Programmable single-pixel-based broadband stimulated Raman scattering

International audienceWe report a simple add-on for broadband stimulated Raman scattering (SRS) microscopes, in order to enable fast and programmable spectroscopy acquisition. It comprises a conventional dispersive spectrometer layout incorporating a fast digital micromirror device (DMD). The approach is validated by acquiring SRS spectra of standard chemicals. We demonstrate DMDs advantage in broadband SRS by showing higher signal-to-noise ratio using multiplexed Hadamard spectral basis , and compressive sensing detection. Our results are applicable to a variety of frequency-domain pump-probe spectroscopy. OCIS codes: (180.4315) Nonlinear microscopy; (290.5910) Scattering , stimulated Raman; (170.5660) Raman spectroscopy. Coherent Raman scattering (CRS) processes provide label-free imaging by exploiting the vibrational spectra of molecules [1]. Amongst various CRS microscopy techniques, coherent anti-Stokes Raman scattering (CARS) is perhaps the most known [2]. In spectroscopy, in opposition to imaging-only, one exploits the inherent molecular vibrational fingerprint to obtain chemically selective quantitative information. However, the presence of a non-chemically specific signal (the non-resonant electronic response) complicates the evaluation of CARS spectra due to coherent spectral interference [3]. In this context, SRS microscopy has been recently proposed [4] as a powerful alternative to CARS, since SRS signals are free from this spurious non-resonant background [5-7], thus leading to cleaner spectral lineshapes and straightforward and robust quantitation (even if small artefacts have been reported [8]). Briefly, a "standard" SRS microscope is based on a pump-probe scheme where a small intensity transfer happens when the energy difference between two overlapping beams matches a vibrational resonance. This intensity transfer is extracted by means of amplitude-modulation and lock-in detection. While most of results in SRS have been focused on microscopy, only more recently microspectroscopy at high acquisition speeds started to be developed. Fast SRS spectroscopy remains challenging as the brute force approach would use an array of lock-in detectors, a device that is not comercially available for MHz modulation frequencies, a high frequency-modulation range, and 100's of channels, to cover the full vibrational spectrum, both aspects necessary for efficient and fast SRS detection and quantitation [5, 6]. There have been various demonstrations of SRS microspec-troscopy of unknown [9] and known spectra [10]. Ozeki and co-workers have mostly used fast wavelength sweeping of one of the laser beams [11, 12]. Rock et al. [13], Seto et al. [14], Cz-erwinski et al. [15], Liao et al. [16], and Lioe et al. [17] used multi-channel detectors in order to detect the multiple SRS wavelengths in a parallel manner. More recently, single-channel-based detection strategies have been reported, either in the time-domain using an interferometer [18], or multiplex frequency-domain approach by modulating each wavelength at a specific radiofrequency [19]. While all of the aforementioned applications have demonstrated SRS spectroscopy, and microspec-troscopy, they all rely on complex technical implementations. A simpler appealing technology that uses multivariate optical element [20] has recently been reported in the context of spontaneous Raman microspectroscopy [21]. It consists of using a DMD in a conventional spectrometer to select photons in spectral bands that will be further detected with a high-speed single channel detector [22]. Here, we demonstrate SRS spectroscopy with fast pro-grammable binary amplitude masks (DMD) set in a Czerny-Turner spectrometer. We build on the standard CRS broadband microscope using only oscillator sources, that is, few nJ pulses, by exploiting the high switching rates of DMDs. Main advantages of our approaches are: (i) optical layout simplicity; (ii) remarkably low price compared to other approaches (although we used here an expensive lock-in amplifier, our method is equally applicable using cheaper tuned amplifier [23]); (iii) compatibility with lock-in detection (phase-sensitive detection); (iv) ability of multiplexing to increase signal-to-noise (see implementation below); and (v) ability to perform compressive sensing. The latter aspect is the most promising one, given that microspectroscopy at video-rate speeds covering the full vibrational spectrum is currently a technical challenge: a compressive sensing approach directly tackles this difficulty. The proposed system is readily applicable as an add-on in current SRS microscopes which use narrow and broadband pulses combination. Figure 1 shows a typical layout for SRS microspectroscopy. Narrow-bandwidth Pump (750 nm, 0.3 nm-bandwidth) and broadband Stokes (815 nm, 7 nm-bandwidth) are provided b