Intensity decay of pulsed laser sources in a highly reflective cavity permits cavity ringdown laser absorption spectroscopy to measure sub-parts-per-million levels of gaseous species.
J. B. Paul, J. J. Scherer, A. O`Keefe, and R. J. Saykally
Cavity ringdown laser absorption spectroscopy is a technique capable of making ultrasensitive (<1 ¥ 10-6 fractional absorption) direct absorption measurements on microsecond timescales with easy-to-use, commonly available pulsed lasers. Invented in its present form for use with pulsed tunable lasers by O`Keefe and Deacon almost 10 years ago,1 cavity ringdown laser absorption spectroscopy (CRLAS) is now being widely used for measuring electronic and vibrational absorption spectra of trace species in gas-phase environments. It is a promising tool for monitoring processes such as combustion dynamics.
Although state-of-the-art continuous wave (CW) absorption techniques such as frequency modulation and intracavity methods can meet or exceed this sensitivity for static absorption measurements, they are commonly difficult to implement, exhibit a seriously degraded sensitivity for transient events, and cannot access many regions of the spectrum available with pulsed lasers. In contrast, CRLAS is easy to implement and interpret, and it is quite generally applicable. Moreover, because CRLAS is a direct absorption technique, the spectra can yield species concentrations in a straightforward manner, subject to the usual conditions of Beer`s Law.2
Conventional absorption techniques typically involve the measurement of a very small change in the total transmitted intensity of a light source through an absorbing medium, leading to a high background condition that commonly results in low sensitivity. However, CRLAS is based on measuring the intensity decay rate of a light pulse trapped in an optical cavity formed by two highly reflective mirrors (R = >0.999). Consider the case in which the light pulse is physically shorter than the cavity round-tri¥length (see Fig. 1). The small amount of light transmitted into the cavity through the entrance mirror is trapped for some period of time as it reflects back and forth, slowly decaying in intensity due to the finite losses--primarily from mirror transmission--of the cavity. A detector (either a photomultiplier tube or a photodiode) monitors this event via the light transmitting through the exit mirror at each reflection.
The decay envelope is exponential due to the constant fraction of the pulse intensity lost for each pass through the cavity, as shown in the following equation:
where R is the mirror reflectivity and tr is the optical round-tri¥time. For a very good set of mirrors (R = 0.9999) the pulse will undergo about 5000 round-trips before the pulse intensity decays to 1/e of its initial value. Due to this large number of passes, Eq. 1 is essentially a first-order differential equation with a first-order exponential solution. Although this differential approximation ignores the high-frequency components of the signal associated with the cavity round-tri¥time, for a typical 0.5-m cavity this 300-MH¥component is easily filtered (assuming the detector is fast enough to observe it) to reveal a smooth decay.
Because tr is just the measured cavity round-tri¥length (2L) divided by the speed of light (c), determining the "cavity ringdown time" (t = L/c(1-R)) gives the per-pass optical loss from the cavity, which for an evacuated cavity is almost entirely due to the finite mirror reflectivity. (Incidentally, this technique was originally developed specifically for measuring very high mirror reflectivities, and to date remains the most effective way to do so.) Note that the decay rate is independent of the the initial pulse intensity, permitting the use of typical pulsed lasers possessing shot-to-shot intensity variations of u¥to 20%.
If an absorbing sample is introduced into the cavity such that the absorption follows Beer`s law for a single pass of the laser pulse through the medium, this absorption (A = 1- e-acl) simply adds to the per-pass cavity loss, resulting in a shorter ringdown time. The solution to Eq. 1 then becomes
Thus, the single-pass absorption spectrum of the sample is obtained by measuring t for each wavelength increment, and solving
A = tr/2 (1/t - 1/to) (3)
where to is the empty-cavity ringdown time. In practice, it is not always necessary to determine to independently because it is slowly varying with respect to most spectroscopic features and is therefore easily subtracted by eye or by fitting to a smooth function.
The more familiar frequency-domain picture of Lorentzian cavity modes separated by c/2L is recovered directly by Fourier transforming the idealized pulse train (see Fig. 2). The temporal pulse spacing transforms to the longitudinal cavity mode spacing, the exponential cavity ringdown decay to the Lorentzian longitudinal mode profiles, and the temporal pulse profile to the frequency envelope encompassing the cavity modes (for transform-limited pulses). This picture is only valid for the case of a length-and temperature-stabilized cavity for which the injected light is perfectly aligned and mode-matched to the longitudinal cavity modes. Therefore, in CRLAS, the pristine cavity mode in the figure is not generally applicable because the regions between the longitudinal modes are somewhat filled in by transverse modes and the mode profiles are considerably broadened. The standard criterion for an optical cavity not acting as a spectrometer requires the incident pulse bandwidth to exceed the cavity free-spectral range (FSR).
The time-domain analog of this condition is that the coherence time of the light source must be shorter than the optical round-tri¥time of the cavity. Satisfying this condition ensures that the per shot laser power accepted by the cavity is essentially independent of the laser center frequency. This allows the laser to be scanned continuously in frequency without the large variations in transmitted intensity normally associated with etaloning.
Scherer and coworkers have shown that for the highly stable cavity geometries (that is, between planar and confocal) used in CRLAS, laser bandwidths that are considerably narrower than the cavity FSR can easily be used with no detrimental effects.3 The reasons for this behavior have been the subject of much recent discussion.4-6 Nevertheless, it is important to realize that even when intensity variations do occur, as long as Beer`s Law holds for a single pass of the laser pulse through the sample, only the peak intensity of the exponential decay is affected, not the decay time itself.
Wide wavelength coverage of CRLAS
Following the initial implementation of CRLAS in the visible region for spectroscopic purposes (see photo on cover and on p. 71), it has been extended into both the ultraviolet (UV) and mid-infrared (IR) regions. In the UV, the primary limitation is the lack of mirrors with reflectivities greater than 99%. Nevertheless, researchers have made impressive use of available resources (in some cases using excimer laser mirrors) to achieve absorption sensitivities of 5-25 ppm. A Mirage optical parametric oscillator (Continuum; Santa Clara, CA) was used to ex tend the technique into the mid-IR (3-4 µm).3 The moderately high power of this laser system proved useful, as the excellent mirrors we obtained (R = 0.9999) transmit through the cavity only 10-8 of the incident light.
While this high reflectivity is generally not a problem in the UV/visible region, where highly sensitive photomultiplier tubes are effective, available laser power can become an issue in the IR where less-sensitive indium antim onide (InSb) and mercury cadmium telluride (MCT) detectors must be used. At present, the current limitation to moving deeper into the IR does not appear to be mirror technology, but rather the availability of either moderately powerful, tunable pulsed light sources or extremely sensitive detectors that could be used in conjunction with less-intense lasers. However, the current rate of development in both of these areas practically ensures that the necessary technology required to build a CRLAS spectrometer that operates continuously from below 200 nm to beyond 15 µm will be developed in the coming years.
Spectroscopic results span spectrum
In its present form, CRLAS has already proven to be a very powerful tool. At Berkeley, our CRLAS spectrometer has a multipass hydrogen Raman shifter in conjunction with a standard Nd:YAG-pumped pulsed dye laser (see Fig. 3). This system will in principle cover the entire spectrum from 400 nm to about 8 µm. Adding a doubling crystal assembly would extend this coverage to 200 nm.
Prior to developing the infrared capability, we operated this system in the visible and near-UV to measure electronic absorption spectra for several metal clusters (Cu2, Cu3, Al2), and metal silicides (CuSi, AgSi , AuSi, PtSi) generated by laser ablation in pulsed supersonic jets (see Fig. 4). In this work we exploited the fast measurement time of CRLAS, as the ringdown time roughly matched the transit time (about 20 ms) of the pulsed molecular beam through the optical cavity, providing a nearly 100% duty cycle.
Recent efforts have focused on the region from 2.6 to 4.0 µm, primarily studying the O-H stretching vibrations of water clusters in supersonic jet ex pan sions. In these experiments, the inherent ability of CRLAS to provide accurate band-intensity information across broad regions of the spectrum has allowed both the determination of the absolute number density of small water clusters [(H2O)n, n = 2-5] in the supersonic expansion and the characterization of a broad feature closely resembling the spectrum of laboratory amorphous ice, which is formed when water is slowly deposited on a cryogenic substrate at 10 K (see Fig. 5).
T. Yu and M. C. Lin at Emory University (Atlanta, GA) have also exploited the microsecond signal-acquisition timescale to pioneer chemical kinetics measurements with CRLAS. These studies, performed in flow tube reactors, de monstrated that as long as the reaction times are slower than the cavity ringdown time, accurate determination of first-order rate constants is straightforward. Yu and Lin initially studied the reactions of phenyl radicals with several molecules by monitoring the absorption of reactants and/or products in the visible for various delay times following a photolysis event that initiates the reactions by creating the radicals. They have since used this method extensively to study various other reactions.
O`Keefe first demonstrated the power of CRLAS for probing plasma environments by measuring the spectrum of N2+ in a glow discharge near 400 nm. Very recently, this work was extended to hollow cathode discharges by Maier and coworkers (Univ. of Basil, Switzerland),7 who reported densities of the N2+ ion near 1011 cm-3. In the UV, Gerard Meijer and coworkers at the Catholic University of Nijmegen, The Netherlands, have used CRLAS to measure the concentration of the OH radical in flames, while at Stanford University (Stanford, CA) the technique has been used to measure the radial distribution of methyl radicals surrounding a 1600 K tungsten filament by monitoring the B-X Rydberg transition at 216 nm, also in the UV.
In the mid-IR, Scherer and associates at Sandia National Laboratories have used the technique to study the rovibrational spectra of OH in flames. The IR-CRLAS provides more detail than Fourier transform spectroscopy for measuring weak absorptions in flames (see Fig. 6), evidencing the impressive sensitivity of the method. Recently, Scherer and Rakestraw detected the HCO radical in a low-pressure methane/oxygen/nitrogen flame via its A-X electronic transition in the visible region.8
One promising application of CRLAS appears to be for trace gas detection. O`Keefe and Lee utilized CRLAS with a 45-cm-long cavity to demonstrate that the ambient concentration of the common pollutant NO2 could be detected at atmospheric concentrations below 1 ppb, certainly far below the 3 ppm limit considered hazardous.9 More recently, Meijer and others demonstrated a detection limit of 1 ppt for mercury, measuring the electronic transition at 216 nm and using a cavity length similar to the one in the NO2 study mentioned above. In this latter work it was also noted that this sensitivity could be further improved using kilohertz-repetition-rate lasers, as the duty-cycle is vastly increased in this case.
The simplicity and versatility of the cavity ringdown method also presents some interesting possibilities for further technical developments. Because the optical cavity does not function as a spectrometer (Fabry-Perot interferometer), the possibility of developing a broadband version of CRLAS is apparent. Meijer and coworkers have made the first progress in this direction, combining a ringdown cavity with a step-scan Fourier-transform spectrometer to spectrally resolve the light as it leaves the cavity.10 Conversely, during a recent visit to his laboratory, Meijer demonstrated a high-resolution version of CRLAS based on measurement of the phase shift in an amplitude-modulated CW laser induced by the cavity, as originally developed by Herbelin and others.11 While much of the convenience and simplicity of the pulsed-laser approach is lost, this phase-shift method may become quite useful for some applications.
Cavity ringdown laser absorption spectroscopy is a rapidly advancing technique, as indicated by the variety of results from recent applications of this method. The simplicity, sensitivity, and good time resolution of this method create exciting possibilities for future applications. Furthermore, CRLAS directly yields species concentrations limited only by the usual considerations that apply to absorption spectroscopy.
The Berkeley CRLAS effort is supported by the Air Force Office of Scientific Research and the National Science Foundation.
1. A. O`Keefe and D. A. G. Deacon, Rev. Sci. Instrum. 59, 2544 (1988).
2. For a more detailed discussion of the material presented here and a complete list of references, please refer to the following review articles: J. J. Scherer, J. B. Paul, A. O`Keefe, and R. J. Saykally, Chemical Reviews (in press, 1996); J. J. Scherer et al., Advances in Metal and Semiconductor Clusters, M. A. Duncan, ed., Chapter 4 (JAI Press, Greenwich, CT, 1996); J. J. Scherer et al., Spectrosc. 11, 46 (1996).
3. J. J. Scherer et al., Chem. Phys. Lett. 245, 273 (1995).
4. J. T. Hodges, J. P. Looney, and R. D. Van Zee, J. Chem. Phys. 105, 10278 (1996).
5. D. Romanini and K. K. Lehmann, J. Chem. Phys. 105, 10263 (1996).
6. J. Martin et al., Chem. Phys. Lett. 258, 63 (1996).
7. M. Kotterer, J. Conceicao, and J. P. Maier, Chem. Phys. Lett. 259, 233 (1996).
8. J. J. Scherer and D. J. Rakestraw, Chem. Phys. Lett. (in press).
9. A. O`Keefe and O. Lee, American Laboratory 21, 19 (1989).
10. R. Englen and G. Meijer, Rev. Sci. Instrum. 67, 2708 (1996).
11. J. M. Herbelin et al., Appl. Opt. 20, 3341 (1980).
In the cavity ringdown laser absorption spectrometer at Berkeley, a frequency-doubled pulsed Nd:YAG laser pumps a Lambda Physik dye laser. This 640-nm red output is Raman-shifted to the infrared to measure parts-per-million concentrations.
FIGURE 1. To perform cavity ringdown laser absorption spectroscopy, a pulse of laser light is injected into a stable optical cavity in which the mirror spacing is longer than the pulse coherence length, and as the pulse circulates in the cavity, a very small fraction of it is transmitted through the mirrors upon each reflection. Detector after the cavity observes an exponentially decaying pulse train that is digitized and fit to a first-order exponential expression, which gives the total cavity losses, including intracavity absorption.
FIGURE 2. Fourier transformation of train of pulses exiting cavity reveals more-familiar spectral picture of narrow cavity modes equally spaced in frequency. Mapping three basic structures between time- and frequency-domain counterparts shows that narrow structures in time lead to broad structures in frequency (and vice versa). Pulse time profile (for a transform-limited pulse) transforms laser frequency bandwidth (a); pulse spacing in time maps to integral multiples of cavity fundamental frequency (b); long exponential decay maps to Lorentzian logitudinal cavity modes (c).
FIGURE 3. The Berkeley infrared cavity ringdown laser absorption spectrometer starts with about 30 mJ of tunable light from a Nd:YAG laser-pumped dye laser. Raman shifting (3rd Stokes) in a 3.5-m 25-pass Herriot cell containing 200 psi of hydrogen gas converts the dye output into the IR, with about 0.4 mJ of tunable IR radiation that is filtered and directed into the CRLAS optical cavity. Cavity mirrors serve as windows of vacuum chamber. Light coming through the rear mirror is focused onto a 2-mm-diameter InSb detector; resultant signal is amplified (50X), digitized, and averaged for 20 subsequent shots in a digital storage oscilloscope, and the signal then is sent to a PC to be fit to an exponential expression in real time. The PC also controls the scanning of the dye laser and plots the spectrum as obtained.
FIGURE 4. Vaporizing a silver target in the presence of silane (SiH4) seeded carrier gas produces diatomic silver silicide (AgSi) in a supersonic expansion; rovibronically resolved C-X band spectrum (R and ¥components) was observed near 380 nm.
FIGURE 5. Pulsed supersonic expansion can create small water clusters; the IR-CRLAS spectrum of O-H-stretches can be attributed to different-sized clusters.
FIGURE 6. More detail is resolvable in the IR-CRLAS spectrum of a 25-torr oxyacetylene flame than in a FT-IR spectrum of the same sample.