Stimulated Raman scattering (SRS) is a well-known interaction between two-color light and molecular vibrations in matter. When the optical frequency difference matches the vibrational frequency, energy is transferred from the high-frequency light to the low-frequency light. Recently, application of SRS to optical microscopy has attracted much attention.1–3SRS microscopy uses two pulsed laser beams of different color, one of which is intensity modulated. The beams are combined with the pulses in synchrony to form two-color pulses, which are focused on a sample. When SRS occurs, the intensity modulation is transferred to the other beam. The transferred modulation is measured through photodetection followed by lock-in detection (the standard technique for extracting such a signal in the presence of noise). Images are taken by scanning the focus point or the sample position.
SRS microscopy has two particularly important features. First, it enables label-free, molecule-specific observation, because SRS provides chemical contrast based on molecular vibration. Second, high-speed imaging up to the video-rate is possible.4 By exploiting these features, one can track in real time various chemicals and drugs in biological samples. Label-free imaging is especially useful for observation of small molecules, which are difficult to label. Furthermore, it is of more general practical interest that one may omit staining and labeling processes, which are time-consuming and may affect samples.
However, the molecular specificity of SRS microscopy remains limited because SRS visualizes molecular vibrations only at the specific frequency determined by the optical frequency difference of the laser pulses. Therefore, we cannot distinguish different molecules when their vibrational spectra overlap at the SRS frequency. This problem could be overcome if we could tune the laser wavelength quickly, to acquire a set of SRS images with different vibrational frequencies. As a result, each pixel would have spectral information (i.e., across a range of frequencies), which may enhance the molecular specificity.
To this end, we have developed a wavelength-tunable laser source for SRS spectral imaging.5 Broadband ytterbium (Yb)-fiber laser pulses are spectrally sliced by an optical filter, and amplified by fiber amplifiers. The transmission wavelength of the optical filter is tuned by changing the direction of a galvanoscanner mirror in the filter, which enables tuning as fast as ∼1ms. The tunability is as wide as ∼300cm−1.
To conduct SRS imaging, we used the Yb-fiber laser along with synchronized, picosecond Ti:sapphire (titanium-doped sapphire) laser pulses. We succeeded in obtaining spectral images of different types of polymer beads. The images have different contrast depending on the vibrational frequency: see Figure 1(a)–(d). From the entire spectral data, we extracted SRS spectra at certain locations. The different spectra of polystyrene and poly(methyl methacrylate) are clear: see Figure 1(e). Although the frame rate in this experiment was as low as 1.3 frames/s, we recently upgraded the frame rate to 15 frames/s, enabling tens of spectral images to be acquired in a few seconds.6 This system would allow fast biological imaging with high molecular specificity.
Figure 1. Spectral imaging of polymer beads by stimulated Raman scattering (SRS) microscopy. (a)–(d) SRS images at different wavenumbers. Scale bar: 10μm. (e) SRS spectra of polystyrene (PS) and poly(methyl methacrylate) (PMMA) reconstructed from the SRS images. Arb.: Arbitrary. (Reprinted with permission.5)
Another important issue is that the solid-state lasers (such as Ti:sapphire) often used for SRS microscopy are bulky and expensive. From a practical point of view, it would be attractive to replace them by compact, cost-effective fiber lasers. When we use fiber lasers for both beams in SRS microscopy, however, it is important to take the intensity noise into account. Because the SRS signal is detected as a small intensity modulation in one of the pulsed beams, excessive intensity noise in that beam will degrade the signal-to-noise ratio. In our experiments imaging polymer beads, that role was played by the Ti:Sapphire laser, which had very low noise (roughly at the shot-noise limit). Typically, the intensity noise of fiber lasers is much higher than the shot-noise limit because optical amplification in the fiber produces amplified spontaneous emission noise.
We mitigated the intensity noise of fiber lasers by developing a collinear balanced detection (CBD) technique,7 where one beam’s optical pulses are split and combined collinearly with a delay. When the original pulse and the delayed replica are detected, the photocurrents that originate from these pulses add up. However, because the replica is delayed in time, the phase of its photocurrent spectrum in the frequency domain is shifted by an amount that is linearly proportional to the frequency. As a result, the two photocurrents add up destructively at specific frequencies, leading to the cancellation of the intensity noise. We can use these frequency regions (i.e., modulate at these frequencies) for lock-in detection of the SRS signal. The CBD technique will be a key for applying fiber lasers for SRS microscopy.
In summary, we have demonstrated a wavelength-tunable fiber laser for SRS microscopy that is suitable for fast biological spectral imaging. We have also shown that the CBD technique can improve the signal-to-noise ratio achievable with dual-fiber-laser SRS microscopy. Although we have been developing laser sources for enhancing molecular specificity and for facilitating the practical implementation of SRS microscopy, many challenges remain: Can we achieve video-rate imaging using two fiber lasers? Can we broaden the wavelength tunability so that we can access the Raman ‘fingerprint region’ (500–1800cm−1), where rich spectroscopic information is obtained? Can we develop ultra-low-noise fiber lasers, which would not require CBD? We are tackling these questions in the next steps of our research. The continuing advances of laser technologies will assist in overcoming these challenges, and may open powerful new biological imaging modalities in the future.