996兲.
关12兴 For a recent review of microcavities, see K. J. Vahala, Nature 关31兴 M. L. Gorodetsky, A. D. Pryamikov, and V. S. Ilchenko, J.
共London兲 424, 839 共2003兲 and references therein. Opt. Soc. Am. B 17, 1051 共2000兲.
关13兴 M. L. Gorodetsky, A. A. Savchenkov, and V. S. Ilchenko, Opt. 关32兴 S. M. Spillane, T. J. Kippenberg, and K. J. Vahala, Nature
Lett. 21, 453 共1996兲. 共London兲 415, 621 共2002兲.
关14兴 D. W. Vernooy, V. S. Ilchenko, H. Mabuchi, E. W. Streed, and 关33兴 Y. Akahane, T. Asano, B.-S. Song, and S. Noda, Nature
H. J. Kimble, Opt. Lett. 23, 247 共1998兲. 共London兲 425, 944 共2003兲.
关15兴 V. Lefevre-Seguin and S. Haroche, Mater. Sci. Eng., B 48, 53 关34兴 W. Hansel, P. Hommelhoff, T. W. Hansch, and J. Reichel, Na-
共1997兲. ture 共London兲 413, 498 共2001兲.
关16兴 D. W. Vernooy, A. Furusawa, N. P. Georgiades, V. S. Ilchenko, 关35兴 A coupling efficiency of 97% between a fiber-taper and a pho-
and H. J. Kimble, Phys. Rev. A 57, R2293 共1998兲. tonic crystal waveguide has been demonstrated 关P. E. Barclay,
关17兴 S. M. Spillane, T. J. Kippenberg, O. J. Painter, and K. J. Va- K. Srinivasan, M. Borselli, and O. J. Painter, Opt. Lett. 29,
hala, Phys. Rev. Lett. 91, 043902 共2003兲. 697 共2004兲兴. For a photonic crystal defect cavity the currently
关18兴 J. I. Cirac, P. Zoller, H. J. Kimble, and H. Mabuchi, Phys. Rev. demonstrated efficiency is 44% 共P. E. Barclay, K. Srinivasan,
Lett. 78, 3221 共1997兲. and O. J. Painter, http://arxiv.org/abs/physics/0405064兲.

013817-10