RAPID COMMUNICATIONS

PHYSICAL REVIEW A 73, 041802共R兲 共2006兲

Observation of multiple thresholds in the many-atom cavity QED microlaser

C. Fang-Yen,1,* C. C. Yu,1 S. Ha,1,† W. Choi,2 K. An,2 R. R. Dasari,1 and M. S. Feld1

1
G. R. Harrison Spectroscopy Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
2
Department of Physics, Seoul National University, Seoul, Korea 151–742
共Received 10 October 2005; published 20 April 2006兲

We report the observation of multiple laser thresholds in the many-atom cavity QED microlaser. Traveling-
wave coupling and a supersonic atom beam are used to create a well-defined atom-cavity interaction. Multiple
thresholds are observed as jumps in photon number due to oscillatory gain. Although the number of intracavity
atoms is large, up to N ⬃ 103, the dynamics of the microlaser agree with a single-atom theory. This agreement
is supported by quantum trajectory simulations of a many-atom microlaser and a semiclassical microlaser
theory. We discuss the relation of the microlaser with the micromaser and conventional lasers.

DOI: 10.1103/PhysRevA.73.041802 PACS number共s兲: 42.50.Pq, 42.55.⫺f

The most fundamental model of light-matter interaction at of 138Ba atoms passes through the TEM00 mode of a high-
the atomic level consists of a two-level atom coupled to a finesse optical cavity 共symmetric near-planar cavity, mirror
single mode of the electromagnetic field, e.g., in an optical separation L ⬇ 0.94 mm, mirror radius of curvature r0
resonator. For an atom-cavity coupling strength greater than = 10 cm, finesse F ⬇ 9.0⫻ 105.兲 The mirror separation and
the decay rates of an atom and cavity 共g Ⰷ ⌫c , ⌫a兲, an excited finesse were determined by measurement of transverse mode
atom exchanges energy coherently with the cavity 共Rabi os- spacing and cavity ringdown decay time. The background
cillation兲 关1兴. pressure in the vacuum chamber is less than 10−6 Torr. Prior
Atom-resonator interaction is also the domain of laser to entering the cavity mode, each atom is excited by a
physics; therefore, it may be somewhat surprising that most continuous-wave pump laser 共Coherent 899-21 Ti:Sapphire
descriptions of laser operation use an incoherent model in- ring laser兲 locked to the 1S0 ↔ 3 P1 transition 共␭ = 791.1 nm,
volving population densities and Einstein A and B coeffi- linewidth ⌫a ⬇ 50 kHz兲 as described in 关9兴. The pump beam
cients. Such a model applies due to gain medium broadening, is focused by a cylindrical lens to a roughly 50 ␮m
laser field nonuniformity, and other statistical effects 关2兴. ⫻ 500 ␮m elliptical Gaussian beam centered approximately
The microlaser is a laser in which coherent Rabi oscilla- 150 ␮m from the cavity axis. This distance between pump
tion is explicitly reflected in the dynamics of the laser. A and cavity mode assures that the pump beam does not over-
controlled atom-cavity interaction and absence of strong sta- lap the cavity, while minimizing the effect of atomic decay
tistical averaging leads to behavior foreign to conventional between the pump and cavity. The total decay rate from the
lasers. In this letter, we report the most dramatic of these 3
P1 excited state corresponds to a decay length ⬇1.3 mm for
effects; the existence of multiple laser thresholds. an atom velocity of 815 m / s. The cavity spacing is adjusted
The microlaser is the optical analogue of the micromaser by a piezoelectric modulator 共PZT兲 to be close to resonance
关3兴, in which a similar bistable behavior has been observed with the 1S0 ↔ 3 P1 transition by monitoring a ␭ = 791.1 nm
for a single atom in the cavity 关4兴. The most significant dif- probe beam through the cavity. The pump beam polarization
ferences between the two experiments are: 共i兲 A large num- direction is set perpendicular to the cavity axis to ensure
ber of atoms is present, 共ii兲 optical frequencies allow direct
detection of generated light, and 共iii兲 the number of thermal
photons at optical frequencies is negligible.
An earlier version of the microlaser experiment 关5兴 was
shown to exhibit laser action with an intracavity atom num-
ber N on the order of 1. In the current setup, N Ⰷ 1; remark-
ably, the many-atom system exhibits very similar behavior to
that predicted from a single-atom theory. This is supported
by quantum trajectory simulations 关6兴 and may be explained
by a semiclassical theory 关7兴. In addition, a theoretical study
of a mesoscopic system of Rydberg atoms in a cavity driven
by a external field 关8兴 demonstrated bistability and multista-
bility via a related mechanism.
The experiment is illustrated in Fig. 1. A supersonic beam

FIG. 1. Schematic of cavity QED microlaser. M: Cavity mirrors,

*Electronic address: [email protected] C: Cavity mode, AB: Atomic beam from oven, A: Collimating ap-
†
Now at: Department of Physics, University of Wisconsin, Madi- erture, P: Pump field 共into page兲, ␪: Cavity tilt angle, LP: Locking
son, WI 53706. probe beam. Dashed line is normal to cavity axis.

1050-2947/2006/73共4兲/041802共4兲/$23.00 041802-1 ©2006 The American Physical Society

 RAPID COMMUNICATIONS

FANG-YEN et al. PHYSICAL REVIEW A 73, 041802共R兲 共2006兲

maximum coupling. A magnetic field of ⬇150 Gauss ori-

ented parallel to the pump beam polarization ensures that
only m = 0 ↔ m = 0 transitions occur.
To measure the state of atoms which have passed through
the pump beam, we monitor fluorescence of the atoms in the
focused beam of a dye laser tuned to the 1S0 ↔ 1 P1,
␭ = 553.5 nm transition. The atoms were found to follow an
adiabatic inversion process due to a Doppler frequency chirp
from pump beam defocus. The effect is similar to that de-
scribed in 关10兴. Incomplete inversion of the atoms by the
pump beam is accounted for by defining an effective atom
number Neff ⬅ N共␳ee − ␳gg兲, where ␳ee, ␳gg are the excited and
ground state populations of the atoms entering the cavity. For
FIG. 2. Gain 共solid line兲 and loss 共dashed line兲 per atom transit
the data presented here, ␳ee − ␳gg ⬇ 0.70.
in the rate equation analysis 关left and right sides of Eq. 共1兲, respec-
The atom-cavity interaction time tint = 冑␲wm / v0 tively兴 including effects of velocity distribution, nonuniform cou-
⬇ 0.10 ␮s, with wm ⬇ 41 ␮m as the mode waist of the cavity pling, and detuning, with Neff = 1000. Closed circles: Stable solu-
and central atom velocity v0. In order to observe multiple tions. Open circles: Unstable solutions.
thresholds, a well-defined atom-cavity interaction is essen-
tial, requiring a redesign of the earlier system 关5兴.
Uniform atom-cavity interaction is achieved via control of
atom-cavity coupling g and interaction time tint. The rapid
Pemit共n兲 = 冕 dgf共g兲 冕 dtinth共tint兲sin2共冑n + 1gtint兲
spatial variation in coupling due to cavity standing wave is
eliminated by introducing a small tilt of the atomic beam is the average emission probability for an atom and n is the
relative to the normal to the cavity axis 共cf. 关11兴兲. In this number of photons in the cavity. The integrals over g and tint
configuration, the microlaser exhibits two Doppler-shifted with respective weighting functions f共g兲 and h共tint兲 reflect
resonances approximately centered at cavity frequencies the distributions in these values. 共An analogous stochastic
␻a ± kv0␪ where ␻a is the atom resonance frequency. averaging is justified in the appendix of 关13兴.兲
The peak traveling-wave atom-cavity coupling is given by The rate equation is expressed graphically in Fig. 2.
g0 = 共1 / 2兲␮ ប 冑␻a / 2⑀0V ⬇ 190 kHz, where V = ␲Lw20 / 4 is the Damping in the oscillatory gain results from averaging over
cavity mode volume and ␮ is the dipole matrix element. g and tint. A solution n⬘ of Eq. 共1兲 is stable if ⳵ / ⳵n共Pemit共n兲
Note that this expression for g0 reflects a factor of 2 reduc- − ⌫ctintn / Neff兲兩n⬘ ⬍ 0. For a sufficiently large atom number,
tion in the peak coupling relative to the standing wave value. there exist more than one stable solution.
The atom beam is restricted to the center of the Gaussian We now describe two experiments to observe multiple
mode by a 25 ␮m ⫻ 250 ␮m rectangular aperture located ap- thresholds, the transitions between solutions of Eq. 共1兲. In
proximately 3 mm from the cavity axis. The aperture is the first experiment, the microlaser cavity was locked on
translated and rotated in order to overlap the cavity axis as resonance 共␻cav = ␻a + kv0␪兲 and the microlaser emission
closely as possible. The resulting total atom-to-atom varia- measured as a function of atom density. The cavity was
tion in peak coupling g is estimated to be ⬇10%. chopped between data collection and locking on resonance
To ensure uniform atom-cavity interaction times, a super- with the atoms. Locking was performed with a 791 nm probe
sonic atomic beam with narrow velocity distribution is used. beam tuned to ␻a + kv0␪ via acousto-optic modulators and
The beam oven is similar to that of 关12兴 and consists of a aligned with the TEM00 mode of the cavity. The cavity trans-
barium-filled resistively heated tantalum tube with a 250 mi- mission was monitored by an avalanche photodiode and a
cron diameter nozzle. The distance between the oven and feedback loop controlled the cavity PZT voltage to hold the
cavity is 45 cm. The longitudinal velocity distribution was cavity at resonance. Photons exiting the cavity were incident
measured via a Doppler fluorescence measurement to be on a cooled photomultiplier tube and pulses are recorded by
maximum at v0 ⬇ 815 m / s and have a width of ⌬vFWHM a photon counter.
⬇ 100 m / s. A total full width at half maximum 共FWHM兲 We collected two sets of data in which: 共i兲 The cavity
variation in gtint of approximately 15% is achieved. Atomic photon number was allowed to develop from an initially
density in the cavity is modulated by a translating knife edge empty cavity 共n = 0兲, and 共ii兲 the cavity was seeded with a
approximately halfway between the oven and beam aperture. large 共n ⬃ 107兲 number of photons by a laser pulse provided
A heuristic rate equation for the microlaser may be ob- by the cavity locking probe beam. In both cases, the micro-
tained by equating the photon gain and loss per atom transit: laser emission was observed for 100 ms.
Results from the cavity locking experiment are shown in
Fig. 3. The solid line represents the solution to Eq. 共1兲 and is
⌫ctintn composed of two branches for the range of Neff shown. The
Pemit共n兲 = , 共1兲
Neff first 共lower兲 and second 共upper兲 branches correspond to the
first and second Rabi oscillations of Fig. 2. The lower half of
the second branch 共the portion of the line with a negative
where slope in this figure兲 is unstable.

041802-2
 RAPID COMMUNICATIONS

OBSERVATION OF MULTIPLE THRESHOLDS IN THE¼ PHYSICAL REVIEW A 73, 041802共R兲 共2006兲

FIG. 3. Photon number n versus effective atom number Neff for

cavity locking experiment. Circles 共䊊兲: “Unseeded” data; Crosses
共⫻兲: “Seeded” data. Solid line: Rate equation model, solution of
Eq. 共1兲 for experimental parameters.

The intracavity atom number was determined by measur-

ing fluorescence at 1S0 ↔ 1 P1, ␭ = 553.5 nm with an imaging
system and charge coupled device 共CCD兲 camera. Uncertain-
ties in cavity transmission, optical reflection coefficients, and
CCD quantum efficiency make it difficult to measure the
absolute density accurately. A fit to the experimental data
was used to obtain a scale factor. A fitting parameter is close
to the value expected from fluorescence measurements. The
number of photons in the cavity is similarly estimated by
considering the cavity parameters, losses in the optics, and
detector quantum efficiency. In both cases, the fitting param-
eters were within ±50% of expected values.
The unseeded results are in good agreement with the
first solution branch. The seeded results agree with the
first branch for densities up to the onset of bistability at
Neff ⬇ 240; above this point, the results agree with the second
branch.
FIG. 4. Cavity tuning line shapes. Estimated photon number n
A calculation for steady-state average photon number us- versus atom-cavity detuning ⌬cav. Solid lines, positive frequency
ing the single-atom micromaser theory of Filipowicz et al. scan. Dashed lines, negative frequency scan. Estimated effective
关13兴 predicts a sharp transition between the first and second intracavity atoms: 共a兲 Neff = 18, 共b兲 Neff = 526, 共c兲 Neff = 657, and 共d兲
branches in at Neff0
⬇ 415. No such transition was observed, Neff = 755.
apparently due to very long metastable lifetimes in the pic-
ture of the Fokker-Planck analysis of 关13兴, due to the large The detuning line shapes display strong asymmetry and
photon number. hysteresis for atom numbers above the second threshold. For
In the second experiment, the cavity detuning ⌬cav = ␻cav densities in which sudden jumps occur, each traveling-wave
− ␻a is varied for constant atom density. Data for the detun- resonance of the cavity scanning line shapes is highly asym-
ing curves for a range of atom densities are shown in Fig. 4. metric and shows hysteresis as a function of scan direction.
The solid and dashed lines represent scans in the positive and This is most likely due to interactions with the two Doppler-
negative detuning directions, respectively. shifted fields at ␻a ± kv0␪. For the general case in which nei-
The two-peaked structure is due to two traveling-wave ther field is dominant, the atoms experience a complex
resonances of the tilted cavity. The Doppler splitting was bichromatic interaction with these two fields. A simple model
measured to be about 27.1 MHz, corresponding to a cavity 关14兴 has been shown to qualitatively describe the line shape
tilt angle ␪ ⬇ 13 mrad. asymmetry and hysteresis observed here. More detailed
As the density is increased, the two resonances broaden quantitative descriptions are in progress.
and increase in amplitude. For Neff ⬇ 630, the second thresh- To examine the application of single-atom theory to our
old appears as a “spike” near the peak of each resonance, and many-atom microlaser, a semiclassical theory 关7兴 of the mi-
at Neff ⬇ 725, a jump to a third branch appears. crolaser analogous to the Lamb theory of the conventional

041802-3
 RAPID COMMUNICATIONS

FANG-YEN et al. PHYSICAL REVIEW A 73, 041802共R兲 共2006兲

laser 关15兴 has been developed. In the case of monochromatic In the microlaser, for any finite degree of broadening in
field and zero atom-cavity detuning, the semiclassical theory gtint, we have Pemit共n → ⬁ 兲 = 1 / 2. Therefore, conventional la-
reduces to the rate equation model. This provides a more ser behavior is recovered in the limit of large photon number.
rigorous justification for Eq. 共1兲. In summary, the microlaser is shown to exhibit multista-
We have also performed quantum trajectory simulations bility due to coherent atom-cavity interaction, which in con-
of a microlaser with up to five intracavity atoms 关6兴. The
ventional lasers is hidden by spontaneous emission and other
many-atom microlaser was shown to be in close agreement
with the predictions from a single-atom theory 关13兴, with a incoherent processes.
perturbation in the width of the photon number distribution We have recently demonstrated that the microlaser field
due to cavity decay during the interaction time. exhibits sub-Poisson photon statistics 关16兴 even for hundreds
The connection between the microlaser with the conven- of intracavity atoms. Measurement of the spectrum of micro-
tional laser may be illustrated by considering the effect of laser emission 关17兴 is also planned.
increased variation in gtint. For simplicity, suppose f共g兲
= ␦共g − g0兲 and interaction times are distributed according to This research was supported by National Science Founda-
atom lifetime. The dimensionless gain is then Pemit tion Grant Nos. 9876974-PHY and 0111370-CHE. K. An
⬁ 2 冑 was supported by a Korea Research Foundation Grant 共KRF-
= ␶−1
a 兰0 dtint exp共−t / ␶a兲sin 共 n + 1gtint兲 = 1 / 2兵关共n + 1兲g ␶a兴 / 关1
2 2

+ 共n + 1兲g ␶a兴其. With Eq. 共1兲, this expression describes a con-

2 2 2002-070-C00044兲. The authors thank J. Thomas of Duke
ventional saturable gain laser, which is monostable and linear University for assistance with the supersonic atom beam, and
above threshold. M. O. Scully for many helpful discussions.

关1兴 E. T. Jaynes and F. W. Cummings, Proc. IEEE 51, 89 共1963兲. Verhaar, and N. F. Verster, Phys. Rev. A 31, 3724 共1985兲.
关2兴 M. O. Scully and S. Zubairy, Quantum Optics 共Cambridge 关11兴 K. An, R. R. Dasari, and M. S. Feld, Opt. Lett. 22, 1500
U.P., Cambridge, UK, 1997兲. 共1997兲.
关3兴 D. Meschede, H. Walther, and G. Muller, Phys. Rev. Lett. 54, 关12兴 K. D. Stokes, C. Schnurr, J. Gardner, M. Marable, S. Shaw, M.
551 共1985兲. Goforth, D. E. Holmgren, and J. Thomas, Opt. Lett. 14, 1324
关4兴 O. Benson, G. Raithel, and H. Walther, Phys. Rev. Lett. 72, 共1989兲.
3506 共1994兲. 关13兴 P. Filipowicz, J. Javanainen, and P. Meystre, Phys. Rev. A 34,
关5兴 K. An, J. J. Childs, R. R. Dasari, and M. S. Feld, Phys. Rev. 3077 共1986兲.
Lett. 73, 3375 共1994兲. 关14兴 C. Fang-Yen, Ph.D. thesis, Massachusetts Institute of Technol-
关6兴 C. Fang-Yen, Opt. Commun. 共in press兲. ogy, 2002.
关7兴 K. An, J. Korean Phys. Soc. 42, 4 共2004兲. 关15兴 W. Lamb, Phys. Rev. 134, A1429 共1964兲.
关8兴 F. Casagrande, L. A. Lugiato, W. Lange, and H. Walther, Phys. 关16兴 W. Choi, J. Lee, K. An, C. Fang-Yen, R. R. Dasari, and M. S.
Rev. A 48, 790 共1993兲. Feld, Phys. Rev. Lett. 96, 093603 共2006兲.
关9兴 K. An, Appl. Phys. Lett. 66, 24 共1995兲. 关17兴 M. O. Scully, H. Walther, G. S. Agarwal, T. Quang, and W.
关10兴 J. P. C. Kroon, H. A. J. Senhorst, H. C. W. Beijerinck, B. J. Schleich, Phys. Rev. A 44, 5992 共1991兲.

041802-4