August 1979 / Vol. 4, No. 8 / OPTICS LETTERS 227

Generation of enhanced coherent anti-Stokes Ramanspectroscopy signals in liquid-filled waveguides

John C. Schaefer and ilan Chabay

National Measurement Laboratory, Center for Analytical Chemistry, National Bureau of Standards, Washington, D.C. 20234

Received April 2, 1979

We have demonstrated enhancement of coherent anti-Stokes Raman spectroscopy signals in liquid benzene con-tained in dielectric waveguide capillaries with angular phase-matching conditions. Enhancement factors as highas 130 were observed relative to a single crossing. These were measured for capillaries of various cross sections asa function of length. The maximum enhancement we observed was in a 50-)Um CX 50-cm X 292-mm capillary witha sample volume of 0.75 pl. Signals increased steadily with capillary length and showed the same dependence onpump-beam crossing angle (phase-matching conditions) as in bulk samples.

We report here the use of optical waveguides to en-hance the coherent anti-Stokes Raman spectroscopy(CARS)' signal of a liquid sample. Optical waveguideshave been employed previously to enhance spontaneousRaman spectra 5 and coherent Raman spectra of gasesand solid fibers.>8 Our experiments differ from thegas-phase experiments in that the nonnegligible fre-quency dispersion of a liquid sample requires that thetwo incident pump beams cross at a certain angle, whichdepends on the pump frequencies. At this phase-matching angle (generally, a few degrees) momentum-conservation relationships among the pump and signalphotons are satisfied, and maximum signal is generated.The use of mode dispersion to allow collinear, phase-matched three-wave mixing as reported for solidwaveguides by Stolen et al. 3 is possible only at a limitednumber of wavelengths, making spectroscopy imprac-tical. Angular phase matching is possible for any twopump wavelengths but does have the disadvantage thatthe volume probed is limited to the intersection of thetwo focused pump beams. The purpose of these ex-periments was to increase the intensity of CARS spectrafrom liquids by increasing the interaction volume whilemaintaining angular phase matching.

Waveguide enhancement of CARS signals in liquidswas demonstrated by measuring the signal that is dueto the 992-cm-' symmetric stretch mode of benzene fora single crossing in a bulk sample and comparing it withthe signal obtained from benzene in bollow waveguides.Pyrex and quartz capillaries of various lengths, cross-sectional areas, and geometries were tested. A squarePyrex capillary with dimensions 50 gm X 50 gm X 292mm (-0.75-gl volume) produced a signal that wasgreater than the bulk signal by more than 2 orders ofmagnitude.

The experimental arrangement was functionally thesame as that described previously.1 The effective fnumber for crossing with the 160-mm crossing lens (L2)was f/16, whereas that for each pump beam separatelywas f/140. For bulk samples, the crossing point was lo-cated within a 1-cm path-length cuvette filled withspectro-grade benzene. In the waveguide experiments,

the crossing point was located at the entrance apertureof the capillary.

Two collecting lenses were placed between the celland the spectrometer. The pump teams and the CARSsignal were recollimated by the first lens, which wasidentical to the crossing lens. The focus of this lenscould be placed at the crossing point when a bulk samplewas run or at the exit aperture of the capillary when awaveguide was being used. The recollimated beamswere focused onto the 0.5-mm-wide slit of a small doublemonochromator SAY DH-10) by the second lens, whose, number was chosen to match that of the monochro-mator (f/3.5). The lens was used at f/8. CARS signalswere detected by a photomultiplier tube and measuredwith a Molectron LP20 photometer. A mirror could beintroduced just upstream of the monochromator, al-lowing the power in the pump beams to be measured bya photodiode.

The waveguide cell is shown in Fig. 1. It consistedof a cylindrical Pyrex cell of the appropriate lengthhaving a long narrow waist and reservoirs at each end.These reservoirs were capped with end windows madefrom microscope slide covers. The capillary itself wasmounted so that it was supported at its ends by aslightly larger Pyrex jacket. The jacket was supportedover most of its length by the narrow waist of the cell.

Air

Fig. 1. Capillary cell, showing pump and CARS beams.

0,146-9592/79/080227-03$0.50/0 ©0 1979, Optical Society of America

228 OPTICS LETTERS / Vol. 4, No. 8 / August 1979

The capillary extended to within a few millimeters ofthe cell windows. The cell was mounted on a two-axisgimbal mount, placed on X, YZ translation stages, al-lowing complete adjustment of position and alignmentof the capillary.

The capillaries used in these experiments were pur-chased from Vitro Dynamics Company. One type wasmade of Pyrex and had a square cross section. Thesewere available in lengths up to 300 mm. A second va-riety was made of quartz with a circular cross sectionand was available only in lengths to 100 mm. Innerdimensions of these capillaries were comparable to thesize of the beam waists at their foci. Capillaries weretested whose inner dimensions were 50, 65, 75, 100, and200 gm. For all the capillaries, the numerical aperturewas assumed to be determined by the largest angle forwhich total internal reflection occurs in benzene. Thisangle was calculated to be about 170, corresponding toa numerical aperture of 0.30.

The experiments were done in the following manner.A cuvette was centered at the crossing point, and pumppowers were measured for the pump beams. TheCARS signal at 2col - w2 was optimized by moving onlythe two collection lenses. The cuvette was then re-placed with a capillary cell that was aligned with the cobeam. The collimating lens was moved downstream torefocus the pump and signal beams onto the mono-chromator slit, and minor adjustments of the collectionlenses were made to reoptimize the CARS signal. Thecrossing angle was set to generate the maximum CARSsignal in the cuvette for a particular xl - w2. Depen-dence of signal on crossing angle was measured for a75-Mm X 75-Mm X 92-mm capillary and was the sameas for the cuvette.

Measurement of the enhancement was made by di-rect comparison of the CARS signal from benzene in thecapillary with that from the cuvette with no change inthe input beams or detection sensitivity. Only thesignal-collection optics and capillary alignment weretweaked. The value of each capillary measurement wasdivided by the cuvette signal measured immediatelybefore. These results are shown in Fig. 2 for the 50-gm-square capillaries and the 65-Mm-diameter roundcapillaries. Thus the enhancement factor of 130 for the50-ym-square by 292-mm capillary shown in Fig. 2 istotally independent of the transmitted pump-beampowers (43% of the incident power for the xl beam and25% for the w2 beam).

Variations in capillary alignment, signal collectionefficiency, and the degree to which the capillary wasobstructed by particles could all combine to reduce oreven eliminate the CARS signal in a given experiment,as is evident in Fig. 2. Cuvette signals, on the otherhand, were constant to within 20% and correlated withvariation in pump powers. In spite of the variations inthe capillary results, however, the results as a wholeshow consistent, significant enhancement and definitetrends with respect to dimensions. As a function oflength, enhancement increased steadily up to 300 mm.Spectra can be generated in the waveguide in spite ofthe variations in signal since pump wavelengths can bechanged without adjustment of capillary alignment orcollection optics. Spectroscopic results will be de-scribed in a future article.

The spatial distribution of pump and CARS signals

0

C)

(2

-r

C0

c)

130120

110

100

90

so

70

60

50

40

30

20

'0

l00 200

length (mm)300

Fig. 2. Waveguide enhancement compared to a singlecrossing in a cuvette as a function of capillary length for 50-gm-square Pyrex capillaries and 65-Mm-diameter roundquartz capillaries.

from the capillary was determined by photographingthe output pattern. The two pump beams formedseparate images, showing maxima and apparent nodesthat were reproducible from one photograph to another.The co, pump beam was aligned with the capillary axisand formed a central spot. The output pattern of around capillary is shown schematically in Fig. 1. Fora square capillary, the w 2 beam formed two verticalbands about the central w, maximum. Each band atw2 had vertical and horizontal nodes. From the dis-tance between W2 maxima and the center of the pattern,it is estimated that the light at Cc2 emerges from thecapillary at an angle of 1.8° relative to the co, axis.This is consistent with the entrance crossing angle (20in air) within the accuracy of the measurement.

The CARS signal was also photographed with a nar-row-band interference filter used at the CARS fre-quency. When both pump beams were present, threemaxima occurred, a central spot with a vertical band oneither side. The patterns observed when the pumpbeams were present individually indicated that thecentral spot was due to stimulated emission at co, + 992cm-' (the anti-Stokes frequency) and the bands on ei-ther side were CARS signal. It is significant that phasematching by angular displacement was maintainedthroughout the length of the capillary. Spatial filteringcan be used to improve the signal-to-noise ratio just asin the bulk measurement.

For convenience, all the results were measured usingpump beams attenuated by approximately 1 order ofmagnitude below the powers that could normally beused with a sample contained in a cuvette. At pumppowers 1.3 orders of magnitude higher than those used,bubble formation within the capillaries interfered withproduction and measurement of CARS signals.

Enhancement showed an overall inverse dependenceon capillary cross-sectional area. To model this de-pendence, we employed the expression given by DeWittet al.9 for anti-Stokes power in the plane-wave limit,

P(WAS) = 4 AS I3X(31 2 [P(W1)] [P(w2)1 L2cnAS A 2

- 50pm x50Opmn

A - 65 prn

A.~~~~~~~~~~~~~~~~~~~~s

A .a I . I . . . .

August 1979 / Vol. 4, No. 8 / OPTICS LETTERS 229

where L is the length and A the cross-sectional area ofthe interaction region. We normalized the results forvarious-diameter capillaries to remove the dependenceof signal on pump power and cross-sectional area. Onewould expect this normalization to yield identical de-pendences of normalized CARS output power on lengthfor all the capillaries. Results indicated that this simplepicture is not adequate to describe the data accu-rately.

The dependence of CARS power on length for a singlecrossing can be inferred from a comparison of the con-focal beam parameter, b, the interaction length, L, andthe waveguide radius, a.7 For our experiments theplane-wave approximation is appropriate, and CARSpower will increase as the square of the interactionlength. This square dependence will also apply to the50-,um capillary since the capillary's inner dimensionis close to the optimum waveguide radius for beams withb _ 10 mm. Our data do not allow a clear distinctionto be made between quadratic and linear dependenceon length, although the 50-Mm capillary appeared toindicate quadratic dependence.

The number of phase-matched crossings per unitlength was found to depend critically on the alignmentof the pump beams with respect to the capillary axis.The maximum number of phase-matched crossingsoccurs when one beam is aligned with the capillary axisand the other beam intersects this axis at the crossingangle. When both beams are tilted in the same direc-tion with respect to the capillary axis, more crossingsoccur, but there are fewer crossings at the correctphase-matching angle. The number of phase-matchedcrossings for a symmetrical alignment is one half of themaximum value. The enhancement measured for thesymmetric case was indeed significantly less than thatfor the optimum alignment used to obtain all the re-ported data.

We have demonstrated a capillary cell that enhancesthe CARS signal by more than 2 orders of magnitudewith a liquid sample volume of -0.75 A1. A cell can bedesigned in which the sample could be made to flowthrough the capillary. Such a cell could be connecteddirectly to a liquid chromatograph to provide continu-ous, nondestructive chemical analysis of the output of

the chromatograph.1 0 Mixing of contiguous eluentfractions would be minimized in a longitudinal capillarycell, and signal is increased by 2 orders of magnitudewith the same pump powers. Effective sample volumesin a longitudinal waveguide cell and a transverse meltingpoint capillary are approximately equal. The wave-guide cell may also prove valuable in studies of vibra-tional and electronic excited states produced in thecapillary and probed by CARS.

The waveguide properties of the capillary require thatthe index of refraction of the sample be greater thanthat of the waveguide material. If silica capillaries mustbe used, this is a serious limitation since it precludes thestudy of aqueous solutions. The development of alow-index capillary or the use of metal or metal-coatedcapillaries could remove this restriction and allow al-most any sample to be used.

We are grateful to Instruments S.A. for the loan of thedouble monochromator and scan control used in theseexperiments. We would like to thank Hank DeLeoni-bus for expert assistance in constructing cells andGregory Rosasco for valuable discussions.

References

1. I. Chabay, G. K. Klauminzer, and B. S. Hudson, Appl.Phys. Lett. 28, 27-29 (1976).

2. R. H. Stolen, E. P. Ippen, and A. R. Tynes, Appl. Phys.Lett. 20, 62-64 (1972).

3. G. E. Walrafen and J. Stone, Appl. Spectrosc. 29, 337-344 (1975).

4. P. Rabinowitz, A. Kaldor, R. Brickman, and W. Schmidt,Appl. Opt. 15, 2005-2006 (1976).

5. G. E. Walrafen and J. Stone, Appl. Spectrosc. 26, 585-589(1972).

6. R. H. Stolen, J. E. Bjorkholm, and A. Ashkin, Appl. Phys.Lett. 24, 308-310 (1974).

7. R. B. Miles, G. Laufer, and G. C. Bjorklund, Appl. Phys.Lett. 30, 417-419 (1977).

8. J. Stone, J. Chem. Phys. 69, 4349-4356 (1978).9. R. N. DeWitt, A. B. Harvey, and W. M. Tolles, Naval

Research Laboratory Memorandum Report 3260 (1976),p. 21.

10. L. B. Rogers, J. D. Stuart, L. P. Goss, T. B. Malloy, Jr., andL. A. Carreira, Anal. Chem. 49, 959-962 (1977).