Raman optical activity (ROA) spectroscopy has been used as an analytical tool in stereochemistry and biochemistry [1–3]. Although it is widely known that ROA spectra provide rich structural information on chiral molecules, its extremely weak signal has made measurements challenging. In ROA spectroscopy, difference Raman spectra between right and left circularly polarized (RCP and LCP) incident and/or scattered light is measured. However, the intensity difference is typically less than ${10}^{-3}$ times that of the parental Raman intensity. In order to improve the signal to noise ratio (SNR) of ROA spectra, many instrumental improvements have been made [4–7]. Recently, we have developed ROA spectroscopy by using coherent anti-Stokes Raman scattering (CARS-ROA) [8,9]. In CARS-ROA spectroscopy, both the phase and amplitude of CARS radiation polarized perpendicularly to the incident polarization is measured with a heterodyne technique, in which the achiral CARS field generated from the sample itself [8] or external reference material [9] was used as a local oscillator. One of the most striking advantages of CARS-ROA spectroscopy over conventional spontaneous ROA spectroscopy is its higher contrast ratio of the chirality-induced signal to the achiral background. In our previous study, it was demonstrated that the contrast ratio of CARS-ROA measurement of $\beta$-pinene is two orders of magnitude higher than that of spontaneous ROA measurement [8]. This improvement is significantly important in view of a future application to time-resolved ROA spectroscopy, which is a great challenge because the time-dependent change of the extremely weak ROA signal needs to be extracted from its huge achiral background. In order to improve the detection sensitivity by compensating the laser fluctuation, scattered circularly polarized (SCP)-ROA spectroscopy has been reported [6]. However, this technique still requires the extraction of the weak ROA signal from the huge achiral background after the signal detection. On the other hand, our CARS-ROA spectroscopy enables us to suppress the achiral background before the signal detection. Measurements less susceptible to the laser fluctuation are, therefore, realized in CARS-ROA spectroscopy. However, the SNR of the spectra obtained in CARS-ROA spectroscopy was not satisfactory in comparison with the conventional ROA spectroscopy due to the weak signal.

The weak signal intensity in our previous setups is partly due to near-infrared (NIR) excitation; we employed 1064 nm and a supercontinuum (SC) ranging from 1.1 to 1.6 μm light source for ${\omega }_1$ and ${\omega }_2$, respectively. Although NIR excitation has several merits, such as low photodamage to the sample and weak fluorescence interference [10,11], it significantly sacrifices the SNR because the ROA signal is proportional to the fifth power of the incident frequency [1,10]. Herein, we have developed a CARS-ROA spectrometer with visible (532 nm) excitation and improved the SNR of ROA significantly.

The schematic of our visible-excited CARS-ROA spectrometer is shown in Fig. 1. A 25 kHz microchip laser (Hamamatsu, L11475), which provides 400 ps pulses at the wavelength of 1064 nm, is used as a light source. The output of the laser is frequency-doubled by a $\mathrm{L}\mathrm{i}{\mathrm{B}}_3{\mathrm{O}}_5$ crystal and the generated second harmonic is separated from the residual fundamental. The major part ($\sim 100\mathrm{mW}$) of the 532 nm radiation is used as narrowband ${\omega }_1$ and the other is used to excite a photonic crystal fiber (PCF). In this study, we employ the dual pumping scheme, in which the PCF is pumped both with 532 and 1064 nm radiation [12,13]. The spectra of SC generated by the PCF with different pumping schemes are shown in Fig. 2. The spectral profile of the SC pumped only by 532 nm is dominated by spiky peaks originating from stimulated Raman processes, which is unsuitable for multiplex CARS spectroscopy. When the PCF is excited both with 532 and 1064 nm, broad and intense SC, which covers almost all of the fingerprint region, is generated in the current experimental condition. The optical configuration around the sample is basically the same as what we previously reported in the NIR-CARS-ROA spectroscopy [8]. Incident ${\omega }_1$ and ${\omega }_2$ polarizations are set parallel to each other with a single Glan–Taylor polarizer. The polarization of the CARS field is selected by the second Glan–Taylor polarizer after the sample, which is slightly tilted from the perpendicular configuration (referred to as $\theta$), so that the small portion of achiral CARS field passes through the polarizer and acts as a local oscillator. Spectra are obtained with a polychromator (Acton SP2300, Princeton Instruments) and a CCD camera (PIXIS 100BR eXcelon, Princeton Instruments). Simulation of the CARS-ROA spectrum is performed by using density functional theory with Gaussian 03 [14]. The minimum energy structure was optimized with the B3PW91 functional and aug-cc-pVDZ basis set and Raman/ROA properties were calculated with the B3LYP functional and 6-31G** basis set. A scaling factor of 0.97 was applied for comparison with the observed spectra.

 

Fig. 1. Schematic diagram of the visible-excited CARS-ROA spectrometer. HWP, half-wave plate; DM, short-pass dichroic mirror (Thorlabs, DMSP805); DM2, long-pass dichroic mirror (Thorlabs, DMLP900); PBS, polarized beam splitter (Thorlabs, PBS201); PCF, custom-made photonic crystal fiber; OD, optical delay; LPF, long-pass filter (Semrock, BLP01-532R-25); GTP, Glan–Taylor prism (Thorlabs, GT5); S, sample; NF, notch filter (Semrock, NF03-532/1064E-25); SPF, short-pass filter (Semrock, SP01-532RU-25).

Download Full Size | PDF

 

Fig. 2. Spectra of the SC generated with the different pump schemes. These are measured as nonresonant CARS spectra of water.

Download Full Size | PDF

The procedure for extracting CARS-ROA spectra is based on that previously reported [8]. We have measured CARS spectra at $\theta =\pm 0.5°$ and the CARS-ROA spectrum can be obtained as the difference spectrum between these. As previously discussed [9], however, CARS-ROA spectra are distorted by optical rotatory dispersion (ORD) of the sample itself. With ORD (here denoted as $\alpha$), the CARS spectrum obtained at $\theta$ is written as

 

Fig. 3. CARS spectra of (+)-$\beta$-pinene measured at $\theta =0.5°$ (red curve) and $\theta =-0.5°$ (blue curve). Dots represent the ratio of these two CARS spectra $I(\theta )/I(-\theta )$ and the dashed curve is from the spline fitting.

Download Full Size | PDF

The CARS and CARS-ROA spectra of $\beta$-pinene measured with the 532 nm setup developed in this work are shown in Figs. 4(a)–4(c). For comparison, the CARS-ROA spectrum of (-)-$\beta$-pinene obtained with the previous 1064 nm setup is shown in Fig. 4(d). In the NIR-CARS-ROA spectrum with 1 h exposure time, weak ROA signals in the higher wavenumber region ($>900{\mathrm{cm}}^{-1}$) were severely buried in noise although several characteristic peaks between $600{\mathrm{cm}}^{-1}$ and $900{\mathrm{cm}}^{-1}$ were detected [Fig. 4(d)]. In the VIS-CARS-ROA spectroscopy, these characteristic peaks are clearly observed only by 1 min exposure [Fig. 4(b)]. With 1 h exposure time [Fig. 4(c)], the VIS-CARS-ROA spectrum was obtained with a high SNR and almost all the peaks are in the mirror image of the enantiomer’s spectrum. Nearly all the spectral features in the measured spectrum also show good consistency with those in the spectrum simulated with quantum chemical calculations.

 

Fig. 4. (a) CARS spectrum of (−)-$\beta$-pinene with 1 h exposure time with the 532 nm setup developed in this research (reduced to 1/10 for comparison). (b) CARS-ROA spectra of (+)-$\beta$-pinene (black) and (−)-$\beta$-pinene (red) obtained with the 532 nm setup. The exposure time was 1 min, and the excitation power was 100 mW for ${\omega }_1$ and 20 mW for ${\omega }_2$. (c) CARS-ROA spectra of (+) and (−)-$\beta$-pinene obtained with 1 h exposure. Measurement conditions other than exposure time are the same as (b). (d) CARS-ROA spectrum of (−)-$\beta$-pinene obtained with the previous 1064 nm setup with the 1 h exposure. The excitation power was 200 mW for ${\omega }_1$ and 70 mW for ${\omega }_2$. (e) Calculated CARS-ROA spectrum.

Download Full Size | PDF

Although it is difficult to quantitatively compare the VIS-CARS-ROA spectrum to the NIR-CARS-ROA spectrum because of the difference in the instrumental performance, the ratio $\mathrm{Re}[{\chi }_{1111}^{*}{\chi }_{2111}]/{|{\chi }_{1111}|}^2$ is independent of instrumental response and can be directly compared. This ratio is calculated as

These ratios for NIR- and VIS-CARS-ROA spectra are shown in Fig. 5. The ratio calculated for the visible excitation is even greater than two times that for the NIR excitation, which is probably due to electronic preresonance. This enhancement, in addition to the reduction of the achiral background in CARS-ROA spectroscopy, makes measurements even easier.

 

Fig. 5. Comparison of the spectra of $\mathrm{Re}[{\chi }_{1111}^{*}{\chi }_{2111}]/{|{\chi }_{1111}|}^2$ calculated from the VIS- and NIR-CARS-ROA measurements.

Download Full Size | PDF

In conclusion, a visible-excited CARS-ROA spectrometer was realized. By employing the dual pumping scheme, we have generated broad and intense SC around 532 nm, which enables multiplex visible CARS-ROA spectroscopy. We have measured CARS-ROA spectra of (+)- and (-)-$\beta$-pinene with the visible excitation as a demonstration of the setup. The obtained spectra in VIS-CARS-ROA are consistent with both the spectrum measured in the previously developed NIR-CARS-ROA setup and that simulated by quantum chemical calculation. In the visible excitation, the SNR higher than that in the NIR excitation was realized mainly due to a large scattering cross section and an optical activity tensor.