High spectral resolution multiplex CARS spectroscopy using chirped pulses

Chemical Physics Letters 387 (2004) 436 441 www.elsevier.com/locate/cplett High spectral resolution multiplex CARS spectroscopy using chirped pulses K.P. Knutsen, J.C. Johnson, A.E. Miller, P.B. Petersen, R.J. Saykally * Department of Chemistry, University of California, D31 Hildebrand, Berkeley, CA 94720-1460, USA Received 6 February 2004 Published online: 10 March 2004 Abstract A simple technique for achieving high spectral resolution coherent anti-stokes Raman scattering (CARS) spectra with a femtosecond laser system is presented. A linearly chirped and stretched (10 ps) pump pulse generates CARS signal only when overlapped in time with the Stokes pulse (90 fs), creating a temporal slit that defines the spectral resolution of the technique. Multiplex CARS spectra for liquid methanol and liquid isooctane are presented, demonstrating a spectral resolution of better than 5cm 1. This new chirped (c-cars) technique should prove useful for chemically-selective imaging applications, as it significantly reduces the non-resonant background contribution. Ó 2004 Elsevier B.V. All rights reserved. 1. Introduction Coherent anti-stokes Raman scattering (CARS) is a third-order process wherein the interaction of three beams, x p and x p 0 (the usually degenerate pump beams), and x s (the Stokes beam) generates the anti-stokes signal at the frequency x as ¼ 2x p x s. By tuning the energy difference x p x s over a molecular resonance, typically a vibrational transition, one can achieve a large enhancement in the signal, relative to the accompanying nonresonant background. The CARS signal is dependent upon both the type and environment of chemical bonds and is therefore a powerful tool for studying chemical properties of a system. Recently there has been much interest in the development of CARS microscopy for chemically-selective imaging of complex (e.g., biological) systems using endogenous chromophores, and both near-field [1] and far-field [2,3] imaging studies of cells have been reported. Current limitations of these techniques include low signal levels and heating of the samples, both suggesting * Corresponding author. Fax: +1-510-642-8369/8566. E-mail address: saykally@uclink4.berkeley.edu (R.J. Saykally). the advantage of utilizing femtosecond lasers due to their low average powers. However, a femtosecond pulse width implies low spectral resolution (e.g., 200 cm 1 relative to sample line widths of ca. 15 cm 1 ), which reduces the resonant to non-resonant CARS signal, and hence, the chemical selectivity. Conventional experimental designs employing narrow bandwidth (picosecond) pulsed lasers achieve sample-limited spectral resolution, but have the disadvantage of requiring frequency scanning to generate a spectrum, increasing acquisition time. Recent compelling demonstrations of multiplex CARS have all employed broadband (femtosecond) light sources to simultaneously excite several vibrational modes in the sample, followed by a probe pulse that is of narrow bandwidth. This can be accomplished using a synchronized femtosecond/picosecond laser combination, wherein the picosecond laser defines the spectral resolution [2,4], but a disadvantage of this technique is the necessity to synchronize the two separate laser systems, which increases the complexity of the experiment. Another approach, demonstrated in sum frequency generation spectroscopy, is to disperse the bandwidth of a femtosecond pulse (e.g., with prisms or gratings), and then introduce a slit in the dispersed beam to decrease its 0009-2614/$ - see front matter Ó 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.cplett.2004.02.049

K.P. Knutsen et al. / Chemical Physics Letters 387 (2004) 436 441 437 bandwidth, effectively converting the femtosecond pulse into a picosecond pulse [5]. Here we describe a simple variation on the latter design, wherein the spectral resolution of our instrument is defined by a temporal slit rather than a mechanical slit. Through the use of chirped laser pulses, the spectral resolution is defined by the temporal overlap of the 90 fs pulse with a temporally chirped pulse stretched to several picoseconds. The advantages of this technique are that the spectral resolution can be easily adjusted via the grating separation to match the natural line widths of the sample transitions, thereby increasing the ratio of resonant to non-resonant CARS signal. Since all of the pulses originate from the same laser, this technique also obviates the need for synchronizing two separate laser systems. The use of chirped pulses in nonlinear optical spectroscopy has been explored previously to study the dynamics of molecules in solution [6,7], and when used in conjunction with a Michelson interferometer, chirped pulses provide a simpler alternative to pulse-shaping [8] for obtaining high spectral resolution four-wave mixing spectra and to control the coherence of the excited states for gas phase [9,10] and condensed phase samples [11]. While similar in concept, the stepping of the delay between the x p 0 and x s pulses and the added analysis of the resultant signal makes those designs more complicated than that which we describe in this Letter. In addition, Cheng et al. [2] described the effect that a chirped pulse has on the observed spectra in their multiplex CARS setup, noting that the overall spectral bandwidth of their CARS signal decreased for increased chirp rate of the Stokes pulse, attributing this to the decreased overlap of the spectral components of the chirped pulse. However, we demonstrate below that more interesting effects occur when the probe pulse is chirped, namely increased spectral resolution and enhanced resonant signal to background. 2. Experimental 2.1. Optical design for c-cars Femtosecond pulses for the pump and Stokes beams are generated by a home-built Ti:sapphire oscillator (40 fs, 300 mw, 88 MHz). Seed pulses are amplified by a regenerative amplifier (90 fs, 1 khz, Spectra Physics, Spitfire) before entering an optical parametric amplifier (OPA) (90 fs, 60 lj at 2100 nm, Light Conversion, TOPAS) which generates the near-infrared idler pulse for the Stokes beam (x s ). A small amount (5%) of the s-polarized 800 nm beam from the amplifier is picked off by a beam-splitter before entering the OPA. This pulse makes four passes between two gratings (800 nm blaze, 830 grooves/mm, Edmund Industrial Optics), stretching the pulse in time due to the induced linear negative temporal chirp, resulting in the blue frequencies of the pulse leading the red frequencies in time. The near-infrared (2100 nm) s-polarized idler beam from the OPA is frequency-doubled by a BBO crystal to a p-polarized 1050 nm (x s ) beam, and temporally overlapped with a fraction of the chirped 800 nm pulse [the degenerate pump (x p ) and probe (x p 0) pulse] using a delay stage. The two beams are then focused to a 10 lm spot by an objective (Olympus, 10x, 0.3 N.A.) and collinearly excite the sample with 3 lj (x p ) and 1 lj (x s ). An identical collection objective collimates the CARS signal into a fiber optic that directs the signal to a spectrograph/ccd (Roper Scientific, spectral resolution 0.1 nm), typically integrating 300 400 laser shots. 2.2. Autocorrelation of chirped pulses In order to characterize the extent of the chirp effected by the grating pair, a homebuilt Michelson autocorrelator was employed to measure the pulse width of the unchirped and chirped x p pulses as a function of grating separation (Table 1). The temporally chirped pulse width is directly proportional to the grating separation [12], described by: 2bðk=dÞdk d k s ¼ cd½1 ðk=d sin cþ 2 Š ; ð1þ where d k s is the variation of group delay, b is the slant distance between the gratings, d is the grating constant, dk is the bandwidth of the pulse, c is the speed of light, and c is the angle of incidence (0 in our design). CARS signal is generated only when the pulses overlap in time on the sample, as defined by the shorter, unchirped 90 fs x s pulse. Therefore, an effective temporal slit is introduced into the optical path since the 90 fs pulse temporally overlaps with only a small portion of the much longer x p pulse. Due to the temporal chirp, Table 1 Autocorrelation results and calculated pulse widths for chirped and unchirped x p pulses Grating separation (cm) Measured pulse width (ps) Calculated pulse width (ps) Spectral resolution (cm 1 ) 0 0.09 0.01 300 7.5 9.4 0.2 9.1 3 13.5 22.4 0.4 22.0 1 34.1 53 1 55.6 0.5

438 K.P. Knutsen et al. / Chemical Physics Letters 387 (2004) 436 441 2000 2x idler, ω s 1800 objective sample objective CARS Fiber optic Spectrometer/ CCD BBO Idler from OPA Unchirped 800 nm from amplifier Grating 2 Chirped 800 nm, ω p Grating 1 CARS signal (a.u.) 1600 1400 1200 1000 800 600 400 200 2940 cm -1 (38 cm -1 FWHM) 2825 cm -1 (25 cm -1 FWHM) 0 3200 3000 2800 2600 Raman shift (cm -1 ) 100 fs 10 ps ω p ω as ωs E ω p v asym Fig. 1. Optical design for chirped CARS. The 800 nm (x p ) pulse from the amplifier passes 4 times between two gratings, stretching the pulse in time due to an induced linear negative temporal chirp that is directly proportional to the grating separation. The transform-limited near-infrared (2100 nm) idler pulse from the OPA is doubled by a BBO crystal to 1050 nm (x s ) and is temporally overlapped with the chirped pulse using a delay stage. A temporal slit where only a fraction of the x p bandwidth generates CARS signals. this overlap also corresponds to only a small fraction of the x p frequency bandwidth, as shown in Fig. 1b. Eq. (2) describes the effective spectral resolution of the CARS output (x as ¼ 2x p x s ) in the limiting case that the chirped (x p ) pulse is much longer than the x s pulse: Dt s Dx p ¼ Dx spec ; ð2þ Dt p where Dt s is the pulse width of the unchirped x s pulse, Dt p is the pulse width of the chirped (x p ) pulse, Dx p is the spectral bandwidth of the chirped (x p ) pulse and Dx spec is the resulting effective spectral resolution of the CARS signal. The spectral resolution determined for each grating separation is given in Table 1. 3. Results and discussion 3.1. c-cars of liquid methanol and liquid isooctane As shown in Fig. 2, the resonance enhancement achieved with high spectral resolution chirped CARS dominates the CARS spectrum of liquid methanol, with t v sym v gs Fig. 2. CARS spectrum of the asymmetric and symmetric C H stretches of liquid methanol at 2940 and 2825 cm 1, respectively. Multiplex CARS energy level diagram for liquid methanol. The chirped pump (x p ) pulse is gated by the 90 fs Stokes pulse ( S ) and thus its effective spectral bandwidth is determined by the extent of the chirp. The spectrally broad Stokes pulse (x s ) simultaneously excites both vibrational levels, and the spectrally narrow the chirped probe pulse (x p ) resolves the two virtual state transitions, generating the observed chirped CARS spectrum. only minimal spectral distortion resulting from interference with the non-resonant electronic CARS background. The energy level diagram in Fig. 2b shows how the observed multiplex chirped CARS spectrum for liquid methanol is generated. Both the symmetric and anti-symmetric C H stretches are excited by the x p and x s pulses, followed by the chirped probe pulse (x p ), which is effectively spectrally-narrowed by the temporal slit, then resolves the two excited state transitions, generating the observed chirped CARS spectrum. The measured line width (FWHM) of the symmetric C H stretch at 2825 cm 1 is 25 cm 1, while that for the asymmetric stretch is 38 cm 1. The spontaneous Raman spectrum of methanol also exhibits a broader asymmetric stretch line width of 30 40 cm 1 [13]. The line widths measured here remain the same for all grating separations used, implying that homogeneous broadening mechanisms limit the spectral resolution. The narrowest observed spectral resolution was achieved for liquid isooctane (Fig. 3). Numerous ali-

K.P. Knutsen et al. / Chemical Physics Letters 387 (2004) 436 441 439 CARS signal (a.u.) 3.5x10 5 3.0x10 5 2.5x10 5 2.0x10 5 1.5x10 5 1.0x10 5 5.0x10 4 0.0 2963 cm -1 2930 cm -1 2875 cm -1 3100 3000 2900 2800 2700 Raman shift (cm -1 ) phatic C H vibrations were observed. Note specifically that the CH 3 symmetric stretch appears as two clearly resolved and reproducible peaks, with a splitting of only 5cm 1 (Fig. 3 inset). These two peaks remain resolvable for all grating separations, and follow the expected positions when the Stokes frequency and relative delay between the pulses are varied. 3.2. Effect of varying x s wavelengths 2906 cm -1 2901 cm -1 2920 2910 2900 2890 Fig. 3. CARS spectrum of liquid isooctane. Inset: two CH 3 symmetric stretch transitions, separated by 5 cm 1, demonstrating that a spectral resolution of ca. 3 cm 1 is achieved. Resonance enhancement for the observed C H vibrations is confirmed by generating the chirped CARS spectra with different x s wavelengths. As shown in Fig. 4a, for x s wavelengths ranging from 1075 to 1025 nm, the maximum CARS signal for liquid methanol remains centered at the same wavelength. The CARS center wavelength does not change as x s is adjusted since it results from the summation of x p with the fixed C H vibrational frequencies, which are not affected by the x s wavelength (see Fig. 4b). However, as the maximum spectral intensity of the x s pulse is shifted across the vibrational energy levels, the relative intensities and shapes of the two C H peaks change due to different excited populations of the two vibrational energy levels and to varying contributions from the non-resonant background, causing the distortion of the resonant CARS peaks [4]. 3.3. Effect of delay on c-cars spectra When the delay between the chirped x p pulse and the x s pulses is varied, the temporal slit selects different x p frequencies, causing the overall CARS spectrum to shift since the center CARS wavelength is determined by the sum of the vibrational frequencies and the pump frequency, as mentioned above (see Figs. 1b and 5). All delay times are reported relative to the delay producing the largest resonant CARS signal (t ¼ 0). The wavelength shift is directly proportional to the relative delay, as determined by the product of the x p bandwidth and the ratio of delay time to x p pulse width. For example, for a relative delay of 3.4 ps, there would be a 109 cm 1 shift in the center CARS wavelength [300 cm 1 * (3.4 ps/ 9.4 ps) ¼ 109 cm 1 ]. This is observed in Fig. 5a by the near overlap of the two resonantly enhanced methanol C H vibrations for 3.4 ps relative delay, as predicted by their Raman spacing (115 cm 1 ). This consistency 40000 CARS signal(a.u.) 35000 30000 25000 20000 15000 10000 5000 0 1075 nm 1062.5 1050 nm 1037.5 nm 1025 nm 640 650 660 670 wavelength (nm) E Fig. 4. Observed liquid methanol CARS spectra for x s between 1025 nm and 1075 nm. CARS spectra are centered at the same wavelength; however, the relative intensities of the two C H peaks change due to different excited populations as the maximum intensity of the x s pulse shifts across the vibrational energy levels. Energy level diagrams for decreasing wavelength of x s (from left to right).

440 K.P. Knutsen et al. / Chemical Physics Letters 387 (2004) 436 441 CARS signal (a.u.) 40000 35000 30000 25000 20000 15000 10000 5000 2940 cm -1 2825 cm -1-1.7 ps delay (x2) t=0 (max intensity) +1.7 ps delay (x2) +3.4 ps delay (x10) 0 640 645 650 655 660 wavelength (nm) E Fig. 5. Scaled spectra of liquid methanol for adjusting the delay between x p to x s. For increased delays, the 2825 cm 1 peak increases in relative intensity compared to the 2940 cm 1 peak, and the center CARS wavelength shifts to the red. A delay of 3.4 ps results in an overlap of the 2825 cm 1 and 2940 cm 1 peaks. Energy level diagrams for adjusting delay. further establishes the mechanism of the observed chirped CARS spectrum with our technique. Recent theoretical studies by Naumov and Zheltikov proposed two different experimental techniques employing chirped pulses to obtain CARS signals [14,15]. The first of these, recently demonstrated by Lang et al. [16], maps the transient CARS spectrum in one laser shot by delaying a chirped probe pulse after the initial pump and Stokes pulses to spatially disperse the temporal evolution of a CARS signal. While this technique can extract dynamical information, it cannot produce a high spectral resolution CARS spectrum. The second method incorporates two chirped pulses and one unchirped femtosecond pulse to obtain high spectral resolution CARS spectra in one shot [14,15]. This proposed scheme is closer to the technique described here; however, their method of achieving high spectral resolution comes primarily from the spatial overlap of large (3 cm beam waist) chirped beams with spatially smaller chirped (or unchirped) beams. The geometrical requirements of such an experiment make it unfeasible for CARS microscopy, thus our design is more suitable for application to biologically relevant systems. The second limitation on the spectral resolution described in their paper results from the temporal overlap of only the first two (pump and Stokes) pulses, similar to earlier studies that used chirped pulses to control the bandwidth of second harmonic generation from a BBO crystal [17]. This bandwidth calculation entirely neglects contributions from the final probe pulse, including the chirp, which through its overlap with the unchirped 90 fs Stokes pulse, defines the spectral resolution of the system. Ultimately, more theoretical work is necessary to rigorously describe the spectral resolution achieved with the experimental design we describe here. 4. Conclusions We have shown that the two-pulse chirped-cars technique allows for very simple control of experimental spectral resolution. The novelty of the technique is that it renders bandwidth reduction unnecessary, instead exploiting only the chirp of the probe to achieve high spectral resolution. As with other multiplex CARS techniques, by dispersing the CARS signal onto a CCD, we can observe the entire (ca. 300 cm 1 ) CARS spectrum in a single laser shot, without the need for adjusting the delay of the pulses, thus allowing for direct comparisons between peak intensities within the broadband region of interest. Through the use of diffraction gratings to induce the linear chirp, we can readily adjust the experimental spectral resolution to match the line width of the vibrations in the sample, increasing the ratio of resonant to non-resonant CARS signals through more efficient use of the laser power. The simplicity of this experimental design enables high spectral resolution CARS spectra to be easily obtained with femtosecond laser systems, and should prove useful for imaging with chemically-selective contrast.

K.P. Knutsen et al. / Chemical Physics Letters 387 (2004) 436 441 441 Acknowledgements This work was funded by the Experimental Physical Chemistry Program of the National Science Foundation. K.P.K. and A.E.M. are supported by graduate fellowships through Lawrence Livermore National Laboratory, and P.B.P. is supported by the Danish Research Agency. J.C.J. is a ChevronTexaco Research Fellow. References [1] R.D. Schaller, J. Ziegelbauer, L.F. Lee, L.H. Haber, R.J. Saykally, J. Phys. Chem. B 106 (2002) 8489. [2] J. Cheng, A. Volkmer, L.D. Book, X.S. Xie, J. Phys. Chem. B 106 (2002) 8493. [3] M. Hashimoto, T. Araki, S. Kawata, Opt. Lett. 25 (2000) 1768. [4] M. M uller, J.M. Schins, J. Phys. Chem. B 106 (2002) 3715. [5] L.J. Richter, T.P. Petralli-Mallow, J.C. Stephenson, Opt. Lett. 23 (1998) 1594. [6] E.T.J. Nibbering, D.A. Wiersma, K. Duppen, Phys. Rev. Lett. 68 (1992) 514. [7] K. Duppen, F. dehaan, E.T.J. Nibbering, D.A. Wiersma, Phys. Rev. A 47 (1993) 5120. [8] N. Dudovich, D. Oron, Y. Silberberg, Nature 418 (2002) 512. [9] V.V. Lozovoy, B.I. Grimberg, E.J. Brown, I. Pastirk, M. Dantus, J. Raman Spectrosc. 31 (2000) 41. [10] R.A. Bartels, T.C. Weinacht, S.R. Leone, H.C. Kapteyn, M.M. Murnane, Phys. Rev. Lett. 88 (2002) 033001. [11] E. Gershgoren, R.A. Bartels, J.T. Fourkas, R. Tobey, M.M. Murnane, H.C. Kapteyn, Opt. Lett. 28 (2003) 361. [12] E.B. Treacy, IEEE J. Quantum Electron. QE-5 (1969) 454. [13] L.K. Iwaki, D.D. Dlott, J. Phys. Chem. A 104 (2000) 9101. [14] A.M. Zheltikov, A.N. Naumov, Quantum Electron. 30 (2000) 606. [15] A.N. Naumov, A.M. Zheltikov, J. Raman Spectrosc. 32 (2001) 960. [16] T. Lang, M. Motzkus, J. Opt. Soc. Am. B 19 (2002) 340. [17] K. Osvay, I.N. Ross, Opt. Commun. 166 (1999) 113.