1. Introduction

CARS microscopy is an indispensable tool for inspecting biological specimen and other materials and thus has been extensively developed along with stimulated Raman scattering microscopy in the past fifteen years [1–3]. Since the performance of a CARS system is mainly dominated by its light source, development of light source has been a main stream of CARS study [4–18]. Especially, recent development of supercontinuum-based ultrabroadband multiplex CARS systems has enabled deep inspection of biological tissues [4, 5]. On the other hand, additional requirements such as compactness, high stability (being alignment free), and low cost are essential for further development of CARS microscopy, while they have been mostly overlooked. Actually, most of the milestones of CARS microscopy have been achieved by an elaborate system that includes a large-scale, expensive laser such as a modelock laser implemented on an optical table and that requires careful alignment of optical system. This fact prevents spread of CARS microscopy to potential users such as biologists, biochemists, and medical scientists. Moreover, endoscopic applications strongly necessitate the above requirements [19–21]. Recently, fiber-based light sources for CARS microscopy that aims at these requirements were reported [6–10]. However, at this moment, their bandwidths are not satisfactorily broad and especially do not fully cover the fingerprint region (500-1800 cm-1), which is an essential region for investigation of biological tissues. Single-pulse CARS systems may have potential of a compact implementation because of its collinear configuration [11–14]. However, they include some bulky components such as prism compressor and pulse shaping system, and thus no compact implementation has been reported.

In this paper, we report a compact light source module for ultrabroadband CARS microscopy on the basis of the previously reported supercontinuum-based configuration [4]. The module has unprecedentedly compact size that enables direct connection to commercial Raman microscopes or spectrometers. Additionally, it is completely alignment-free and relatively low-cost compared with standard light sources for CARS such as modelock laser or optical parametric oscillator. The point of the development is to lead pump light into a single mode fiber for compact implementation. As will be shown below in detail, pump light can cause (unwanted) nonlinear effects in the fiber. Thus we investigated conditions of fiber lengths and incident light powers where nonlinear effect in a single mode fiber is negligible and a photonic crystal fiber generates sufficiently broad Stokes light.

This manuscript is organized as follows. In Section 2, basic configuration, detailed design, and packaging of the light source module are described. In Section 3, observation of ultrabroadband CARS spectra and CARS images from a murine adipocyte will be reported, which is a performance test of the developed light source. Finally, in Section 4, further improvement of the light source is discussed.

2. Development of a compact light source

2.1 Basic configuration

A schematic of a newly developed light source is shown in Fig. 1. The basics of the setup are common to those of the previously reported ultrabroadband multiplex CARS microscopy [4]. An output from a microchip laser (HorusLaser HLX-I-F040, ~1.1 ns pulse duration, 1064 nm center wavelength, ~495 mW average power, ~27 kHz repetition rate) is lead into a photonic crystal fiber (PCF, NKT Photonics SC-5.0-1040-PM) and a polarization-maintaining single mode fiber (PMF, Nufern PM980-XP) after split by a polarization beam splitter (PBS). The power ratio was adjusted by a half waveplate that was inserted on the beam pass in front of the PBS. An isolator was also inserted on the pass to reject return light to the laser. A supercontinuum light is generated in the PCF, which is used as a Stokes light for CARS generation, and a longpass filter inserted on an output beam path of the PCF to reject the original laser and anti-Stokes light components. A laser light propagates the PMF without major temporal/spectral change and is used as pump light for CARS generation. Tiny wavelength components that can be generated by nonlinear effects in the fiber were rejected by a bandpass filter placed at the output beam path of the fiber. The pump and the Stokes light beams are combined by a dichroic mirror and are set to be collinear. Incident light polarization directions for the PCF and PMF were set to be parallel with one of the propagation axes directions of the fibers. The polarization of the pump and the Stokes light beams were set to be the same.

 

Fig. 1 A schematic of experimental setup. HWP, half-wave plate; PBS, polarization beam splitter; PCF, photonic crystal fiber; LPF, longpass filter; PMF, polarization maintaining fiber; BPF, bandpass filter; DM, dichroic mirror.

Download Full Size | PDF

2.2 Optical design

Initially we investigated proper fiber lengths and incident light powers, which dominantly determine the performance of the present setup. As a starting point, it should be noted that the lengths of the PCF and the PMF should be nearly the same to assure temporal overlap of the pump and the Stokes light beams after combination. We confirmed that the group velocities of the pump light in the PMF and the Stokes light in the PCF were almost the same, and thus the fiber lengths were set to be the same. The required accuracy of the optical path length difference is a few centimeters so that it is much shorter than pulse lengths of the pump and the Stokes light beams in a free space (~30 cm).

Next we measured Stokes light powers generated from the PCF with various lengths to investigate a proper length. The powers were measured at the output of the PCF through the longpass filter and the incident light power was set as maximum (~480 mW). As shown in Fig. 2(a), against intuition, it was found that shorter length is beneficial for higher efficiency of Stokes light generation, as long as the length is in the range of 3-20 m. It is deduced that this result is caused by dissipation of the Stokes light in the PCF. An observed spectrum of the supercontinuum light for 3 m fiber length was sufficiently broad as shown in Fig. 2(b). However, too short fiber length may cause insufficient broadening of supercontinuum, which results in limited bandwidth of CARS microscopy. Actually, we observed such a spectrum with 1 m PCF and 200 mW incident power as shown in Fig. 2(c). Additionally, the degree of broadening depends on incident light power, too. Thus the length should be determined by considering possible incident light power so that sufficient broadening is obtained.

 

Fig. 2 (a) Stokes light powers for various PCF lengths. (b) A power spectrum of supercontinuum after 3 m PCF. (c) A power spectrum of supercontinuum after 1 m PCF.

Download Full Size | PDF

In turn, we investigated characteristics of output of the PMF for various incident light powers. Powers of incident light wavelength component, temporal waveforms of incident light wavelength components, and spectra of a 3 m PMF output for various incident light powers are shown in Fig. 3(a)-3(c), respectively. A bandpass filter (bandwidth: 10 nm) was inserted in the output beam path to obtain data in Fig. 3(a) and 3(b). A 3 GHz detector was used to observe waveforms in Fig. 3(b). In Fig. 3(a), it was observed that the output power saturates for larger incident light powers. It is obvious from Fig. 3(b) and 3(c) that this saturation is caused by wavelength conversion at high power temporal region [22, 23]. This result means that there is a power limitation of incident light powers for the PMF.

 

Fig. 3 (a) Pump powers, (b) temporal waveforms of pump light component, and (c) observed spectra, at 3 m PMF output for various incident light powers.

Download Full Size | PDF

From the above results, we determined the design parameters as follows. First we set the fiber lengths of the PCF and the PMF as ~3 m. Then we set the incident light power for the PMF as the maximum one where no saturation occurs. Then the rest of the light was led into the PCF. At this condition, the Stokes light had a sufficiently broad spectrum that covers > 3000 cm⁻¹.

2.3 Packaging

On the basis of the above investigations, we designed and assembled a light source module. A design drawing and a picture of the module are shown in Fig. 4(a) and 4(b), respectively. By using the fiber configuration, the part of laser beam coupling to the fibers and that of beam combining were separately packaged. The former had a dimension of 22 × 8.5 × 11.4 cm³. The latter had a SM1 internal thread, which enables direct connection to standard optical holders. All of the components were fixed by epoxy adhesive or YAG laser welding, which realized completely alignment-free setup. After the assembly, we confirmed that the temporal overlap of the pump and the Stokes light beams was nearly optimal as shown in Fig. 4(c), which is essential for efficient CARS generation.

 

Fig. 4 (a) Design drawing of the light source module. (b) Assembled module. (c) Observed temporal waveforms of the pump and the Stokes light beams from the assembled module.

Download Full Size | PDF

3. Application to CARS microscopy

We applied the developed light source module to CARS microscopy. A schematic of the experimental setup is shown in Fig. 5. An output beam of the light source is focused on a sample by an objective (Nikon Plan Apo IR 60x NA1.27WI). The CARS light generated from the sample is collected by another objective (Nikon S Plan Fluor ELWD 40x NA0.60), passes thorough a shortpass filter and a notch filter, and is incident on a spectrometer (Princeton Instruments Acton SP-2358). Theoretical lateral and depth resolution of a CARS image for a low-contrast sample calculated from these objectives are ~0.2 μm and ~0.6 μm, respectively, which are defined as the inverse of the maximum lateral and depth components of cutoff frequencies of optical transfer function [24]. The power spectrum of the CARS light is then measured by a CCD camera (Princeton Instruments PIXIS 400BR) attached to the spectrometer. The sample was placed on a piezo stage (Mad City Labs Nano-LPS100) and the focused beam position was scanned by the stage. The sample was polystyrene and an adipocyte differentiation of the C3H10T1/2 cell line. An observed CARS spectrum was normalized by that of medium around the cell (alphaMEM). A Raman spectrum was reconstructed form the normalized CARS spectrum by the maximum entropy method [25]. Proper spline interpolation was performed to reject phase offsets in the reconstructed Raman spectra [25].

 

Fig. 5 Schematic of experimental setup for CARS microscopy with the light source module.

Download Full Size | PDF

First we observed a CARS spectrum and reconstructed a Raman spectrum of polystyrene as shown in Fig. 6(a) and Fig. 6(b), respectively. Known Raman peaks of the polystyrene were successfully observed in the range of 500-3200 cm⁻¹ [26]. Next we reconstructed a Raman spectrum of a lipid droplet in a cell in the range as shown in Fig. 6(c). The exposure time was 90 ms. Not only a strong peak around 2900 cm⁻¹ (mixture of CH₂ and CH₃ stretches) but also resonances in the fingerprint region such as peaks at ~1650 cm⁻¹ (amide I) and ~1440 cm⁻¹ (C-H bending) were observed [27]. Relatively higher noise level at lower wavenumber region may be caused by imperfect spatial overlap of the pump and the corresponding Stokes light components at the focus, which can be improved by correcting axial chromatic aberration of the focusing objective. A lipid concentration map generated by peak values at ~2900 cm⁻¹ in reconstructed Raman spectra is shown in Fig. 7. The field of view, pixel numbers, and exposure time for each pixel were 16 × 16 μm², 64 × 64, 30 ms, respectively. Strong signals from a lipid droplet near the center and relatively weak signals representing the cell morphology that may attribute to lipid bilayer were successfully observed.

 

Fig. 6 (a) An observed CARS spectrum of polystyrene normalized by nonresonant CARS spectrum of water. (b) A reconstructed Raman spectrum of polystyrene. (c) A reconstructed Raman spectrum of a murine adipocyte.

Download Full Size | PDF

 

Fig. 7 CARS image of a murine adipose cell. Scale bar: 5 μm.

Download Full Size | PDF

4. Discussion

For further improvement of CARS generation efficiency, a straightforward way is to increase the pump/Stokes light power levels. Choosing fiber lengths shorter than 3 m may result in such improvement. The best lengths depend on the maximum power of a microchip laser and required bandwidth. Simply choosing a microchip laser with a higher power would also be beneficial. Although we used a commercial PCF that was designed to generate supercontinuum light with 1064 nm wavelength of the pump light, fine tuning of design parameters of the PCF on the basis of target wavelength range may result in improvement of the Stokes light power.

On the other hand, especially when a biological sample is observed, sample damage should be considered at the same time. In the present implementation, total light power in front of the focusing objective was ~100 mW at maximum, which can easily damage a biological sample. Therefore other strategies such as higher numerical aperture or reduction of axial chromatic aberration of a focusing objective should be taken for higher efficiency when fragile samples are assumed.

Although unprecedentedly compact implementation of ultrabroadband CARS light source has been achieved in this work, the present configuration may have some possibilities of further downsizing. First, an isolator could be removed if angled endfaces for the fibers are applied. Second, size of a microchip laser can potentially be downsized by appropriate thermal design. Moreover, beam splitting by a PBS and beam combining by a dichroic mirror may be replaced by fiber couplers, which enables all-fiber configuration. For this implementation, couplers should be carefully designed to minimize power loss at joints between a coupler and a fiber.

5. Conclusion

We developed a compact light source for ultrabroadband CARS microscopy. It is alignment-free and relatively low-cost compared with previous light sources of CARS microscopy. Critical design parameters such as fiber lengths and incident light powers were determined to minimize unwanted nonlinear effect and to obtain sufficiently broad Stokes light. By using an assembled module, we successfully observed an ultrabroadband CARS spectrum and a CARS image of a murine adipocyte. The module is expected to greatly spread the CARS microscopy to various fields by its extreme easiness to handle.