Microscopy of Living Cells

CARS Microscopy

Figure 1: Pump (ωP) and Stokes (ωS) are tuned so that their difference is resonant with a vibration of interest (ωR). Anti-Stokes (ωAS) is then detected.

All of the advantages of nonlinear optical microscopy are available with coherent anti-Stokes Raman scattering (CARS) microscopy. This is a third order (four-wave mixing) process in which the vibrational spectrum of a sample is probed. This is done by using a pump (ωP) and Stokes (ωS) beam to generate an anti-Stokes signal (2ωP - ωS = ωAS). If the energy of the pump and Stokes matches the Raman vibration of interest (ωP - ωS = ωR), the molecules of interest are coherently driven, leading to a strong CARS signal. This offers the possibility of functional imaging, i.e. imaging specific chemical groups and following them inside the sample. By proper choice of the pump and Stokes wavelengths it is possible to tune the CARS signal to a particular vibrational mode, and hence to obtain a chemical-specific image. The energy diagram illustrating the CARS process is sketched in Figure 1.

Traditionally CARS spectroscopy is performed using a noncollinear geometry due to the phase matching condition required. In CARS microscopy, however, this condition is relaxed because of the tight focussing of the objective. This opens the door to using collinear sources for CARS microscopy. In this case, the CARS signal adds constructively in the forward direction, where its collection is known as F-CARS. One problem with F-CARS is that a nonresonant signal due to other molecules present in the focus is also generated in the forward direction and reduces contrast.

There are many methods used for improving contrast of resonant to nonresonant signal for CARS microscopy, but the easiest method is to correctly choose the excitation pulse length. Longer pulses (1-10 picoseconds) offer improved contrast but at the expense of signal strength compared to shorter pulses (50-100 femtoseconds). This trade-off requires a compromise for optimal performance: achieve maximum contrast with minimal loss of signal. As may be expected, the optimal pulse length depends on the molecule being studied (on the Raman linewidth of the molecule) and hence the ideal laser system should allow user variable pulse length.

Beyond the ability to achieve optimal CARS performance, using a variable pulse length offers a significant advantage for multimodal CARS microscopy where CARS microscopy is performed simultaneously with other microscopy techniques such as two-photon excitation fluorescence (TPEF) microscopy, second harmonic generation (SHG) microscopy, etc.. For these other nonlinear optical microscopy techniques, minimal pulse lengths are preferred. If the pulse length of the excitation lasers can be varied, CARS, TPEF and SHG signals can be optimized. This allows the simultaneous collection of multimodal signals while minimizing sample exposure to laser light.

Our Approach to CARS Microscopy

We have built a CARS microscope using a single femtosecond oscillator and a photonic crystal fibre, as is detailed on the facilities page. To achieve variable pulse length, we make use of a technique known as spectral focussing. Starting with short femtosecond pulses, we stretch the pulses in time (we chirp the pulses). By carefully choosing the dispersion in both the pump and Stokes, the combined beams behave like transform-limited (i.e. unchirped) pulses with a different effective pulse length. This is best illustrated in figure 2 below. Here, we use time-frequency plots to illustrate the pump and Stokes pulses. On this plot, transform-limited fs pulses are vertical ellipses while ps pulses are horizontal ellipses. Chirped fs pulses are spread out in time which leads to a change of frequency as a function of time (diagonal ellipse). If the pump and Stokes are chirped the same way (i.e. they have the same slope), the difference between them is a constant as a function of time and hence the CARS resonance is a constant (CARS depends on ωP - ωS = ωR). By varying the slope of pump and Stokes, the effective pulse length is changed and allows us to easily tune for optimal performance. An advantage of this approach is that by changing the temporal overlap between the pump and Stokes, we can change the frequency difference (i.e. in Figure 2 we change from Ω1 to Ω2). This allows us to tune to different Raman resonances and probe different molecules. A key point is because the microscope optics can shape the pulses considerably, that the chirp-matching must be done at the microscope focus. This can be done with autocorrelation and cross-correlation measurements, as shown in Figure 3. Figure 4 demonstrates the considerable effects that dispersion and chirp-matching have on the spectral resolution in a CARS measurement.

Figure 2: Frequency time plot of the CARS process. By using chirped pulses we can vary the effective excitation pulse length and achieve optimal CARS performance. By changing the time delay between the pump and Stokes, we can change the frequency difference and change which molecule we probe.

Figure 3: Chirp characterization in-microscope and sub- sequent optimization of CARS spectral resolution. (a) Interferometric autocorrelation of pump pulse using second harmonic generation in KDP powder. The 14 nm bandwidth 805 nm pulse has been chirped to 0.34 cm-1/fs at the microscope focus. (b) The chirp in the Stokes beam is characterized by cross-correlation with the pump beam using sum frequency generation (SFG) in KDP. The SFG frequencies can be converted to anticipated Stokes frequencies and the chirp in the Stokes can be mapped (c).

Figure 4: Methanol CARS spectral resolution as a function of pulse stretch (dispersion) and inter-pulse chirp matching. Black: unmatched chirps; Green: matched chirps; Red: highly stretched pulses with matched chirps


We have used our system for several applications. Some selected images are highlighted below.

Figure 5: Multimodal CARS microscopy of live Human Hepatoma (HuH) Cells. Green: Two photon excitation fluorescence (TPEF) (stained with ER-Tracker); Red: CARS from lipids [with the Pezacki Group, SIMS-NRC, Ottawa]

Figure 6: Rat dorsal nerve section. CARS images the fatty myelin sheath; Schmidt-Lanterman Clefts are clearly visible [with Peter Stys, OHRI, Ottawa]

Figure 7: Multimodal CARS image of rabbit aorta section. Green: endogenous TPEF (mostly elastin); Blue: SHG in forward direction (mostly collagen); Red: CARS (mostly lipids). [with Michael Sowa and Alex Ko, IBD-NRC, Winnipeg]

We have continued to work on laser-source development for CARS microscopy. Figure 8 was taken using an all-fibre laser system to generate the CARS image. A fibre-based laser operating at 800 nm with pulse widths approximately 100 fs provided the pump. Part of the pump was used to generate the Stokes in a low-threshold nonlinear suspended-core fibre. Total power from the femtosecond laser was less than 100 mW.

Figure 8: All-fibre multimodal CARS microscopy of Huh-7 cells: Green: TPEF (stained with Hoechst); Red: CARS [with IMRA America, Inc.]

CARS Spectral Imaging of Myelin: Evidence for Structural Domains

with Mahmud Bani, Institute Biological Sciences, National Research Council

Myelin surrounds nerves and consists of many wraps of lipid bilayers. It is necessary for nerve function, and as it is a lipid-rich material, it provides a good CARS signal.

A spectral image of Myelin:

Figure 9: CARS images at different frequencies.

Trigeminal nerve from Mouse Brain:

Figure 10: Left: Average CARS signal from C-H region. Right: Ratio of2880:2845 cm-1

Possible Significance:
Condensed state Raman spectra show both chemical and physical structure.
Samples are probably not healthy (could measure action-potential for functional measure).
CARS spectra may reveal beginning of damage not seen by CARS, or other imaging modes.

More information on CARS Microscopy can also be found at www.CARSLab.ca