1. Introduction

Coherent Anti-Stokes Raman Scattering (CARS) spectroscopy is very effective in enhancing the weak and incoherent Raman signals which are a limiting factor in the application of conventional Raman microscopy. A pump laser and a Stokes laser are able to create coherent populations in a given vibrational state and result in coherent addition of the CARS field to increase the signal quadratically with respect to the number of molecular oscillators [1]. The CARS signal is generated by exciting a Raman-active vibrational mode (ω_v) of a chemical species, for example, the symmetric CH₂ stretching and C=C stretching, could be excited with a pump (ω_p) and a Stokes (ω_s) laser through a four-wave mixing process. By tuning the frequency difference between these near-infrared (NIR) pump and Stokes lasers to match the vibrational frequency of a molecule, ω_p - ω_s = ω_v, the resonance-enhanced anti-Stokes signal (ω_p + ω_v) can be produced in the easily detected visible spectral region. Since it relies on the native chemical species of the specimen, CARS microscopy has been one of the key tools in providing spatial distributions of chemical information at molecular level without additional labeling [2,3]. Moreover, CARS is a χ⁽³⁾ nonlinear process, so it provides intrinsic optical sectioning capability for three-dimensional imaging, similar to other nonlinear imaging modalities [4,5].

It is obvious that a light source system with two or more wavelengths is required to perform CARS imaging. To sufficiently cover the Stokes shift for chemical identification, the light source should provide a broad range of wavelength tunability to access the full CARS spectral range and to extend the feasibility to variety samples. Meanwhile, to observe this nonlinear phenomenon, a pulsed laser with strong peak intensity is required. The linewidth of the laser should be limited to several cm⁻¹, therefore a picoseconds laser is adequate to ensure the spectral resolution. To generate CARS efficiently, it is vital to temporally overlap the pump and Stokes lasers, whose wavelengths are significantly different.

There has been much effort in developing a suitable picosecond light source suitable for CARS microscopy [6–18]. Two electronically synchronized picosecond mode-locked Ti:Sapphire lasers with timing jitter less than 0.5 ps were first used. A picosecond synchronously pumped optical parametric oscillator (OPO) was latter adopted to eliminate the timing jitter [6–9]. Petrov and Yakovlev [10,11] who combined the red-shifted supercontinuum output and a home-made, 2-4 ps, several MHz of 1,064 nm laser demonstrated that this setup was very suitable for recording the multiplex CARS spectra. The ultrabroadband multiplex CARS spectra and imaging were also shown by using a microchip laser of 6.6 kHz repetition rate, ~ <1 ns pulse width and a supercontinuum light source from a photonic crystal fiber [12]. But with this 6.6 kHz repetition rate, a longer exposure time for each pixel is needed for acquiring the CARS images. Meanwhile, there are a lot of multiplex CARS spectroscopy and imaging done by combining femtosecond Ti:Sapphire lasers with photonic crystal fibers [13–16]. Until recently, Pegoraro et al. [17] reported a compact all fiber-based CARS laser pumped by a femtosecond laser to demonstrate the capability for the multimodal nonlinear microscopy. However, these femtosecond pumped CARS lasers can not provide a better spectral resolution than picosecond lasers.

Jurna et al. [18] applied an efficient, picosecond, tunable, narrow-bandwidth, noncritical phase-matched lithium triborate based OPO laser to CARS spectroscopy and microscopy. Ganikhanov et al. first demonstrated CARS microscopy by generating the signal and idler beams in the same optical path from a non-critically phase-matched synchronously pumped OPO cavity as the pump and Stokes beams [19]. This design based on the use of OPO’s was made commercially available (picoEmerald, APE GmbH, Germany), which requires no extra efforts to overlap the pump and Stokes lasers [20–22]. However, optical resonant cavities are still required to increase the optical conversion efficiency and add additional complication. Krauss et al. [23] presented a very compact, widely wavelength tunable, Er:fiber laser, which had covered the vibrational frequencies from 1150 cm⁻¹ to 3800 cm⁻¹.

Recently, Tzeng et al. [24] demonstrated a sub-picosecond single-pass supercontinuum-seeded optical parametric amplifier (SCOPA) which could generate a tunable NIR laser source from 0.7 to 1.9 μm by adjusting the temperature of the magnesium oxide doped periodically poled lithium niobate (MgO:PPLN) crystal. The approach naturally overlaps the pump and Stokes beams without using resonant cavities and this is very suit for CARS microscopy. Compared to the electronically synchronized Ti:Sapphire lasers, the SCOPA approach eliminates the timing jitter and thus is free from associated complicated electronic control. The tunable range of SCOPA extends from 700 nm to 2000 nm, which is much larger than a picosecond mode-locked Ti:Sapphire laser, sufficient to access a wider range of chemical compounds.

In this work, we describe the new constructed supercontinuum-seeded optical parametric amplifier laser driven by a picosecond mode-locked laser to generate pairs of pump/Stokes beams for CARS spectroscopy. With this cavity-free and multi-band NIR laser source, we developed a CARS imaging system to demonstrate a microscopy system providing a rapid frequencies tuning, and high efficiency in spatial and timing overlapping for CARS signal generation. This novel imaging system exhibits a significantly reduced of the requiring total laser power with a narrow bandwidth and illustrates the capability of frequency doubling imaging by a sample of 2 μm polystyrene beads, in addition to the labeling-free advantage, this system will enable imaging of subcellular components, pharmaceuticals in delivery constructs, distributions in monolayer functionalized surfaces.

2. Experimental setup

The experimental setup of a SCOPA-based CARS microscope is shown in Fig. 1 . The fundamental laser was a passively mode-locked Nd:YVO₄ laser with 7.5 ps pulse width, average power 4.5 W, and 80 MHz repetition rate (IC-1064-10000, High-Q laser, Austria). This fundamental was first frequency-doubled by a 2.5 cm, type-I noncritical phase-matching lithium triborate crystal (Castech Inc., China) to generate a 1W second harmonic (SH) as the pump laser of SCOPA. The residual 1,064 nm laser of 2W was coupled into a 2.5 m photonic crystal fiber (SC-5.0-1040, Crystal Fibre, Denmark) to generate a supercontinuum (SC) as a seeding source of SCOPA. The optical path difference between the SH beam and SC beam was optimized by an optical delay line to maximize the output power of this SCOPA.

 

Fig. 1 The optical layout of a SCOPA-based CARS microscope. Note that CARS imaging can be implemented by simply guiding the output of SCOPA into a microscope. O.I.: optical isolator; HWP: half wave plate; LBO: lithium triborate; DM: dichroic mirror; PCF: photonic crystal fiber.

Download Full Size | PDF

In this SCOPA, the nonlinear crystal was a 2-cm-long, temperature-controlled MgO:PPLN crystal (HC Photonics, Taiwan) with high nonlinearity and controllable phase-matching condition. The poling period of the MgO:PPLN crystal varies from 6.8 to 7.8 μm with a 0.2 μm step increment. The temperature of the MgO:PPLN crystal was adjusted from 30 to 200 ^oC to select the wavelengths of the signal wave from 750 to 1,021 nm and the idler wave from 1,111 to 1,830 nm. The tuning curve of this MgO:PPLN crystal is shown in Fig. 2 . The black solid lines were experimental fitting curves according to the published Sellmeier equation in recent literature [25]. Under this setup, the detectable Raman-active vibrational frequencies are from 396 cm⁻¹ to 7,868 cm⁻¹.

 

Fig. 2 The complete SCOPA tuning curve versus temperature with six different MgO:PPLN grating periods at temperatures from 30 °C to 200 °C. The black solid lines were experimental fitting curves according to the Sellmeier equation [25].

Download Full Size | PDF

The maximum output power of SCOPA with 1W SH beam pumping is about 100 mW, which corresponds to 10% conversion efficiency (signal + idler beams) and is equivalent to 1.2 nJ/pulse. The pulse width and spectral width of the signal beam at 795 nm are measured to be ~1 ps and 12 cm⁻¹. Mainly, the SCOPA laser provide three pairs of frequencies for CARS excitations in the system with the Nd:YVO₄ fundamental frequency. The first pair is the signal and residual 1,064 nm beams, of which the latter was extracted from the dichroic mirror. The second pair is the signal and idler beams through the OPA process. The other pair is the residual 1,064 nm and idler beams, generating inherently the CARS frequency of the signal of parametric process, thus it is difficult to extract the CARS signal from the signal OPA background. When the output power of SCOPA was maximized, the spatial and temporal overlapping between 1,064 nm, signal and idler waves were also optimized. The SCOPA laser beam was steered into an inverted microscope (TE2000U, Nikon, Japan) with two broadband high-reflection mirrors. Then, an oil-immersion objective (Plan Apo VC, 60X, NA 1.4, Nikon, Japan) focused the laser beam to the sample which was placed on a XY piezo driven specimen stage (NS821, Nano Control, Japan). The forward anti-Stokes signals were collected by another objective (UPLSAPO, 20X, NA 0.75, Olympus, Japan), sent through a shortpass filter (FES700, Thorlabs), dispersed by a monochromator with a 300 grooves/mm grating (iHR320, Horiba Jobin Yvon, Japan), and recorded by a charge-coupled device (DU940, Andor Technology, Ireland) or a photomultiplier tube (R928, Hamamatsu Photonics, Japan). The data acquisition program was written in a graphically-based programming language of Labview 8.6 (National Instruments). The average power of total lasers for acquiring the CARS images of 2 μm polystyrene beads was less than 10 mW.

3. Results and discussion

By tuning the wavelength of the SCOPA output signal from 825 nm to 797 nm by pumping the MgO:PPLN crystal of 7.4 μm period with SH beam, the CARS spectrum of polystyrene was recorded from 2,725 cm⁻¹ to 3,150 cm⁻¹. After considering the wavelength-dependent output powers of signal beam, the spectrum is shown in Fig. 3(a) , with a wavelength-fixed Stokes beam coming from a residual 1,064 nm laser, revealing the aliphatic C-H symmetric stretching mode at 2,910 cm⁻¹ and C-H stretching mode on the phenyl ring around 3054 cm⁻¹. By selecting the signal wavelength at 803 nm, which is resonant with the aromatic stretching mode, Fig. 3(b) shows a spectrum of the anti-Stokes signals. Note that two peaks at 645 nm and 539 nm, can be observed simultaneously. If the 1,064 nm beam was blocked, the 645 nm peak disappeared completely and left the 539 nm peak unchanged. In contrast, if the idler beam at 1,576 nm was removed, the peak at 539 nm diminished while the 645 nm peak was still observable. Therefore, we conclude that the 645 nm was the CARS signal involving a 803, 1,064 nm beam and the 539 nm was the CARS signal related to the 803 nm beam and idler beam of 1,576 nm.

 

Fig. 3 A CARS spectrum (a) and anti-Stokes signals (b) of a polystyrene bead.

Download Full Size | PDF

We note that the energy difference between the 803 nm and 539 nm signals was 6,108 cm⁻¹, double the fundamental aromatic C-H stretching frequency. We suggest the 539 nm is the result of the non-resonant overtone excitation of the aromatic C-H band by the 803 nm and idler beam of 1,576 nm. This finding implies that a pump beam with wavelength from 750 to 1,021 nm coupled with a Stokes beam of 1,830 to 1,111 nm or 1,064 nm, can perform CARS imaging of molecular vibrational bands in a spectral range as broad as 396 to 7,868 cm⁻¹.

The CARS images of 2 µm polystyrene beads in Fig. 4(a) and 4(b) were obtained by raster-scanning with the MgO:PPLN crystal setting for poling period 7.4 μm and 130 ^oC, from which a signal beam of 803 nm and an idler beam of 1,576 nm were generated. Fig. 4(a) shows the CARS image of the fundamental aromatic C-H band of 2 μm polystyrene beads at 645 nm. A blurred CARS image, Fig. 4(b), was observed at 539 nm. We attributed this image at 539 nm to the non-resonant excitation by 803 and 1,576 nm. By adjusting the MgO:PPLN temperature to 116 ^oC, a better signal-to-noise ratio image was obtained, as shown in Fig. 4(c). These observations lead us to conclude that the overtone CARS image is indeed the aromatic C-H stretching mode due to resonant enhancement of the overtone state. At 116 ^oC, the SCOPA generates a signal of 808 nm beam and an idler of 1,556 nm beam, then they work as pump and Stokes beams, respectively, to create the CARS signal at 539nm equivalent to the energy difference of 5,945 cm⁻¹, which was consistent with the reported value of the first overtone aromatic C-H stretching mode of polystyrene [26]. In comparison with Fig. 4(a), the slight loss of contrast in Fig. 4(c) may partially results from the reduced cross section of vibrational overtone, decreased Raman effect with a longer wavelength quadratically, which are inherent of the optical properties of the Raman scattering. The chromatic aberration of the objective could contribute to part of the contrast loss in the vibrational overtone imaging due to the longer wavelength implemented. The latter could be improved with a NIR-compensated objective or a reflective objective.

 

Fig. 4 CARS images of 2 μm polystyrene beads with field-of-view of 10x10 μm² and a 0.1 μm step size. The CARS images for the 645 nm and the 539 nm are shown in (a) and (b), respectively. The overtone CARS imaging is shown in the image(c). Scale bar: 2 μm.

Download Full Size | PDF

 

Fig. 5 The in vivo CARS image of C. elegans with field-of-view of 80x80 μm² and 0.2 μm step size. Scale bar: 20 μm.

Download Full Size | PDF

The in vivo CARS image of pharynx of the wild type of nematode Caenorhabditis elegans (C. elegans) at C-H stretching mode (2845 cm⁻¹) in Fig. 5 was also acquired. We used 1% agar pad to immobilize the C. elegans. The solution of the pad is M9 which doesn’t contain any anesthetics. The function of M9 buffer is like the normal saline. The stage of C. elegans of CARS image is L3 to L4 stage. The nerve system in the pharynx could be clearly observed. This CARS image was recorded with a total average power of 40 mW and the acquisition time of 10 minutes.

4. Conclusion

We demonstrated that a picosecond version of a single-pass SCOPA can produce the required light source to perform CARS microscopy. This cavity-free laser source is appropriate for CARS generation by providing the pump and Stokes beams simultaneously. No extra optical setup is required for overlapping these beams temporally and spatially. Simple incorporation of two planar mirrors to send the SCOPA into a microscope is sufficient to acquire CARS signals. This SCOPA system is tunable easily to produce parametric pairs in the wavelength range from 700 to 1900 nm of which is sufficient to probe most of molecular vibrational levels. Hence, this system is providing an opportunity to probe multiple vibrational modes at once for CARS imaging. Its operation is quite robust, requires no cavity or additional mirror adjustment during wavelength tuning. Moreover, the picosecond laser pulses provide not only a reasonable high peak intensity for efficient CARS generation, but also a narrower linewidth for better spectral resolution. With these advantages, we can therefore conclude that this compact SCOPA laser driven CARS microscopy can be a valuable tool for noninvasive, label-free biomedical imaging applications.