6. Single Wavelength Detection Spectroscopy for gas phase Raman spectroscopy
Peter C. Chen
Department of Chemistry,
Atlanta, GA 30314
In comparison to other forms of spectroscopy, Raman spectroscopy relies on a relatively weak scattering process. Consequently, Raman spectrometers typically use an intense fixed single wavelength source (e.g., a laser) to illuminate the sample, followed by sensitive multiwavelength detection such as a wavelength tunable system, interferometer, or multichannel system. During the 1960s, HeNe lasers, diffraction gratings, and photomultiplier tubes became commercially available, paving the way for Raman spectroscopy using a tunable monochromator with sensitive photoelectric detection. During the 1980s, Raman spectral analysis using a Fourier Transform spectrometer was introduced, primarily to overcome difficulties with fluorescence interference by working in the near infrared region. The multiplex advantage is important in Fourier Transform Raman spectroscopy because of the low sensitivity of detectors in this region. Spectrometers consisting of a single monochromator with a sensitive multichannel array detector (e.g., diode array or charged coupled device (CCD)) have also grown popular in recent years. The use of a simple single monochromator provides superior throughput over the double or triple monochromator, while the CCDs provide high sensitivity, low noise, and a multiplex advantage. These advantages and improvements in multiwavelength detection are particularly beneficial in systems where the Raman signal is weak.
One disadvantage of the multiwavelength detection approach in Raman spectroscopy is the susceptibility to spectral interference from a number of different sources and over a broad range of wavelengths. Raman signals are typically orders of magnitude weaker than fluorescence, which can originate from the cell windows or optics, the analyte itself, or other species in the sample. Several methods have been developed for overcoming fluorescence, including Fourier Transform Raman, quenching additives, burning out, and fast gating methods. However, these methods are not universal, are limited in effectiveness, and may introduce unwanted effects. A second source of spectral interference is background light, which may be problematic when working in environments where control of ambient light is not possible. This situation is typical in combustion or flame analysis, as well as monitoring applications in the presence of solar or room light. The intensity of light at various wavelengths may vary dramatically with the source and over time, and may be sufficient to damage sensitive photoemissive detectors. The third source of spectral interference is elastic scattering (i.e., Rayleigh scattering), which can cause interference for low frequency spectra. Peaks from low frequency motions appear very close to the Rayleigh peak and are easily dwarfed by its size. Unlike the triple monochromator, more recently developed techniques often require an additional optical device to remove Rayleigh scattering. Although these filters may efficiently reduce Rayleigh scattered light, they often remove part or all of the nearby low frequency Raman scattering as well.
During the 1960s, coherent Raman spectroscopy emerged as an alternative method for generating Raman spectra. The most widely used form, Coherent Antistokes Raman Spectroscopy (CARS) was recognized as being fluorescence-free because the signal is blue-shifted relative to the lasers. The selection rules and detection limits in the gas phase (>0.1%, due to a non-zero nonresonant background) are similar to that of conventional Raman. The fact that the output is generated as a high intensity beam provides a high signal-to-interference ratio that makes the technique attractive for applications such as gas phase combustion diagnostics. However, in addition to requiring multiwavelength detection, the technique requires a tunable laser. Therefore, the technique has suffered from the paucity of broadly tunable lasers; dye lasers are tunable over only a few hundred wavenumbers.
The purpose of this paper is to discuss an alternative approach called single wavelength detection spectroscopy (SWDS) for obtaining gas phase Raman spectra. Instead of using a single fixed wavelength laser and multiwavelength detection, this approach uses a unique broadly tunable source (the optical parametric oscillator) with detection at a single fixed wavelength. The SWDS approach reduces the risk of spectral interference because light is no longer detected over a broad range of wavelengths. SWDS can be used in conjunction with coherent Raman spectroscopy as a means for generating interference-free Raman spectra from gas phase samples, where Raman scattering is particularly weak.
Optical Parametric Oscillator
Conventional vibrational Raman spectroscopy covers a range of roughly 3000cm-1 by scanning the detection system. SWDS, however, involves scanning the wavelength of the light source to cover this range while the detection system (e.g., monochromator w/ PMT) remains fixed at a single wavelength. In order to accomplish this feat, a broadly tunable source of coherent light is required.
The most widely tunable optical source of coherent light is the optical parametric oscillator (OPO). The OPO is a device that was first demonstrated  in 1965, but has suffered from the poor quality of optical materials and pump lasers. Improvements in laser and optical technology have solved such problems, leading to rapid commercial development of OPO’s during the 1990s. In fact, the OPO is not a laser but a nonlinear optical device that is pumped by a laser (e.g., a Nd:YAG or a Ti:Sapphire laser). For example, several companies offer beta-barium borate (BBO) OPOs that, when pumped at 355nm (3nNd:YAG ,the third harmonic of a Nd:YAG laser), produce a narrow bandwidth (<0.2cm-1) output that is tunable over a range of roughly 450nm to 1800nm. Tuning may be achieved by changing the angle or temperature of the BBO crystal in the OPO. The resulting range is almost 2 orders of magnitude greater than that covered by that of the traditional dye laser, and provides light in regions inaccessible by laser dyes. New spectroscopic applications of the OPO have recently been reviewed elsewhere [2,3].
Unlike most other sources of light, the OPO actually generates two complementary beams called the signal beam (higher frequency) and the idler beam (lower frequency). For a 355nm pumped BBO OPO system, the signal covers the region from around 450 – 710nm, while the idler covers from around 710nm – 1800nm. These two beams have a unique inverse synchronous behavior; any change in the frequency of the signal beam results in a corresponding equal but opposite change in the frequency in the idler beam. The sum of the idler and signal frequencies must equal the (constant) pump frequency, as described by the equation 3nNd:YAG = nsignal + nidler.
SWDS with coherent Raman spectroscopy
The inverse synchronous behavior of the OPO provides the means for producing a fixed single wavelength output when used to drive a CARS signal. CARS is a multi-beam technique that differs experimentally from conventional Raman spectroscopy in three aspects. First, instead of using a single fixed wavelength input source, CARS uses three intense input beams (with frequencies n1, n2, and n3) that are spatially and temporally overlapped in the sample to generate Antistokes Raman light at n4 = n1 – n2 + n3. A spectrum is produced by monitoring the intensity of n4 as a function of the difference frequency n1 – n2 (see Figure 1). Second, the angles of the input beams are carefully controlled to ensure conservation of momentum in a process called phasematching. Proper phasematching of the 3 input beams allows the signal n4 to be emitted as an intense laser-like output beam. The third difference is that CARS requires the use of at least one tunable laser, since the peaks are achieved when n1 – n2 = nRaman. In the past, pulsed tunable dye lasers have been used to produce n1 or n2. The limited tuning range of dye lasers, however, has unfortunately prevented CARS from being used to produce complete Raman spectra.
Figure 1.Energy level diagrams for Raman and CARS.
Raman involves monitoring the intensity of the emitted light at frequency n2 as a function of n1 – n2. CARS involves monitoring the intensity of the emitted light at frequency n4 as a function of n1 – n2.
Simultaneous use of both OPO signal and idler beams for CARS extends the tuning range to cover the entire vibrational region  and allows the output signal to be generated at a single fixed wavelength for SWDS. Using the idler beam for n1 and the signal beam n3, the CARS equation for the output becomes n4 = nidler – n2 + nsignal = 3nNd:YAG – n2. Using a fixed source (e.g., the fundamental 1064nm beam from the Nd:YAG laser) for n2 causes the output n4 to be fixed at 3nNd:YAG – nNd:YAG = 2nNd:YAG. Therefore, the inverse synchronous behavior of the OPO signal and idler beams allows the CRS process continuously emit light at a frequency of 2nNd:YAG, corresponding to a wavelength of 532nm, regardless of the frequency of the signal and idler beams.
The experimental layout for the spectrometer is shown in Figure 2, and is described in more detail elsewhere [5, 6]. The main components include a Nd:YAG pumped OPO system and a (double) monochromator that is set to 532nm. The signal and idler beams from a Spectraphysics MOPO 730 are focused into the sample, along with a fixed wavelength beam (e.g., light from the Nd:YAG pump laser (1064nm), from a dye laser, or from another OPO). The shot-to-shot fluctuations of the laser system is around 10%, so spectral averaging is used to reduce the effects of such fluctuations. A 0.257m double monochromator (Oriel MS257) in additive mode with two 2400lines/mm gratings is set at 532nm, with entrance and exit slits set at 10 microns in order to act as a narrow bandpass filter. The resulting spectral width (FWHM) of the monochromator is 0.025nm (0.9cm-1). The signal is detected and monitored using a 1p21 photomultiplier tube, a Stanford Research Systems preamplifier and boxcar integrator, and a personal computer.
Figure 2. A simplified experimental layout of the SWDS spectrometer with collinear phasematching (top), involving an OPO, double monochromator, temporal delay, focusing lens, sample, and single wavelength detection system. If a folded BOXCARS phasematching configuration is used (bottom), the output may be spatially filtered to remove elastically scattered light.
For gas phase samples, the beams of light may be aligned in two ways (see Figure 2). The first option is to align all three input beams so that they are collinear. The output beam at 532nm is then generated in a direction that is collinear with the input beams. This approach allows tuning of the OPO over a broad range (>>4000cm-1) while monitoring the desired signal at 532nm, and is appropriate for vibrational spectroscopy. As the wavelength of the OPO approaches 532nm (for probing low frequency shifts, i.e. rotational spectroscopy or Brillouin scattering) elastically scattered light from the OPO may begin to enter the detection system. To prevent this problem, the laser beams may be aligned and overlapped in the sample at angles that are sufficiently large to permit spatial separation of the output beam from the input beams. Strategies such as BOXCARS and folded BOXCARS [7, 8], use two dimensional and three dimensional phasematching approaches for calculating such angles.
Overcoming spectral interference
The use of the OPO for SWDS not only fixes the tunability problem, but also provides a way to improve rejection of spectral interference. First, as with CARS, fluorescence is no longer a problem because the output signal is blue-shifted relative to the input lasers. The output beam is also intense and coherent, making it possible to use a spatial pinhole filter to remove unwanted light. The high intensity of the coherent output beam makes it feasible to use high rejection devices such as the double monochromator, despite its low throughput efficiency. Unique to the SWDS approach is the ability to use extremely narrow slits (10m) on a double monochromator set at a single wavelength to remove unwanted light both spatially and spectrally. The Raman signal is an intense laser-like beam at 532nm and can be tightly focused through such narrow slits of a monochromator that is fixed at 532nm. Unwanted light that may be present is reduced both spatially by the narrow slits of the monochromator and spectrally by the resulting narrow bandwidth (532 + 0.025nm).
Using extremely narrow slits in SWDS helps to address applications where spectral interference would otherwise hinder conventional Raman spectroscopy that uses a multiwavelength detection approach. For example, this approach provides high level rejection of ambient room light and simulated sunlight  for applications that could include environmental Raman monitoring of gas phase species. Light present at wavelengths outside of the narrowly defined detection bandwidth (532 + 0.025nm) are subsequently rejected, so unwanted spectral peaks (e.g., due to Hg vapor lines in fluorescent lights) are never detected. Rejection of unwanted light also reduces risk of damage to the detector.
Figure 3. Comparison between the “band-stop” detection approach in conventional Raman spectroscopy and the “band-gap” detection approach in SWD spectroscopy. In conventional Raman (left side) the laser remains fixed and the scattered signal is detected over a range of wavelengths. In SWD spectroscopy (right side) the Raman signal remains at a single fixed wavelength while tuning the OPO. Unlike conventional Raman, the resolution in SWD spectroscopy is determined solely by the laser, and rejection depends upon the relatively narrow bandpass of the monochromator.
This technique is also capable of reducing incoherent Rayleigh scattering to low levels, permitting the acquisition of low frequency spectra. For conventional Raman spectroscopy, Rayleigh scattering is encountered as the detection system wavelength approaches that of the fixed wavelength laser. A common problem encountered when using a “band-rejection” filter to remove Rayleigh scattering is due to the filters’ broad bandwidth and broad cutoff range. Use of such a filter can remove light close to the Rayleigh peak, causing a loss of low frequency Raman information near the Rayleigh line (see Figure 3). On the other hand, SWDS is based upon continuously generating and detecting light a single fixed wavelength and uses no filters or mechanical devices that may reduce the desired Raman signal. The double monochromator acts as an extremely narrow 532nm “band-pass” filter with a bandwidth that can be controlled through the choice of gratings and slit widths. Since the Raman signal is always generated at 532nm, it is never attenuated, and no Raman information is lost.
Figure 4 shows results from a low frequency scan using the SWDS spectrometer on air, yielding Stokes and Antistokes rotational peaks from oxygen and nitrogen. A Rayleigh peak appears in the center of the spectrum when lOPO signal beam = lmonochromator = 532nm. In conventional Raman spectroscopy, Rayleigh scattering is much more intense than the Raman scattering. Because the Rayleigh scattering is incoherent and the Raman signal is coherent, however, the intensity of the Raman peaks can be greater than that of the Rayleigh peak.
Figure 4. Results from SWD with SSOPO CARS in air. Four spectra were averaged in order to reduce the effects of shot-to-shot fluctuations from the OPO.
The linewidths of the Raman peaks shown in Figure 4 differ from that of the Rayleigh peak. The Stokes and Antistokes Raman peaks have linewidths of <0.2 cm-1, corresponding to that of the OPO beams. These linewidths do not depend upon the monochromator. The central Rayleigh feature, however, is produced as the OPO signal beam scans through the 532nm position and has a profile (0.9 cm-1 FWHM) that is determined by the monochromator.
Since it is possible to scan the OPO across 532nm without using additional filters or mechanical devices to remove the elastic scattering, it is possible to probe down to the zero wavenumber shift level. Spectral information at near-zero frequencies include Brillouin scattering and motions with energy levels in the microwave region. Such information may be obscured by the Rayleigh peak. In order to remove the Rayleigh peak, one scan is performed using all 3 input beams, and a second scan is performed after blocking either the OPO idler beam or the 1064nm beam (but not the OPO signal beam). Spectral subtraction of these two scans can then be used to remove any effects that do not depend upon the combination of all three beams, including incoherent Rayleigh scattering. For extremely low frequency studies approaching 0cm-1 shifts, it is also important to verify that photons from the Nd:YAG laser do not cause unwanted seeding of the OPO at 532nm.
Finally, in addition to eliminating fluorescence, background, and Rayleigh interference, SWDS reduces other potential problems that may affect the quality of the spectra. For example, it eliminates the need for detector wavelength-sensitivity calibration, since the photodetector senses light at a single fixed wavelength. Self-absorption of the emitted light by the sample should also be constant if light is detected at a single fixed wavelength.
Single Wavelength Detection using an OPO is a method for producing gas phase Raman spectra that are free from spectral interference. Spectral interference is often the primary limitation to applications of Raman spectroscopy, a scattering technique that is easily overwhelmed by other processes such as fluorescence. The use of the OPO allows exceptionally broad tunability that has been much needed for coherent Raman spectroscopy, a technique that produces intense Raman signals. The unique properties of the OPO also provide a means for generating light at a single fixed wavelength, resulting in improved rejection of interference and ease of detection. Using this approach, high resolution (Dn = 0.2cm-1) interference-free Raman spectra may be obtained over a range of above 4000cm-1 down to 0cm-1, suitable for vibrational, rotational, and Brillouin studies.
Drawbacks to this approach include the shot-to-shot variations in the laser system and the relatively low repetition rate (10Hz). The acquisition of several long spectra (e.g., 3600cm-1) taken with high resolution (e.g., a change of 0.1cm-1 per laser pulse) therefore takes several hours. Other drawbacks include the high cost of the instrumentation (OPO and pump laser) and experimental complexity required to overlap the input beams in space and time. For applications of Raman where spectral interference is not a problem, more conventional approaches are probably more appropriate. However, as OPO technology continues to improve, resulting in higher performance (e.g., better stability and higher repetition rates) and lower costs, this approach may grow increasingly attractive, especially in applications involving low frequency studies or problems with spectral interference.
This material is based upon work supported by the National Science Foundation under Grant No. CHE-9702087 and the NASA FAR program under grant NAG3-1974. Additional support was provided through NASA Cooperative Agreement Award MIE NCCW-0078 and NIH RIMI grant RR11598.
1. J. A. Giordmaine and R. C. Miller, Phys. Rev. Lett. 14, 973 (1965).
2. J. X. Zhou, X. Hou, K. X. Yang, S.-J. J. Tsai, and R. G. Michel, Lasers Based on Optical Parametric Devices: Wavelength Tunability Empowers Laser-Based Techniques in the UV, Visible, and Near-IR, Appl. Spectrosc. 52, 176A (1998).
3. B. J. Orr, M. J. Johnson, and J. G. Haub, in Tunable Laser Applications, F. J. Duarte, ed., Marcel Dekker, Inc., New York, pp.11-82 (1995).
4. P. C. Chen, Improved Scanning Range for CARS Using a Tunable Optical Parametric Oscillator, Analytical Chemistry, 68, 3068 (1996).
5. P. C. Chen, Rejection of Background Light Using Single Wavelength Detection in Nonlinear Raman Spectroscopy, Appl. Spectrosc. 51, 376 (1997).
6. P. C. Chen and M. Pinnix, Low-Frequency Raman Spectra Using Synchronously Scanned Optical Parametric Oscillator CARS, Appl. Spectrosc. 51, 1678 (1998).
7. Y. Prior, Three-dimensional phase matching in four-wave mixing, Appl. Opt. 19, 1741 (1980).
8. J. A. Shirley, R. J. Hall, and A. C. Eckbreth, Folded BOXCARS for rotational Raman studies, Opt. Lett. 5, 380 (1980).
Received 6th August 1998, received in revised format 15th September 1998, accepted 15th September1998.