Content uploaded by Richard Engeln
Author content
All content in this area was uploaded by Richard Engeln
Content may be subject to copyright.
Cavity enhanced absorption and cavity enhanced magnetic rotation
spectroscopy
Richard Engeln,a) Giel Berden, Rudy Peeters, and Gerard Meijer
Department of Molecular and Laser Physics, University of Nijmegen Toernooiveld,
6525 ED Nijmegen, The Netherlands
~Received 23 June 1998; accepted for publication 11 August 1998!
It is experimentally demonstrated that a narrow band continuous wave ~cw!light source can be used
in combination with a high-finesse optically stable cavity to perform sensitive, high-resolution direct
absorption and optical rotation spectroscopy in an amazingly simple experimental setup, using ideas
from the field of cavity ring down spectroscopy. Light from a scanning narrow band cw laser is
coupled into the cavity via accidental coincidences of the laser frequency with the frequency of one
of the multitude of modes of the cavity. The absorption and polarization rotation information is
extracted from a measurement of the time-integrated light intensity leaking out of the cavity as a
function of laser wavelength. © 1998 American Institute of Physics. @S0034-6748~98!01811-5#
I. INTRODUCTION
In the field of cavity ring down ~CRD!spectroscopy
there is a trend to use narrow band continuous wave ~cw!
light sources1–4 rather than pulsed light sources, for which
the technique has originally been developed.5,6 Not only is
the data acquisition rate in a pulsed CRD experiment limited
by the repetition frequency of the laser ~typically 10–100
Hz!, and is the spectral bandwidth in the best case Fourier
transform limited ~but typically 0.1 cm21), but the pulsed
laser systems are also rather bulky and therefore less suited
for, for instance, trace gas measurements outside of the labo-
ratory. Although pulsed lasers do have the advantage that
radiation can be produced over a wide wavelength range ~in
principle!, it should be realized that the pulse energies deliv-
ered by commercially available laser systems are actually at
the high end for what is required for CRD experiments. One
could use less powerful, and preferably more compact
~cheaper ?!, light sources that are optimized for specific
wavelength regions for these experiments as well. Ideally,
the ever more readily available single-mode tunable diode
lasers which can cover the visible and near-infrared spectral
regions are used for these experiments.
When using narrow band cw lasers for CRD spectros-
copy, one might at first think that thereby one of the biggest
advantages and probably the main reason for the enormous
success of CRD spectroscopy, i.e., the simplicity of the
highly sensitive absorption detection setup, is thrown over-
board, as now the cavity will have to be frequency locked to
the laser ~or vice versa!. Although various schemes for lock-
ing of narrow band cw lasers to optical cavities have been
successfully implemented in the past and although it is there-
fore well known how to proceed, the resulting experimental
setup will certainly always be more involved than the ‘‘con-
ventional’’ pulsed CRD setup.
Locking of the laser frequency to an external optical
cavity is the only way to proceed when a large light intensity
in the ‘‘build-up’’ cavity is needed and when efficient use of
the light source is required. This is for instance the case
when background-free detection schemes @laser induced
fluorescence ~LIF!, photoacoustic ~PA!detection, bolometer
detection, etc.#are used to monitor or detect, with the highest
possible sensitivity, species present inside ~or passing
through!the cavity. The commonly used approach for this is
to make a confocal cavity, i.e., to make the distance dbe-
tween the mirrors that build up the optical cavity identical to
the radius of curvature rof the two mirrors, and to actively
adjust the length of the cavity such as to match the frequency
of one of the longitudinal modes of the high finesse cavity to
the laser frequency. When the optical cavity is merely used
to perform efficient multipassing along a well-defined line,
as in CRD absorption spectroscopy, the only requirement for
the light intensity inside the cavity is that a sufficient amount
of light leaks out of the cavity ~per unit time!such that it can
still be detected with a good signal-to-noise ratio, without
being limited, for instance, by the noise of the detection sys-
tem. In this case, one is allowed to use the accidental coin-
cidences of the frequency of one of the modes in the optical
cavity with the frequency of the laser to couple light into the
cavity. Obviously, the efficiency of coupling light into the
cavity is determined by the multitude of modes that can be
excited in the cavity in combination with the frequency jitter
of these modes relative to the laser frequency. A cavity with
a quasicontinuum mode structure is obtained by choosing the
mirror separation dwithin the stability regime, i.e., 0,d
,r,r,d,2r. It has been shown previously that an unsta-
bilized cavity of this type enables semiefficient injection of
~narrow band!light into the cavity.1,7
In this article we report on the use of narrow band cw
light sources to perform cavity enhanced absorption ~CEA!
and cavity enhanced magnetic rotation ~CEMR!spectros-
copy. The radiation from a scanning narrow band cw laser is
coupled into the cavity via the accidental coincidences of the
laser frequency with the frequency of one of the multitude of
modes of the cavity. The absorption and polarization rotation
a!Electronic mail: richarde@sci.kun.nl
REVIEW OF SCIENTIFIC INSTRUMENTS VOLUME 69, NUMBER 11 NOVEMBER 1998
37630034-6748/98/69(11)/3763/7/$15.00 © 1998 American Institute of Physics
information is extracted from a measurement of the time-
integrated light intensity that leaks out of the cavity. From a
plot of the inverse of this intensity versus wavelength, the
direct absorption and/or optical rotation spectra are obtained.
Using diode lasers as well as a ring-dye laser, both the CEA
and the CEMR technique are demonstrated by recording the
appropriate spectra for oxygen, water, and ammonia in a cell,
as well as by recording the spectra of molecular oxygen and
ammonia in a slit-jet expansion.
II. EXPERIMENTAL SETUP AND MEASUREMENT
PROCEDURE
A scheme of the experimental setup is depicted in Fig. 1.
As a light source we have used two single-frequency cw
diode lasers ~EOSI LCU 2010A and New Focus Model
6262!as well as an Ar1-laser pumped single-frequency cw
ring dye laser ~Spectra Physics 380!. The diode module in
the EOSI laser covers the 750–780 nm region ~up to 15
mW!. The New Focus diode laser is tunable in the 1506–
1595 nm region with 5 mW maximum power. The ring dye
laser is operated on DCM to cover the spectral region of the
g
band of molecular oxygen around 628 nm and delivers
typically 200 mW. Both diode lasers have an external cavity
and can be scanned mode-hop free by piezo tuning the end
mirror. The ring dye laser is scanned mode-hop free by
changing the cavity length with two galvo-driven plates.
During all the experiments reported here the lasers are re-
peatedly scanned over a spectral range of typically 1 cm21at
a rate on the order of 5–100 Hz.
The narrow band cw laser radiation is coupled into a
high finesse stable optical cavity, formed by two planocon-
cave mirrors with a diameter of 25 mm and a radius of cur-
vature of 21 m and a specified optimum reflectivity of typi-
cally R50.9999. In the cell experiments, the mirrors act as
windows for the cell. When using short cells ~5–20 cm!, the
mirrors are directly flanged onto a stainless steel tube and no
further alignment of the mirrors relative to each other is
needed. Obviously, the cell as a whole is adjusted such as to
couple in the laser-light efficiently. To avoid optical feed-
back from the cavity to the laser, the cell is positioned under
a small angle with respect to the incoming laser beam or a
Faraday isolator is used. When using longer cavities ~25–90
cm!, independent alignment of the two mirrors is required.
The light that leaks out of the cavity is detected by either a
photodiode or a photomultiplier tube ~PMT!. The detector
signal is amplified and displayed on a digital oscilloscope.
To record wavelength spectra, the oscilloscope is used in x-y
mode, in which the horizontal axis is triggered by the voltage
ramp used to scan the laser and is therefore proportional to
the laser wavelength. In the inset of Fig. 1 the light intensity
leaking out of a 45-cm-long empty cavity is shown as a
function of the wavelength of the diode laser ~around 765
nm; bandwidth several MHz!. The horizontal axis corre-
sponds to a total frequency range of 0.115 cm21, which is
about ten free spectral ranges ~FSRs!of the optical cavity.
During the wavelength scanning, which is linear in time,
different transverse cavity modes are excited. As the ~unsta-
bilized!cavity will drift during scanning, the observed mode
pattern is not expected to repeat perfectly. It is observed that
light is coupled into the cavity at an approximate rate of 104
times per second. This rate strongly depends on the detailed
mode structure of the cavity,7in combination with the scan-
ning rate of the laser and the frequency jitter of the cavity
modes, i.e., the ‘‘unstability’’ of the cavity.
If one were to make a single scan, and were to probe the
absorption spectrum of species inside the cavity that way, it
is evident from the inset, and it has been pointed out by
others,8–10 that spectral features that are narrower than the
spacing between the modes ~approximately 50 MHz in this
particular example!would escape observation. In repeating
this procedure over the same spectral region, and summing
up the observed results, these ‘‘gaps’’ in the spectrum can be
filled up, however, via ‘‘random interleaved sampling.’’ It
should be noted at this point that each of the cavity modes is
rather narrow ~around 100 kHz in this case!, and that the
apparent width of the modes as observed in the inset is a
mere reflection of the effective bandwidth of the laser during
scanning, convoluted with the time response of the optical
cavity and the time response of the detection system. The
integrated intensity at each of the different modes depends
linearly on the finite ‘‘resonance’’ time of, a fraction of, the
spectral profile of the laser with the cavity mode. As in this
particular example the resonance time is mainly determined
by the scanning rate of the laser, which should therefore be
identical for all the different cavity modes, the remaining
observed intensity differences have to be attributed to differ-
ent ~diffraction!losses or in-coupling efficiencies for differ-
ent transverse modes. This also explains the observed repeti-
tion of the intensity pattern, in which the FSR of the cavity
can still be recognized.
In a ‘‘standard’’ cw CRD experiment, the absorption
information is deduced from the time dependence of the in-
tensity buildup and/or decay of an optical cavity when tuned
into and/or away from resonance with the laser frequency.
For this, a triggering system is required that actively controls
when data taking has to start, together with fast detection
FIG. 1. Scheme of the experimental setup. Narrow band radiation from a
rapidly scanning cw laser is coupled into an optical stable cavity when
accidentally coincident with one of the cavity modes. The time-integrated
intensity is detected and displayed on a digital oscilloscope. The inset shows
a typical time-resolved intensity pattern as recorded at the exit of a 45-cm-
long cavity, while scanning the laser over a 0.115 cm21wide spectral re-
gion.
3764 Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.
electronics.2–4 From the observed time dependence the abso-
lute value of the ‘‘ring-down’’ time
t
~
n
!is determined,
which can be expressed as11,12
t
~
n
!5d
c~12R1
k
~
n
!d!.~1!
In an experiment, the total cavity loss 1/c
t
~
n
!is plotted as a
function of frequency, as this is directly proportional to the
absorption coefficient
k
~
n
!, apart from an offset which is
mainly determined by the finite reflectivity of the mirrors. To
record a complete spectrum, the laser is slowly scanned or
stepped from one frequency to the other, much like in a
pulsed CRD experiment.
In the cavity enhanced absorption method reported here,
the laser is rapidly scanned in time and the signal on the
detector is integrated over a time that, with a given scanning
rate of the laser, is matched to the expected width of the
spectral lines. Scanning the laser over 1 cm21ata10Hz
repetition rate with an integration time of 1 ms, corresponds
to a spectral integration over only 0.01 cm21. Several iden-
tical scans are summed on the oscilloscope to improve the
measurement statistics. The data on the scope are transferred
via general purpose interface bus ~GPIB!to a PC for further
analysis. When the inverse of the time integrated detector
signal is plotted versus the wavelength of the laser ~vide
infra!absorption spectra over the full scanning range, spectra
that are of comparable quality to those obtained via ‘‘con-
ventional’’ pulsed or cw CRD spectroscopy, appear directly
on the screen in a fraction of a second!
In choosing the optimum experimental parameters sev-
eral considerations are important. The width of the individual
cavity modes, D
n
cavity , is rather small ~in the tens of kHz
range!for the high-Q optical cavities used in these experi-
ments, and is actually considerably smaller than the width of
the spectral profile of the scanning laser. We assume in the
following that the spectral profile of the laser is a block func-
tion with a width D
n
laser@D
n
cavity and is scanned linearly in
time. When the laser is slowly scanned into resonance with a
cavity mode, light intensity will gradually build up in the
cavity. The exact time dependence of this process can be
described using, for instance, the formalism as outlined by
Zalicki and Zare in Appendix B of their article,8and depends
on the details of the spectral profile of the laser, the proper-
ties of the optical cavity, and the scanning rate of the laser.
The maximum intensity that can be reached inside the cavity
is proportional to the spectral overlap of the laser profile with
the profile of the cavity mode. As the spectral profile of the
cavity mode is well approximated by a Lorentzian profile
with a width proportional to the total cavity losses, i.e., pro-
portional to 1/
t
, and with an intensity proportional to
t
2, the
maximum intensity in the cavity is directly proportional to
the ring-down time
t
. The light intensity inside the cavity
will converge to this limiting value, provided that the laser
stays in resonance with the cavity mode sufficiently long,
i.e., provided that the scanning rate is small compared to the
ratio of D
n
laser to
t
. When the laser is tuned out of resonance,
the intensity in the cavity will exponentially decay in time,
again governed by the time constant
t
.
In the upper part of Fig. 2 the calculated time depen-
dence of the light intensity behind a 15-cm-long cavity, hav-
ing round trip losses of either 231024~
t
55
m
s!or 4
31024~
t
52.5
m
s!, is shown. For these calculations a block
profile with a width of 5 MHz and a scanning rate of 0.25
MHz/
m
s('8cm
21
/s) is assumed for the laser. As the inten-
sity decay when the laser is tuned out of resonance with the
cavity mode follows a strict exp(2t/
t
) time dependence, it is
evident that the time-integrated signal behind the cavity
would be exactly proportional to
t
if the intensity buildup
would follow a @12exp(2t/
t
)#dependence. Although the
latter is not strictly true, it is explicitly shown in the lower
part of Fig. 2 that the time-integrated intensity nevertheless
follows a linear
t
dependence to a good approximation, and
that this approximation gets better with lower scanning rates.
Experimentally there is also a lower limit to the scanning
rate, which is set by the requirement that all cavity modes
should be in resonance with the laser more-or-less equally
long, as otherwise large intensity fluctuations will occur.
This implies that the scanning rate has to be significantly
higher than the rate at which the cavity modes are jittering,
or, alternatively, that one has to stabilize the cavity suffi-
ciently to fulfill this requirement also for lower scanning
rates. Even with the unstabilized optical cavities we used,
FIG. 2. Upper part: Calculated light intensity at the exit of a 15-cm-long
optical cavity as a function of time. In the calculation, the spectral profile of
the laser is assumed to be a block profile with a width of 5 MHz, scanned at
a rate of 0.25 MHz/
m
s. Calculations are performed for round trip cavity
losses of 231024~
t
55
m
s!as well as for losses that are twice as big.
Lower part: Calculated integrated light intensity exiting the 15-cm-long cav-
ity as a function of the photon lifetime
t
in the cavity, for three different
values of the scanning rate of the laser.
3765Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.
there is a large ‘‘window’’ of scanning rates available in
which both requirements are fulfilled. It is evident from the
curves in the upper part of Fig. 2 that the light is actually
coupled into the cavity as efficient as when the cavity is
actively locked to the laser. This, combined with the strongly
relaxed requirements on the light intensity behind the cavity
as now the total time-integrated signal is used to extract the
absorption information from, makes the power levels of
commercially available diode lasers more than sufficient for
these experiments.
III. RESULTS AND DISCUSSION
In Fig. 3 a part of the absorption spectrum of the
b1(g
1(v852)←X3(g
2(v950) band of 16O2~
g
band!, show-
ing the bandheads of the RRand RQbranches, is shown as
recorded with the cw ring dye laser in a 12-cm-long cell
filled with 200 mbar of molecular oxygen at room tempera-
ture. The spectrum is a compilation of three partly overlap-
ping measurements, each covering about 1.5 cm21averaged
over 100 scans. With a scanning rate of the laser of around 5
Hz, this implies an effective recording time of one minute. In
the vertical direction, the inverse of the time-integrated in-
tensity behind the cavity is plotted, with the baseline denoted
as zero. The value of the baseline is proportional to (1
2R)/d, but, contrary to CRD experiments, the absolute
value of this quantity is not directly determined in this ex-
periment. Therefore, the intensity scale of the spectrum is
expressed relative to the baseline intensity. This relative ab-
sorption spectrum can obviously be put on an absolute scale
if the ring-down time of the empty cavity is known via some
other way. If we do the reverse, and extract the effective
mirror reflectivity Rfrom the measured spectrum using the
calculated population distribution of ground-state molecular
oxygen and the known absorption cross sections for these
transitions,13 we deduce a reflection coefficient R50.9998,
in good agreement with independent CRD measurements.12
The relative line intensities as observed in the spectrum
match calculated absorption spectra very well, thereby ex-
plicitly demonstrating the viability of the data extraction pro-
cedure. It is worth noting that the noise level on the baseline
of the spectrum is at the 1023level, as good as can be ob-
tained in standard CRD experiments. As mentioned in the
previous section, the time axis of the digital oscilloscope
which is used to average the traces is triggered by the ramp
voltage used for scanning the laser. To accurately determine
the absolute frequency position as well as to be able to cor-
rect for possible nonlinearities in the scanning, the well-
known absorption spectrum of I2is recorded simultaneously.
The linewidth of the oxygen transitions is 0.050 cm21,
which is as expected from convoluting a Doppler broadened
profile with a width of 0.035 cm21with a pressure broad-
ened profile with a width of, under our experimental condi-
tions, 0.024 cm21.14
One can of course always deduce absolute absorption
coefficients from the CEA spectra, when the species whose
absorption cross section or density is not known is measured
simultaneously with a species with a known absorption co-
efficient. In Fig. 4 part of the CEA spectrum of a mixture of
1.0 mbar H2O and a trace amount of NH3is shown, as mea-
FIG. 3. Cavity enhanced absorption spectrum as recorded with 200 mbar
molecular oxygen in a 12-cm-long cell at room temperature, displaying the
bandheads in the RRand RQbranches of the b1(g
1(v852)←X3(g
2(v9
50) transition.
FIG. 4. Part of the CEA spectrum of a mixture of H2O with a trace amount
of NH3at a total pressure of 1.0 mbar, measured in a 90-cm-long cell at
room temperature. Transitions I and II originate from H2O the others origi-
nate from NH3.
3766 Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.
sured with the diode laser around 1518 nm in a 90-cm-long
cell. Due to the lower reflectivity of the mirrors used in this
setup (R'0.9992), the laser could be scanned over the ap-
proximate 0.7 cm21region at a rate of 100 Hz. The light that
leaks out of the cavity is detected by a InGaAs photodiode
detector and amplified with a 0.1 ms rise time amplifier. The
spectrum shown is the average over 1000 consecutive scans
and is acquired in 10 s. The H2O transitions are the 651
←542 ~I!and the 652←541 ~II!rotational transitions of the
~1,1,3!←~0,0,0!vibrational band, and have integrated ab-
sorption cross sections of 1.13310224 and 3.39
310224 cm21cm2/mole, respectively.15 Peaks III to V in the
spectrum are all due to NH3but are as yet unassigned. The
absorption cross section of peak V has been reported, how-
ever, as 1.21310221 cm21cm2mole,16 and a partial pressure
of ammonia in our cell of 1.731023mbar is thus deduced.
There are other NH3lines in the tuning range of this diode
laser with an order of magnitude larger cross section. In view
of the signal-to-noise ratio of the current spectrum, partial
pressures of 1026mbar of ammonia will be detectable in this
setup on these more favorable transitions.
The applicability of the CEA detection technique to mo-
lecular jets, is demonstrated here by presenting absorption
spectra of individual rotational lines of the b1(g
1(v850)
←X3(g
2(v950) band of expansion cooled molecular oxy-
gen, using a diode laser around 762 nm. A detailed spectro-
scopic study on jet-cooled ammonia, of which CEA spectra
have been recorded in the 1.5
m
m region, will be reported
elsewhere. A planar jet is formed by expanding a gas mixture
of 30% 16O2in Ar through a 40 mm30.03 mm slit nozzle.
At a stagnation pressure of 760 Torr a 1200 m3/h rootspump
~Edwards EH1200!backed by a 180 m3/h rotary pump ~Ley-
bold SV180!reaches a background pressure of 0.5 Torr.17
The mirrors that form the optical cavity are now positioned
10 cm apart, with the cavity axis being along the long axis of
the slit nozzle, intersecting the jet within a few mm from the
orifice. In Fig. 5 the CEA spectrum of the PP1(1) transition
of molecular oxygen as measured in this planar jet is shown.
The spectrum is recorded by averaging over 500 scans with
the laser scanning the 0.25 cm21spectral region at a 15 Hz
rate, i.e., the spectrum is recorded in approximately 30 s. A
narrow line with a full width at half maximum of 270 MHz
is seen on top of a broad background. The background signal
is mainly attributed to thermalized oxygen gas in the optical
cavity outside of the expansion region. Contrary to CRD
absorption spectroscopy, the CEA method is sensitive to the
intensity profile of the diode laser over the spectral region
that is scanned, which in the present case gives an additional
minor contribution to the structured background signal. The
observed linewidth is in accordance with the width of 80
MHz that is observed in direct absorption measurements on
C2H4in a similar planar jet at roughly a factor of four longer
wavelengths.17 The observed line width of 270 MHz is less
than 20% of the FSR of the optical cavity that is used, and
the spectrum therefore exemplifies that the random inter-
leaved sampling due to the frequency jitter of the unstabi-
lized, multimode, cavity can be rather efficient indeed.
The CEA detection technique reported here, can be used
to measure optical rotation spectra, when a polarization ana-
lyzer is placed in front of the detector. We have recently
theoretically outlined and experimentally demonstrated that
polarization spectroscopy can be combined with pulsed CRD
spectroscopy. In particular, we demonstrated that the optical
rotation of molecular oxygen placed in a magnetic field can
be sensitively and quantitatively determined in such a pulsed
polarization dependent CRD detection scheme.12 Using this
same model system, i.e., the PP1(1) transition of the
g
band
of molecular oxygen in a magnetic field, we here demon-
strate that sensitive cavity enhanced magnetic rotation spec-
troscopy can be performed as well. The experimental CEMR
setup is only slightly different from the CEA setup described
earlier. Prior to entering the optical cavity the laser beam
passes through a Glan–Thompson polarizer with an extinc-
tion of 1025to better define the polarization state of the
incoming light. The 12-cm-long optical cavity is filled with
200 mbar of oxygen and placed inside a magnet which pro-
duces a homogeneous magnetic field up to 0.88 T over the
whole length of the cell, perpendicular to the axis of the
optical cavity ~Voigt configuration!. The polarization of the
light that enters the cavity makes an angle
f
B545° relative
to the direction of the magnetic field B. A second Glan–
Thompson polarizer is placed between the end mirror of the
cell and the detector. The polarization direction of the light
that is transmitted by this analyzer makes an angle
f
D, rela-
tive to the polarization direction of the incoming light, and
can be set with an accuracy of 0.1°.
In Fig. 6 the inverse of the time-integrated light intensity
passing through the analyzer is shown as a function of fre-
quency in the spectral region of the PP1(1) transition of the
b1(g
1(v852)←X3(g
2(v950) band of molecular oxygen
for seven different values of the angle
f
D. Each spectrum is
an average over 100 scans, measured witha5Hzrepetition
FIG. 5. CEA spectrum of the PP1(1) transition of the b1(g
1(v850)
←X3(g
2(v950) band of 16O2measured in a 10-cm-long optical cavity
positioned around a slit-jet expansion. The observed line width of 270 MHz
is due to residual Doppler broadening in the jet, and is less than 20% of the
FSR of the optical cavity.
3767Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.
rate, and takes thus only 20 s to record. The vertical scale is
the same for all seven spectra, and is only indicated as such
for the uppermost spectrum, again in units relative to the
baseline intensity. We demonstrated in our previous study
that for detection angles
f
Dsufficiently far away from the
crossed geometry, exponentially decaying transients result
which are exclusively sensitive to polarization dependent ab-
sorptions ~magnetic dichroism!.12 Both the upper two and the
lower two spectra shown in Fig. 6 meet this requirement
reasonably well. To interpret these spectra, we can therefore
directly adapt Eq. ~6!of Ref. 12 to our present situation, as
the time-integrated intensity will again be proportional to the
ring-down time for the specific analyzer direction
f
Din this
case. We thus find that the inverse of the time-integrated
intensity is given by a term that is approximately propor-
tional to @
k
i
(
n
)2
k
'(
n
)#/cos(
f
D) on top of a baseline, in
which
k
i
(
n
) and
k
'(
n
) correspond to the absorptions polar-
ized parallel (DM561; outer two peaks!and perpendicular
(DM50; central peak!to the magnetic field, respectively.12
This equation explains the observed sign difference between
the two types of transitions, the intensity increase upon ap-
proaching
f
D590° as well the overall sign change upon
passage through the crossed polarizer geometry.
The central spectrum in Fig. 6, recorded with the crossed
polarizer geometry, exclusively shows the effect of optical
rotation due to dispersion. If the polarizers are indeed exactly
crossed, it is obvious that any polarization rotation is ex-
pected to lead to an increase in the time-integrated intensity
behind the cavity, contrary to the observation where both
increasing and decreasing signals are observed. The observa-
tions can therefore only be explained when a slight polariza-
tion rotation induced by the mirrors is taken into account as
well. If we assume that, apart from the contribution of the
magnetic circular birefringence of molecular oxygen, the
mirrors introduce a small phase shift between the two mutu-
ally orthogonal polarization directions, the observed spec-
trum can be well understood. The remaining two spectra,
recorded 1° away from the crossed polarizer configuration,
contain all the possible contributions to polarization rotation
that have been mentioned up to now, and are a weighed sum
of the spectra that have been discussed.
We have demonstrated a simple yet sensitive experimen-
tal scheme to record high-resolution optical absorption
and/or optical rotation spectra of molecules in open air, cells
or jets, using ideas from the field of CRD spectroscopy.
Light from a narrow band cw laser is coupled into a high-
finesse stable optical cavity when accidentally coincident
with one of the multitude of modes of this cavity. While
rapidly scanning the narrow band laser, the time-integrated
intensity exiting the cavity is monitored. Under the condi-
tions that the scanning laser is sufficiently long in resonance
with one of the cavity modes that the light intensity inside
the cavity approaches its limiting value, the time-integrated
light intensity exiting the cavity is in a good approximation
proportional to the ring-down time
t
. Direct absorption spec-
tra and/or optical rotation spectra can therefore be obtained
by plotting the inverse of the time-integrated intensity versus
the laser frequency. With a single mode laser scanning over
typically a 1 cm21spectral region scanning rates of 5–100
Hz have been employed in this study, yielding high quality
spectra in a matter of seconds. The noise equivalent detec-
tion limit readily approaches 1023of the baseline intensity,
fully comparable to the best CRD spectra reported to date.
As the spectral information is deduced from the time-
integrated signal rather than from the time dependence of the
signal it is possible to perform these measurements with rela-
tively low power lasers and cheap detection systems.
Note added in proof. A fully equivalent experimental
approach is to scan the laser slowly and to rapidly scan the
length of the cavity, linearly and repeatedly, by mounting
one cavity mirror on a piezo electric crystal. The same con-
ditions as mentioned in the paper, i.e., that the laser fre-
quency is sufficiently long in resonance with one of the cav-
ity modes that the light intensity inside the cavity approaches
its limiting value, has to be fulfilled. Although this method is
experimentally slightly more involved, it circumvents pos-
sible problems associated with averaging nonperfectly
matched frequency scans. Since the submission of this paper
we have experimentally verified that the latter approach can
be used as well.
Obviously, measurement of the time-integrated light in-
tensity behind the cavity can also be used in a pulsed CRD
experiment to determine
t
,18 since
*
0
`I0exp(2t/
t
)dt5I0
t
.
However, the large pulse-to-pulse fluctuations in the inten-
sity behind the cavity (I0) will limit the SNR ratio that can
be obtained. In the recently reported ‘‘~time-!integrated cav-
ity output analysis’’ scheme18 normalization is accomplished
FIG. 6. Cavity enhanced magnetic rotation spectra of the PP1(1) transition
of the b1(g
1(v852)←X3(g
2(v950) band of molecular oxygen as recorded
in a magnetic field of 0.88 T ~Voigt configuration!using a 12-cm-long
optical cavity filled with 200 mbar oxygen, measured as function of the
angle
f
D.
3768 Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.
by measuring I0in a time-resolved measurement, making
this scheme identical to existing schemes in which the cavity
decay time is determined using a two-channel boxcar
averager.19
ACKNOWLEDGMENTS
This work is part of the research program of the ‘‘Stich-
ting voor Fundamenteel Onderzoek der Materie ~FOM!,’’
which is financially supported by the ‘‘Nederlandse Organi-
satie voor Wetenschappelijk Onderzoek ~NWO!.’’ The work
is supported in part by the National Institute of Public Health
and the Environment ~RIVM!. We acknowledge Dr. J. Oo-
mens for making the slit-expansion setup available to us for
these experiments.
1R. Engeln, G. von Helden, G. Berden, and G. Meijer, Chem. Phys. Lett.
105, 262 ~1996!.
2D. Romanini, A. A. Kachanov, N. Sadeghi, and F. Stoeckel, Chem. Phys.
Lett. 264, 316 ~1997!.
3B. A. Paldus, J. S. Harris, J. Martin, J. Xie, and R. N. Zare, J. Appl. Phys.
82, 3199 ~1997!.
4Y. He, M. Hippler, and M. Quack, Chem. Phys. Lett. 289, 527 ~1998!.
5A. O’Keefe and D. A. G. Deacon, Rev. Sci. Instrum. 59, 2544 ~1988!;A.
O’Keefe, J. J. Scherer, A. L. Cooksy, R. Sheeks, J. Heath, and R. J.
Saykally, Chem. Phys. Lett. 172, 214 ~1990!.
6M. D. Wheeler, S. M. Newman, A. J. Orr-Ewing, and M. N. R. Ashfold,
J. Chem. Soc., Faraday Trans. 94, 337 ~1998!, and references therein.
7G. Meijer, M. G. H. Boogaarts, R. T. Jongma, D. H. Parker, and A. M.
Wodtke, Chem. Phys. Lett. 217, 112 ~1994!.
8P. Zalicki and R. N. Zare, J. Chem. Phys. 102, 2708 ~1995!.
9J. T. Hodges, J. P. Looney, and R. D. van Zee, J. Chem. Phys. 105, 10 278
~1996!.
10 K. K. Lehmann and D. Romanini, J. Chem. Phys. 105, 10 263 ~1996!.
11 R. T. Jongma, M. G. H. Boogaarts, I. Holleman, and G. Meijer, Rev. Sci.
Instrum. 66, 2821 ~1995!.
12 R. Engeln, G. Berden, E. van den Berg, and G. Meijer, J. Chem. Phys.
107, 4458 ~1997!.
13 The absolute intensities of the rotational lines in Fig. 3 can be calculated
using the absorption cross section of the PP(1) transition determined by
CRD spectroscopy @from Fig. 3~a!of Ref. 12#and the relative intensities
from a calculation of the rotational line strengths ~Ref. 14!.
14 K. J. Ritter and T. D. Wilkerson, J. Mol. Spectrosc. 121,1~1987!.
15 HITRAN ~acronym for high resolution transmission molecular absorption!
database ~http://www.hitran.com!.
16 L. Lundsberg-Nielsen, F. Hegelund, and F. M. Nicolaisen, J. Mol. Spec-
trosc. 162, 230 ~1993!.
17 J. Oomens, L. Oudejans, J. Reuss, and A. Fayt, Chem. Phys. 187,57
~1994!.
18 A. O’Keefe, Chem. Phys. Lett. 293, 331 ~1998!.
19 D. Romanini and K. K. Lehmann, J. Chem. Phys. 99, 6287 ~1993!.
3769Rev. Sci. Instrum., Vol. 69, No. 11, November 1998 Engeln
et al.