Gianfrani et al. Vol. 16, No. 12 / December 1999 / J. Opt. Soc. Am.

B 2247

Cavity-enhanced absorption spectroscopy

of molecular oxygen

Livio Gianfrani
Dipartimento di Scienze Ambientali della Seconda Università di Napoli
and Istituto Nazionale per la Fisica della Materia, Unità di Napoli Via Arena, 22, I-81100, Caserta, Italy

Richard W. Fox and Leo Hollberg

National Institute of Standards and Technology, 325 Broadway, Boulder, Colorado 80303

Received February 3, 1999; revised manuscript received May 28, 1999

A high-finesse optical cavity was employed to perform highly sensitive spectroscopy of molecular oxygen at
wavelengths near 763 nm. An equivalent absorption length of !1 km was obtained by a 26-cm-long optical
cavity with a finesse of 6000. An extended cavity diode laser was frequency locked to the cavity, and pure
absorption profiles were recovered by monitoring of the cavity transmission during continuous scans of the
cavity resonance through O2 rotational lines, allowing a detailed investigation of the line shapes. Phase
modulation of the laser at a frequency equal to the cavity free-spectral-range frequency was employed for de-
tection of weak absorption signals inside the cavity. A minimum detectable absorption coefficient of 6.9
! 10"11 cm"1 Hz"1/2 was measured. Finally, a test of the symmetrization postulate in 16O nuclei was dem-
onstrated. [S0740-3224(99)00112-5]
OCIS codes: 300.1030, 300.6360, 120.2230.

1. INTRODUCTION tivity, and this becomes even more of a problem as the ef-
fective path lengths are increased.
High-finesse cavities are a powerful tool for sensitive la-
A significant reduction in the spectrometer’s detection
ser spectroscopy, since the effective absorption path
limits can be obtained by RF heterodyne techniques in the
length can be 100–1000 times greater than commonly
used multipass cells. A number of methods have been detection of the transmitted power. In particular, a het-
used to monitor absorption in the cavity, including mea- erodyne cavity-enhanced technique has been recently
surement of the cavity decay and the cavity transmission demonstrated to detect weak saturated absorption signals
and the use of radio-frequency (RF) techniques. Cavity inside an optical cavity.9,10 This method, which has been
ring-down spectroscopy has been widely used to measure called NICE-OHMS, or noise-immune cavity-enhanced
cavity decay times and hence absorption.1–3 Initially in- optical heterodyne molecular spectroscopy,10 is based on
vestigated with pulsed lasers, cavity ring-down spectros- frequency modulation of the laser beam at the cavity free-
copy has been extended to single-mode cw dye lasers4 and spectral-range (FSR) frequency or a multiple of the FSR.
cw extended-cavity diode lasers.5,6 More recently, the An absorption detection limit of 10"14 cm"1 in the detec-
application of heterodyne detection to cavity ring-down tion of narrow saturated absorption features has been
spectroscopy has dramatically increased the sensitivity to reported,10 with a cavity finesse of 105 . This result is one
the point of approaching shot-noise-limited detection of the most sensitive absorption measurements ever re-
levels.7 ported, to our knowledge, by any technique to date.
The most direct approach to measure absorption in a As with optical heterodyne spectroscopy without the
high-finesse cavity is simply to measure a change in the optical cavity, the signal arises when a sideband is differ-
transmitted power. With this method a detection limit of entially absorbed or when dispersion causes a phase shift
10"8 cm"1 has been reported for C2H2 detection8 in a non- of the carrier or sidebands. Residual fluctuations of the
tunable 46-cm-long Fabry–Perot cavity, with a finesse of laser power and frequency will affect the transmitted car-
!18 000. In that experiment the cavity transmission rier but will also lead to exactly the same variations in
was monitored while the laser was tuned through the cav- the intensity and phase of the transmitted sidebands.
ity modes that were coincident with the absorption line, The signal derived from the synchronous detection of the
allowing the acquisition of a few points over a Doppler- transmitted power at the FSR is relatively immune to la-
broadened profile. As we show in this paper, with the la- ser frequency fluctuations, as compared with the direct
ser frequency locked to a tunable cavity, the transmitted approach.
power can be recorded during continuous frequency scans Doppler-broadened absorption profiles of C2HD over-
of the cavity resonance over the absorption line. How- tone transitions near 1.064 "m have been measured in
ever, the residual fluctuations of the laser frequency with this manner by temperature tuning a Nd:YAG laser and
respect to the cavity resonance are converted to ampli- locking onto successive cavity transmission peaks around
tude noise, which makes it difficult to achieve good sensi- an absorption line.11 The molecular linear dispersion

0740-3224/99/122247-08$15.00
2248 J. Opt. Soc. Am. B / Vol. 16, No. 12 / December 1999 Gianfrani et al.

was recorded by counting the frequency of the RF source with different actuators. To monitor absorption in the
that was locked to the cavity FSR. This scheme was also cavity, we synchronously detect the transmitted power at
proposed in Ref. 12, and numerical simulations on CH4 at the FSR frequency. Continuous cavity scans are made
1.651 "m, CO2 at 1.563 "m, CO at 1.567 "m, and NO2 at through absorption lines, with the laser source always
0.8 "m demonstrated how gas concentrations at the part- locked to the cavity resonance. We employed a 26.4-cm-
per-trillion level could in principle be measured with an long half-symmetric cavity in a stainless-steel chamber
optical cavity with a finesse of 10 000. that could be pumped down and filled with O2. Weak
An extension of this heterodyne cavity-enhanced tech- magnetic-dipole transitions of the b 1 # #g ( v ! $ 0)
nique to measure the full Doppler or pressure-broadened ← X 3 # "g ( v " $ 0) band,13 around 762 nm, were mea-
shape may find applications in the field of molecular spec- sured. We performed a test of accuracy and sensitivity
troscopy, in trace-gas detection, and in the detection of by recording heterodyne signals at very low oxygen pres-
forbidden transitions. However, there are complications sures. Recordings of 16O18O and 16O17O magnetic-dipole
associated with extension of the extreme detection limit of lines in natural abundant samples provided a further
10"14 cm"1 to wider wavelength scans required for demonstration of the low detection limits of our spectrom-
Doppler- and pressure-broadened lines. The widths of eter. We also report a detailed experimental investiga-
such lines (!1 GHz) are of the same magnitude as spuri- tion of pure absorption profiles provided by the direct ob-
ous interference fringes common in laser spectroscopy servation of the cavity transmitted light. The theory of a
(100 MHz to 10 GHz). It can be expected that any such Fabry–Perot cavity in the presence of an absorbing me-
fringes will negatively affect the detection limit compared dium is also reviewed. Finally, we discuss the possibility
with that of a narrow saturated feature. In addition, of using our spectrometer to improve the upper limit to a
even modest wavelength scans require an optical cavity possible violation of the symmetrization postulate in the
capable of scanning microns of length. Measuring wider oxygen spectrum, of several orders of magnitude with re-
spectra may require repeated relocking of the laser to dif- spect to the values recently set.14,15
ferent cavity fundamental modes and acquisition of the
spectra in a piecewise fashion.
Here we report an implementation of the heterodyne 2. EXPERIMENTAL SETUP
cavity-enhanced (CE) technique in which we continuously The experimental apparatus is shown in Fig. 1. We use a
measure spectra up to 8 GHz (0.27 cm"1) wide. Although single-mode antireflection (AR) coated diode laser that
this is broad enough for our measurement purposes in O2, emits at a wavelength of 776 nm at room temperature,
the present scan range is limited by the piezoelectric ac- mounted in a grazing-incidence extended-cavity configu-
tuators of the cavity, and wider scans should be possible ration that was 13 cm long.16 By lowering the diode tem-

Fig. 1. Heterodyne CE experimental setup. The cavity reflection is used to lock the laser to the mth cavity mode and to lock the RF
source to the cavity free spectral range. The acousto-optic modulator (AOM) is used to reduce low-frequency-intensity noise. The side-
bands from electro-optic modulator (EOM) #2 are transmitted through the ( m " 1) and ( m # 1) fundamental cavity modes and detected
with a high-speed photodetector, in a manner analogous to RF heterodyne spectroscopy.
Gianfrani et al. Vol. 16, No. 12 / December 1999 / J. Opt. Soc. Am. B 2249

perature and adjusting the grating, we could tune the la- scanned (by approximately 2 kHz/"m of mirror motion),
ser wavelength down to 762 nm, hence covering most of this necessitated locking of the 568-MHz RF source to the
the oxygen band. The spectral characteristics of FSR. The reflected power from the cavity is higher when
extended-cavity lasers are well known. Typically they the sidebands are reflected, thus dithering the RF fre-
have an inherent fast linewidth of !10 kHz, but they also quency about the FSR and demodulating the cavity reflec-
have a large amount of frequency noise or jitter at acous- tion at the dither frequency results in a lock signal.10
tic frequencies. Relatively slow (!1-MHz bandwidth) The 568-MHz frequency was dithered at 70 kHz, and the
electronic feedback can tightly lock such a laser to the feedback loop bandwidth was !10 kHz. Both the dither
center of a high-finesse cavity, as described below. rate and the loop bandwidth were limited by the RF
The optical path from the laser to the cavity included sweep generator employed. We note that a dither rate as
two phase modulators, two optical isolators, and an opti- low as 70 kHz is not optimum in this case, since the diode
cal fiber. The 50-cm-long single-mode optical fiber was laser exhibits a relatively large amount of noise in this
used in the beam path as a spatial filter in order to facili- frequency region. It is also possible to derive an error
tate the mode matching of the laser light into the cavity. signal for this lock from the sum or difference frequencies
The use of the fiber decoupled any adjustments of the la- of the two phase modulators.19 In our case that would
ser from the alignment into the cavity and more impor- entail demodulation of the cavity reflection at 554 MHz,
tantly, the alignment through the phase modulators. which is the difference between the FSR and the Pound–
The fiber ends were angle polished to avoid reflections, Drever–Hall lock frequency. This is a much lower noise
and the fiber was followed by a polarizer to fix the output region of the spectrum and should result in improved per-
polarization angle. formance.
The optical resonator had a FSR of 568 MHz and was The cavity transmission was monitored with a wide-
constructed with a 100-cm-radius input coupler and a flat band (1-GHz) detector. The high-frequency portion of
output mirror. Scanning was accomplished by means of the photocurrent from the cavity transmission was band-
three piezoelectric transducer stacks mounted between pass filtered, demodulated at 568 MHz, lowpass filtered,
the input mirror and the invar cavity spacer. Each stack and sent to a data-acquisition system. The phase of the
was capable of 8 "m of motion with 150 V applied, and 568-MHz local oscillator was adjusted so that dispersion-
the stacks were arranged on a diameter 120° apart, oper- like signals were observed when the cavity mode was
ated in compression, and driven in parallel. No change scanned over an absorption line. This is the heterodyne
in the resonance peak height that would indicate a mis- CE signal. As with heterodyne detection without a cav-
alignment was observed as up to 14 longitudinal cavity ity, the signal level will fluctuate with changes in the
modes were scanned past the laser frequency. An empty- phase modulation imparted to the light beam. We can
cavity finesse of 6000 was found from a measurement of suppress resulting noise and baseline changes by using a
the cavity-mode linewidth. The cavity was mounted in a modulation of the CE signal and then demodulating sub-
stainless-steel vacuum chamber with Brewster-angled sequent to the high-frequency RF demodulation. We did
windows, into which high-purity oxygen gas could be this by using a dither of the cavity. As the cavity was
filled at adjustable pressures and pumped out with an oil- slowly swept over the spectral region of interest, a sine
free vacuum pump system. We monitored the pressure modulation was superimposed on the cavity piezo drive
with an uncertainty of %10 mTorr using a capacitance signal. The rate was typically 50 Hz, with a modulation
manometer (and also using a Pirani gauge as a check). depth corresponding to approximately 500 MHz of optical
The laser source was frequency locked to the cavity frequency. The modulation depth was limited by broad-
with the Pound–Drever–Hall method,17 with sidebands ening of the observed lines, and the rate was limited by
at !14 MHz with an amplitude of !5% of the carrier. the servo gain bandwidth of the loop holding the RF
Phase-sensitive detection of the cavity-reflected beam modulation on the FSR frequency. Using a lock-in am-
provided a dispersion signal that was used to feed back to plifier, we observed the first derivative of the dispersion-
the laser injection current and the external cavity length, like profile by demodulating at 50 Hz.
by means of a fast and a slow servo amplifier, respec- High-quality optical components with good AR coatings
tively. We measured a bandwidth of the fast loop of !1.5 were used. They were carefully mounted in order to
MHz. With the laser locked to the cavity, sweep and avoid spurious interference effects caused by residual
dither signals used to modulate the cavity were also sent backreflections.
to the laser in a feedforward manner. Unfortunately, when the laser frequency was locked to
In the presence of an absorbing gas, direct absorption the cavity, the electronic corrections of the servo on the
profiles could be recorded with the dc output of the fast laser injection current increased the intensity noise inside
photodetector monitoring power transmitted through the the loop bandwidth.20 In principle, this should not affect
cavity. This served as a means to calibrate accurately the noise background of the signal demodulated at the
the effective path length of the cavity by scanning across FSR from the cavity transmission. In practice, the phase
lines with a known absorption coefficient. modulator produces some residual amplitude modulation
The RF sidebands at 568 MHz were impressed by at the FSR, the magnitude of which is dependent on the
electro-optical modulator 2, a resonant single-loop LC laser intensity. We reduced this source of noise using an
tank circuit driven by 300 mW of RF power.18 The ca- acousto-optic modulator in a feedback loop (!100-kHz
pacitance was dominated by the crystal, an AR-coated bandwidth) to control the laser power. A beam splitter
5 mm ! 5 mm ! 50 mm long piece of LiNbO3. Because and detector before the cavity sampled !1/3 of the power,
the FSR changes by a small amount as the cavity is and the acousto-optic modulator was used to deflect
2250 J. Opt. Soc. Am. B / Vol. 16, No. 12 / December 1999 Gianfrani et al.

power from the fiber input so as to keep the sampled 4. RESULTS

power constant.
A. Pure Absorption Spectroscopy
Pure absorption profiles, corresponding to the P P (9) line
at 763.8434 nm, were recorded for several values of the
3. THEORY oxygen pressure, always keeping the laser source fre-
In the presence of an absorbing medium the optical field quency locked to the cavity. The results are shown in
transmitted through a Fabry–Perot cavity is given by Fig. 2. In spite of the large frequency scans (greater than
5 GHz), a flat background was observed in each spectrum.
E t$ % &
$ t 1 t 2 exp$ i 2 ' c t /( & exp$ i 2 ' n l /( & This is the case even though the laser power was chang-
E0 ing appreciably over the scan owing to the characteristics
of the extended-cavity laser. Indeed, the intensity con-
exp$ ") $ % & L /2& trol feedback loop was able to cancel any change in the la-
! . (1)
1 " r 1 r 2 exp$ i 4 ' n L /( & exp$ ") $ % & p L & ser power during the cavity scans. We also note that the
laser remained locked to the cavity even in the case in
Here, E 0 is the field amplitude incident upon the cavity, ( which the transmittance was less than 10%. The total
is the laser wavelength, L is the optical cavity length, and fractional change in the transmittance at the line center
n is the refractive index. ) (%) is the pressure-dependent was measured as a function of the O2 pressure, and is
absorption coefficient of the absorbing medium in shown in Fig. 3. The saturation of the signal at higher
cm"1 Pa"1, p is the pressure, and t 1 , t 2 , and r 1 , r 2 are pressure demonstrates that spectrometers that employ
the amplitude-transmission and reflection coefficients for cw-cavity-enhanced spectroscopy are capable of a large
the input and output mirrors. dynamic range and still enjoy a linear response at low lev-
When the laser frequency is resonant with the cavity, els of absorption.
in the limit of ) ( % ) p L & 1 and r 1,2 * 1, the fraction of the Using Eq. (3), we found a nearly perfect fit of the ex-
power transmitted over the incident power may be writ- perimental data, as shown in the same figure. Since the
ten as absorption coefficient of this line at line center and room
temperature, ) ( % 0 ), is well known,21 we were able to de-
P t$ % & t 12 t 22 1 duce from the fit an accurate measurement of L eq : 962

! "
$ . (2)
) $ % & p L !r 1 r 2
2 2
P0 $ 1 " r 1r 2 & ('2) m. This value is necessary to determine an accu-
1# rate calibration of the spectrometer sensitivity. Also,
1 " r 1r 2 knowledge of L eq enables the measurement of unknown
The equation above, after a simple algebraic calcula- absorption coefficients. For instance, we repeated the
tion, provides the fractional change in the transmitted same experimental procedure, as shown in Figs. 2 and 3,
power caused by the lossy medium: for the P Q (13) line at 764.6304 nm. Using the above re-
ported value of L eq , we found an absorption coefficient
+ P t$ % & 1 ) ( % 0 ), at room temperature, of 4.54(6) ! 10"8 cm"1

! "
$1" . (3) Pa"1 (6.06 ! 10"6 cm"1 Torr"1), which is in agreement
) $ % & p L !r 1 r 2
2
Pt with the value reported in Ref. 22.
1# The absorption line shape of the transmitted power
1 " r 1r 2
versus wavelength is provided by Eq. (2). This is demon-
If the one-pass absorption loss ) ( % ) p L is much lower
than the empty cavity loss 1 " r 1 r 2 , Eq. (3) becomes

+Pt !r 1 r 2
$ % & $ 2)$ % &pL . (4)
Pt 1 " r 1r 2
Comparing Eq. (4) with the Beer–Lambert law, we may
conclude that the absorption path length is enhanced by a
factor 2 F / ' , where F is the empty-cavity finesse, given by

' !r 1 r 2
.
1 " r 1r 2
Thus an equivalent interaction length L eq can be defined
as

2
L eq $ FL. (5) Fig. 2. Directly detected transmitted power from the cavity as
'
the system is scanned over an oxygen absorption line. The six
It is worth noting that the Doppler-broadened absorp- traces are with pressures of 13.3, 106.7, 173.3, 400, 800, and
1333 Pa. The shift in the absorption is primarily the cavity
tion profile is a Gaussian function only in the limit of va- shifting that is due to the increasing refractive index with O2
lidity of Eq. (4). Otherwise, Eq. (2) must be used to pre- pressure. The departure from a Doppler line shape is due to the
dict the line shape observed. decrease in the effective path length with increasing cavity loss.
Gianfrani et al. Vol. 16, No. 12 / December 1999 / J. Opt. Soc. Am. B 2251

shift of the cavity resonance with respect to the absorp-

tion line increases with the oxygen pressure inside the
chamber. The refractive index increases with the pres-
sure, leading to a displacement of the cavity resonance to-
ward the red. The pressure shift of the line center makes
a contribution much smaller than the observed shift (the
pressure-shift coefficient24 is !3.8 kHz/Pa (0.5 MHz/
Torr). Although thermal drift of the cavity length could
contribute to the observed shift, we experimentally veri-
fied that it is negligible on the time scale of our measure-
ment. From a fit of each spectrum the line center was
determined as a function of the oxygen pressure. A lin-
ear fit of the experimental points shown in Fig. 5 provides
a displacement rate of 0.910(3) MHz/Pa. We performed
these measurements also for the P Q (13) line and found a
shift of 0.899(9) MHz/Pa. The uncertainties correspond
to one standard deviation. These shifts correspond to a
Fig. 3. Peak absorption of the 16O2 P P (9) line as a function of change in the index of refraction of approximately
pressure. The system response is linear if the absorption that is
due to the gas is much less than the mirror loss and saturates 2.3 nPa"1 (3 ! 10"7 Torr"1).
with increasing absorption as the effective path length is re-
duced. A fit of these data with the theory (see text) results in an
accurate determination of the effective path length, nearly 1 km
in this case.

strated in Fig. 4, where a good fit of the spectrum corre-

sponding to an O2 pressure of 533 Pa (4 Torr) is shown.
An attempt to fit the experimental points with a Gaussian
function shows a clear disagreement. Indeed, in this
case the Beer–Lambert law [Eq. (4)] cannot be used be-
cause the single-pass absorption ) ( % 0 ) p L , equal to 7.3
! 10"4 , is comparable to 1 " r 1 r 2 , which is !5
! 10"4 . However, a slight asymmetry in the line shape
occurs, as evidenced in the residuals shown in the upper
part of Fig. 4. This may well be due to nonlinearities of
the scan owing to the piezoelectrics moving the mirror.
A more accurate determination of line shapes would re-
quire some sort of linearization of the cavity scan. We
Fig. 4. Dots are data points taken from one trace of Fig. 2. The
note that, in principle, this information is available if the
oxygen pressure is 533 Pa. A least-squares fit of the theory [Eq.
FSR frequency is monitored because a linear scan in op- (2)] to the data is shown as a thin solid curve nearly on top of the
tical frequency results in a linear scan of the FSR. How- data points. The residuals from the fit are shown above with a
ever, the deviations of the data from Eq. (2) could also be 5! magnified scale. A least-squares fit of a Doppler line shape
due in some part to thermally induced changes in the re- to the data is also shown.
fractive index of the absorbing gas. If the refractive in-
dex is changed by , n , the cavity resonance is shifted by

, % m $ "% m , n . (6)

Therefore even if we attempt to change linearly the

cavity resonance frequency % m about an absorption line
by using a linear voltage ramp and well-behaved piezo-
electric actuators, further changes in % m are introduced
by the absorbing medium, giving rise to a slight nonlinear
scan. Similar effects are often observed owing to nonzero
values of , n /, T in frequency-doubling crystals in
build-up cavities. Also, this effect has been investigated
in the near-IR spectrum of oxygen and acetylene.23 In
that case a free-running laser source was employed and
asymmetric cavity resonances were observed when the la-
ser frequency was resonant with the sample gas. How-
ever, these effects were much smaller in oxygen, likely be- Fig. 5. Data points are the frequency shift of the apparent ab-
sorption line shape as a function of the oxygen pressure and are
cause of the slow collisional energy relaxation. shown along with a best-fit straight line. The data points were
A more evident effect, owing to refractive-index determined from best fits of the Fig. 2 data series to Eq. (2). The
changes, can be noted in Fig. 2, in which the frequency shift is primarily due to the increasing index of refraction.
2252 J. Opt. Soc. Am. B / Vol. 16, No. 12 / December 1999 Gianfrani et al.

We note that simply evacuating the cavity to determine

the baseline would not be sufficient, because oxygen ab-
sorption in the ambient air path to the cavity is readily
observed. An analysis of the detection sensitivity can be
done accurately only if the laser source is tuned to a
wavelength region where no absorption features occur.
Trace D was taken at a wavelength free from any over-
lapping magnetic-dipole lines of oxygen isotopomers, i.e.,
16 18
O O, 16O17O, 17O18O, 18O2, and 17O2, as can be verified
with the data reported in Refs. 25 and 26.
The noise is clearly due to spurious interference
fringes, mainly caused by the residual reflectivity of the
AR-coated facets of EOM#2. An accurate signal-to-noise
(S/N) measurement requires the estimation of the noise
level in the background signal. If the background noise
were random, we could measure the ratio between the
amplitude of the signal and the root-mean-square of the
background. However, the background is still dominated
by systematic fringes, and we measure a S/N ratio of 8000
by graphically comparing trace C, divided by the S/N, and
trace D. More explicitly, the peak in trace C may still be
recognized as a signal when trace C is divided by 8000
and added to trace D. Since trace C represents an ab-
Fig. 6. Trace A is the direct photocurrent detected from the cav-
sorption at line center of 3 ! 10"7 cm"1, the measured
ity transmission while scanning over the P Q (13) line of 16O2 with
50 mTorr of pressure. Trace B is the heterodyne CE signal, the absorption detection limit was ) min $ 3.8 ! 10"11 cm"1.
demodulation of the photocurrent at the free-spectral-range fre- Traces B, C, and D were recovered with a cavity-
quency. Trace C is the demodulation of the CE signal with a transmitted power of 500 "W, a low-pass filter bandwidth
lock-in while the cavity was slowly modulated and swept over the B of 3 Hz, and 10 coadditions. Coadditions refers to av-
absorption. Trace D is the noise baseline in a nearby spectral
region that has no known absorptions. Traces B, C, and D are eraging of N spectral traces together to reduce the
10 coadditions (see text). random-noise component by !N . Each individual data
trace exhibited more noise fluctuations than is evident in
B. Heterodyne Measurements trace D. In fact, it was only by averaging that we could
The heterodyne CE measurement should be far superior observe the fringes clearly. Therefore we find the
to a direct absorption measurement, since the laser has equivalent absorption detection limit on a per hertz basis
less noise at higher frequencies and less residual fre- with no coadditions by multiplying the measured limit by
quency noise is converted to amplitude noise by the cav- !N and dividing by !B . This measured absorption de-
ity. Four signal traces are shown in Fig. 6 that illustrate tection limit is 6.9 ! 10"11 cm"1 Hz"1/2.
the heterodyne versus the direct measurement. The first Under shot-noise-limited operation, the theoretical
three were recorded with 6.67 Pa (50 mTorr) of oxygen in minimum detectable single-pass absorption is given by10
the cavity, while trace D was recorded with four times
that pressure.
Trace A shows the direct absorption corresponding to
the P Q (13) line of 16O2. The absorption coefficient at
+Pt
Pt
# min

single
pass
$ ) min L $
'
2F
! 2eB 1
- P t J 0$ . & J 1$ . &
!2, (7)

line center is 3.0 ! 10"7 cm"1, measured by recording of

an absorption of 2.9% over the 962-m path. where the modulation index . was approximately 0.5, the
Trace B is the heterodyne CE signal from the cavity- empty-cavity finesse F $ 6000, the cavity-transmitted
transmitted power. The cavity output was detected with power P t $ 500 " W, and the photodiode responsivity -
a wideband detector, and the resulting photocurrent was $ 0.5 A/W. Normalizing Eq. (7) with respect to the
synchronously detected at the modulation frequency, bandwidth yields a theoretical shot-noise-limited detec-
which was locked to the cavity’s FSR. tion level of 5.8 ! 10"11 Hz"1/2, equivalent to a minimum
Trace C also represents the heterodyne CE signal but is detectable absorption coefficient of 2.2 ! 10"12 cm"1
recorded by a slow (50-Hz) modulation of the cavity fre- Hz"1/2. Thus our present detection limit is approxi-
quency as the resonance is swept over the absorption line. mately a factor of 30 above the shot-noise limit.
After synchronous detection at the FSR frequency, a
lock-in amplifier is used to demodulate the 50-Hz signal. C. O2 Forbidden Transitions
This second demodulation serves to reject fluctuations of Trace D of Fig. 6 was recorded by scanning of the cavity
the modulation index, for instance, caused by noise of the resonance around the calculated position of the oxygen
RF power source. The resulting line shape is the deriva- 16
O2 P P (12) forbidden line at 764.5088 nm. In the O2
tive of Trace B. spectrum, transitions starting from even rotational quan-
Finally, trace D is a magnified portion of the baseline, tum numbers are missing. This is a consequence of the
away from any known signals. It is evident the system is symmetrization postulate of quantum mechanics. This
limited by systematic noise. phenomenon has been used to perform a test of the sym-
Gianfrani et al. Vol. 16, No. 12 / December 1999 / J. Opt. Soc. Am. B 2253

metrization postulate,14,15 and an upper limit to a pos-

sible violation in O2 has previously been set at 5
! 10"7 .
The spectral region around the 16O2 P P (12) line was
explored in the presence of 2.67 ! 104 Pa (200 Torr) of
oxygen inside the chamber. No evidence of a signal in
the expected position of the P P (12) line was found. Thus
after a comparison with trace C of Fig. 6, considering the
S/N ratio, scaling the observed signals for the different
values of the pressure, and also taking into account the
pressure broadening of the lines, we found an upper limit
to the intensity of the P P (12) forbidden line with respect
to the P Q (13) line of 5 ! 10"8 . This is factors of 10 and
16 smaller than the respective limits reported in Refs. 14
and 15.
We also observed weak lines of 16O18O and 16O17O iso-
topomers in natural abundance oxygen samples (atomic Fig. 9. Comparison of the signal from a 16O2 line with a known
abundance: 18O $ 0.2%, 17O $ 0.037%). In Fig. 7 the absorption coefficient allows the determination of the absorption
heterodyne CE signal corresponding to the P Q (13) line at coefficient of the 16O17O line. This is given by the ratio between
764.4890 nm of 16O17O isotopomer is recorded for four val- the amplitude of the two signals (!0.16) times the absorption co-
ues of the oxygen pressure, from 667 Pa to 1.33 ! 104 efficient of the 16O2 line (4.54 ! 10"8 cm"1 Pa"1) and is divided
by the partial pressure of the 16O17O measurement relative to the
Pa. In Fig. 8 the peak amplitude of these signals is plot- 16 16
O O measurement (0.38). The natural molecular fractional
abundance of 16O17O is 7.6 ! 10"4 .

ted as a function of the pressure, and a slight nonlinear

behavior can be noted. This is due to two factors: the
onset of pressure broadening and the decreasing modula-
tion width from the constant dither amplitude. In Fig. 9
a comparison between the P Q (13) lines in 16O17O and
16
O2 enables the measurement of the absorption coeffi-
cient ) ( % 0 ), at room temperature, for the 16O17O line. It
was found to be 1.91(3) ! 10"8 cm"1 Pa"1.

5. CONCLUSIONS
We have extended the cavity-enhanced (or NICE-OHMS)
technique to the Doppler-limited regime by recording sig-
nals continuously over frequency spans up to 8 GHz. Us-
Fig. 7. NICE-OHMS signal from absorption of the P Q (13) line ing an equivalent path length of 962 m, we observed weak
of 16O17O at 764.489 nm for four different pressures of natural- magnetic-dipole transitions of molecular oxygen around
abundance oxygen (from 660 Pa up to 1.3 ! 104 Pa). 762 nm. Absorption profiles were continuously recovered
with the laser locked to the cavity by scanning of the cav-
ity resonance frequency through the line. We also pre-
sented an investigation of the line shapes from the inter-
action of the intracavity laser radiation with Doppler-
broadened lines.
Our sensitivity corresponds to a minimum detectable
absorption coefficient of 6.9 ! 10"11 cm"1 Hz"1/2. Al-
though this is the lowest detection limit obtained in the
laboratory for the molecular oxygen near-IR absorption
spectrum, if we consider the recent experiments in Refs.
14 15, 25, and 26, it is still a factor of 30 from the shot-
noise level that should be achievable. The dominant
noise in this experiment was due to residual interference
fringes from optical components. An improvement may
be provided by incorporation of a Brewster-cut crystal in
the FSR phase modulator instead of the AR-coated one
used here. Another possible source of the systematic
Fig. 8. Peak signal levels from Fig. 7 plotted as a function of background is deviations of the 568-MHz sidebands from
pressure. A best-fit simple polynomial curve is shown also.
The slight nonlinearity is due to pressure broadening and the de-
the cavity FSR. Realizing the RF lock servo in a differ-
creasing modulation width, not to saturation of the NICE-OHMS ent manner may result in a higher S/N ratio and better
signal. performance.
2254 J. Opt. Soc. Am. B / Vol. 16, No. 12 / December 1999 Gianfrani et al.

Further improvements to the sensitivity are possible by C2HD overtone transition,’’ Opt. Lett. 21, 1000–1002
an increase in the mirror reflectivity. Available low-loss (1996).
10. J. Ye, L.-S. Ma, and J. L. Hall, ‘‘Ultrasensitive detections in
mirrors easily support a cavity finesse of 6 ! 104 , which atomic and molecular physics: demonstration in molecu-
would provide an order-of-magnitude improvement of the lar overtone spectroscopy,’’ J. Opt. Soc. Am. B 15, 6–15
sensitivity. Although the resulting cavity linewidth (!10 (1998).
kHz) would be comparable with the inherent laser line- 11. J. Ye, ‘‘Ultrasensitive high resolution laser spectroscopy
width, there is enough electronic gain available to col- and its application to optical frequency standards,’’ Ph.D.
dissertation (University of Colorado, Boulder, Colorado,
lapse nearly all the power into the cavity lock. 1997).
Our spectrometer enabled an improvement of the up- 12. A. Mugino, T. Tamamoto, T. Omatsu, M. A. Gubin, A. Mori-
per limit to a possible violation of the symmetrization pos- naga, and N. Takeuchi, ‘‘High sensitive detection of trace
tulate in the 16O2 spectrum. In this paper, we have set a gases using optical heterodyne method with a high finesse
intra-cavity resonator,’’ Opt. Rev. 3, 243–250 (1996).
value of 5 ! 10"8 , which is a factor 10 better than in pre- 13. H. D. Babcock and L. Herzberg, ‘‘Fine structure of the red
vious work with O2. system of atmospheric oxygen bands,’’ J. Astrophys. 108,
167–190 (1948).
14. M. de Angelis, G. Gagliardi, L. Gianfrani, and G. M. Tino,
‘‘Test of the symmetrization postulate for spin-0 particles,’’
ACKNOWLEDGMENTS Phys. Rev. Lett. 76, 2840–2843 (1996).
The authors are grateful to J. Ye, L.-S. Ma, and J. L. Hall 15. R. C. Hilborn and C. L. Yuca, ‘‘Spectroscopic test of the
for their helpful assistance, including the designs of the symmetrization postulate for spin-0 nuclei,’’ Phys. Rev.
Lett. 76, 2844–2847 (1996).
resonant modulator and high-speed detectors used in this 16. The laser was an SDL 5400C (mentioned to specify experi-
work. This work is a contribution of the U.S. Govern- mental parameters; other sources may be suitable).
ment, and is not subject to copyright. 17. R. W. P. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M.
Ford, A. J. Munley, and H. Ward, ‘‘Laser phase and fre-
quency stabilization using an optical resonator,’’ Appl.
Phys. B. 31, 97–105 (1983).
REFERENCES AND NOTES 18. L.-S. Ma, JILA, University of Colorado, Boulder, Colorado
1. A. O’Keefe and D. A. G. Deacon, ‘‘Cavity ring-down optical (personal communication, 1997).
spectrometer for absorption measurements using pulsed la- 19. R. G. DeVoe, C. Fabre, K. Jungmann, J. Hoffnagle, and R.
ser sources,’’ Rev. Sci. Instrum. 59, 2544–2551 (1988). G. Brewer, ‘‘Precision optical-frequency-difference mea-
2. D. Romanini and K. K. Lehmann, ‘‘Ring-down cavity ab- surements,’’ Phys. Rev. A 37, 1802–1805 (1988).
sorption spectroscopy of the very weak HCN overtone 20. R. W. Fox, L. D’Evelyn, H. G. Robinson, C. S. Weimer, and
bands with six, seven, and eight stretching quanta,’’ J. L. Hollberg, ‘‘Amplitude modulation on frequency-locked
Chem. Phys. 99, 6287–6301 (1993). extended-cavity diode lasers,’’ Proc. SPIE 2378, 58–62
3. J. B. Paul, J. J. Scherer, A. O’Keefe, and R. J. Saykally, (1995).
‘‘Cavity ringdown measures trace concentrations,’’ Laser 21. M. de Angelis, L. Gianfrani, F. Pavone, A. Sasso, and G. M.
Focus World 33, 71–75 (1997). Tino, ‘‘Temperature dependence of self-broadening in
4. D. Romanini, A. A. Kachanov, N. Sadeghi, and F. Stoeckel, molecular-oxygen spectrum,’’ Nuovo Cimento 18, 557–564
‘‘CW cavity ring down spectroscopy,’’ Chem. Phys. Lett. (1996).
264, 316–322 (1997). 22. K. J. Ritter and T. D. Wilkerson, ‘‘High-resolution spectros-
5. D. Romanini, A. A. Kachanov, and F. Stoeckel, ‘‘Diode laser copy of the oxygen A-band,’’ J. Mol. Spectrosc. 121, 1–19
cavity ring down spectroscopy,’’ Chem. Phys. Lett. 270, (1987).
538–545 (1997). 23. P. Dubé, L.-S. Ma, J. Ye, P. Jungner, and J. L. Hall, ‘‘Ther-
6. B. A. Paldus, C. C. Harb, T. G. Spence, B. Wilke, J. Xie, J. mally induced self-locking of an optical cavity by overtone
S. Harris, and R. N. Zare, ‘‘Cavity-locked ring-down spec- absorption in acetylene gas,’’ J. Opt. Soc. Am. B 13, 2041–
troscopy,’’ J. Appl. Phys. 83, 3991–3997 (1998). 2054 (1996).
7. M. D. Levenson, B. A. Paldus, T. G. Spence, C. C. Harb, J. 24. A. J. Phillips and P. A. Hamilton, ‘‘Pressure-shift of the
S. Harris, and R. N. Zare, ‘‘Optical heterodyne detection in (0, 0) and (1, 0) bands of the oxygen b 1 # #g " X 3 # "g transi-
cavity ring-down spectroscopy,’’ Chem. Phys. Lett. 290, tion from Fourier transform spectroscopy,’’ J. Mol. Spec-
335–340 (1998). trosc. 174, 587–594 (1995).
8. K. Nakagawa, T. Katsuda, A. S. Shelkovnikov, M. de La- 25. H. Naus, A. de Lange, and W. Ubachs, ‘‘b 1 # #g " X 3 # "g
bachelerie, and M. Ohtsu, ‘‘Highly sensitive detection of (0, 0) band of oxygen isotopomers in relation to tests of the
molecular absorption using a high finesse optical cavity,’’ symmetrization postulate in O-16(2),’’ Phys. Rev. A 56,
Opt. Commun. 107, 369–372 (1994). 4755–4763 (1997).
9. J. Ye, L.-S. Ma, and J. L. Hall, ‘‘Sub-Doppler optical fre- 26. G. Gagliardi, L. Gianfrani, and G. M. Tino, ‘‘Investigation of
quency reference at 1.064 "m by means of ultrasensitive the b 1 # #g ( % $ 0) % "X 3 # "g ( % $ 0) magnetic-dipole tran-
cavity-enhanced frequency modulation spectroscopy of a sitions in O-18(2),’’ Phys. Rev. A 55, 4597–4600 (1997).