1. Introduction

Coherent anti-Stokes Raman scattering (CARS) imaging has been demonstrated as a powerful tool for label-free optical imaging applications. It possesses many merits such as chemically selective contrasts based on Raman vibrational activity, high sensitivity, rapid acquisition rates, inherent three-dimensional sectioning and sub-micron spatial resolutions [1–3]. Consequently, CARS microscopy has been widely employed for a variety of important biomedical applications, including imaging of viruses, cells, tissues, and live animals, using signals originated from CH₂-rich molecular structures [4–10]. In a CARS process, two laser beams are used for excitation, a pump beam at frequency ω_p and a Stokes beam at frequency ω_s (ω_s<ω_p). They are tightly focused onto the sample, creating a composite electric field beating at the frequency Ω=ω_p − ω_s. When this beating frequency matches a vibrational frequency of the molecules, it synchronizes the molecular oscillators, producing strong excitation across the entire focal volume. The resulting emission signal at the anti-Stokes frequency (ω_CARS=2ω_p - ω_s) is orders of magnitude stronger than the conventional spontaneous Raman signals, enabling its superior signal-to-noise ratio and video rate imaging speed for various applications [10].

Recently, optical fiber delivered CARS imaging system has attracted major research attention because of its flexibility for optical alignments as well as potentials for CARS endoscopy applications [11–15]. For biomedical applications, it is desirable to implement CARS imaging through an endoscope for examination of tissues structures in real-time for clinical diagnosis. Hence, one important directions of CARS imaging is to develop a flexible, stable, and easy-to-use optical fiber based system as those used in confocal fluorescence endoscopy [16], optical coherence tomography [17, 18], and in vivo endoscopy imaging systems [19]. However, the progress in developing a fiber-based CARS probe is still slow due to several challenges [12–15, 20, 21]. One of the major challenges is the four-wave mixing (FWM) effect in optical fibers. When the pump and Stokes waves travel in the same fiber with temporal overlapping, strong FWM signals are generated at the anti-Stokes frequency, ω_FWM, exactly matching CARS frequency (ω_CARS=2ω_p - ω_s). If this FWM signal enters the detection system together with the desired CARS signal, the CARS image quality will be severely affected due to the FWM background noises. To address this question, a longpass dichroic mirror, or a longpass/bandpass filter is usually employed to block the unwanted FWM signals for CARS microscopy imaging [14, 15]. Nevertheless, most endoscopy systems utilize the same fiber for delivery of excitation lasers and collection of returned emission signals for detection. In such case, the implementation of a dichroic mirror or a filter will block not only FWM signals but also the returned emission signals for detection. Therefore, it is desirable to build a FWM-suppressing mechanism to reduce the FWM background noises to realize a fiber-based CARS imaging system.

In this manuscript, we demonstrated an optical fiber delivered CARS microscopy imaging system with a polarization-based mechanism for suppression of FWM signals in delivery fiber. Polarization maintaining fibers (PMF) were used as the delivery fiber to ensure stability of the state of polarization (SOP) of lasers. The pump and Stokes waves were coupled into PMFs at orthogonal SOPs along the slow and fast axes of PMFs, respectively, resulting in a significant reduction of FWM signals generated in the fiber. At the output end of PMFs, a dual-wavelength waveplate was used to realign the SOPs of the two waves into identical SOPs prior to their entrance into the CARS microscope. Therefore, it allows the pump and Stokes waves with identical SOPs to excite samples at highest excitation efficiency. Our experimental results showed that this polarization-based FWM-suppressing mechanism can dramatically reduce FWM signals generated in PMFs up to approximately 99%. Meanwhile, the PMF-delivered CARS microscopy system with this mechanism can still produce high-quality CARS images. Moreover, because this polarization-based mechanism only changes SOP of light, unlike aforementioned filter-based mechanism for FWM suppression, it won’t block the returned emission signals in the delivery fiber for detection in most endoscopy designs. Consequently, our PMF-delivered CARS microscopy imaging system with the polarization-based FWM-suppressing mechanism potentially offers a new strategy for building fiber-based CARS endoscopes with effective suppression of FWM background noises.

2. Materials and methods

The schematic of our CARS microscopy system is shown in Fig. 1 . The light source is a broadly tunable picosecond optical parametric oscillator (OPO) based on a periodically poled LiB₃O₅ crystal (Levante, APE, Berlin). The OPO is pumped by the second harmonic (532 nm) output of a mode-locked Nd:YVO4 laser (High-Q Laser, Hohenems, Austria). The laser delivered a 7-ps, 76-MHz pulse train at both 532 nm and 1064 nm. The 1064 nm pulse train is used as the Stokes wave. The 5-ps OPO signal is used as the pump wave with a tunable wavelength ranging from 670 nm to 980 nm. The beating frequency between the pump and Stokes beams covers the entire chemically important vibrational frequency range of 100–3700 cm⁻¹. The pump and Stokes beams are overlapped by a time-delay line and lenses in both temporal and spatial domains to satisfy the precondition for producing a CARS signal. The narrow-bandwidth pump and Stokes pulses (~3.5 cm⁻¹) with durations of 5 ps can effectively reduce the non-resonant CARS background [10], and thus ensures a high signal-to-noise ratio as well as a sufficient spectral resolution. Meanwhile, the light source also provides excellent power stability, allowing high-sensitivity ultrafast imaging. A combination of one ¼-λ waveplate (Thorlabs Inc.) and one linearly polarizer (Thorlabs Inc.) is placed in both pump and Stokes arms, which is used to rotate the orientation of the linearly polarization states of two waves before they are coupled into fibers. A customized dual-wavelength multi-order waveplate (Sinoceramics (USA), LLC.) is placed at the output end of PMFs to realign SOPs of two waves into identical SOPs in prior to their entrance into the CARS microscope. The dual-wavelength waveplate is designed to achieve that orthogonally polarized components’ phase delay difference between the pump and Stokes waves is π, for instance, ½-λ @ pump & 1-λ @ Stokes, or 1-λ @ pump & ½-λ@Stokes. In our experiments, we adopted the ½-λ @ pump & 1-λ @ Stokes waveplate for all experiments because there was no discernable difference in terms of CARS image quality between these two types of waveplates in our tests. The pump and Stokes wavelengths were tuned to 817 nm and 1064nm, respectively, resulting in a 2845 cm⁻¹ Stokes shift, matching with the Stokes shift of the CH₂ stretch vibration.

 

Fig. 1 Schematic of our PMF-delivered CARS microscopy system with (a) pump and Stokes beams coupling into a PMF and a dual-waveplate to rotate polarization states of output lasers; (b) configurations of the CARS microscope.

Download Full Size | PDF

The microscopy system was modified from a FV300 confocal laser scanning microscope (Olympus, Japan). The modified microscopy subsystem has three PMT detection channels. They are able to detect backward (Epi) CARS signals, forward CARS signals and Rayleigh scattering transmission signals, which are used as reference. In our experiment, the pump and Stokes waves were coupled into PMFs using a 10× (NA=0.25, Newport) microscopy objective and then collimated using another 10× objective. The dichroic mirror (DM2) used in the microscope was 770dcxr from Chroma Technology Corp. The bandpass filter before PMTs was hq660/40m-2p from Chroma Technology Corp. A 1.2-NA water immersion objective lens (×60, IR UPlanApo, Olympus, Melville, NJ) was used, yielding a CARS resolution of ~0.4 μm in the lateral plane and ~0.9 µm in the axial direction [15].

Nufern panda-style PMF (PM1300-HP, Thorlabs Inc.) was used in our experiments. Its microscopic picture is illustrated as an inset of Fig. 1a. PM1300-HP is a MMF for the pump (817nm) and Stokes (1064nm) beams because its cut-off wavelength is ~1200nm, and we have previously demonstrated the use of MMF for laser delivery for CARS imaging [15]. It had a mode field diameter of ~9.5μm and a cladding diameter of 125μm. In our experiments, the coupling efficiencies were about 70% and 60% for the pump and Stokes beams coupled into either the fast or slow axis of PMF by a 10× objective. The polarimeter used to measure SOPs of lasers was TXP Polarimeter with external measurement heads (Thorlabs Inc.). The autocorrelator used to measure auto/cross-correlation function curves was an autocorrelator for APE Levante Emerald OPO (High-Q Laser, Hohenems, Austria). The optical spectrometer used to measure the optical spectra was 86142B optical spectrum analyzer (Agilent Technologies Corp., USA) and HR4000 (Oceanoptics Inc., USA).

3. Experimental results and discussions

To maintain the stability of SOPs of the polarized laser lights during their travel in fibers, we employed PMFs to deliver the laser lights. PMFs have been demonstrated as a good choice to maintain the SOP of linearly-polarized lights launched into the fiber during their propagation, with little or no cross-coupling of optical power between the two orthogonal linear polarization modes [22, 23]. We examined the orientation of the main axes of PMFs both passively and actively. First, we visualized the well-cleaved input fiber end of PMFs using a digital camera through coupling with a 10× objective to identify the position of the stress-inducing elements and visually adjusted the geometric orientation of main axes as shown in the inset of Fig. 1a. Then, we used the calibrated polarimeter to measure the output SOPs of the lasers at the output end of PMFs to ensure that lasers were well coupled into the fast and slow axes of PMFs. Meanwhile, we also measured the polarization extinction ratio (PER=−10logP_min/P_max, P_min and P_max are minimum output power and maximum output power, respectively) of PMFs using a linear polarizer and a powermeter at the output of PMFs. The PER were 21.17dB and 21.27dB when the pump wave coupled into the fast or slow axes respectively, while the PER were 22.66dB and 21.38dB when the Stokes wave coupled into the fast or slow axes, respectively. Figure 2 illustrated the polarization state paths (red star points connected by green curves) of the pump (817nm) and Stokes (1064nm) on the Poincare sphere traced at the output of 1-meter PMFs as the linear polarizer at the input of PMFs was rotated through 180°. We noted that all points were nearly located on the surface of the sphere, which means the degree of polarization of lights output from PMFs was nearly 100% [24]. It indicated that little or no depolarization effect occurring in PMFs. Great-circle paths to be traced out on the Poincare sphere were obtained by rotating the linear polarizer through 180°. The paths intersect the equators (blue curves) twice on opposite sides of the sphere. Since equators represent linear polarizations, these paths indicated that there were two positions of the linear polarizer that led to linearly polarized lights emerging from PMFs. The intersections with equators occurred on opposite sides of the sphere further showed that these two linearly polarized lights were orthogonal to each other [25]. In fact, they were linearly polarized lights along the fast and slow axes of PMFs.

 

Fig. 2 Polarization state paths (red star points connected by green curves) of (a) 817nm and (b) 1064nm on the Poincare sphere traced at the output of 1-meter PMFs as the linear polarizer at the input of PMFs was rotated. Blue curves indicated equators which represents linear polarizations.

Download Full Size | PDF

We examined the dispersion-induced broadening of the pulse width using PMFs. We measured the autocorrelation function curves of the pump (817nm, 100mW) and Stokes (1064nm, 100mW) beams either directly at the output from OPO or laser, or after passing through 1-meter PMFs. The normalized autocorrelation curves are shown in Fig. 3a . By measuring the FWHM bandwidth of the curves, we found there are no discernable differences in terms of broadening percentage when the pump and Stokes beams travel in either fast or slow axis. The pulse broadening percentage was 14%/meter (5.647ps spread to 6.589ps) for 817nm and 7%/meter (9.764ps spread to 10.471ps) for 1064nm. Because the resulting pulse widths were still at the picosecond level, the broadening effect of the pulses could be ignored when using PMFs for the fiber delivery for narrow-band CARS imaging. In addition, the walk-off length between the pump and Stokes beams was 2.1mm/meter or 0.07ps/meter for PMFs, which was obtained by measuring the cross-correlation function curves before and after the lasers pass though an 1-meter PMFs.

 

Fig. 3 (a) Normalized measured autocorrelation function curves of the pump (817nm, 100mW) and Stokes (1064nm, 100mW) waves either directly at output from OPO or laser, or after passing through 1-meter PMFs; (b) Normalized measured pump (817nm) and (c) Stokes (1064nm) wave spectra as a function of power in PMFs.

Download Full Size | PDF

We examined the self-phase modulation (SPM) effect induced by the pump (817nm) and Stokes (1064nm) waves propagating in a 1-meter PMFs. Figure 3b and 3c illustrate normalized measured pump and Stokes beams’ spectra as a function of propagating power in PMFs. Ripples in spectra of pump beams originated from artifacts of the 2nd order grating inside the grating-based Agilent spectrometer [15]. We noted that FWHM bandwidth of both pump and Stokes beams increased with power and no discernable differences were observed when the pump and Stokes beams travel in either fast or slow axis. In Fig. 3b, at 200 mW, FWHM bandwidth broadened by ~34.7% but was still far from 2π phase shifts (peak splitting in central wavelength) [23, 26]. Here, the power of 200mW has exceeded the average power (i.e. less than tens of milliwatts) usually required for CARS microscopy imaging. In Fig. 3c, at 200 mW, FWHM bandwidth broadened by ~36.5%, which was far from 2π phase shifts as well. These results indicated that 1-meter PMFs can be used to deliver individual pump or Stokes beams without generating serious SPM-induced phase shifts.

It is well-known that the efficiency of a FWM process in fibers highly depends on the SOPs of input lights [22, 23, 27–29]. Because the pump and probe beams are identical (i.e. ω_p = ω_p’) in our CARS system, our case is defined as partially degenerate FWM. When SOPs of two input lights are orthogonal to each other in partially degenerate FWM, the efficiency of FWM process in fibers will be approaching to zero because of polarization effect [27]. In our experiments, we first examined the polarization effect on FWM generations in the 1-meter PMFs. To separately control input SOPs of the pump and Stokes waves, a ¼-λ waveplate was used to transfer the linearly polarized light into circularly polarized light in both pump and Stokes arms. Then, a linear polarizer was used to select a desired input orientation of linearly polarized light coupling into PMFs for both pump and Stokes beams. In our first experiment, once the pump beam was coupled into the fast axis of PMFs, we rotated the linear polarizer at the Stokes arm to change the orientation of linear polarizations of the input Stokes beam with a step of 10°. Meanwhile, we recorded output FWM spectra from PMFs using an Oceanoptics HR4000 optical spectrometer. Similarly, we examined the FWM spectra in three additional scenarios: the pump beam coupling into the slow axis, the Stokes beam coupling into the fast and slow axes.

Measured results of normalized FWM signals in four different cases are shown Fig. 4a . According to Fig. 4a, we noted as expected that the intensity of FWM signal was a function of the relative angle between the linear polarization orientation of the pump wave and that of the Stokes wave. All four curves resembled sinusoidal curves. The maximal FWM intensity was obtained when linear polarization orientations of the pump and Stokes are same, i.e., the relative angle is around 0°, 180° and 360°. The minimal FWM intensity was obtained when linear polarization orientations of the pump and Stokes are orthogonal, i.e., the relative angle is around 90° and 270°. Therefore, FWM signals can be effectively suppressed by controlling the relative angle between the linear polarization orientations of the pump and Stokes waves. To obtain the minimal FWM signals, the relative angle should be set at 90°, i.e., the pump and Stokes waves propagates along fast and slow axes of PMF respectively or vice versa. Figure 4b shows measured spectra of FWM signal peaks when the pump (817nm) and Stokes (1064nm) waves propagated in either the fast or slow axis of 1-meter PMFs. We noted that when the pump and Stokes waves were input at the slow and fast axis respectively, FWM intensity was suppressed by ~99% compared to the FWM intensity when the pump and Stokes waves were both input at the fast axis. There was no big difference in terms of performance of FWM suppression when the pump and Stokes waves coupled into either combination of orthogonal axes (i.e. 817nm at fast and 1064nm at slow, or 817nm at slow and 1064nm at fast), suggesting that we can employ either combination to suppress FWM in our system.

 

Fig. 4 (a) Optical intensity of FWM signals emitted at 661.33nm as a function of the relative angle between the linear polarization orientation of the pump (817nm) and that of the Stokes (1064nm) coupled into 1-meter PMFs measured by an Oceanoptics HR4000 optical spectrometer; (b) Measured spectra of FWM signal peaks when the pump (817nm) and Stokes (1064nm) propagated in either the fast or slow axis of 1-meter PMFs.

Download Full Size | PDF

Afterwards, we demonstrated that the PMF-delivered CARS microscopy system with the polarization-based FWM-suppressing mechanism can still produce high-quality CARS images. The setup is shown in Fig. 1. The linearly polarized pump and Stokes waves were coupled into either slow or fast axis of 1-meter PMFs. After passing the fiber, the two waves were then collimated by another 10× objective. Prior to the entrance of the CARS microscope, a dual-wavelength waveplate designed for the pump and Stokes wavelengths was used to realign SOPs of two waves back to identical SOPs. The pump and the Stokes waves were tuned to 40 mW and 20 mW for CARS imaging. We tested this setup by imaging calibrated 10μm polystyrene beads (PEB) spin-coated on a glass slide, which generated strong resonant CARS signals at the aliphatic symmetric CH₂ stretch (Δω = 2845cm⁻¹). In our experiments, we employed the ½-λ @ pump & 1-λ @ Stokes waveplate for the following experiments.

In our first experiment, the linearly polarized pump and Stokes waves were both coupled into the fast axis of 1-meter PMFs. At this point, the FWM intensity was about 8550 counts (blue peak in Fig. 4b). We captured CARS images of PEBs with and without placing the dual-wavelength waveplate into the PMF-delivered CARS microscopy system. The results were showed in Fig. 6 . Figure 6a and 6b illustrate forward and Epi CARS images of PEBs respectively when the dual-wavelength waveplate was not placed into the CARS system. Figure 6c and 6d show forward and Epi CARS images of PEBs respectively when the dual-wavelength waveplate was being placed into the CARS system. We noted that clear CARS images of PEBs (Fig. 6a and 6b) were achieved when the dual-wavelength waveplate was not placed into the CARS system. It was because the pump and Stokes waves possessed identical SOP after emerging from the fast axis of PMFs such that they could produce strong CARS signals and clear CARS images using the microscope. However, only weak PEBs CARS images (Fig. 6c and 6d) were obtained when the dual-wavelength waveplate was placed into the system. It was because the pump and Stokes waves possessed identical SOP after emerging from the fast axis of PMFs. Once passed through the waveplate, their SOPs were thus changed to be orthogonal to each other. At this point, CARS signal was proportional to depolarization ratio of the Raman line [30–32]. Since the symmetric CH₂ stretching vibration of PEBs was probed in our case and its depolarization ratio is typically quite small [32], only weak CARS signals were obtained when orthogonally polarized pump and Stokes waves excited PEBs. Another reason for the weak CARS signals in Fig. 6c and 6d or weak background in Fig. 6 may be because the control voltage and gain of PMT might be set low when capturing CARS images. In experiments for Fig. 6 and Fig. 7 , we kept all parameters of PMT constant when capturing CARS images regardless of orthogonality of polarized pump and Stokes waves. Since CARS signals of PEBs were very strong (e.g. in Fig. 6a and 6b) when pump and Stokes waves excited PEBs at identical SOP, to avoid over-saturation in this scenario, we might set the control voltage and gain of PMT a little low.

 

Fig. 6 CARS images of 10μm PEBs captured by 1-meter PMF-delivered CARS microscopy system. The linearly polarized pump and Stokes waves were both coupled into the fast axis of PMFs. Forward (a) and Epi (b) CARS images of PEBs when the dual-wavelength waveplate was not placed into the CARS system; forward (c) and Epi (d) CARS images of PEBs when the dual-wavelength waveplate was placed into the CARS system. Scale bar is 10μm.

Download Full Size | PDF

 

Fig. 7 CARS images of 10μm PEBs were captured by 1-meter PMF-delivered CARS microscopy system. The linearly polarized pump and Stokes waves were coupled into the slow and fast axis of PMFs respectively. Forward (a) and Epi (b) CARS images of PEBs when the dual-wavelength waveplate was not placed into the CARS system; forward (c) and Epi (d) CARS images of PEBs when the dual-wavelength waveplate was placed into the CARS system. Scale bar is 10μm.

Download Full Size | PDF

In our second experiment, the linearly polarized pump and Stokes waves were coupled into the slow and fast axis of PMFs, respectively, where the FWM intensity was about 97 counts (green peak in Fig. 4b). Compared to the results when the pump and Stokes waves were both coupled into the fast axis (i.e. case of Fig. 6), the FWM intensity was suppressed by ~99%. We captured CARS images of PEBs with and without placing the dual-wavelength waveplate into the PMF-delivered CARS microscopy system. The results are showed in Fig. 7. Figure 7a and 7b are forward and Epi CARS images of PEBs when the dual-wavelength waveplate was not placed into the CARS system. Figure 7c and 7d, on the other hand, are forward and Epi CARS images of PEBs when the dual-wavelength waveplate was placed into the CARS system. We noted that only weak CARS images of PEBs (Fig. 7a and 7b) were observed when the dual-wavelength waveplate was not placed into the CARS system. It resulted from the fact that the pump and Stokes waves possessed orthogonal SOPs after emerging from the slow and fast axis of PMFs. At this point, only weak CARS signals of PEBs could be obtained for aforementioned reason. Another reason for the weak CARS signals in Fig. 7a and 7b or weak background in Fig. 7 may be because the control voltage and gain of PMT might be set low to avoid over-saturation when capturing CARS images. However, clear CARS images of PEBs (Fig. 7c and 7d) were obtained when the dual-wavelength waveplate was placed into the CARS system. It was due to the fact that SOPs of the pump and Stokes waves were changed back to identical polarization states after passing through the waveplate, resulting in strong CARS signals and clear CARS images of PEBs. Collectively, the two experiments above demonstrated that the dual-wavelength waveplate can be used to achieve conversion between identical and orthogonal SOPs of the pump and Stokes waves.

Finally, we further assessed the performance of this PMF-delivered CARS microscopy system, with the polarization-based FWM-suppressing mechanism, by imaging two types of mouse tissues ex vivo. Figure 8 shows Epi CARS images of mouse skin and liver tissues. We can clearly observe the cellular structure of the tissues. These results demonstrated that this CARS microscopy system can be used to image tissues accompanied with dramatically reduced FWM background noises generated in the fiber.

 

Fig. 8 Epi CARS images of the mouse skin and liver tissues. Scale bar is 10μm.

Download Full Size | PDF

5. Conclusion

The results presented in this paper demonstrate that PMFs can be used for delivery of ultrafast lasers with negligible dispersion and SPM as well as with reliable polarization maintaining capability. The intensity of FWM signals generated in fibers highly depends on the relative SOPs between the pump and Stokes waves. This mechanism can be utilized to suppress FWM noises for fiber-delivered CARS imaging system. Best suppression of FWM noises is achieved when SOPs of the pump and Stokes waves are orthogonal to each other. A dual-wavelength waveplate can be used to change the orthogonality of the two excitation waves and realign their SOPs back to identical SOPs for high-efficiency CARS imaging. Our results indicate that the PMF-delivered CARS microscopy system with this mechanism can generate high-quality CARS images. In addition, because this polarization-based mechanism only changes SOP of light, unlike aforementioned filter-based mechanism for FWM suppression [14, 15], it won’t block the returned emission signals in the delivery fiber for detection in most endoscopy designs. Therefore, our PMF-delivered CARS microscopy imaging system with the polarization-based FWM-suppressing mechanism will potentially provide a new strategy for building fiber-based CARS endoscopes with significant suppression of FWM background noises.