Coherent anti-Stokes Raman scattering (CARS) microscopy [1, 2, 3] enables spatially resolved spectroscopic investigations of microscopic objects. The most widely used implementations use point-scanning techniques, where the sample is raster-scanned with a focused, short- pulsed collinear pump and Stokes beams. This approach enables high three-dimensional resolution. However, widefield configurations, where an extended area of the object is simultaneously exposed to the laser radiation, are possible [4, 5]. One advantage of this approach is the possibility for single-shot image acquisition [6, 7].

In the widefield (WF) approach, phase matching plays a more important role than in scanning approaches of CARS microscopy. Phase-matched widefield CARS [4, 6] optimizes the angles of incidence of the laser beams such that they optimally fulfill the wave-matching condition for the vibrational resonance of interest, whereas non-phase-matched widefield CARS [5, 7] relies on the structures in the sample to redistribute the light to achieve phase matching. The latter has the advantage that phase matching is not satisfied in the solvent around the particles, and thus contributions from the solvent are largely suppressed. On the other hand, this means that one has only crude control over phase matching, especially in a heterogeneous sample with particles of different size and chemical composition.

It would be desirable to have the advantage of both approaches: the high efficiency provided by optimized phase matching for the targeted structures and at the same time the possibility to fine-tune phase matching to highlight particles of a certain size and content.

Here we show that it is possible to “customize” phase matching by including a liquid crystal spatial light modulator (SLM) in the pump/probe-beam path of a widefield CARS microscope, which is used to precisely control the pump/probe-beam incidence angles. The angular resolution of approximately $0.5°$ is practically only limited by the angular spread and the pointing instability of the laser beams. We present data from experiments and numerical simulations that confirm that already for $\mathrm{\mu m}$-sized objects, the obtained signal strength crucially depends on these angles and that this sensitivity increases with increasing object size. We explore the possibilities this behavior offers, namely to influence the image contrast between objects of different size by controlled detuning of the pump/probe-beam angles, and we show that this procedure can also be an effective approach to suppress the mostly undesired anti-Stokes signal from the medium that embeds the small target objects. Our approach is conceptually related to epi-CARS microscopy [8], where small particles become visible by detecting the back- scattered anti-Stokes signal, because large objects that are more sensitive to phase-mismatch only emit into the forward direction.

The setup is shown in Fig. 1. We have implemented an extremely folded box-CARS geometry [4, 9], where the collimated pump/probe beams impinge from two directions and with incidence angles around $\pm 75°$. These angles correspond to a numerical aperture (NA) of roughly 1.28 in water. An oil condenser (Olympus U-AAC, $\mathrm{NA}=1.4$) was used to deliver the beams to the sample, where they illuminated an area of $80\mathrm{\mu m}$ diameter. The laser system (Infinity from Coherent, Inc.) delivers synchronous $3\mathrm{ns}$ pulses at $1064\mathrm{nm}$ and $355\mathrm{nm}$ at repetition rates of up to $30\mathrm{Hz}$. The UV pulse is used to pump a midband optical parametric oscillator from GWU- Lasertechnik. The relatively long pulses ease the temporal pulse alignment. The Stokes beam was directly coupled through the objective lens such that it formed a collimated beam in the specimen plane, where it covered an area of approximately $25\mathrm{\mu m}$ diameter. As a consequence, the generated anti-Stokes signal was emitted toward the objective lens. The images were captured with an electron-multiplying CCD camera (Andor iXon X3 888).

Our experiments were performed on polystyrene (PS) beads that were immersed in various media. The wavelengths we used to excite the C-H stretch resonance at $3052{\mathrm{cm}}^{-1}$ were $1064\mathrm{nm}$ for the Stokes and $803\mathrm{nm}$ for the pump/probe beams. The pulse energies were about $100\mathrm{\mu J}$ for the Stokes and $150\mathrm{\mu J}$ for the pump/probe beams, with corresponding spectral widths of $0.05{\mathrm{cm}}^{-1}$ and $5{\mathrm{cm}}^{-1}$, respectively. The power values were measured before the objective and condenser lenses, which establishes an upper limit for the power in the sample plane. The linear polarization states of all beams were chosen to match at the position of the specimen in order to maximize the CARS signal.

The SLM (LC-R 3000, HOLOEYE Photonics AG) was used to shape the pump/probe beams by diffracting them from a displayed binary phase grating (Fig. 1). The grating constant, which can be set by the computer, determines the beam separation in the condenser back aperture and thus the incidence angle of the beams at the sample. This approach allows one to set the beam angles with high precision and reproducibility. Diffraction artifacts due to noninteger pixel values of the grating period were found to have negligible influence, but could be further reduced by adding a shear to the SLM grating [10]. The angles of incidence were measured by recording images of the standing wave patterns formed by the pump and probe beams in the sample plane and determining their grating periods. Using $\mathrm{NA}=n_{{\mathrm{H}}_2\mathrm{O}}\sin \alpha$, the incidence angles can be expressed as corresponding values of the illumination NA.

We measured the dependence of the total CARS signal on the illumination NA for beads with $2.8\mathrm{\mu m}$, $1.4\mathrm{\mu m}$, and $500\mathrm{nm}$ diameters. To get rid of even the remaining, very low solvent background of ${\mathrm{H}}_2\mathrm{O}$, the beads were immersed in ${\mathrm{D}}_2\mathrm{O}$ or a index matching fluid with the refractive index of water (Zeiss Immersol W 2010), which was found to produce very little background. The obtained data points (all curves normalized) are plotted in Fig. 2. For the sake of clarity, the data of the $1.4\mathrm{\mu m}$ beads are not shown. The data were corrected for the NA- dependent condenser transmission as well as for the Fresnel transmission coefficients for the refraction at a glass/water interface. We also performed basic numerical simulations where we estimated the total generated anti-Stokes signal according to the relation $I_{\mathrm{AS}}\propto L^2{\mathrm{sinc}}^2(\mathbf{\Delta }\mathbf{k}L/2)$. Here $\mathbf{\Delta }\mathbf{k}$ is the wavevector mismatch, which is determined by the pump beam NA, and L, the interaction length of the pump and Stokes beams inside the material. To accommodate the spherical shape of our objects, which has a variety of interaction lengths ranging from zero to the bead diameter, we modeled a bead by a bundle of axial “rods” of different lengths, which together approximate a sphere. The total anti-Stokes intensity was then calculated as the sum of the intensities generated by the individual rods.

The experimentally obtained curves show maxima at NA values of around 1.275 ($2.8\mathrm{\mu m}$ beads) and 1.265 ($500\mathrm{nm}$ beads), respectively. Both values differ slightly from the theoretically expected value of 1.29, which can be directly calculated from the excitation geometry in Fig. 1a. A part of this peak shifts might be due to a systematic error in the calibration of the NA-axis. Also the angular spread of the pump and Stokes beams contribute to a peak shift toward lower NA values. The difference in peak positions between the data of the large and small beads might be caused by size-dependent refractive effects which are not considered in our numerical model. The simulated curves were multiplied by constant factors and shifted along the NA-axis to match the experimental data for better visual comparison. It is found that our simple model delivers quite accurate results for the $500\mathrm{nm}$ beads, but increasingly underestimates the curve widths for larger bead sizes. This trend is somewhat expected considering broadening effects like the angular distribution and pointing instability of the laser beams. Generally, the measurements confirm the expected reciprocal relation between bead size and curve width. The measured/simulated widths (FWHM) were determined to be $0.023/0.011$, $0.032/0.022$, and $0.051/0.062$ for the $2.8\mathrm{\mu m}$, $1.4\mathrm{\mu m}$, and $0.5\mathrm{\mu m}$ beads. Converted into degrees, these values correspond to $3.5°/1.7°$, $4.7°/3.9°$ and $6.6°/7.9°$. In summary, our measurements showed that phase matching is less critical but nevertheless still important for relatively small objects.

The ability to control the pump/probe-beam NA allows one not only to optimize the CARS signal but also to control the contrast between objects of different sizes. For instance, the CARS intensity ratio for beads of $1.4\mathrm{\mu m}$ and $2.8\mathrm{\mu m}$ diameter is more than doubled when the NA is changed from 1.27 to 1.24. This is demonstrated by the images in Fig. 2. Note the individual calibration bars, which have been adapted according to the peak signals of the larger bead. The ability to control the contrast also allows one to effectively suppress the signal from the solvent, which might otherwise exceed the signal of small particles. The large volume of the solvent makes its signal overwhelmingly strong, but also makes it much more sensitive to wave vector mismatch than the signal of small sample structures.

In order to quantify the attainable background suppression, we measured the CARS signals from a $500\mathrm{nm}$ PS bead in Immersol W 2010 and—in a separate measurement—from the bulk of 1% agarose gel. The agarose layer where CARS occurred had an estimated thickness of $50\mathrm{\mu m}$. The normalized signals and their ratio (at $3052{\mathrm{cm}}^{-1}$) are plotted on the left of Fig. 3. The ratio shows a minimum around $\mathrm{NA}=1.28$ and strongly increases for detuned angles. At a NA of 1.24, where the bead still emits about 90% of its peak signal, the signal to background ratio has improved by a factor of roughly 100. Further evidence for the observed background suppression is provided by a real-time CARS movie of a single PS bead ($500\mathrm{nm}\text{\hspace{0.17em} dia.}$) embedded in 1% agarose gel [Fig. 3, (Media 1)]. The gel produces a strong resonant CARS signal at $3052{\mathrm{cm}}^{-1}$. One movie frame contains the accumulated signal of three laser pulses. During the recording, the pump/probe-beam NA was repeatedly switched between 1.27 and 1.24. Single movie frames for each NA value are shown in Fig. 3. The CARS signals along a line sectioning the bead in vertical direction are plotted below these images. The agarose signal completely masks that of the bead at $\mathrm{NA}=1.27$. The structure of the bead is still visible, however, because it scatters/ refracts the anti-Stokes light generated in the agarose. This was confirmed by slightly detuning the pump wavelength from the narrow PS resonance. The image hardly changed, which proves that the majority of the signal originates from the agarose. At $\mathrm{NA}=1.24$, the agarose signal effectively vanishes, leaving the bead in an almost background-free environment. Detuning the pump- wavelength again at this lower NA made the bead practically indistinguishable from the background noise, which proved that its signal is almost exclusively composed of a resonant vibrational contribution. Evaluation of the movie data showed that the NA change from 1.27 to 1.24 improved the signal ratio of bead to agarose by a factor of 70 ($+18\mathrm{dB}$), which is in good agreement with the data shown in the diagram.

We have introduced a novel implementation of a CARS widefield microscope, which, by using a SLM for pump/probe-beam steering, enables dynamic, accurate and repeatable computer-controlled tuning of the pump/ probe-beam angles and thus phase matching. This option allows not only to maximize the anti-Stokes signal for specific target resonances, but also to manipulate the image contrast in a way to highlight small particles against larger ones. We showed that this technique also effectively eliminates the dominant signal of a surrounding solvent by controlled NA-detuning.

This work was supported by the Austrian Science Fund FWF (Project P22085), the ERC Advanced Grant 247 024 catchIT and the Higher Education Commission (HEC) of Pakistan (S. Khan).

 

Fig. 1 (a) extremely folded box-CARS beam geometry; (b) sample cell; (c) CARS WF microscope with an SLM.

Download Full Size | PDF

 

Fig. 2 Left, comparison of the NA-dependence of the total CARS signal of PS beads with $2.8\mathrm{\mu m}$ and $0.5\mathrm{\mu m}$ diameter; right, the CARS images show beads of 1.4 and $2.8\mathrm{\mu m}$ diameter, taken at $\mathrm{NA}=1.27$ and 1.24; the plots below show vertical sections through the beads.

Download Full Size | PDF

 

Fig. 3 Left, NA-dependence of the normalized signal strengths and the signal ratio of a $500\mathrm{nm}$ PS bead and agarose gel; right, $500\mathrm{nm}$-bead in agarose gel, imaged at $\mathrm{NA}=1.27$ and 1.24. Reducing the NA improved the signal ratio bead/agarose by a factor of 70 (Media 1); the plots below the images show vertical sections through the bead.

Download Full Size | PDF

1. M. D. Duncan, J. Reintjes, and T. J. Manuccia, Opt. Lett. 7, 350 (1982). [CrossRef]   [PubMed]  

2. A. Zumbusch, G. R. Holtom, and X. S. Xie, Phys. Rev. Lett. 82, 4142 (1999). [CrossRef]  

3. M. Müller and A. Zumbusch, Chem. Phys. Chem. 8, 2156 (2007). [CrossRef]   [PubMed]  

4. C. Heinrich, S. Bernet, and M. Ritsch-Marte, Appl. Phys. Lett. 84, 816 (2004). [CrossRef]  

5. I. Toytman, K. Cohn, T. Smith, D. Simanovskii, and D. Palanker, Opt. Lett. 32, 1941 (2007). [CrossRef]   [PubMed]  

6. C. Heinrich, S. Bernet, and M. Ritsch-Marte, New J. Phys. 8, 36 (2006). [CrossRef]  

7. I. Toytman, D. Simanovskii, and D. Palanker, Opt. Express 17, 7339 (2009). [CrossRef]   [PubMed]  

8. A. Volkmer, J. X. Cheng, and X. S. Xie, Phys. Rev. Lett. 87, 023901 (2001). [CrossRef]  

9. M. Müller, J. Squier, C. A. De Lange, and G. J. Brakenhoff, J. Microsc. 197, 150 (2000). [CrossRef]   [PubMed]  

10. R. Fiolka, M. Beck, and A. Stemmer, Opt. Lett. 33, 1629 (2008). [CrossRef]   [PubMed]