Broadband cavity ringdown spectroscopy of the NO3 radical

Article abstract of DOI:10.1016/S0009-2614(01)00573-5

Cavity ringdown spectroscopy (CRDS) has been demonstrated using a broadband (20 nm) laser source and a two-dimensional clocked detector array. Absorption spectra of dilute samples (50-500 parts per trillion) of the nitrate radical, NO₃, have been obtained between 650 and 670 nm by monitoring simultaneously the time and wavelength resolved output of a ringdown cavity. The potential of broadband CRDS for making measurements on samples containing multiple absorbers (e.g., atmospheric samples) is shown by applying analysis methods from differential optical absorption spectroscopy to quantify the NO₃ concentration in the presence of nitrogen dioxide impurities.

Full text of DOI:10.1016/S0009-2614(01)00573-5

6
July 2001
Chemical Physics Letters 342 +2001) 113±120
www.elsevier.com/locate/cplett
Broadband cavity ringdown spectroscopy of the NO radical
3
*
Stephen M. Ball , Ian M. Povey, Emily G. Norton, Roderic L. Jones
University Chemical Laboratory, University of Cambridge, Lens®eld Road, Cambridge, CB2 1EW, UK
Received 30 March 2001; in ®nal form2 May 2001
Abstract
Cavity ringdown spectroscopy +CRDS) has been demonstrated using a broadband +20 nm) laser source and a two-
dimensional clocked detector array. Absorption spectra of dilute samples +50±500 parts per trillion) of the nitrate
3
radical, NO , have been obtained between 650 and 670 nm by monitoring simultaneously the time and wavelength
resolved output of a ringdown cavity. The potential of broadband CRDS for making measurements on samples
containing multiple absorbers +e.g., atmospheric samples) is shown by applying analysis methods from di€erential
optical absorption spectroscopy to quantify the NO
ties. Ó 2001 Elsevier Science B.V. All rights reserved.
3
concentration in the presence of nitrogen dioxide impuri-
1
. Introduction
Cavity ringdown spectroscopy +CRDS) is now
resolved ringdown signals. Coupled with analysis
techniques fromconventional long path di€eren-
tial optical absorption spectroscopy [6] +DOAS),
this development will provide the basis for ®eld
measurements of a wide range of atmospheric
trace species.
well established as a method for making sensitive
and quantitative measurements of the absorption
due to a host of gas phase species [1±3]. The var-
ious ways in which CRDS has been incorporated
into experimental methodologies have recently
been reviewed by Berden et al. [1], who also pro-
vide a comprehensive tabulation of the species
probed by the technique up to June 2000. Gener-
ally these methods employ narrow band lasers as
probes, and necessarily so where the sensitivity of
CRDS is optimised by matching the laser radia-
tion to a single Fabry±Perot mode of the cavity
2. Broadband CRDS
The electronic transitions used to probe certain
atmospheric species [6] give rise to absorption
spectra much broader +several nanometres) than
the spectral width of laser sources commonly em-
À1
ployed [1] in CRDS +<1 cm ). In order to use
[
4,5]. In this Letter we report on a novel develop-
conventional CRDS to monitor such absorbers, it
would be necessary to acquire ringdown decays
sequentially at a number of discrete wavelengths
+for example, [7]). However, the atmosphere is a
complex mixture of absorbing and scattering spe-
cies, which may vary rapidly in time, and whose
overlapping absorptions must be identi®ed and
separately quanti®ed if measurements are to yield
ment of CRDS which instead uses a broadband
probe laser and a charge coupled device +CCD)
camera for the simultaneous detection of spectrally
*
Corresponding author. Fax: 44-1223-336362.
E-mail address: smb45@cam.ac.uk +S.M. Ball).
0
009-2614/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 0 9 - 2 6 1 4 + 0 1 ) 0 0 5 7 3 - 5

114
S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
valid results. As DOAS has very e€ectively dem-
onstrated, the con®dence with which the various
absorbers can be quanti®ed is greatly increased
when absorption data are obtained simultaneously
at a large number of points distributed over a wide
wavelength range.
A few investigators have employed broadband
sources in their CRDS studies. Engeln and Meijer
À1
[
8] injected the output of a broadband ꢀ400 cm †
dye laser into a CRDS cavity and employed a
Fourier transforminterfero me ter to record the
resulting multi-exponential intensity decays as a
function of the displacement of the interferome-
ter's mirror. Several hours of acquisition was re-
quired to yield time dependent spectral intensity
distributions. A second group, Crosson et al. [9],
used multiple pulses from a free electron laser +25
nmFWHM centred at 5.38 lm). A monochro-
mator was employed to resolve the cavity output
into separate narrow +0.03 nm) wavelength ranges
incident sequentially on a single HgCdTe detector.
Finally, Scherer [10] devised a technique termed
Fig. 1. The wavelength dependence of the broadband laser
source +left axis) and the measured mean re¯ectivity of the
CRDS mirrors used in this study +right axis). The heavy solid
3
line is the NO absorption cross-section [11] and dashed line is
the zeroth order resolution of the instrument +both left axis).
3. The broadband CRDS measurement procedure
and apparatus
`
ringdown spectral photography' in which a ro-
The broadband laser source and the detector
array used here are more usually employed in a
di€erential absorption LIDAR experiment [16] to
probe the altitude dependence of water vapour,
tating mirror disperses the cavity output onto a
two-dimensional detector array. However, whilst
the potential for using a broadband source and
dispersing the cavity output in both time and
wavelength was noted, the technique was demon-
strated using only a sequence of discrete wave-
lengths froma narrow band laser.
This Letter reports a novel variation of CRDS
that resolves the cavity output in both time and
wavelength simultaneously. Absorption spectra
have been recorded under laboratory conditions
over a range of 20 nmemploying a broadband dye
laser source and a two-dimensional detector array.
The species chosen with which to demonstrate the
3
NO and other absorbers in the troposphere.
Pulsed broadband radiation was generated
froma dye laser +Spectron SL4000B) pumped by a
Nd: YAG laser +Spectron SL803-G) operating at
20 Hz. The dispersive element was removed from
the dye laser cavity permitting lasing in a contin-
uous range of some 20 nm according to the gain
curve of the dye, a mixture of DCM and LDS698
dyes dissolved in dimethyl sulfoxide and methanol.
The spectrally resolved output of the source is also
shown in Fig. 1. Less than 5 mJ of broadband
radiation was incident on ringdown cavity's input
mirror. Since the cavity was optically stable with a
quasicontinuum mo de structure [17], no special
precautions were taken to match the characteris-
tics of laser to the cavity.
technique is the nitrate radical, NO , which pos-
3
sesses a broad absorption band [11±13] centred at
6
nighttime oxidation [13] and hence removal of a
range of gases released into the atmosphere, and
has previously been studied in situ by broadband
optical absorption techniques including DOAS
[6,14], zenith sky absorption spectroscopy [15] and
broadband LIDAR [16]. NO has also been pro-
bed by CRDS in the laboratory using a narrow
band diode laser [7].
3
62 nm, see Fig. 1. Atmospheric NO initiates the
The ringdown cavity consisted of two plano-
concave mirrors separated by 95 cm. Samples en-
tered the cavity through a side armin a glass tube
+34 mm internal diameter) positioned between the
mirrors, and exited through gaps of <3 mm be-
tween the ends of the tube and the mirror mounts.
3

S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
115
Relatively large sample ¯ows +ꢁ5 l/min) were
employed to prevent ambient air from contami-
nating the cavity mirrors. The mirrors +Los Gatos
Research) had a diameter of 20 mm and a 6 m
radius of curvature. The average re¯ectivity of the
mirrors measured in this work, Rꢀk†, ranged from
where ‰x
i
Š is the concentration of species x
i
and
r
i
ꢀk) is its wavelength dependent absorption +or
scattering) cross-section. The quantity r Š is
ꢀk†‰x
the absorbance of species x per centimetre path.
i
i
i
Calculation of sꢀk) and hence Rꢀk) assumed that
the only attenuation introduced by the nitrogen
purge gas was Rayleigh scattering, assumed to be
the same as in air [18].
9
9.995% at 650 nmto 99.997% at 670 nm, and is
shown in Fig. 1. The re¯ectivity data were ob-
tained using the equation
The noise seen in the re¯ectivity data of Fig. 1
re¯ects the accuracy with which ringdown times
were measured for each of the detector's 512
wavelength bins. Note that the noise is greatest
towards 670 nmwhere the laser output is lowest.
Certain other spikes arise fromisolated defects in
the detector and appear with the same coordinates
on the detector array for all spectra. A more so-
phisticated analysis procedure would increase the
quality of the measurements by excluding data
fromdefective pixels.
Simultaneous wavelength and time detection
was achieved using an astigmatic imaging Czerny
Turner type spectrometer +Chromex 250 IS) and a
charge coupled device +CCD) detector. The CCD
chip +EEV 37-10) was an array of 512 Â 1024
square pixels and was masked to protect unex-
posed regions. Light fromthe cavity was collected
d
sꢀk† ˆ
ꢀ1†
cj ln Rꢀk†j
fromvalues calculated for the ringdown ti me of
the empty cavity as a function of wavelength, sꢀk).
The other quantities in Eq. +1) are the cavity
length, d, and the speed of light, c. No experiments
were performed on an evacuated cavity; rather,
empty cavity ringdown times were calculated from
0
ringdown times, s ꢀk), measured when the cavity
was purged with dry, ®ltered nitrogen +see also
Fig. 2a). The presence of an absorbing +or scat-
tering) sample reduces the ringdown time of the
cavity according to the equation
X
1
1
ˆ
‡ c
r ꢀk†‰x Š;
ꢀ2†
i
i
s0ꢀk† sꢀk†
i
Fig. 2. The measured time and wavelength dependent ringdown spectrum for +a) the cavity ¯ushed with dry, ®ltered nitrogen and +b)
for sample 1 +470 pptv of NO ). The arrow marks the position of the 662 nm peak of the NO absorption feature.
3
3

116
S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
by a lens positioned directly behind the output
mirror and conveyed through an optical ®bre into
the spectrometer where it was dispersed in wave-
length. The spectrumwas then directed to the
CCD detector where, owing to the imaging prop-
erties of the spectrometer, it fell on 10 rows of
pixels. The detector converts photons to charge
with an eciency of approximately 40% at 650 nm.
Suitable clocking of the CCD pixel voltages [16]
then allowed the charge accumulated in the ex-
posed pixels to be moved down the chip to unex-
posed regions, allowing subsequent CRDS spectra
to be captured in the exposed rows of the CCD. At
the clocking rate used +0.5 ls per row), the 10
exposed pixel rows thus correspond to a time res-
olution of 5 ls, at least a factor of 10 shorter than
ringdown times measured here. The combined ef-
fect of the dispersion and the clocking was to
create a two-dimensional +time/wavelength) spec-
trumon the CCD +see Fig. 2). An additional
bene®t of the approach is that by reverse clocking
the charge, the two-dimensional spectrum can be
reset to its starting position and the process re-
peated allowing many spectra to be integrated di-
rectly on the CCD chip prior to readout.
NO
composition of N
3
was prepared in situ fromthe thermal de-
within the glass tube enclos-
2
O
5
ing the ringdown cavity. A ¯ow of 0.5±5 l/min of
dry, ®ltered nitrogen was passed through a glass
trap containing N O at 196 K. It is unlikely that the
2
5
partial pressure of N O entrained into the ¯ow
2
5
reached its saturated vapour pressure because the
residence time of the carrier gas in the trap was a few
2 5
seconds at most. The N O mixing ratio in the
sample ¯ow was adjusted by diluting the trap ¯ow
with further dry, ®ltered nitrogen to give a total ¯ow
2 5
of 5 l/min. The N O trap was bypassed for exper-
iments where the cavity was purged with nitrogen.
The N O sample was prepared by a standard
2
5
method of the oxidation of NO +99.0%, Linde
Gases, UK) by excess ozone +ca. 5% in oxygen) in
2 5
a glass ¯ow reactor. The N O product was col-
lected and stored in the glass trap at 196 K. The
trap was ¯ushed for at least 30 min prior to ex-
periments, but otherwise the sample was used
without puri®cation.
4. Results and discussion
The spectral resolution of the instrument was
measured with the spectrometer set to zeroth or-
der. The representative zeroth order spectrumof
the source shown in Fig. 1 is approximated well by
a Gaussian function of 0.45 nmFWHM, and is
substantially narrower than the 3.5 nmFWHM of
Figs. 2a and b are three-dimensional represen-
tations of the measured cavity output versus time
and wavelength for, respectively, the cavity ¯ushed
with nitrogen and containing a sample of NO in
3
nitrogen +sample 1). Each data set took approxi-
mately 12 min to acquire and is the average of 25
spectra, each spectrumbeing the result of 255 laser
shots integrated on the CCD chip. For clarity of
presentation, the signals recorded by the 512
wavelength columns of the CCD and the 96
clocked rows containing the ringdown decay have
3
the NO band. Any artefact attributable to in-
strument resolution can be further excluded [2]
because data were acquired only for the ®rst 50 ls
0
of the decay: s ꢀk† P 50 ls at all wavelengths
studied. This limited clocking procedure is a ves-
tige of the detector's use in tropospheric LIDAR
experiments [16].
been
averaged
into
50 Â 24
bins
of
0:398 nm  2:00 ls. The zero time spectra are es-
sentially the same as the spectrally resolved output
of the laser shown previously in Fig. 1.
The CCD clocking rate was determined from
the number of pixel rows illuminated by the 655
nmoutput of a laser diode mo dulated with a
square wave of various frequencies. A clocking
rate of 0:502 Æ 0:002 ls per row was deduced from
a linear ®t to the number of illuminated pixel rows
plotted against the modulation period of the laser
diode. The 10±90% rise/fall times of the laser diode
were measured to be <5 ns using a photodiode and
a digital oscilloscope.
It is clear fromFig. 2a that the cavity output
decays with time after the laser pulse. Further-
more, the decay is slower at long wavelengths
where the re¯ectivity of the CRDS mirrors is
higher +see also Fig. 1). The same basic behaviour
is repeated in Fig. 2b, which additionally exhibits
an absorption feature around 662 nmthat grows
more prominent at later times. The feature present

S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
117
in both ®gures near 654 nmresults froma defect in
the CCD detector's mask, and occurs in all time
bins for the a€ected wavelengths. Ringdown times
were calculated fromlinear least squares ®ts to the
natural logarithmof the wavelength resolved in-
tensity versus time. Thus three-dimensional spec-
tra such as those in Fig. 2 yield sets of 512
wavelength resolved ringdown times. The CDRS
absorption spectrumof the sample is then calcu-
lated using Eq. +2) where sꢀk) represents the set of
and long wavelength tails. The spectrumof the
underlying absorber must therefore also be struc-
tured: if it were unstructured, the NO feature
3
would merely be superposed on a wavelength in-
dependent o€set and would have a di€erential
structure identical to the NO absorption band
3
alone. Furthermore, the continuum seems to be
due to another absorber within the sample. The
above analysis has already accounted for the
wavelength dependence of the mirror re¯ectivity,
and the e€ect is unlikely to be caused by contam-
ination of the mirrors because the mean value of
the continuumabsorption correlates with the
sample concentration and not the order in which
the spectra were recorded.
0
ringdown times for the ¯ushed cavity and s ꢀk) the
set for the cavity plus sample. Note that the con-
tribution of Rayleigh scattering cancels because
nitrogen is employed as both the purge and the
carrier gas.
Fig. 3 shows a representative broadband CRDS
+sample 2). The data
NO
absorber, it and NO
in the ringdown cavity. Indeed, the concen-
tration of NO is likely +and is indeed found be-
low) to substantially exceed that of NO because
2
is the obvious candidate for the continuum
spectrumof a sample of NO
3
3
being in equilibriumwith the
were placed on an absolute wavelength scale by
comparison with the wavelength dependence of
the NO absorption cross-section [11]. The di€er-
2 5
N O
2
3
3
ential structure in the measured spectrum is clearly
attributable to the characteristic 662 nmband of
NO superposed on a continuumabsorbance of
approximately 1:5 Â 10 cm . Preliminary anal-
ysis of several CRDS spectra indicated that the
di€erential structure did not reproduce the exact
+1) the reactive NO radical is lost readily on col-
3
lisions with surfaces and +2) the N O precursor
2
5
almost certainly contains an NO2 impurity re-
maining from its preparation and the vapour
3
À8
À1
pressure of the NO
order of magnitude greater than that of N
at 196 K. The absorption cross-section [19] of
NO
has values in the range 0.7±1.5 Â
2 2 4
dimer, N O , is roughly an
2
O
5
3
shape of the NO absorption band. In fact, the
di€erential was always too ¯at in the band's short
2
À20
2
À1
10
cm molecule between 650 and 670 nm.
Thus absorption due to the NO impurity should
2
only comprise a small part of the measured
absorbance at 662 nmwhere the absorption cross-
section of NO
larger than that of NO
found to account for the majority of the absor-
bance at wavelengths in the tails of the NO band.
N O itself and another probable impurity, nitric
3
is some three orders of magnitude
2
, but the NO impurity was
2
3
2
5
acid, both have absorption bands [12] peaking at
k < 200 nm, but their absorption between 650 and
670 nmare likely to be completely negligible.
Methods analogous to those commonly used to
analyse DOAS measurements [6,16] were em-
ployed to remove the contribution of the under-
2 3
lying NO absorption and to determine the NO
concentration. First, the wavelength dependent
absorption due to an initial amount of NO was
Fig. 3. The upper trace is the CRDS absorption spectrumof
sample 2 overlaid by the absorption spectrum calculated for a
2
calculated using literature values [19] of the NO2
3 2
mixture of 220 pptv NO and 31 ppbv NO +heavy solid line).
The lower trace is the residual spectrumo€set by 4  10À8 cmÀ1
.
absorption cross-section and was subtracted from

118
S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
the measured absorbance per centimetre of the
sample. The result was plotted against the NO
the small uncertainty in the CCD clocking rate
+0.4%), or the measurement precision of the ring-
down times +typically 1% around 655 nm, but 5%
3
absorption cross-section, and a linear ®t to the
data obtained. The ®tting procedure was iterated
until the value selected for the NO concentration
2
at 670 nm). The NO mixing ratios that produce
the zero intercept in Fig. 4 are 46, 31 and 26 ppbv
for samples 1, 2 and 3, respectively.
Fig. 3 also shows the absorption spectrumcal-
2
resulted in a ®tted intercept of zero indicating that
the NO2 contribution had been completely re-
moved from the CRDS spectrum. The gradient of
culated for a mixture of 220 pptv NO
NO overlaid on the CRDS spectrumof sample 2.
3
and 31 ppbv
this ®nal ®t yields the NO concentration directly
in units of molecules cm . The results for three
representative samples are presented in Fig. 4,
3
2
À3
+The calculation assumed mirror re¯ectivities de-
rived fromsmoothing the data in Fig. 1.) There is
excellent agreement between the measured and
calculated spectra in Fig. 3. The residual spectrum
shows no evidence of other absorbers, and has a
3
where the ®tted NO concentrations have been
converted to mixing ratios at 298 K and 760 Torr.
The uncertainties quoted for the NO mixing ra-
3
À9
À1
tios are merely those in the ®tted gradients. They
contain no contributions fromthe uncertainty in
standard deviation of 4:7 Â 10 cm . This value
is almost a factor of ®ve greater than the baseline
À9
À1
the NO
3
cross-section +Æ 12.5% at 662 nm[12]),
loss of 1 Â 10 cm quoted by King et al. [7] for
their CRDS measurement of NO . But the best
3
sensitivity shown by the present broadband
method is somewhat better than that anticipated
for the diode laser study [7] +1.6 versus 3 pptv)
because here the NO concentration is deduced
3
from512 points spanning the whole absorption
band rather than fromthe di€erence in absorption
measured at two wavelengths on/o€ resonance. In
À9
À1
other words, the value of 4:7 Â 10 cm quanti-
es the scatter in the data of sample 2 about its
best ®t line in Fig. 4, but the precision with which
NO ] is determined from the gradient of the ®t is
®
[
3
far better than the uncertainty in the absorbance
measured at any individual data point.
The standard deviations of residual spectra
2
calculated assuming as above that NO was the
continuumabsorber were al mo st always signi®-
cantly smaller than values returned from an al-
Fig. 4. Plots of the measured CRDS absorbance for three
samples versus the NO absorption cross-section [11] at the
3
ternative analysis conducted assuming
a
corresponding wavelength. The absorbances of samples 1 and 2
have been o€set vertically so that data at wavelengths where the
structureless, sloping continuum. Moreover, the
standard deviation of the residual fromthe usual
analysis also tended to its minimum value for the
3
NO absorption is weak are not obscured. The absorbances of
samples 1 and 3 are shown on the upper and lower portions of
the left axis, respectively, and those of sample 2 on the right
axis. The wavelength dependent contribution fromabsorption
2
same amount of NO that produced the zero in-
tercept in the plots like Fig. 4. This evidence,
whilst compelling, does not conclusively identify
by an NO
the data sets +heavy lines) have an intercept of zero ± see text.
The gradients of the ®tted lines give the NO concentrations
directly +shown converted to mixing ratio at 298 K and 760
Torr). The uncertainty in the NO concentration is the standard
2
impurity has been removed so that the linear ®ts to
3
NO
scale di€erential structure in the NO
2
as the continuumabsorber because the ®ne
absorption
2
3
spectrumis too weak to be observed for the [NO 2]
of the present study. The use of DOAS analysis
methods does however demonstrate that broad-
band CRDS can quantify the overlapping ab-
error in the gradient of the ®t, and contains no contribution
from the measurement precision of the ringdown times or the
larger Æ12:5% uncertainty in the absolute value of the NO
3
absorption cross-sections.

S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
119
sorption features of multiple species, and could be
used to resolve for example, water vapour, aerosol
and other absorbers present in atmospheric sam-
ples that would be dicult to separate by an on/o€
resonance technique.
To further validate the analysis procedure, the
full time and wavelength dependence of the
broadband CRDS measurement was reconstructed
3 2
using the ®tted NO and NO amounts. Fig. 5a
shows the time and wavelength resolved ratio of the
measured output intensity of the ¯ushed cavity and
the cavity containing sample 1, i.e., the ratio of the
data shown in Figs. 2a and b. The spectrumat time
t is equivalent to the transmission of sample 1 for a
path length of t  c +where c is the speed of light).
The data in Fig. 5a are normalised to unity in the
®
rst time bin to remove the e€ects of a small change
in the laser intensity between Figs. 2a and b,
though such intensity ¯uctuations do not a€ect
measurement of the ringdown times. Fig. 5b shows
the output intensity ratio calculated fromEq. +3),
ratioꢀt; k† ˆ expꢀÀt  c Â
X r
i
ꢀk†‰x
i
Š†;
ꢀ3†
i
where the summation includes the absorbances
due to 470 pptv NO and 46 ppbv NO . Fig. 5c
3
2
shows the residual, i.e., Fig. 5a minus Fig. 5b. The
residual contains only noise which, like the resid-
ual spectrumof Fig. 3, is most pronounced at long
wavelengths where the laser output is weak. Again,
it would seemthat the wavelength dependent los-
ses due to the cavity mirrors and the NO
sample have been correctly quanti®ed.
3 2
=NO
5. Application to a ®eld instrument
Broadband CRDS has been shown able to
quantify NO in laboratory samples +50±500 pptv)
3
with a precision of a few pptv. In addition, over-
lapping absorptions due to multiple species, in this
Fig. 5. +a) The time and wavelength resolved ratio of the
measured output intensity of the ¯ushed ringdown cavity and
the cavity containing sample 1, and +b) the ratio calculated for a
3 2
mixture of 470 pptv NO and 46 ppbv NO . The residual dif-
ference between the data of +a) and +b) is shown in +c).

120
S.M. Ball et al. / Chemical Physics Letters 342 (2001) 113±120
case NO
vidually quanti®ed. Measurements of NO
ambient air report [14] values in the range 5±20
pptv, although considerably higher amounts, ꢂ100
pptv, have been observed in urban environments
3
and NO
2
, have been separated and indi-
in clean
porting this work in the formof a NERC Ad-
vanced Research Fellowship +SMB) and a `Small
Research Grant' GR9/04411.
3
[
15,16]. The sensitivity of the current instrument is
therefore already sucient to detect atmospheric
NO provided that ringdown times in atmospheric
References
[1] G. Berden, R. Peeters, G. Meijer, Int. Rev. Phys. Chem. 19
+
3
2000) 565.
samples are similar to those in laboratory samples.
However, scattering by atmospheric aerosol is
likely to reduce signi®cantly the e€ective path length
accessible with a CRDS ®eld instrument. Prelimi-
nary measurements on samples of urban air indicate
that ringdown times of the present 95 cm cavity are a
factor of 3 to 4 smaller than those of the ¯ushed
cavity between 650 and 670 nm. Nevertheless,
ringdown times of 35 ls +equivalent to a 10.5 km
path) have been demonstrated around 660 nm in
samples of urban air contained in a 140 cm cavity
employing the same mirrors as used above. Indeed it
may be possible to determine ringdown times more
accurately than above by employing a dedicated
CCD clocking procedure to acquire the complete
ringdown decay: the present instrument collects
data for only the ®rst 50 ls of the decay which is
approximately one ringdown period at 650 nm and
just over half a ringdown period at 670 nm. The
authors estimate that a ®eld instrument with a 50 mJ
broadband laser source, a 1.5 mcavity and a defect-
[
[
2] P. Zalicki, R.N. Zare, J. Chem. Phys. 102 +1995) 2708.
3] S.M. Newman, I.C. Lane, A.J. Orr-Ewing, D.A. Newn-
ham, J. Ballard, J. Chem. Phys. 110 +1999) 10 749.
4] R.D. van Zee, J.T. Hodges, J.P. Looney, Appl. Optics 38
+1999) 3951.
[
[
5] T.G. Spence, C.C. Harb, B.A. Paldus, R.N. Zare, B.
Willke, R.L. Byer, Rev. Sci. Instrum. 71 +2000) 347.
6] U. Platt, Phys. Chem. Chem. Phys. 1 +1999) 5409.
7] M.D. King, E.M. Dick, W.R. Simpson, Atmos. Environ.
[
[
3
4 +2000) 685.
[8] R. Engeln, G. Meijer, Rev. Sci. Instrum. 67 +1996)
708.
2
[
9] E.R. Crosson, P. Haar, G.A. Marcus, H.A. Schwettman,
B.A. Paldus, T.G. Spence, R.N. Zare, Rev. Sci. Instrum. 70
+
10] J.J. Scherer, Chem. Phys. Lett. 292 +1998) 143.
1999) 4.
[
[11] W.J. Marinelli, D.M. Swanson, H.S. Johnston, J. Chem.
Phys. 76 +1982) 2864.
[
12] W.B. DeMore, S.P. Sander, D.M. Golden, R.F. Hampson,
M.J. Kurylo, C.J. Howard, A.R. Ravishankara, C.E.
Kolb, M.J. Molina, Chemical kinetics and photochemical
data for use in stratospheric modeling, Evaluation Number
12, JPL publication 97-4, Jet Propulsion Laboratory,
Pasadena, 1997.
[
13] R.P. Wayne, I. Barnes, P. Biggs, J.P. Burrows, C.E.
Canosa-Mas, J. Hjorth, G. Le Bras, G.K. Moortgat, D.
Perner, G. Poulet, G. Restelli, H. Sidebottom, Atmos.
Environ. 25A +1991) 203.
free CCD would sample atmospheric NO with the
3
same 2 pptv sensitivity demonstrated above but in a
reduced acquisition time of 2.5 min. The reduced
acquisition time is made possible principally be-
cause the use of a more intense light source would
allow more photons to be collected per CCD pixel
per unit time, especially near 670 nm where the
output of the present source is weak.
[14] B.J. Allan, N. Carslaw, H. Coe, R.A. Burgess, J.M.C.
Plane, J. Atmos. Chem. 33 +1999) 129.
[
15] S.R. Aliwell, R.L. Jones, J. Geophys. Res. 103 +1998)
719.
5
[
16] I.M. Povey, A.M. South, A. t'Kint de Roodenbeke, C.
Hill, R.A. Freshwater, R.L. Jones, J. Geophys. Res. 103
+
1998) 3369.
[
17] G. Meijer, M.G.H. Boogaarts, R.T. Jongma, D.H. Parker,
A.M. Wodtke, Chem. Phys. Lett. 217 +1994) 112.
18] M. Nicolet, Planet. Space Sci. 32 +1984) 1467.
19] A.C. Vandaele, C. Hermans, P.C. Simon, M. van Rooz-
endael, J.M. Guilmot, M. Carleer, R. Colin, J. Atmos.
Chem. 25 +1996) 289.
Acknowledgements
[
[
The authors would like to acknowledge the
Natural Environment Research Council for sup-