1. Introduction

Raman spectroscopies, including spontaneous Raman spectroscopy and coherent Raman spectroscopy (CRS), attract significant attentions, as they allow label-free, chemical-selective and sensitive imaging and sensing [1,2]. With recent expansions in the field of biomedical imaging [3,4], material science [5] and remote sensing [6], there is a growing demand for new instrumentation development, which allows efficient coherent and broadband excitation of Raman spectra. In this report, we present a novel approach to construct an electronically tunable laser system, which is capable of attaining high fidelity stimulated Raman scattering (SRS) and coherent anti-Stokes Raman scattering (CARS) spectra without any moving parts.

Optical processes based on coherent Raman scattering are often favorable, as they could achieve efficient and sensitive nonlinear optical interactions [1,7]. Moreover, their multi-photon nature improves the axial spatial resolution, an essential feature for imaging and sensing applications [8]. In common coherent Raman instrumentation, picosecond pulses are favored due to their appropriate linewidth (i.e. a few wavenumbers) and a moderate excitation power [1,7]. Nevertheless, other approaches, including continuous-wave laser to femtosecond-pulse excitation coherent Raman spectroscopies, stimulated Raman photoacoustic imaging, spectrally tailored excitation stimulated Raman scattering (STE-SRS) and coherent molecular normal modes generation have also been explored and experimentally demonstrated [9–18]. CRS requires two laser sources with independently tunable wavelengths (pump and Stokes beams). Typically, to generate wavelength-tunable, high-energy picosecond laser pulses, an optical parametric amplifier (OPA) and a tunable seed are required. The seed should be conveniently tuned to cover most of the vibrational Raman fingerprint and high-frequency regions (~800 - 1800 cm⁻¹ and around 3000 cm⁻¹ [19,20]).

The purpose of this article is to introduce and demonstrate an alternative concept for generating picosecond tunable laser source suitable for SRS / CARS applications. Similar with Ozeki et al.’s work [21], we adopt the seeding + amplification strategy to prepare a tunable pulsed laser source. However, instead of the diffractive grating, we adopted an electronically tunable acousto-optical filter (AOTF) to select narrow-band emission with a desired center wavelength. In this way, the moving part (i.e., diffractive grating) was eliminated from the system. Acousto-optic devices have recently gained growing interests, as they could work as tunable filters / switches in microspectroscopy imaging applications [22] and fiber-optic systems [23,24]. A typical acousto-optic device is based on acoustic wave induced light diffractions in an anisotropic medium. The device usually consists of a piezo transducer attached to a birefringent crystal. When a radio-frequency (RF) driving voltage is applied to the piezo transducer, acoustic waves are generated within the crystal, producing a phase grating that diffracts part of the incident light beam under phase-matching conditions. When tuning the RF driving frequency of the piezo transducer, the diffraction window of the crystal will be changed according to the phase matching condition [25,26]. Compared with other wavelength tuning techniques [21,27,28], AOTFs provide a pure electrical solution that eliminates all the moving parts (e.g., diffractive gratings or galvo-mirrors) within the system.

2. Experimental approach

Figure 1(a) outlines the optical setup. For all the experiments, we employed a home-built picosecond Nd:YVO₄ laser, which was described elsewhere [29]. For this particular set of studies, the repetition rate was adjusted to 200 kHz, and the pulse duration was set at 5 ps. The energy of the fundamental beam (1064.20 nm) was split into two parts. The first part was sent to a single-mode optical fiber (~2 meters, UHNA3; Thorlabs, Inc.). A stable supercontinuum emission ranging from ~1090 to ~1600 nm was generated through stimulated Raman scattering induced by its GeO₂ fiber core (Stage 1) [30]. An AOTF (TF1650-1100-2-3-GH40, Gooch & Housego Inc.) was inserted in the beam path of the collimated fiber output. In this way, an angular dispersion, which is controlled by the AOTF’s driving frequency, was created. A narrow band of the supercontinuum was selected to pass the mechanical iris behind the AOTF (Stage 2) and sent to an optical parametric amplifier (OPA, a heated 20-mm-long LiB₃O₅, LBO) as seed photons (idler). The second part of the fundamental beam was sent to generate the second-harmonic radiation (532.1 nm), which was used as the pump source of the temperature-tuned OPA. Due to the nature of parametric process, “signal” and “idler” wavelengths emerge at the OPA’s output (Stage 3), so that ω_signal + ω_idler = ω_pump. The OPA emission was focused onto the sample by an aspheric lens (N.A. = 0.5; Thorlabs, Inc.) to allow efficient coherent Raman interaction inside the sample volume. Prior to this, a notch filter was used to block the remaining 532 nm emission. Therefore, only the “signal” and the “idler” pulses could reach the sample (Stage 4). After transmitting through the sample, the scattered photons, including the pump, probe, and CARS components, were collected by another high numerical aperture aspheric lens (Stage 5). The forward-scattered CARS signal was collected by a spectrometer (Shamrock; Andor Technology Inc.) with the attached TE-cooled CCD (Newton 970; Andor Technology Inc.). The stimulated Raman gain / loss signal contained in the “signal” and “idler” beams could also be collected by a photodiode (Large area photoreceiver 2031, Newport Inc.). The corresponding spectra for each stage labeled in Fig. 1(a) are provided in Fig. 1 (b) for illustration purpose.

 

Fig. 1 (a) Schematics of the experimental setup. SHG: Second Harmonic Generation; OPA: Optical Parametric Amplifier; AOTF: Acousto-Optical Tunable Filter; (b) The spectrum in each stage on the optical beam path.

Download Full Size | PDF

In order to acquire a coherent Raman spectrum over a broad range, we scanned the AOTF driving frequency. In this way, the wavelength of the amplified signal / idler could be sequentially swept over a certain wavelength range. The temperature of the OPA crystal could be adjusted when necessary to avoid the seed wavelength falling substantially outside the gain bandwidth. When recording the CARS signal, the shutter of the spectrometer was kept open with an appropriate short-pass optical filter placed in front of the entrance to block both the pump and the Stokes beams. The spectrometer ran in an accumulation mode so that the CARS signal could be completely recorded by the end of the sweeping procedure. When recording the SRS spectra, we modulated the AOTF output at around 17 kHz using its internal modulation function. Meanwhile, the photodiode output was demodulated by a lock-in amplifier (SRS 810, Stanford Research Systems), which was synchronized with the AOTF’s modulation frequency.

3. Results and discussions

Figure 2 shows typical spectral outputs of the OPA. Here we only illustrate the “signal” beam. When acquiring the data, the OPA crystal was heated to 122 °C, 132 °C and 145 °C, respectively. When heating the OPA to 122 °C, the self-generation of the parametric crystal ranges from 790 to 860 nm. Seed photons could be efficiently amplified to the saturation level when falling into this window. In the plot shown in the upper half of Fig. 2, we swept the AOTF driving frequency from 35 MHz to 39.5 MHz. The corresponding OPA output was tuned from ~800 nm to 860 nm, covering the entire self-generation window at this temperature. A typical line-width of the OPA output was ~1 nm (14 cm⁻¹), in agreement with previous studies [31]. The OPA’s self-generation wing was suppressed by ~20 dB as compared to the peak of the amplified signal. The bottom part of Fig. 2 exhibits the correspondence between the driving frequency and the output wavelength. The OPA crystal needed to be heated to three different temperatures in order to cover all the desired wavelength range. For each temperature, the correspondence between the diving frequency and the output wavelength exhibited a linear relationship. Since the wavelength of the supercontinuum source ranges from ~1090 nm to 1600 nm, the spectral separation between the OPA’s “signal” and “idler” beams could be tuned from ~400 cm⁻¹ to 6000 cm⁻¹ with appropriate AOTF driving frequency. The coherent Raman spectra within this range could be recorded accordingly. In practice, the minimum step size of the AOTF’s driving frequency can be as small as 10 kHz. Therefore, the minimal spectral step size of the OPA output could be as narrow as ~0.13 nm.

 

Fig. 2 Typical output of the OPA as a function of the AOTF driving frequency. Top: spectrometer reading under different AOTF driving frequency. Lower: The correspondence between AOTF driving frequency and output wavelength. Only the “signal” peaks are shown in this figure. Dots represent the position of the central peak, and the error bars represent the FWHM of the lines.

Download Full Size | PDF

Figure 3 illustrates typical coherent Raman spectra collected with the above-described system. In this demonstration, we selected solutions of ethanol and DMSO (both Sigma-Aldrich, Inc.) as representative samples. The input power of the supercontinuum source was less than 100 mW. The output power of the AOTF ranged from 1 μW to 10 μW. The output power of the OPA ranged from 1 to 50 mW, depending on the wavelength. No saturation was observed in the single-pass OPA arrangement; however, the pulse-to-pulse stability remained under 3% due to the exceptional stability of our fundamental pump laser. The parametric crystal’s temperature was set as 142 °C in order to cover Raman shift ranging from 2800 cm⁻¹ to 3100 cm⁻¹. The AOTF driving frequency was swept from 44 MHz to 46 MHz with its step size of 10 kHz. The dwelling time for each step was set rather arbitrarily to 30 ms. Since the modulation frequency was ~17 kHz, the dwelling time could be further reduced. In practice, the scanning procedure could also be accelerated by increasing the step size of the driving frequency.

 

Fig. 3 CARS, SRS and spontaneous Raman spectra for (a) ethanol and (b) DMSO at around 3000 cm⁻¹. The insets show the corresponding CARS spectra as a function of the pump wavelength.

Download Full Size | PDF

Figure 3 (a) shows the acquired CARS and SRS for pure ethanol. The spontaneous Raman spectra were collected elsewhere using a back-scattering geometry with 2-s integration time. The inset of this figure represents the CARS spectra as a function of its original wavelength. The main features are similar between the three data sets. However, the CARS and SRS spectra are weighted by the amplification efficiency of the OPA. Therefore, the signal strength at ~2600 cm⁻¹ – 2800 cm⁻¹ is stronger than the ~3000 cm⁻¹ peak. Meanwhile, the background in this region (~2600 cm⁻¹ – 2800 cm⁻¹) is still strong after the phase-retrieval procedure [20], making the overall signal-to-noise ratio (SNR) lower than ~5 for the 2710 cm⁻¹ peak. The other three peaks are recognizable and possess the SNR similar to spontaneous Raman spectra. With the advantage brought by coherent Raman process, the integration time per channel was substantially lowered (30 ms for CARS / SRS vs. 2 s for spontaneous Raman scattering). Figure 3(b) illustrates the CARS / SRS and spontaneous Raman spectra for DMSO, which presents strong Raman scattering cross-section at ~3000 cm⁻¹. Based on the 2910 cm⁻¹ Raman peak, we find the line-width for spontaneous Raman spectrum is 15.83 cm⁻¹ (fitted by the Lorentzian function). Both SRS and CARS spectra are slightly broader than spontaneous Raman (16.63 cm⁻¹ for SRS, and 19.33 cm⁻¹ for CARS, respectively).

Microscopic coherent Raman imaging is paramount for biomedical researches and material sciences. Various methods and techniques have been proposed and extensively employed to prepare suitable laser sources in CRS applications. In this report, by combining an AOTF and an OPA, we have demonstrated another approach allowing experimental arrangement with no moving parts. This setup consists only one laser source. The typical energy per seeding pulse in this experiment was ~50 pJ (10 μW in power), which was equivalent to a 10 W cw seeding laser. Therefore, the OPA could efficiently generate “signal” and “idler” emissions without significant side lobes. Meanwhile, the entire setup contains no moving parts, making the system laborsaving and cost-effective. Moreover, the pump and the Stokes beams automatically overlap with each other. No manual adjusting is needed to overlap the two beams. This feature substantially simplifies the difficulty in system maintenance.

We note that an electro-optical (EO) filter and an optical parametric oscillator (OPO) [32, 33] can be also used for low-energy applications. However, many remote sensing applications (see, for example [14, 19],) would benefit from higher energies available at the output of the OPA system described in this report. Due to the rapid response of the acousto-optical crystal, the AOTF setup possesses a faster slew rate (< 40 μs) when compared with EO filters (~100 μs). Nevertheless, this responding speed also limits our modulation frequency (up to ~50 kHz). In our current setup, the background noise still limits its applicability. Although the SNR is greater than 800 when probing pure samples, the noise can lead to some issues in practical applications including biomedical imaging. The noise is mainly generated by the unsaturated amplification, which is induced by the relatively short interaction range within the parametric crystal. A longer interaction path or a double-pass OPA will help to suppress such noise while employing other nonlinear crystals increase the gain bandwidth to avoid relatively slow temperature tuning. In the same time, the spectral resolution of SRS and CARS measurements can be further improved by narrowing down the line-width of the AOTF, for example, by means of employing a double-pass geometry [31].

4. Summary

In summary, we have demonstrated a simple and efficient approach to preparing laser excitation pulses suitable for CRS applications. We anticipate a wide use of this experimental arrangement in future applications of coherent Raman microspectroscopy.