152 OPTICS LETTERS / Vol. 1, No. 5 / November 1977

Stimulated Raman spectroscopy using low-power cw lasers

Adelbert Owyoung and Eric D. Jones

Sandia Laboratories, Albuquerque, New Mexico 87115Received June 24, 1977

It is demonstrated that stimulated Raman spectroscopy (SRS) can be performed using cw dye lasers at power levelsover six orders of magnitude smaller than those generally associated with pulsed stimulated Raman studies. Thepreliminary results suggest that cw SRS is a potentially powerful alternative to conventional spontaneous Ramanscattering, with resolution limited solely by laser linewidth and sensitivity independent of resolution require-ments.

Spontaneous Raman spectroscopy is a powerful anddiverse tool for the study of material excitations. Thelimitations imposed by small scattering cross sections,however, have resulted in the development of nonlinearoptical techniques, such as coherent anti-Stokes Ramanscattering (CARS),' the Raman-induced Kerr effect(RIKE),2 and stimulated and inverse Raman scatter-ing,3-5 to increase both the signal levels and the infor-mation obtained in the excitation of Raman modes. Inthis Letter we report the demonstration of cw stimu-lated Raman spectroscopy (SRS), whereby the advan-tages offered by pulsed stimulated Raman scatteringas a spectroscopic tool are obtained using low-power cwlasers. The technique provides a strong coherent out-put signal, which is linear in both scattering numberdensity and pump power, contains no contributionsfrom nonresonant background nonlinearities, requiresno phase matching, and has no inherent limitationsresulting from minor beam depolarization in the opticalsystem. Because of the coherent properties of theoutput signal, a high degree of discrimination againstluminescence and background fluorescence is obtained,in contrast to conventional Raman techniques. Fur-thermore, the technique is expected to be particularlyuseful in Raman studies in gases because of its intrinsichigh resolution and wide dynamic range. The resolu-tion is limited only by the laser linewidth, and the dy-namic range is limited by the spectral tuning range. Weshall first describe the technique, then outline sometheoretical considerations, and finally conclude by de-scribing the preliminary experimental results that es-tablish the feasibility of the technique.

The cw stimulated Raman scattering experiment isimplemented by using two cw laser sources whose dif-ference frequency is in or near resonance with theRaman frequency (Stokes or anti-Stokes). The pumplaser power is amplitude modulated, and hence, throughthe third-order nonlinear susceptibility, an ampli-tude-modulated gain (loss) is introduced to the probebeam. Thus, if standard synchronous detection tech-niques are utilized, the Raman gain (loss) spectrum canbe displayed by sweeping one of the laser wave-lengths.

A schematic representation for the implementationof SRS using two cw lasers is shown in Fig. 1. A CRL

Model 490 Rhodamine 6G cw dye laser supplies thepump beam, which is continuously tunable from 5700to 6400 A with a linewidth of 0.5 cm-'. Modulation ofthe pump is accomplished by using an 8-kHz chopper.A Spectra-Physics Model 125 He-Ne laser with an rfquieting option and intracavity etalon is operated on asingle longitudinal mode to provide a stable probesource at 6328 A. The two beams are combined on adichroic mirror and subsequently passed through aspatial filter, which assures their collinearity as they arefocused through the sample cell. Here the ampli-tude-modulated beam imposes a small Raman gain(loss) on the probe beam in the vicinity of the Stokes(anti-Stokes) mode. The beams are then convenientlyseparated by a diffraction grating and bandpass filters.The probe signal is detected on a biplanar photodiodeand demodulated by a lock-in amplifier, which is ex-ternally referenced to the 8-kHz chopper. The resultis an ability to measure induced gains (losses) signifi-cantly less than 1 part in 106. Frequency tuning of thedye laser then results in a direct display of the Ramanspectrum. Since the process is coherently driven, thesignal is independent of spectral resolution require-ments, and total efficiencies exceeding those obtainedby spontaneous Raman scattering techniques may beexpected for resolutions higher than 0.1 cm-'.

In order to understand quantitatively the beam in-teraction in the Raman medium, consider two linearlypolarized monochromatic beams of the form Re-(Eue -itt) and Re(E,,e -iwt). The nonlinear polarization

SINGLE MODEHeNe LASER

BEAMSPLITTER -

RH INE 6DYE LASER

PAR MODEL HR-8 BIPLANARLOC K-IN AMPL I F I ER PHOTODIODE

l FILTERS

REFERENCEDETECTOR SPATIAL \ 3o

FILTER -- 5954A16328A

SAMPLE RATING- | 8kHz CELL

I- 5954A I

CHOPPER

,/

Fig. 1. Schematic diagram of the cw stimulated Ramanspectroscopy experiment.

November 1977 / Vol. 1, No. 5 / OPTICS LETTERS 153

component, which results from treating EQ as the spa-tially varying pump field, is given by convention as6X 3ijkl(-,w, W,Q,-Q)EukEuj*E,,, where X3 iikl is thethird-order nonlinear susceptibility tensor and repeatedspatial indices are assumed to be summed. 6 Of par-ticular interest in the present study are the cases inwhich EQ and E,, are polarized parallel or perpendicularto each other. In these cases the spatial indices ijklbecome iiij, with i = i and i id j, respectively, for theparallel and perpendicular cases. By direct substitu-tion into Maxwell's equations, it may be shown that fora medium of length L, the peak intensity gain g expe-rienced by a field of frequency X along any "ray" definedby the vector grad E,, may be written in the form7-8

g = [Im x3iiji(_WWQ,_Q)] IE (

where n is the linear refractive index and the integralis taken along the ray path, s. Furthermore, if one as-sumes that the beams have identical Gaussian-beamparameters and are exactly coincident through thesample, differences in their diffraction properties maybe neglected and Eq. (1) may be evaluated analytically.Assuming that the focal region of the beams is containedwithin the sample and integrating over the cross-sec-tional area of the beams, Eq. (1) yields a power gain5P/P for the field E,, of

(2)bP(w) 3847r4 .

P(C) X2 c [Im X3iijj(-@C0S,C0 -Q)1P(Q),

where P(Q) and P(co) are the pump and probe inputpowers, respectively, and X2 = (2rc) 2 W'-1Q-. It is tobe noted that Eq. (2) is independent of the degree offocusing and is linear in both Im X3iiii and P(Q). Also,since the gain is independent of Re X3, the signal willgive a direct measurement of the Raman spectrum withno interference from nonlinear dispersion in the vicinityof the Raman mode or nonresonant background non-linearities. 3 This result is to be contrasted with similaranalysis for the CARS technique where phase matchingis required and terms proportional to I X3 12 contributeinterference effects between the real and imaginaryterms of x3. As an extension of the cw SRS capability,we note that information on Re X3iiii is contained in thephase of the modulated output probe beam and thus canalso be extracted by using an interferometer underconditions similar to those outlined in the presentwork.3

In order to demonstrate the feasibility of cw SRSwithout adding the stringent requirements on laserlinewidth that would be needed for high-resolutionspectra in gaseous media, we have chosen to performthese initial studies on a well-characterized Ramanmode in liquid benzene.3 Referring again to Fig. 1, theTEMOOq single-mode He-Ne laser provides a 5-mWprobe source at the sample. The Rhodamine 6G dyelaser delivers 50 mW of average power to the sample toprovide the pump source. The 0.5-cm'1 linewidth(FWHM) of this source currently is the limitation to thespectral resolution. Substitution of these parametersin Eq. (2) with Im X3 1111 = 15.9 X 10-14 cm,3 /erg resultsin a value of 0P(w)/P(w) = 1.75 X 10-5 for the peakRaman gain arising from the 992-cm-' mode of ben-

Fig. 2.spectramode.

DEPOLARIZED E,,iE,

980 990SMTKES SHIFT (cm )

1000

Polarized and depolarized cw stimulated Ramanin liquid benzene in the vicinity of the 992-cm-1

zene.3 A typical spectral scan in the vicinity of thismode is shown in Fig. 2, which shows the central peakat 992 cm-1 and 6vidences the weaker surroundingmodes in this region.3' 9 The magnitude of the observedsignal is in good agreement with the predicted value.The noise fluctuations are due primarily to amplitudefluctuations in the single-mode He-Ne probe laser.This source of noise can be eliminated by amplitude-stabilization techniques. For comparison, a depolarizedscan across the same wavelength region in which EQ isperpendicular to E,, is also shown. Since the 992-cm-1mode is polarized, the line is missing in this configura-tion. The quantum limit of detection in this experi-ment is posed by shot noise; hence, an ITT FW-114Abiplanar photodiode is used for optical detection toallow average signal currents as high as 100 gA. It isthus expected that a gain of 1 part in 106 should bemeasurable with signal-to-noise ratios in excess of 100under these conditions.

In summary, it has been demonstrated that cw SRSpresents a viable alternative to spontaneous Ramanscattering and nonlinear optical techniques, such asCARS and RIKE, in spectroscopic applications. Be-cause of the coherence of the probe beam, techniquessuch as spatial filtering can be used to eliminate spuri-ous noncoherent noise signals resulting from lumines-cence, background fluorescence, and scattering. Also,the technique is both self phase matching and insensi-tive to minor beam depolarization in the sample. Theexperimental requirements for high-resolution mea-surements utilizing this technique preclude the neces-sity of using monochromators, Fabry-Perot interfer-ometers, and other apparatus associated with conven-tional Raman measurements. Background non-linearities give no contribution to the measured signalsince only Im X3 is measured and the Raman spectrumis thus obtained directly.1 0 Since we are dealing witha stimulated scattering process, the limitations on res-olution are entirely determined by the laser sources, and

154 OPTICS LETTERS / Vol. 1, No. 5 / November 1977

no loss of sensitivity results as resolution is increased.With currently available dye-laser systems, resolutionsin excess of 0.003 cm-' (i.e., 10 MHz) are realizable.This feature makes cw SRS particularly appealing forhigh-resolution applications in gaseous media.

From the above discussion, it is obvious that thesekinds of detection techniques can be applied to otherareas of high-resolution spectroscopic work such asBrillouin scattering and two-photon absorption."

The expert technical assistance of R. E. Asbill isgratefully acknowledged. This work was supported bythe U.S. Energy Research and Development Adminis-tration.

References

1. R. F. Begley, A. B. Harvey, and R. Byer, Appl. Phys. Lett.25, 387 (1974).

2. D. Heiman, R. W. Hellwarth, M. D. Levenson, and G.Martin, Phys. Rev. Lett. 36, 189 (1976).

3. A. Owyoung and P. S. Peercy, J. Appl. Phys. 48, 674(1977).

4. A. Lau, W. Werncke, M. Pfeiffer, K. Lenz, and H. J.Weigmann, Sov. J. Quant. Electron. 6,402 (1976).

5. P. Lallemand, P. Simova, and G. Bret, Phys. Rev. Lett.17, 1239 (1966).

6. X 3 ijkl is equivalent to the Cijkl tensor defined by P. D.Maker and R. W. Terhune in Phys. Rev. 137A, 801(1965).

7. A. Owyoung, Ph.D. Thesis (California Institute of Tech-nology, 1971, unpublished).

8. R. W. Hellwarth, "Third Order Susceptibilities of Liquidsand Solids," in Progress in Quantum Electronics (Per-gamon, Oxford, England, 1977), Vol. 5, No. 1.

9. J. E. Griffiths, M. Clerc, and P. M. Rentzepis, J. Chem.Phys. 60, 3824 (1974).

10. Although a small contribution to Im X3 can arise fromtwo-photon absorption, this is generally small comparedwith the nonresonant electronic nonlinearities that con-tribute to Re X3 and provide interference effects in theCARS measurements.

11. W. K. Bischel, P. J. Kelly, and C. K. Rhodes, Phys. Rev.Lett. 34, 300 (1975).