Coherent anti-Stokes Raman scattering (CARS) microscopy is a branch of nonlinear microscopy that allows chemical imaging of targeted vibrational transitions in unstained samples. A resonantly enhanced blue-shifted CARS signal is generated from NIR or visible light, thus the method is more sensitive than normal Raman microscopy and offers better resolution than IR microscopy. CARS microscopes are mostly set up as confocal scanning microscopes, but wide-field approaches are possible as well.

Brief Introduction to CARS-microscopy

The beating between two laser beams overlapping at a sample in space and time can excite vibrations of molecules in the sample, if their difference in frequency matches a vibrational transition. In this case inelastic coherent anti-Stokes Raman scattering (CARS) takes place and a blue-shifted (anti-Stokes) signal beam is emitted. As this is a nonlinear optical process, high intensities, i.e. pulsed lasers, are required. The CARS signal grows proportional to the Stokes intensity and quadratically with the pump intensity and the number of resonant scatterers. CARS-microscopy allows labelling-free mapping of spatial distributions of organic molecules as shown in figure 1. As no staining with fluorescent dyes is necessary, one avoids problems of photobleaching and phototoxicity alltogether.

Since pioneering work in the early eighties [1], over the last 10 years CARS microscopy has developed into a powerful new microscopic tool, triggered especially by the work of Zumbusch, Holtom, and Xie [2] in the late nineties. Many technical variations of CARS microscopy have been developed (see [3] for an overview) most of them motivated by the necessity to increase the signal to noise ratio, which is often affected by a nonlinear CARS background that is not chemically selective.

Setup and Instrumentation

Most CARS microscopes are based on confocal scanning microscopy, working with tightly focussed beams of NIR laser pulses in the pico- or femtosecond regime and with oil-immersion objectives of high numerical aperture (NA). The confocal (CF) setup has the advantage of high spatial resolution, but requires scanning of the sample, which is avoided in wide-field (WF) CARS microscopy where one can take an image of the whole sample "all at once".

In contrast to scanning CARS-microscopy, where the tight focussing of the lasers leads to large momentum uncertainty and thus makes momentum conservation rather uncritical, in wide-field CARS the phase-matching condition imposes a restriction that has to be met. In our case a special excitation geometry, which satisfies momentum conservation among the beams in the sample, enables efficient build-up of a CARS signal from the scatterers [4-5]. The idea for the implementation of wide-field CARS in the "extremely-folded box-CARS geometry" is schematically shown in figure 2. We use nanosecond pulses and wavelengths in the NIR, for example 1064 nm for the Stokes pulse and a wavelength around 800 nm for the pump pulse which is tuned by an optical parametric oscillator (OPO). This gives access to the aliphatic stretching vibrations between 2850 cm^-1 and 2950 cm^-1, which are abundant in organic molecules and give a particularly strong CARS signal. Other non-scanning apporaches [6] are based on collinear geometries and rely on the scattering structures themselves to redirect the beams to satisfy phase-matching. As phase-matching is not satisfied in the bulk solvent, the nonlinear background from the solvent is effectively prevented.

Performance

As a consequence of the photon energy conservation ω_{_aS} = 2ω_{_P} - ω_{_S}, the anti-Stokes wavelength can be considerably shorter than the beams exciting the sample. The combination of NIR excitation and detection at shorter wavelength makes it possible to have deep penetration depth into the sample volume and good optical resolution at the same time. The lateral spatial resolution of a WF-CARS microscope is on the same order as for an ordinary WF-microscope. The NA of the ­microscope objective and the CARS-wavelength determine the diffraction-limit of the spatial resolution, whereas the NA of the dark-field condensor is adapted to fulfil phase-matching, and thus determines the efficiency, rather than the resolution. The actually achieved imaging performance depends not only on the wavelengths and the NA of the microscope objective and the dark-field condensor, but also on the details of the coherent nonlinear interaction in the sample (e.g.on the local refractive index and its influence on the phase-matching). In practice, we have been able to image spherical polymer particles as small as 500 nm in diameter. Axially one can achieve optical sectioning, since the region where all interacting beams have sufficiently high intensity is only a few µm wide. The nonlinear dependence on the pump field favors optical sectioning, as in other types of nonlinear microscopes. We have measured the optical sectioning power of our system to be on the order of 3-4 µm.