Appl. Phys. B 56, 287-293 (1993) Applied physics PhysicsB and Laser Chemistry © Springer-Verlag 1993 Determination of Temperature by Stimulated Raman Scattering of Molecular Nitrogen, Oxygen, and Carbon Dioxide G. Millot, B. Lavorel, G. Fanjoux, C. Wenger Laboratoire de Spectronomie Mol6cutaire et Instrumentation Laser, Universit6 de Bourgogne, U.R.A. CNRS, 6, Bd. Gabriel, F-21000 Dijon, France (Tel.: 33-80/39 59 81, Fax: 33-80/39 59 71) Received 11 November 1992/Accepted 9 February 1993 Abstract. We have determined the temperature from SRS spectra of N2-N 2, N2-CO 2, 02-02, and CO2-CO 2 recorded in wide pressure and temperature ranges. The fitting pro- cedure takes simultaneously into account the Dicke effect and motional narrowing. We have quantified the accuracy of the MEG and ECS-P models for rotational relaxation. The temperature extracted from each model is compared with thermocouple measurements. The influence of vibra- tional broadening and shifting is discussed in detail. PACS: 33.70, 42.65.Dr Modem optical techniques like coherent anti-Stokes Raman spectroscopy (CARS) or stimulated Raman spectroscopy (SRS) have been proven to be effective for temperature measurement in real combustion environments [1]. Whereas CARS has demonstrated its usefulness in many applications, SRS has been especially used for fundamental studies [2]. These fundamental studies are devoted to the determination of line broadening and line shifting coefficients over a wide temperature range and to the study of interference effects for overlapping lines. Self collisions and collisions with minor species define the Q-branch profile for the probe molecule. The main collisional systems investigated up to now are N2-N 2 [3-13], Na-CO 2 [14], N2-H20 [15], N2-CO2-H20 [16], Oa-O 2, O2-N 2 [17, 18], CO2-CO 2 [19,20], and CO- CO [21,22]. Among these collisional systems, we have been particularly interested in Nz-N2, N2-CO2, 02-02, and CO 2- CO 2. The Q-branch Raman spectra have been recorded with the high-resolution stimulated Raman spectrometer devel- oped at Dijon a few years ago [3, 4, 6]. The main charac- teristics of this laser spectrometer are, first, its high spectral resolution (2.3 x 10 -3 cm-~), and second, its accurate abso- lute frequency calibration (better than 10 .3 cm -1) [23]. An- other characteristic is the absence of the nonresonant part of the Raman signal, since the stimulated Raman process only depends on the imaginary part of the susceptibility. We have taken advantage of these properties for developing and testing different models describing rotational relaxation rates, by comparison between calculated spectra and SRS experimental data over a wide density range. Among these models, we have focused our attention on the MEG law [7] used in practical applications and on the ECS-P law [9] which has a more physical background. The nonlinear susceptibility gives a spectral shape which depends on pressure and temperature. The synthetic shape depends on modeling of the rotational relaxation rates and consequently measurements of medium properties will also depend on the model used for these relaxation rates. In our previous studies [9, 11, 14, 17-20], we have com- pared the MEG and ECS-P models and we have given promi- nence to their different behavior for modeling of line broad- ening coefficient and collisional line narrowing in the Q- branch. The main goal of this report is to explicitly test the accuracy of these models for the temperature determi- nation. So, we have developed a computer routine which allows us to fit the temperature in wide pressure and tem- perature ranges, which works for a mixture of several gases and which takes simultaneously into account the Dicke effect and motional narrowing. The temperature is fit to SRS Q- branches recorded in a temperature-controlled cell equipped with a thermocouple. A large part of the data comes from previous studies, except for pure nitrogen, for which a spec- trum at 730 K has been especially recorded for this work, and also except for spectra of N2-N 2 at 140 K and CO 2- CO 2 at temperatures above 500 K (unpublished results). We have not tried to modify the models to obtain the exact tem- perature, our aim being to quantify the precision of each simple model for temperature fit from a more physical point of view. Many studies have been reported up to now in CARS thermometry ([24-28], and references therein), show- ing that several precautions should be taken in order to avoid systematic errors in the determination of the temperature. Among the parameters which may induce systematic errors let us note the nonresonant third-order susceptibility - (3~ /'((NR)' the absolute frequency calibration, the spectral dispersion of the detection system, cross coherence effects between ele- mentary CARS polarizations when a multimode pump laser source is used, nonlinearities in the response of intensified diode array detectors, effects of finite pump-laser bandwidth,