Multi-color stimulated Raman scattering microscopy with arapidly tunable optical parametric oscillator

Lingjie Kong†, Minbiao Ji†, Gary R. Holtom, Dan Fu, Christian W. Freudiger, and X. SunneyXie*

Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA

AbstractStimulated Raman scattering (SRS) microscopy allows label-free chemical imaging based onvibrational spectroscopy. Narrowband excitation with picosecond lasers creates the highest signallevels and enables imaging speeds up to video-rate, but sacrifices chemical specify in sampleswith overlapping bands compared to broadband (multiplex) excitation. We develope a rapidlytunable picosecond optical parametric oscillator with an electro-optical tunable Lyot filter, anddemonstrate multi-color SRS microscopy with synchronized line-by-line wavelength tuning toavoid spectral artifacts due to sample movement. We show sensitive imaging of three differentkinds of polymer beads and live HeLa cells with moving intracellular lipid droplets.

Coherent Raman scattering microscopy, as a label-free, highly sensitive, chemically specificimaging technique, has recently found wide applications in biomedical imaging [1–7].Stimulated Raman scattering (SRS) microscopy [4–7] has emerged as an alternative tocoherent anti-Stokes Raman scattering (CARS) microscopy, as it is free from spectraldistortions, straightforward in contrast interpretation, and is linearly dependents on analyteconcentration.

Narrowband excitation with picosecond lasers for SRS microscopy is commonly used due toits spectral resolving capability and high sensitivity, and video-rate molecular imaging invivo with epi-detected SRS has been demonstrated [7]. In such implementation, a singleRaman band is probed at a time. However, to distinguish M chemical components withoverlapping Raman bands in complex biological samples, N (N ≥ M) frequencies should bemeasured. Then the concentration maps of individual chemical species are numericallydecomposed from the multi-colored images based on multivariant analysis. Reliable multi-color SRS imaging methods are required to quantitatively resolve multiple chemical species.

In the narrowband implementation of SRS microscopy for complex samples, a single Ramanfrequency is probed throughout an image frame, and the laser frequency is sequentiallytuned frame-by-frame [8,9]. However, for live cell or tissue imaging with samplemovements, such as random organelle drifting on a time scale shorter than the image framerate, numerical decomposition will inevitably introduce spectral artifacts due to samplemotion, limiting the usefulness of frame-by-frame wavelength tuning based multi-color SRSmicroscopy [10].

Alternatively, femtosecond SRS [11] and multi-color SRS microscopy [12] were presentedwith simultaneous multi-channel spectral detection, similar to multiplex-CARS [13,14].

© 2012 Optical Society of America*xie@chemistry.harvard.edu.†These authors contributed equally to this work.

NIH Public AccessAuthor ManuscriptOpt Lett. Author manuscript; available in PMC 2013 April 15.

Published in final edited form as:Opt Lett. 2013 January 15; 38(2): 145–147.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Spectrally tailored excitation SRS microscopy [15] was demonstrated for highly specificmolecular imaging, implemented through the collective excitation of selected vibrationalfrequencies with a spatial light modulator based pulse-shaper. Modulation multiplexingbased multiplex SRS microscopy [10] was achieved by modulating individual wavelengthbands at different radio-frequencies with an acousto-optical tunable filter device. In all thesetechniques, the intensity information for multiple frequencies is obtained on a pixel-by-pixelbasis, making them less susceptible to sample motion induced spectral artifacts. But inpractice, their sensitivities are lower as part of the excitation power is wasted on spectralbands that do not have information, which decreases the imaging speed. Also, the high speedelectronics necessary to synchronize the independent femtosecond and picosecond sources issusceptible to timing jitter.

In this paper, we demonstrate a line-by-line wavelength tuning based SRS microscopy withan optical parametric oscillator (OPO) rapidly tuned by an electro-optical modulator, whichretains the high sensitivity and reduces spectral artifacts due to sample motion.

Figure 1(a) shows the diagram of the fast tuning OPO. It is based on a commercial OPO(Levante Emerald, APE-Berlin), which is synchronously pumped by a frequency-doubledNd:YVO4 mode-locked laser (76 MHz, 5 ps pulses at 532 nm, High-Q laser GmbH). Theoriginal Lyot filter is replaced by a homemade electro-optical tunable Lyot filter. Followingthe design shown in Ref.[16], it is made up of a static multi-order crystalline quartzwaveplate (United Crystal Company), a LiNbO3 based Pockel's cell (EO-AM-NR-C1,Thorlabs), and two Glan-Brewster prisms (MGLBAS5, Karl Lambrecht).

Parametric down-conversion is based on the provided temperature tuned LBO crystal (20mm long), with type-I noncritical phase matching. It is found that even though the phasematching condition selects the intracavity polarization, the Glan-Brewster prisms are stillnecessary to avoid discontinuities in tuning the wavelength. The temperature of the LBOcrystal and the cavity length are set to generate near-infrared signal at the desired centralwavelengths, and fast wavelength tuning is accomplished by applying sequential DCvoltages to the electric-optical modulator.

To achieve fast wavelength tuning, voltages on the Pockel's cell are applied stepwise withhomemade circuits. The tuning speed is limited by the slow slew rate of the operationalamplifier (PA97DR, Cirrus Logic), which takes ~100 μs to go from −400 V to +400 V. Thisis much longer than the timescale (~1 μs) for building-up of the steady-state operation,determined by cavity dynamics intrinsic to the OPO [16,17]. Figure 1(b) shows thewavelength tuning performance of the OPO, monitored with a fast spectrograph of 0.2 nmspectral resolution (USB 4000, Ocean Optics). Stepped voltage sweeps, increasing from−400 V to 400 V monotonically with increments of 50 V and durations of 15 ms, arerepeated 100 times, and the peak wavelengths are recorded. It is shown that the outputwavelengths change linearly and reproducibly with the applied voltages. With a 50-λwaveplate, the achieved tuning span around 805 nm is ~7.5 nm (covering ~115 cm−1), withthe free spectral range determined by the sum of the waveplate and the residual staticbirefringence of the electro-optical crystals. For every output, the spectral bandwidth is ~0.3nm, resulting in a spectral resolution of ~5 cm−1. Unlike the results of Ref.[18], hysteresis isobserved when the applied stepwise voltages are first increased and then decreased over thefull voltage range (data not shown).

The SRS signal is detected using a high-frequency modulation-transfer scheme [4]. Thetunable OPO signal is used as the pump beam for SRS. The 1064 nm beam is used as theStokes beam, and is amplitude modulated at 10 MHz. The spatially and temporallyoverlapped pump and Stokes beams are sent through a laser scanning microscope (IX81/

Kong et al. Page 2

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

FV1000, Olympus) and focused onto the sample. The SRS (stimulated Raman loss) signal isdetected with a homemade lock-in amplifier [7].

The FV1000 is operated in the `Line Kalman' mode, in which the galvos repeatedly scaneach physical line for multiple times before moving to the next. We utilize the line triggersignal from FV1000 to synchronize the driving voltages for the Pockel's cell by an analogoutput card (PCI-6229, National Instruments) with a custom Labview program, so that eachphysical line is scanned N times at N pre-selected OPO wavelengths sequentially.Wavelength switching occurs at falling edges of the line trigger, during the retrace of galvoswhen the image data is not collected. During each frame of the 2D galvo scan, N images atthe corresponding Raman frequencies are constructed from the simultaneously recorded rawSRS signals (PCI-6132, with another custom Labview program) in accordance to the linetrigger voltages from FV1000. Then the multivariant calibration based on least-squaresanalysis is applied to convert the raw image data into concentration maps of each chemicalcomponent [10,12].

We first evaluate the validity of spectral decomposition by imaging mixed beads coated on acover glass, including melamine, poly(methyl methacrylate) (PMMA), and polystyrene.Their spontaneous Raman spectra are shown in Fig. 2(a), from which the desired probingwavelengths and corresponding applied voltages can be chosen. Figures 2(b–d) show theSRS images (330×330 pixels) at 3054 cm−1, 2979 cm−1, 2959 cm−1 respectively, with a60× objective lens (NA 1.2, water). The calibration matrix is constructed [10] from the insitu imaging data of the three different beads, and is used to reconstruct the concentrationimage. Figure 2(e) shows the RGB composite image with each color for each beadcomposition. It can be seen that this scheme is well suited to distinguish multiple chemicalcomponents with overlapping Raman bands.

Next we image live HeLa cells to show the ability to minimize artifacts from organelledrifting during imaging. For cells or tissues containing primarily lipid and protein, probingat two Raman bands is sufficient to map their distribution. Here, the frequency shifts of 2838cm−1, and 2914 cm−1 (corresponding to pump wavelengths of 817.2 nm, and 812.1 nmrespectively) are used. As shown in Fig. 3(a), the 2914 cm−1 channel is attributed to bothlipid and protein, and the 2838 cm−1 channel is mainly attributed to lipid. The powers at theobjective focus are: 28 mW of Stokes, 81 mW at 817.2 nm, and 78 mW at 812.1 nm. Thepixel dwell time is 8 μs, and the wavelengths are switched every 5.184 ms. By pickingspecific regions with features corresponding to protein and lipid, the calibration matrix iscreated [10]. Figure 3(b) shows the retrieved image of the HeLa cell, which shows thedistribution of protein and lipid. Lipid droplets are shown in green, and protein is shown inred throughout the cytoplasm and nucleus. In comparison, the retrieved image fromsequential frame-by-frame wavelength tuning is shown in Fig. 3(c). After decomposition,the moving lipid droplets introduce artifacts as protein signals, such as the red dropletsmarked with arrows, which is obviously due to lipid droplet motion when compared withFig. 3(b).

In conclusion, we develop a rapidly tunable OPO with an elctro-optical tunable Lyot filterfor multi-color SRS microscopy. With this system, all the necessary spectral information foreach image line is recorded on the time scale of 10 milliseconds instead of seconds,substantially minimizing spectral artifacts due to sample motions compared to conventionalframe-by-frame sequential tuning. Also, the current system can be used for hyperspectralimaging, to get detailed information of a vibrational spectrum [8,9]. With sequential line-by-line wavelength tuning, imaging of mixed beads and live HeLa cells are demonstrated.

Kong et al. Page 3

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

AcknowledgmentsWe thank Fa-ke Lu, and Wenlong Yang for help on preparing HeLa cells and helpful discussions. This work wassupported by the National Institute of Health's T-R01 (1R01EB010244-01) awarded to X.S.X.

Referrences1. Zumbusch A, Holtom GR, Xie XS. Three-dimensional vibrational imaging by coherent anti-Stokes

Raman scattering. Physical Review Letters. 1999; 82:4142–4145.

2. Cheng JX, Volkmer A, Book LD, Xie XS. An epi-detected coherent anti-stokes raman scattering (E-CARS) microscope with high spectral resolution and high sensitivity. Journal of Physical ChemistryB. 2001; 105:1277–1280.

3. Evans CL, Xie XS. Coherent Anti-Stokes Raman Scattering Microscopy: Chemical Imaging forBiology and Medicine. Annual Review of Analytical Chemistry. 2008; 1:883–909.

4. Freudiger CW, Min W, Saar BG, Lu S, Holtom GR, He C, Tsai JC, Kang JX, Xie XS. Label-FreeBiomedical Imaging with High Sensitivity by Stimulated Raman Scattering Microscopy. Science.2008; 322:1857–1861. [PubMed: 19095943]

5. Ozeki Y, Dake F, Kajiyama S. i. Fukui K, Itoh K. Analysis and experimental assessment of thesensitivity of stimulated Raman scattering microscopy. Optics Express. 2009; 17:3651–3658.[PubMed: 19259205]

6. Nandakumar P, Kovalev A, Volkmer A. Vibrational imaging based on stimulated Raman scatteringmicroscopy. New Journal of Physics. 2009; 11:033026.

7. Saar BG, Freudiger CW, Reichman J, Stanley CM, Holtom GR, Xie XS. Video-Rate MolecularImaging in Vivo with Stimulated Raman Scattering. Science. 2010; 330:1368–1370. [PubMed:21127249]

8. Suhalim JL, Chung C-Y, Lilledahl MB, Lim RS, Levi M, Tromberg BJ, Potma EO. Characterizationof Cholesterol Crystals in Atherosclerotic Plaques Using Stimulated Raman Scattering and Second-Harmonic Generation Microscopy. Biophysical Journal. 2012; 102:1988–1995. [PubMed:22768956]

9. Ozeki Y, Umemura W, Sumimura K, Nishizawa N, Fukui K, Itoh K. Stimulated Ramanhyperspectral imaging based on spectral filtering of broadband fiber laser pulses. Optics Letters.2012; 37:431–433. [PubMed: 22297376]

10. Fu D, Lu F-K, Zhang X, Freudiger C, Pernik DR, Holtom G, Xie XS. Quantitative ChemicalImaging with Multiplex Stimulated Raman Scattering Microscopy. Journal of the AmericanChemical Society. 2012; 134:3623–3626. [PubMed: 22316340]

11. Ploetz E, Laimgruber S, Berner S, Zinth W, Gilch P. Femtosecond stimulated Raman microscopy.Applied Physics B-Lasers and Optics. 2007; 87:389–393.

12. Lu F-K, Ji M, Fu D, Ni X, Freudiger CW, Holtom G, Xie XS. Multicolor stimulated Ramanscattering microscopy. Molecular Physics. 2012; 110:1927.

13. Cheng JX, Volkmer A, Book LD, Xie XS. Multiplex coherent anti-stokes Raman scatteringmicrospectroscopy and study of lipid vesicles. Journal of Physical Chemistry B. 2002; 106:8493–8498.

14. Wurpel GWH, Schins JM, Muller M. Chemical specificity in three-dimensional imaging withmultiplex coherent anti-Stokes Raman scattering microscopy. Optics Letters. 2002; 27:1093–1095.[PubMed: 18026371]

15. Freudiger CW, Min W, Holtom GR, Xu B, Dantus M, Xie XS. Highly specific label-freemolecular imaging with spectrally tailored excitation-stimulated Raman scattering (STE-SRS)microscopy. Nature Photonics. 2011; 5:103–109. [PubMed: 23015809]

16. Saar BG, Holtom GR, Freudiger CW, Ackermann C, Hill W, Xie XS. Intracavity wavelengthmodulation of an optical parametric oscillator for coherent Raman microscopy. Optics Express.2009; 17:12532–12539. [PubMed: 19654655]

17. Godard A, Rosencher E. Energy yield of pulsed optical parametric oscillators: A rate-equationanalysis. Ieee Journal of Quantum Electronics. 2004; 40:784–790.

Kong et al. Page 4

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

18. Ewbank MD, Rosker MJ, Bennett GL. Frequency tuning a mid-infrared optical parametricoscillator by the electro-optic effect. J. Opt. Soc. Am. B. 1997; 14:666–671.

Kong et al. Page 5

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig. 1.(a) Schematic of the fast tuning OPO. DM: dichroic mirror, M: mirror, G: Glan-Brewsterprism, W: waveplate, EO: electro-optical crystal, LBO: lithium triborate crystal. (b) Tuningcharacteristic of the OPO with an electric-optical tunable Lyot filter. The dots show averagevalues, and the error bars show the standard deviations after 100 repetitions.

Kong et al. Page 6

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig. 2.(color online) (a) Spontaneous Raman spectra of melamine, PMMA, and polystyrene beads.(b–d) Raw images taken at 3054 cm−1, 2979 cm−1, and 2959 cm−1. (e) Reconstructed SRSimage (Red: 2 μm polystyrene, Green: 1–10 μm PMMA, Blue: 2.9 μm melamine).

Kong et al. Page 7

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

Fig.3.(color online) (a) Spontaneous Raman spectra of protein and lipid. (b) Reconstructed SRSimage by line-by-line tuning, free of spectral artifacts due to lipid droplet driftings. (c)Reconstructed SRS image by frame-by-frame tuning. The two arrows point to artifacts dueto lipid droplet driftings. Red: protein, green: lipid. Both images are 460×318 pixels. Thescale bar is 10 μm. (60×water objective lens used)

Kong et al. Page 8

Opt Lett. Author manuscript; available in PMC 2013 April 15.

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript

NIH

-PA Author Manuscript