Probing vibrational levels of molecules with coherentanti-Stokes Raman spectroscopy (CARS) has severalapplications ranging from the analysis of flames and com-bustion processes [1–3] to chemically sensitive micro-scopy of biological or chemical samples [4]. CARS hasseveral advantages compared to its linear counterpart,namely, generation of the signal on the blue side ofthe excitation pulse away from fluorescence. When theanalyte concentration and laser intensity are sufficientlyhigh, its coherent nature allows faster measurements forreal-time studies. As a nonlinear effect, the use of pulsedultrashort lasers with high peaks powers enhances dra-matically the signal for a given average power. However,as pulses get shorter, unwanted nonlinear effects alsohave a deteriorating impact, e.g., nonlinear photodamageand generation of a nonresonant background, which in-terferes with the signal [5].A wide range of setups has been proposed, trying to
minimize such effects while maximizing the collected in-formation. Picosecond CARS uses spectrally narrow ex-citation to excite specific Raman levels. This techniqueallows fast imaging biological probes, with the energydifference between the two lasers tuned to the strongC-H stretching vibrational mode [6]. In spite of that,picosecond CARS requires the synchronization of twolasers or parametric optical converters. Furthermore,detection covers just a constrained spectral windowat a time, giving only limited contrast and thereforehampering its applicability in the identification ofmaterials such as explosives or in chemically selectivemicroscopy.Such drawbacks can be easily overcome in the femto-
second regime, where an entire CARS spectrum can bedirectly measured using broadband excitation. A singlefemtosecond oscillator [7] or a supercontinuum [8] canseed the whole setup, as shown in single-beam CARS ormultiplex CARS (MCARS) [9–11] experiments. The in-creased signal generation due to shorter pulses andthe amount of spectral information acquired in a one-shotexperiment make femtosecond CARS well suited in caseof robust probes such as polymers. However, as a part ofthe excitation light does not match any Raman levels,such photons do not contribute to a specific molecular
signal. They just cause potential photodamage and in-crease the nonresonant contribution to the CARS signal.Furthermore, the readout times of CCD detectors usedto record spectra are much longer than those of single-channel ones and span the acquisition of a single imageover several minutes, limiting its application to fixedsamples.
In this Letter we combine the advantages of picose-cond with broadband femtosecond excitation, namely,fast acquisition and low photodamage with the capabilityof single-channel CARSmeasurement. In a MCARS setup,the broadband supercontinuum Stokes pulse is ampli-tude shaped to be resonant to specific Raman transitions[Fig. 1(a)]. The addition of the shaper gives access tohigh flexibility: specific Stokes-wavelengths are rapidlyswitched on and off all over the transmitted region andwith a resolution better than the pump bandwidth. Thebackground from regions free of vibrational levels is effi-ciently suppressed. In contrast to picosecond dual-pumpexperiments, where a second tunable source is added tothe setup, the broadband continuum can be tailored in anarbitrary number of Stokes pulses, limited only by thenumber of shaper pixels. With spectrally selective maskssensitive to just one chemical component, the scheme ispotentially suitable for one-channel detection with asso-ciated rapidity and lock-in detection ability. Chemicalmapping with tailored excitation and one-channel detec-tion is demonstrated.
A scheme of the experimental setup is presented inFig. 1(b). The 1:0 W (12 nJ per pulse) output of aTi:sapphire femtosecond oscillator is driving the setup.The 5-nm-wide spectrum initially available is split intoa 1-nm-narrow pump (at 780 nm), and an amount ofabout 100 mW is reflected into an end-sealed photoniccrystal fiber [(PCF) NL-PM-750, Crystal Fibre A/S]. Thebroadband supercontinuum is then sent through the 4fpulse shaper with a liquid-crystal mask.
In the presented setup, only the Stokes pulse is con-trolled by shaping, whereas the pump/probe pulse re-mains unchanged. This approach, with just one shapedpulse, allows a straightforward control over the probabil-ity of population for a Raman mode [AðΩÞ]:
where Ep and Es are the pump and Stokes electricfields, respectively. Mðω0Þ is an amplitude-only shapingmask. It avoids the complexity of previous CARSschemes with two or even three tailored interactions[7,12], where the effects of the applied phase or ampli-tude modulations are entangled and have to be takenin account by scanning additional parameters to retrieverelevant molecular information. A novel approach is theuse of evolutionary strategies to obtain molecule-specificexcitation pulses [13].
The selective excitation of a particular Raman-shift re-gion is demonstrated in Fig. 2. In Fig. 2(a), the CARSspectrum of CH2I2, obtained when all pixels are set tofull transmission, is compared with the one obtained byselecting a three-pixel window. It indicates that we canaddress any Raman transition from the fingerprint regionup to over 3000 cm−1. The excitation spectrum can beselectively varied according to Raman transition(s) usingthe spatial light modulator (SLM). Figure 2(b) exempli-fies that for poly(ethylene) terephthalate (PET) we areable to select simultaneously part of the fingerprint re-gion (1590 cm−1 to 1820 cm−1) and the band at 3049 cm−1
or to excite exclusively the band at 1780 cm−1. The widthof the amplitude shaping window is a compromise be-tween the resolution and Raman signal. In our setup,on average each pixel corresponds to about 7 cm−1,which is much narrower than the bandwidth of the pump:Δωp ¼ 16 cm−1. In general, it is not interesting to makethe Stokes spectrum narrower than the pump spectrumbecause the minimum bandwidth of the signal is limitedby the bandwidth of the pumpΔωp. Additionally the non-resonant background is suppressed outside the selectedregion with a factor more than 200, as shown in the insetof Fig. 2(a). The nonresonant background is, however,not suppressed in the selected spectral regions and willstill contribute to the well-known interferences with theresonant CARS signal. Nevertheless, this interferencedoes not hinder the ability of exciting specific Ramantransitions.
Such an ability can be now exploited to obtain chemicalselectivity in a polymer blend. In general, the microscopicstructure of polymers can be critical for their mechano-chemical properties but is difficult to investigate becausethe addition of dyes may modify the arrangement in theblend. Therefore, the ability of mapping these polymersat the micrometer scale without staining is of great inter-est for characterization of polymer blends, and it was suc-cessfully demonstrated with MCARS [9,11]. Our samplecontains polystyrene (PS), polyethylene (PE), and poly(methyl) methacrylate (PMMA). The CARS spectra ofpure polymers show several interesting characteristics:information is concentrated around 3000 cm−1, the bands
Fig. 2. (Color online) (a) CARS spectrum of CH2I2 with broad-band Stokes, i.e., full transmission (black curve) and with tai-lored Stokes, i.e., three pixels transmit light at the position ofthe traced curve (red). Inset, magnification of the CARS spec-trum with tailored Stokes shows background suppression (logscale). (b) Selection of regions of interest in the spectrum ofPET. Black curve, full spectrum; red curve, excitation with awindow in the fingerprint region and a narrow part around3049 cm−1; blue curve, narrowband excitation of the band at1780 cm−1.
are partly overlapping, and PE produces much strongersignals than PMMA.Nowwe demonstrate chemoselectivemapping of polymers in a ternary blend by combiningsingle-channel detection with amplitude tailored Stokesexcitation.To optimize the amount of information for the classi-
fication of polymers, we select the most relevant Ramanshifts. To this end, a mask is assigned to each polymer,cutting out a specific band and eliminating other wave-lengths from the Stokes pulse. Then the signal isintegrated over the pixels of the CCD camera from2800 cm−1 to 3500 cm−1 to simulate single-channel detec-tion in the region of interest. Based on the spectra of purepolymers, the chosen masks were carefully selected andcentered at 2835 cm−1 for PE, 2905 cm−1 for PS, and3022 cm−1 for PMMA [Figs. 3(a)–3(c)]. For each mask,the same square of 100 μm × 100 μm is scanned, with astep of 1 μm. The acquisition time per pixel is of200 ms. After integration, the image obtained with thePE mask is assigned the color blue, the PS mask red,and green for PMMA, as shown in Figs. 3(e)–3(g). Therecombination of these three images without further pro-cessing already gives a false-color mapping of the chosenarea in the blend [Fig. 3(h)]. Moreover, by tailoring theStokes spectrum, we were able to reduce the Stokes ra-diated power by almost a factor of 400. A moderate over-lap in the regions mapped [between PE and PMMAregions seen in Fig. 3(g)] shows how robust the methodis even when Raman bands with different intensitiesoverlap. Chemical selectivity in the case of overlappingspectra or for minor species can be improved by usingmore complicated amplitude tailored Stokes excitationspectra or further processing. This does not jeopardizethe speed of acquisition or induce significantly morephotodamage.In conclusion, we have shown successfully the appli-
cation of tailored Stokes excitation in MCARS to achievechemoselective imaging using single-channel detection.
Our approach combines the spectral resolution of anarrowband CARS setup (7 cm−1) with the large acces-sible bandwidth of a broadband CARS setup (1000–3500 cm−1), while guaranteeing low photodamage. Withthis scheme it was possible to distinguish three distinctchemical components of a polymer in a ternary blendwithout the use of any additional analysis algorithm.Furthermore, this shaper-assisted MCARS setup is akey advance toward real-time chemoselective imagingof sensitive samples, by taking full advantage of the com-plete vibrational spectrum of a molecule.
This work is supported by the Bundesministerium fürBildung und Forschung within the MEDICARS project.
References
1. S. A. Akhmanov, N. I. Koroteev, S. A. Magnitskii,V. B. Morozov, A. P. Tarasevich, and V. G. Tunkin,J. Opt. Soc. Am. B 2, 640 (1985).
2. A. M. Zheltikov, J. Raman Spectrosc. 31, 653 (2000).3. T. Lang, M. Motzkus, H. M. Frey, and P. Beaud, J. Chem.
Phys. 115, 5418 (2001).4. J. X. Cheng and X. S. Xie, J. Phys. Chem. B 108, 827 (2004).5. Y. Fu, H. Wang, R. Shi, and J. X. Cheng, Opt. Express 14,
3942 (2006).6. C. L. Evans and X. S. Xie, Annu. Rev. Anal. Chem. 1, 883
(2008).7. N. Dudovich, D. Oron, and Y. Silberberg, Nature 418, 512
(2002).8. B. von Vacano, W. Wohlleben, and M. Motzkus, Opt. Lett.
31, 413 (2006).9. T. W. Kee and M. T. Cicerone, Opt. Lett. 29, 2701 (2004).10. H. Kano and H. Hamaguchi, Opt. Express 86, 121113 (2005).11. B. von Vacano, L. Meyer, and M. Motzkus, J. Raman
Spectrosc. 38, 916 (2007).12. S. Postma, A. C. W. van Rhijn, J. P. Korterik, P. Gross, J. L.
Herek, and H. L. Offerhaus, Opt. Express 16, 7985 (2008).13. S. D. McGrane, R. J. Scharff, M. Greenfield, and D. S. Moore,
New J. Physics 11, 105047 (2009).
Fig. 3. (Color online) Chemical selectivity with single-channel detection. (a)–(c) CARS spectra with broadband Stokes (blackdotted curves) and with tailored Stokes spectra for each mask [respectively, for PE (blue), PS (red), and PMMA (green)]. (d) CARSspectra of pure polymers. (e)–(g) Chemical maps obtained with the respective amplitude masks. (h) Resulting RGB image ofpolymer.
