2.1.

Samples

2.2.

Optical Parametric Oscillator System

The master oscillator is a picosecond mode-locked frequency-doubled [TeX:] \documentclass[12pt]{minimal}\begin{document}${\rm Nd}\!:\!{\rm YVO\hbox{-}laser}$\end{document} $\mathrm{Nd}:\mathrm{YVO}\text{-}\mathrm{laser}$ (HighQ laser⁴) operating at 532 nm. A second amplification stage enables the laser to deliver 8 W of green output power that is equally divided to pump two OPOs.

Each individual OPO (Emerald, APE, Berlin, Germany⁵) uses a temperature-tuned, noncritically phase-matched LBO (LiB₃O₅) crystal for parametric downconversion. The temperature of the crystal sets the wavelengths of the two generated NIR signal and idler. Precise wavelength tuning is accomplished through the use of a stacked Lyot filter positioned in the OPO cavity at Brewster's angle. Because the LBO crystal is noncritically phasematched, the signal and idler beams exit the laser cavity in a collinear fashion, with pulse trains that are temporally synchronized. It is nevertheless possible to spatially separate the signal and idler beams with a supplied beam splitter. In this case, the two exits of the OPO cavity are operated. The OPO can be easily and continuously tuned over a wide spectral range from 690 to 990 nm for the signal and between 1150 and 2450 nm for the idler output.

2.3.

Optical Parametric Oscillator Control Software

The controller of the OPO gives full diagnostics of the system (power, spectrum, etc.). Moreover, a set of commands enables the control of the signal wavelength (and idler) over three parameters. The coarse and fine tunings are gained via the temperature of the crystal and the tilt of the Lyot filter, respectively. The cavity length insures the synchronous pumping of the OPO cavity. The system can be manually set to any desired signal wavelength with a commercial user-friendly interface. Automated wavelength scanning has been performed through a home-built software to facilitate acquisition of broadband CARS spectra. First, a so-called probe program iteratively investigates the working parameters for the OPO on the desired spectral range. Working parameters include the crystal temperature, the Lyot filter tilt, and the cavity length (piezostage position). They are subsequently saved into a file, which constitutes a database of accessible wavelengths and corresponding OPO settings. Then, the main CRS microscope-driving software loads this database and applies the parameters therein into the OPO. The probe program works as follows. Temperature is adjusted to reach a sufficient OPO power, and then both Lyot filter and piezostage are optimized to find the typical parameters for each signal wavelength. The whole database (corresponding to 400–3400 cm⁻¹) is created in a few hours. As described in Sec. 5, it takes typically 60 min to obtain a spectrum on the 900–3200 cm⁻¹ range with 10-cm⁻¹ resolution.

2.4.

Nonlinear Scanning Microscope

Our custom nonlinear microscope is arranged around a commercial microscope stand (Nikon Eclipse Ti-U). The pump and Stokes beams necessary for CRS microscopy are temporally synchronized, recombined, and sent onto a couple of galvanometric scanning mirrors (6200 H, Cambridge Technology). These excitation beams enter the microscope through the backward excitation port and are directed onto the filter cube. A telescope with a magnification of 3 is used to image the scanning mirrors onto the rear aperture of the infinity-corrected focusing microscope objective. With this arrangement, the microscope provides diffraction-limited resolution.

Forward detection is performed thanks to a dedicated mechanical unit manufactured by Tokyo Instruments (Japan) that directly bounds to the Nikon Eclipse stand. This unit insures a three-axis manual control for the forward collection. The forward emitted light is collected using a long working distance 40 × air objective (LUCPLFLN, 40 ×, 0.6 NA, Olympus) and focused on a channel photomultiplier (CPM) tube (Perkin Elmer Optoelectronics, CPM C972). The backward emitted light (Epi-CARS) is collected by the focusing objective lens and focused on a photomultiplier (Perkin Elmer Optoelectronics, CPM C972) or on a compact spectrometer (Ocean Optics, USB 4000 VIS NIR), both connected on side ports. Note that both forward and backward detection work in the so-called nondescanned geometry and therefore must have large active areas. The magnification of the imaging optics has been carefully chosen to image on the detectors a scanning area as large as 200 × 200 μm². Forward and backward CPMs work here in the photon-counting regime. This is an advantage when low signals are detected but care must be taken when strong CARS signals are detected as (i) it limits the maximum count rate to the repetition rate of the lasers and (ii) photomultipliers are subjected to nonlinearities at high count rates. In the present case, we always work well below these spurious regimes by reducing the incident laser power. A set of high- and low-pass wavelength filters are used that ensure that the detected signal is free of laser background and is solely associated to the desired contrast mechanism (TPEF, SHG, or CARS).

The dual-axis driver (Micromax 67320, Cambridge Technology) of the galvanometric mirrors is controlled by analog voltages generated by a high-speed analog output PCI board (NI 6731, National Instruments). The TTL pulses generated by both forward and backward CPMs are recorded by a PCI counting board (NI 6602, National Instruments). A detailed description of the synchronisation method is given in Ref. 18. The whole system is connected to a personal computer and controlled by a home-built program (LABVIEW, National Instruments).

3.1.

Collinear Signal and Idler Optical Parametric Oscillator Output

This is the easiest way to perform CARS with one OPO. The signal and idler pulse trains are collinear, overlapped, in time, and used as the pump and Stokes wavelengths, respectively, for the CARS process. Figure 1a gives the schematic of this CARS setup, where the signal and idler beams are directly sent on the galvanometer mirrors. Because these two beams are collinear, power adjustment is performed simustaneously on the two colors, which is far from being optimal. Neutral density filters or a polarizer (if polarization at the focal point is not an issue) can be used. Because of Gaussian beam diffraction, the divergences associated with the two colors at the OPO output are not the same and impossible to tune separately. Figure 1b gives the signal and idler wavelength associated to each Raman shift together with the corresponding anti-Stokes CARS wavelength. The CARS wavelength is >850 nm for Raman shifts of <1500 cm⁻¹, a spectral range where the sensitivity of standard PMT drops.¹⁹ On the opposite, for large Raman shifts, the CARS wavelength is <800 nm and can be thus easily detected. However, the associated idler beam falls in a spectral range (> 1100 nm) where a standard microscope objective lens exhibits a poor transmission. Note that for deep-tissue penetration, IR wavelengths are less absorbed and scattered, a situation that is in favor of this collinear signal and idler configuration.

As a simple illustration of the problem that one may face using a collinear signal and idler, we focus here on longitudinal chromatic aberrations. Chromatic (and, to a lower extent, geometrical) aberrations are present in any imaging system and are difficult to control over a broad spectral range. This is an issue in CRS because the pump and Stokes beams must overlap in space, ideally for any Raman shift. It is interesting to note that the collinearity is extremely good between the signal and idler beams and, as far as we could notice, the two beams always overlap in the transverse plane (x, y) at the microscope objective focus. On the contrary, this is not always the case along the optical axis (z-axis). To illustrate this point, we have used the TPEF signal generated by TTB crystals (see Sec. 2) to check the signal and idler focal plane along the z-axis (alternatively, one could use fluorescent beads and two-photon absorption). TTB crystals exhibit fine spatial structures above and below the diffraction limit that nicely shows the imaging properties of a specific objective lens. Figure 2a shows the result using a 60 × water-immersion Plan Apo (Plan ApoVC, 60 ×, 1.2 NA, WI, Nikon) objective lens manufactured by Nikon. Sending the collinear idler and signal beams, we find a difference of 11 μm between the focal planes of the two beams. Optimizing the TPEF image obtained with the λ_exc = 916 nm signal results in a poor TPEF image obtained with the λ_exc = 1270 nm idler (corresponding to a 3045-cm⁻¹ Raman shift). By moving the focal plane 11 μm farther, the idler image can be restored. With such an objective lens, it is clear that no CRS imaging is possible in this collinear configuration. Using the recently released 40 × water-immersion CFI Apo (CFI Apo LWD, 40 ×, 1.15 NA, WI λ_S, Nikon) objective lens manufactured by Nikon gives very different results, as can be seen in Fig. 2b. With this well-corrected objective lens, the λ_exc = 916 nm signal and λ_exc = 1270 nm idler focal planes nicely overlap, as can be seen from the TTB crystal fine structure. Therefore, we could directly perform CARS imaging of polystyrene beads on resonance [see bottom right image in Fig. 2b].

Fig. 2

Chromatic aberrations in CRS imaging using the OPO with collinear idler and signal output. (a) TPEF TTB crystal imaging at λ_exc = 916 nm (signal) and λ_exc = 1270 nm (idler) showing a difference of 11 μm in the focal planes of the Nikon Plan ApoVC 60 × WI objective lens. TPEF image is restored for the idler beam moving the focal plane by 11 μm. (b) Using the Nikon CFI Apo LWD 40 × WI λ_S objective lens, the two focal planes nicely overlap and allow contrasted CARS imaging of polystyrene beads at 3045 cm⁻¹ Scale bar is 20 μm. Laser power was 5 mW for both pump and Stokes.

When dealing with chromatic aberrations, it is possible to separate the signal and idler beams at the OPO exit to manage their divergence separately via two independent telescopes in order to match their focal planes (see Sec. 3.2). Nevertheless, this scheme can be used only for small focal-plane mismatches and rapidly proves to be very inefficient in terms of image quality. Transverse chromatic aberration may also play a role especially near the edges of the image and consequently produces intensity falloff. In our case and with the objectives used in this work, we have not noted such an effect over the 200 × 200 μm² field of view. Table 1 summarizes the assets and constraints of using an OPO with a collinear signal and idler output.

Table 1

Pros and cons for “collinear signal and idler OPO output.”

We shall now move to another well-used mode of operation when only the OPO’s signal and the 1064 nm output from the Nd: YVO are used.

3.2.

Using the Optical Parametric Oscillator Signal and the 1064-nm Beam from the Nd: YVO Oscillator

Another advantageous mode of operation for CRS is to use the OPO signal beam from the OPO and the fundamental 1064-nm beam from the Nd: YVO oscillator as the pump and Stokes beams, respectively, for the CARS process. Figure 3a shows the schematic of the setup where only the signal from the OPO is used (the idler beam is stopped within the OPO housing with a dedicated filter) and is further recombined with the fundamental 1064-nm beam that does not enter the green module (SHG unit incorporated into the Nd: YVO housing).

Because the paths of the two beams are now well separated, power and divergence correction are adjusted independently with “halfwave plate/polarizer” units and telescopes, respectively. Spatial overlap between the two beams is controlled with a gimbal mirror mount while temporal overlap adjustment requires a dedicated delay line (mounted on the fundamental beam in our setup). In this configuration, it is possible to control the polarization state of the two beams independently (note that: if a retroreflector is used in the delay line, the outcome polarization state can be rotated). Figure 3b gives the fundamental and signal wavelength associated with the Raman shift together with the corresponding anti-Stokes CARS wavelength. The CARS wavelength falls in a range that can be detected with photomultiplier tubes if the targeted Raman shift is >1000 cm⁻¹. The fundamental and signal beams are now well transmitted by most of the commercial IR-corrected objective lenses.

As an example of the effectiveness of this operation mode, we report fresh mouse ear imaging on the CH₂ symmetric stretching mode (2845 cm⁻¹) of lipids with a 40 × long working distance standard air objective lens manufactured by Olympus (LUCPL FLN, 40 ×, 0.6 NA, air, Olympus). Because this lens is not fully corrected in the IR, we had to slightly tune the beam divergence to match the focal planes [see Fig. 4a]. They were distant by 3 μm when injecting parallel incoming beams. Figure 4b presents the forward-detected CARS signal for various penetration depths corresponding to the stratum corneum (z = 0 μm), the sebaceous gland (z = 30 μm) and the deep adipocytes of the dermis (z = 55–85 μm).

Fig. 4

(a) z overlap in the focal plane using a conventional long working distance lens (LUCPL FLN 40 × 0.6NA air Olympus) is possible by slightly correcting the incoming beam divergence. (b) CARS images at 2845 cm⁻¹ of freshly removed mouse ear sample at different penetration depths: stratum corneum (z = 0 μm), sebaceous gland (z = 30 μm), and deep dermis adipocytes (z = 55–85 μm). Laser powers were 15 mW for the Stokes and 25 mW for the pump. Scale bar is 40 μm; pixel dwell time is 50 μs. To improve the signal-to-noise ratio, we average 10 frames of 250×250 pixel images.

This “OPO signal and fundamental 1064 nm” configuration is certainly the best trade-off when using a single OPO (see Table 2). Such a configuration has proved to be quite versatile to perform CRS imaging in various configurations, including CARS (Ref. 6) and SRS.¹³ Note that hyperspectral imaging can also be done in this configuration (see Sec. 5). With the current available detectors, this is the operation mode that we recommend. We shall now move to a more exotic, nevertheless, very powerful operation mode using two OPOs.

Table 2

Pros and cons for “OPO signal and 1064 nm fundamental.”

4.1.

Wave Mixing as Multiple Contrast Mechanisms Imaging

When coherent Raman, sum frequency generation (SFG), SHG, and TPEF can be addressed on the same sample with three incoming colors, quite a few wave-mixing lines appear. Figure 6a summarizes the wave-mixing combinations that generate CARS, SFG, SHG, and TPEF. These wave-mixing combinations are readily observed on TTB crystals that provides a test sample for all these contrast mechanisms [Fig. 6a]. The spectrum depicted in Fig. 6b results from nonlinear interactions between the three wavelengths (OPOI: 816.8 nm; OPO2: 888.7 nm, and fundamenta11064 nm) in a TTB sample. Incoming beams (black) interact to generate CARS lines (red), a TPEF broad spectrum (green), and SHG and SFG lines (blue).

Focusing on the CARS contrast mechanism, three different Raman shifts can be advantageously used to image various chemical bonds. We illustrate this ability on Fig. 6c for 20 μm polystyrene beads dispersed in olive oil. 2845 cm⁻¹, 990 cm⁻¹, 1855 cm⁻¹ Raman shifts are used to image oil, beads and the nonresonant background (NRB) respectively. Although the different contrast mechanisms have been illustrated on test TTB samples, they can be readily used for background free CARS and SHG imaging on fresh tissue samples.

4.2.

Tissue Imaging with Two Optical Parametric Oscillators

One of the key features of this three-beam configuration is the ability to detect a “background-free” CARS signal as previously demonstrated.^{21, 22} This is simply done by subtracting two CARS images that have been separately obtained. It should be noted here that this straightforward background subtraction is not free of artifacts and give satisfactory results only if the peak CARS signal is subtracted with a nonresonant background located far enough from the considered vibrational resonance. Although this is the case here, special care must be taken as reported in Ref. 23. Figure 7 illustrates such a scheme where the forward CARS anti-Stokes beam at 2845 cm⁻¹ [Fig. 7a, F-CARS R: lipid CH₂ resonance] and 1855 cm⁻¹ [Fig. 7b, F-CARS NR: nonresonant] are acquired and subtracted (F CARS R-NR). In Fig. 7e, this resonant CARS image (F-CARS R-NR) has been superimposed with the SHG image [E-SHG on Fig. 7c and associated spectrum Fig. 7d]. We attribute the CARS signal from lipids and the SHG’s from collagen. One can clearly distinguish the lipid-rich sebaceous gland (4), one hair hole (2), and the stratum corneum (1); these lipid features are well separated from the collagen matrix (3).

Fig. 7

Resonant CARS and SHG signals from a transversal skin mouse cryosection (20 μm thick). (a) Forward CARS CH₂ 2845 cm⁻¹, (b) Non resonant CARS 1855 cm⁻¹, (c) SHG, and (d) associated spectrum. (e) (CARS 2845 cm⁻¹)−(CARS 1855 cm⁻¹) difference super imposed with the SHG signal. 1: Stratum corneum; 2: hair hole; 3: collagen matrix; and 4: sebaceous gland. Scale bar is 40 μm; pixel dwell time is 50 μs; field of view is 200 × 200 μm² (250 × 250 pixel); average is mo frames; Stokes and pump powers were 15 and 25 mW, respectively.

1 shows the resonant CARS and SHG images versus the penetration depths into the ear mouse sample. Quite noticeable is how the collagen matrix delineates the sebaceous gland near z = 30 μm.

Video 1

Movie of the resonant CARS and SHG signals from an ear mouse on the 0–50 μm depth range (QuickTime, 2.19 MB).

Up to now, CARS images have been obtained for at most three wavenumbers (see Fig. 6). This might not be informative enough to tackle a specific molecular recognition problem. In Sec. 5, we investigate the possibility of performing hyper-spectral CARS imaging using OPOs.

Fig. 6

(a) Wave-mixing processes observed in nonlinear TTB crystals. Red: CARS processes; green: TPEF; dark blue: SFG; underlined blue: SHG. TTB sample as imaged with SFG, SHG, TPEF, and CARS. (b) OPOI: 816.8 nm (plain dark); OPO2: 888.7 nm (plain dark), and fundamental 1064 nm (dashed black) interact to generate CARS lines (red plain and dashed), a TPEF broad spectrum (green), and SHG and SFG lines (blue). (c) 250 × 250 pixel CARS images of 20 μm polystyrene beads in olive oil for three Raman shifts (2845 cm⁻¹, oil; 990 cm⁻¹, polystyrene; 1855 cm⁻¹, NRB). Scale bar is 20 μm; pixel dwell time is 50 μs. Stokes and pump powers were 5 and 10 mW, respectively. (Color online only.)