Coherent Raman Scattering (CRS) Microscopy
High-Speed Label-Free Chemical Imaging
Raman spectroscopy allows non-destructive characterization of the chemical composition of a sample based on the intrinsic spectroscopic properties of the molecular bonds. Raman imaging is possible by raster scanning the excitation laser through the sample and generating a point-by-point three-dimensional chemical map. While the chemical specificity of traditional Raman spectroscopy is excellent, the signal is extremely weak. This limits spectral analysis to high-concentration species and the acquisition rates in imaging applications to minutes and hours. Stimulated Raman scattering (SRS) provides amplification of the weak spontaneous Raman signal by >10,000x enabling high-speed, label-free chemical imaging at speeds up to video-rate (30 frames/s).
The SRS Advantage:
No dyes or labels needed
High-speed (up to 30 frames/s)
Three-dimensional resolution without physical sectioning
Deep tissue imaging with near-IR lasers
Reduced auto-fluorescence background
Raman imaging at speeds up to video-rate (30frames/s). This movie shows a capillary with quickly moving red blood cells in the skin of a living mouse. Research use only. Image courtesy: Xie Group (Harvard University).
Stimulated Raman Scattering Explained
In traditional spontaneous Raman scattering, the sample is excited with a single monochromatic light source, the so-called "pump." Incident photons inelastically scattered by molecules excite/de-excite molecular vibrations and generate new radiation at the Stokes/anti-Stokes wavelengths, respectively, such that the total energy is conserved (Figure A). By dispersing the sample emission, it is possible to analyze the vibrational spectrum of the sample. As each chemical bond has a specific stiffness (e.g., C=C is stiffer than C-C) and associated mass (e.g., C-C is heavier than C-H), each bond has a characteristic vibrational frequency. The combination of the vibrational frequencies and their respective abundance creates a chemically specific "vibrational fingerprint" of a molecule that can be obtained non-destructively using Raman spectroscopy.
In the quantum-mechanical picture (Figure B), the molecular population is excited from the ground state to a virtual excited state by absorbing an excitation photon. From the virtual state it then transitions to an excited vibrational state by emitting a red-shifted photon at the Stokes wavelength. Because the virtual state does not typically correspond to an actual energy level of the molecule, it has an infinitely short lifetime and the spontaneous transition rate is very low, making spontaneous Raman scattering a very weak process.
In stimuated Raman scattering, the sample is illuminated with a second light source at the Stokes wavelength of a targeted vibration. In the quantum-mechanical picture, the presence of the Stokes field results in a stimulated rather than a spontaneous transition from the virtual to the vibrational excited state. Similar to the principle of light amplification by stimulated emission of radiation (LASER), this results in an amplification of the molecular transitions rate, boosting the weak Raman signal and reducing the signal integration time for high-speed imaging. The amplification rate depends on the exact laser parameters. Under biocompatible excitation conditions, an imaging speed increase >10,000x can be achieved routinely.
Principle of CRS. (A) Spontaneous Raman scattering. The sample is excited with light at a single frequency ωp. The output spectrum contains new radiation on both the Stokes (ωS) and anti-Stokes sides (ωaS) due to inelastic light scattering off molecular vibrations. (B) In CRS the combined action of pump and Stokes beams transfers molecules in the sample from the ground state into the targeted vibrational state. As a consequence a pump photon is absorbed and a Stokes photon is generated. This allows signal amplification by >10,000x compared to spontaneous Raman scattering.
Zumbusch, Andreas; Holtom, Gary R.; Xie, X. Sunney "Vibrational Microscopy Using Coherent Anti-Stokes Raman Scattering," Phys. Rev. Lett. 82, 4142 (1999). link
Freudiger, Christian W.; Min, Wei; Saar, Brian G.; Lu, Sijia; Holtom, Gary R.; He, Chengwei, Tsai, Jason C.; Kang, Jing X.; Xie, X. Sunney "Label-Free Biomedical Imaging with High Sensitivity by Stimulated Raman Scattering Microscopy" Science, 322, 1857-1861 (2008). link
Saar, Brian G.; Freudiger, Christian W.; Reichman, Jay; Stanley, C. Michael; Holtom, Gary R.; Xie, X. Sunney "Video-Rate Molecular Imaging in Vivo with Stimulated Raman Scattering" Science 330, 1368-1370 (2010). link
Label-free imaging of cell-division of a mammalian cell. (red) DNA. (blue) lipids. Research use only. Image courtesy: Xie Group (Harvard University)
Raman spectral of DNA, protein and lipids. Even though vibration bands are overlapping, chemically specific imaging (see video above) can be generated by hyper-spectral imaging based followed by spectral unmixing. Research use only. Image courtesy: Xie Group (Harvard University)
The best signal amplification, and therefore highest imaging speeds, are achieved with picosecond lasers that have a linewidth narrower than the typical Raman bandwidth, as all the excitation energy is focused into a single molecular transition. This approach known as "narrowband SRS" works extremely well for isolated vibrational frequencies. It has, however, a limited spectral selectivity, because single frequency imaging cannot distinguish individual molecular species with overlapping vibrational spectra (see Figure). To overcome this limitation, one can image the same frame at different vibrational frequencies consecutively by synchronizing the image acquisition and automated laser tuning. Once such a hyper-spectral image stack is acquired one can then apply various spectral un-mixing techniques to separate the individual color channels. For example, the video shows how the shifted vibrational response from DNA can be separated from the partially overlapping spectral response of lipids and proteins to create a chemically specific signal (red color channel) showing cell division in vivo without disruption of this sensitive cell process by the laser excitation.
For samples with unknown vibrational signatures, one would first acquire a hyper-spectral image stack at a fixed frequency step-size to extract the maximal spectral information. For example, it is possible to acquire a hyper-spectral image stack of the entire high-wavenumber region from 2800 cm-1 - 3100 cm-1 in steps of 10 cm-1 within 1s, if the laser tuning is synchronized with video-rate CRS imaging. Once optimized vibration frequencies are known (e.g. 2850 cm-1, 2920 cm-1 and 2970 cm-1 in the example of DNA imaging) the image speed can be further improved by acquiring data from only the relevant transitions and not "wasting" time and excitation power on spectrally nonspecific regions. As SRS spectra are identical to those of spontaneous Raman, one can also use databases of spontaneous Raman spectra for the choice of optimized excitation frequencies.