Filament-based stimulated Raman spectroscopy J. H. Odhner, D.A. Romanov, and R. J. Levis Department of Chemistry and Center for Advanced Photonics Research Temple University Philadelphia, PA 19122

ABSTRACT A new multiplexed stimulated Raman spectroscopic technique encompassing a single-shot spectral measurement range of over 3900 cm-1 is presented. Impulsive excitation of all Raman active vibrational modes present in a medium is achieved by self-compression of a laser pulse undergoing filamentation in air, creating coherent vibrational wave-packets. These wave-packets create a macroscopic polarization of the medium that imparts sidebands on a delayed narrowband probe pulse. The background-free measurement of impulsively excited Raman modes in gas-phase N2, O2, H2, CO2, toluene, ammonia, and chloroform with a spectral resolution of 25 cm-1 is presented. Keywords: filamentation, nonlinear optics, Raman scattering, spectroscopy.

1. INTRODUCTION Laser-based remote chemical identification is an important and rapidly growing field1, 2. Vibrational spectroscopy is particularly suitable for molecular detection due to its functional specificity and sensitivity35 . Spontaneous Raman spectroscopy has the ability to acquire complete Raman spectra, but has inherently low signal strength, making application to the case of remote, in situ trace chemical detection difficult6. Narrowband stimulated Raman spectroscopy (SRS) can lead to much higher signal intensities in comparison with spontaneous Raman, but the need to scan the frequency of the probing beam to acquire the full Raman spectrum makes SRS a challenging technique to implement. Multiplexed femtosecond vibrational spectroscopy alleviates the need for tunable sources7. However, the dispersion accompanying femtosecond pulse propagation over long distances must be compensated to enable application of femtosecond lasers to remote detection8. Here we present initial results for the development of a spectroscopic technique based on pulse self-compression through femtosecond laser filamentation, with the potential for applications in remote detection. When an ultrafast laser pulse propagates in a molecular medium, any Raman-active vibrational modes with an oscillation period longer than the shortest temporal feature of the intensity envelope of the laser pulse will be excited9, 10. This phenomenon has been termed impulsive stimulated Raman scattering (ISRS)11. To measure a complete Raman spectrum for any given sample, i.e. to excite all possible vibrations of a sample up to the hydrogen stretch at 4152 cm-1, the laser pulse must contain a temporal feature on the order of 8 fs in duration. Once a vibrational coherence has been excited in the medium, a narrowband (picosecond), time-delayed probe pulse traveling collinearly or non-collinearly with the exciting pulse can be used to probe the system. Interference between the probe pulse and the coherently excited medium results in the generation of Raman sidebands at frequencies p ± v, where p is the probe pulse frequency and v is the frequency of a vibrational mode of the medium. The spectral resolution of the sidebands in this multiplexed scheme (also known as time-delayed coherent Raman scattering12) depends only on the probe pulse bandwidth, the natural linewidth of the transition, and the spectrometer resolution13. Delivery of a dispersion-compensated laser pulse of duration on the order of 10 fs to remote targets generally requires a sophisticated optical system, prior knowledge of the dispersion imposed by the intervening distance, and/or active control for dispersion compensation8. Using a laser filament as the excitation source circumvents these requirements by relying on the intrinsic pulse-compression, which

occurs during the filamentation process. Femtosecond laser filamentation is initiated whenever a laser pulse propagates in a medium with a peak power above the critical power for self-focusing, Pcrit=3.7702/8n0n2, where 0 is the laser central wavelength, n0 is the index of refraction, and n2 is the second order, intensitydependent index of refraction for the material14. Self-focusing leads to an increase in the on-axis intensity of the beam, which is accompanied by self-phase modulation, self-steepening, and, at high power (on the order of several Pcrit), ionization of the medium. Ionization leads to the creation of an underdense plasma that spatially defocuses the pulse. Refocusing can occur of the on-axis intensity falls below the ionization threshold (due to dispersion or energy depletion) and plasma defocusing is halted. The dynamic interplay between these effects (self-focusing and plasma de-focusing) can result in confined propagation of the laser pulse over distances many times the Rayleigh length of the beam15. Femtosecond filamentation results in pulse compression to a duration approaching the single-cycle regime16, 17, with corresponding spectral broadening from the ultraviolet to the near infrared. The formation distance of the filament and the filament length (as characterized by the fluorescence emission in the plasma channel) from the laser can be controlled optically by manipulating the initial pulse chirp using pulse-shaping techniques18, 19. Finally, filament formation can be controlled using deformable mirrors20. Application of filaments to remote Raman spectroscopy has been suggested21 and impulsive Raman detection of carbon dioxide and molecular hydrogen has been demonstrated in a filament using 10 fs driving pulses22, as well as in air using 45 fs driving pulses23. Here we present a study of a filament-based, multiplexed vibrational Raman spectroscopic technique and apply the method to a series of molecules including N2, O2, CO2, H2, toluene, ammonia, and chloroform. We also investigate the dependence of the Raman signals as a function of filament-probe time delay.

2. EXPERIMENTAL The experimental setup is shown in figure 1. A Ti:sapphire oscillator (KM Labs) seeds a chirped-pulse amplifier (Legend, Coherent), which delivers 2.5 mJ, 40 fs pulses at a 1 kHz repetition rate with a central wavelength of 800 nm. The pulse is split 80:20 to generate a filament and a Raman probe pulse, respectively. The filament pulse (1.9 mJ) is focused in ambient air using a 2 m lens, generating a >65 cm long plasma filament. The lens is mounted on a 30 cm long stage, which provides control over the spatial overlap between the filament and the Raman pump without significantly changing the filament-probe temporal delay.

b e

d

g

a f

c

Figure 1. Experimental setup for filament-based Raman spectroscopy. (a) beamsplitter, (b) 2m lens and 30 cm stage, (c) 4-f spectral filter, (d) delay stage, (e) sample, (f) 4-f Rayleigh line filter, (g) spectrometer.

The optimum distance for impulsive excitation under the conditions described above was found to be ~250 cm from the lens. We attribute this to the nonlinear propagation of the pulse undergoing filamentation, namely tight spatial confinement over the length of the filament, coupled with pulse self-compression dynamics. Filamentation of the probe pulse is inferred from the extended visible fluorescence channel in air, and also by the clean spatial mode of the white-light spectrum observed after the fluorescence channel. The filament is optimized by maximizing the spectral blue shift in the pulse spectrum after filamentation,

which is achieved by making fine adjustments in the grating compressor of the amplifier. The probe pulse polarization is rotated with a half-wave plate and spectrally filtered to reduce the bandwidth using a reflective zero-dispersion 4-f compressor with an adjustable slit in the Fourier plane to control the bandwidth. The resulting 0.6 ps, 10 PJ pulse (~25 cm-1 FWHM at 795 nm) passes through a variable delay stage and is focused using a 0.5 m lens onto the filament at a small angle (T=1.5q) with respect to the filament axis. A positive delay is introduced between the filament and the probe pulses to minimize crossphase modulation between the filament and the probe beam. The orthogonal polarizations of the filament and probe beams serves to further minimize cross-phase modulation effects. To reduce air turbulence, the filament is generated in an open-ended tube preceding the overlap region (not shown in figure 1). The volume of the probed region was measured by spatially scanning a carbon dioxide jet through the overlap region and measuring the Raman signal as a function of the CO2 jet position. The result is plotted in figure 2. The filament core has an estimated diameter of ~70 micrometers24. Using these values, the interaction volume is estimated to be 1.8x10-4 cm3, corresponding to a 7.4 nM sample size at standard temperature and pressure. The low vapor pressure of the liquid samples used here correspond to even smaller sample sizes of detected species (213 pM for toluene, assuming a saturated vapor pressure of 22 mmHg).

Figure 2. Measured intensity profile of the CO2 1285 cm-1 and 1385 cm-1 modes as a function of longitudinal distance along the beam intersection of the filament and probe beams (squares, bottom axis). The Raman spectrum of CO2 (red spectrum), and air (oxygen and nitrogen, black spectrum , 1560 cm-1 and 2340 cm-1, respectively) is shown on the, upper axis).

The Raman signal travels nearly collinearly with the probe beam and is measured, after filtering the probe wavelength in the Fourier plane of a second zero-dispersion 4-f compressor, in a spectrometer (USB4000, Ocean Optics). Raman spectra are processed by subtracting a baseline reference spectrum from the collected data. Both the Stokes and anti-Stokes lines were detected with comparable intensity and exhibited similar behavior with respect to probe and filament intensity. For simplicity, we present here only the antiStokes spectra here. The apparatus was aligned to optimize the intensity of Raman lines present in ambient air (nitrogen, oxygen, carbon dioxide and water). Vapor phase samples are prepared by bubbling argon gas through liquid samples at room temperature. Evaporative cooling is avoided by use of a hotplate to maintain ambient temperature conditions, as needed. The vapor is then introduced into the interaction region through a nozzle as shown in figure 1. This arrangement displaces the air in the interaction region, which accounts for the lack of signal from air in the detected spectra. An exhaust line opposite the nozzle removes the vapor from the air and maintains steady flow conditions in the interaction region.

3. RESULTS To characterize the spectral range and quantitative abilities of this method, the Raman spectrum of an air-

hydrogen mixture was measured and is presented in figure 3. The expected peak ratios of nitrogen and oxygen in air can be calculated using the relative concentrations of species (~80% and ~20% for nitrogen and oxygen, respectively) and the relative Raman gain coefficients (0.071 and 0.016 GW/cm at >10 atm [25] for nitrogen and oxygen, respectively. The measured nitrogen-oxygen peak ratio of 17.6:1 is in good agreement with the calculated peak ratio of 17.8:1, suggesting that all molecules with vibrations up to 2331 cm-1 (corresponding to a period of 14.3 fs) are excited by a pulse shorter than the vibrational period of nitrogen. A similar estimation of the pulse duration using hydrogen was not possible due to experimental limitations (lack of quantitative data on hydrogen concentration used in the experiment), but the measurement of the hydrogen vibrational (4152 cm-1) and rotational (357 cm-1, 592 cm-1, 819 cm-1, and 1039 cm-1) Raman spectra demonstrate the spectral range of the measurement to be extremely broad (~3800 cm-1).

Figure 3. Raman spectrum of air and hydrogen measured using filament-based SRS. The dotted line at 263 cm-1 denotes the filter cutoff used for this measurement.

Several factors must be taken into account to make measurements using our impulsive Raman scattering scheme quantitative. The stimulated Raman scattering process is phase-matched, which leads in our (noncollinear) case to a slight angular spread of signal frequencies, resulting in alignment-dependent variation of the observed peak ratios at the extremes of the Raman spectrum. This is partially corrected for in the 4-f filter used to remove the Raman probe from the signal, and could be eliminated entirely by the use of an integrating sphere or more precise dispersion compensation. Another consideration is the time-dependence of the Raman peaks as a function of pump-probe time delay, which must be taken into account for quantitative analysis. As reported previously23, a decrease in scattering intensity is observed as a function of delay between the filament and probe beams (figure 4, next page). This is due to dispersion of the rovibrational wavepacket initially prepared by the filament pulse. The decay time of the vibrational coherence excited by a filament has previously been shown to depend on the temperature of the system and on the relevant molecular parameters. The time-dependence of the nitrogen/oxygen ratio at room temperature is also plotted in figure 4, and reveals that the ratio increases with increasing time delay. This change in peak ratios must be taken into account when probing at longer delay times in order to retain the quantitative nature of the measurement. In the following sub-sections we present a theoretical description of the time-dependence of the Raman signal, experimentally measure the time-dependent dynamics for air, and present Raman spectra of several other common solvents to further demonstrate the utility of the filament-based Raman spectroscopic method.

Figure 4. Time dependence of the Raman intensities of nitrogen (upper curve, left) and oxygen (lower curve, left). The dotted line shows the calculated nitrogen/oxygen ratio based on the calculated Raman intensity profiles.

3.1 Theory of time and temperature dependence of Raman signal The time-dependence of the Raman response of a sample depends on the ro-vibrational structure of the molecules and the temperature. The time-dependence can be used to study the dynamics in a medium after excitation, in this case caused by the filament pulse. The initial thermal population defines the coherent superposition of excited ro-vibrational states, which in turn determines the temporal propagation of the wavepacket. Previous work showed that excited ro-vibrational states will initially be in phase, but will destructively interfere as time proceeds due to the differences in spacing between ro-vibrational energy levels in an anharmonic potential well26. The dispersion of the ro-vibrational coherence in the sample is observed as a decrease in signal intensity as a function of pump-probe time delay. The two major causes of the observed wavepacket dispersion are the anharmonicity of the molecular potential energy surface and the rotation-vibration coupling. The former reduces vibrational level spacing with increasing n, and the latter decreases the ro-vibrational level spacing with increasing n. In the filament, all thermally populated states are excited, which leads to dispersion for oscillators of differing frequencies. We test this hypothesis by comparing the observed decay with that predicted by a simple model that includes first-order corrections for anharmonicity and rotation-vibration coupling,

En , J

!Ze  n  1 2  !Ze xe  n  1 2  2

 Be  D e n J  J  1 ,

where Ze is the vibrational mode

frequency, Ze xe is the anharmonicity parameter, Be is the rotational constant, and D e is the rotationvibration coupling parameter. Restricting our analysis to the Q-branch, the impulsive excitation of the system induces all inter-level transitions with 'n 1 and 'J 1 , thereby creating a time-dependent polarization whose evolution is determined by oscillations at frequencies, Zn, J Ze  2Ze xe  n  1  D e ! J  J  1 :

¦ A T cos ª¬Z

P t

n, J

n, J

n, J t º ¼

°­ °½ Re ® An , J T exp ª¬iZn , J t º¼ ¾ , ¯° n , J ¿°

¦

(1)

where the temperature-dependent amplitudes, An , J T , are determined by the initial populations of the J-th sublevel of n-th vibrational manifold. Over a wide range of temperatures, the double inequality, Be