1. Introduction

Rapid detection of bacterial spores becomes an increasingly important task, and new approaches and methods based on nonlinear optical spectroscopy have been recently introduced as a promising new route for rapid and remote detection and recognition of certain molecular species [1–2]. Vibrational spectroscopy is particularly useful, since unlike fluorescence spectroscopy it provides precise fingerprint information on molecular species [3–4]. The analysis of vibrational spectra, in principle, allows for the identification of different molecules in the gas mixture and in solution. However, the major vibrational spectroscopy tools based on Raman spectroscopy and infrared absorption spectroscopy are often hardly applicable for remote sensing. One of the major problems of a wider spread of Raman spectroscopy, for example, is a very low Raman scattering cross-section resulting in a weak signal. Although the technique of enhancing this cross-section via plasmonic resonances of specially made substrates (so-called Surface Enhance Raman Spectroscopy, SERS [5]) exists, it requires a direct contact of solution or gas mixture under study with the substrate. In this respect, coherent anti-Stokes Raman scattering (CARS) is an elegant nonlinear optical approach to enhance the signal by means of coherent nonlinear optical excitation of vibrational levels. CARS spectroscopy has been extensively used for combustion diagnostics [6–7], and, recently, CARS spectroscopy has attracted a significant attention of microscopists [8–14] since it allows real-time microscopic imaging with vibrational contrast. While CARS spectroscopy has been widely used to study molecules in solution, it has been mostly limited to investigations of simple molecules, for which vibrational spectrum is mostly concentrated in a narrow spectral region, and observations are limited to a single molecular band or a single vibrational line. To be able to detect molecules and, especially, to identify bacterial spores, it is necessary to go beyond that and be able to record the whole vibrational spectrum, which can be then analyzed for specific signatures. In this report we used a broadband CARS spectroscopy to detect Bacillus subtilis (B. subtilis) spores in aqueous solution.

2. Raman spectroscopy of B. subtilis

The choice of the system is driven by the necessity of detecting anthrax spores [1]. B. substilis spores chemically resemble anthrax spores and are, unlike anthrax, harmless. It is well known that 2,6-pyridinedicarboxylic acid (or dipicolinic acid (DPA)) is the major chemical component of bacterial spores accounting for over 15% of its molecular weight [1,15]. The Raman spectra of DPA have been recently revisited and mode assignment was provided [16]. The Raman spectra of different bacteria have been also measured [17]; however, since we are using a specific strain PS533 [18], we decided first to measure this spectrum for future comparison. We used a commercial Raman microscope (Renishaw-2000, Renishaw, Inc.), which uses 632.8-nm line of He-Ne laser for excitation of Raman spectrum. The Raman spectrum is collected in a backscattering geometry and, after passing through a set of two notch filters, is directed to a spectrometer with the attached TE-cooled CCD for detection. The spectrometer and the whole detection system are intensity and wavelength calibrated. The accuracy of the wavelength for the center line determination is better than 0.5 cm^-1, and the spectral resolution is about 3 cm^-1.

B. subtilis spores were dispersed on a glass slide and the incident laser light was focused on those spores with a 50x microscope objective (N.A. = 0.55). The incident power was properly attenuated to ensure the noninvasiveness of excitation. The obtained spectrum is largely dominated by a strong fluorescent background (see Fig. 1(a)), which is rather typical for biological systems. To obtain a high quality Raman spectrum we set the acquisition for continuous accumulation and averaging of 1000 spectra in the spectral region from 400 to 4000 cm^-1. After subtracting the fluorescent background using a standard procedure described elsewhere [19], we obtained the spectrum shown in Fig. 1(b), which resembles the Raman spectrum of DPA and the earlier recorded spectrum of B. subtilis. We note that in the high-frequency region (>2700 cm^-1) the Raman spectrum of spores differs significantly from that of DPA. We attribute this difference to the presence of other organic compounds in spores. Clearly, the region from 700 to 1700 cm^-1 appears to be the most attractive to identify the presence of spores, given the number of strong isolated spectral lines and well-pronounced bands.

 

Fig. 1. Raman spectrum from B. subtilis spores recorded as it is (a), and with digital background subtraction (b).

Download Full Size | PDF

We prepared a water solution of those spores by adding 10 mg of the spores’ powder to 1 mL of double distilled deionized water. As a result, a colloidal solution of those spores was formed, which was used for further analysis. We unsuccessfully tried to detect Raman spectrum of those spores in solution. The signal was at least two orders of magnitude weaker, while the Brownian motion of spores in solution and light scattering in solution provided an additional source of noise, so that no Raman spectra could be collected in a reasonable amount of time. That is why we use CARS spectroscopy thus getting stronger signal and isolating ourselves from the strong fluorescent background.

3. CARS experimental set-up

The CARS signal is the product of nonlinear optical interaction of three waves resulting in the coherent generation of the fourth wave at the frequency ω_CARS = 2ω ₁ - ω ₂, where ω ₁ and ω ₂ are the frequencies of two pump waves, which have to be temporarily and spatially overlapped in the sample. In a typical CARS experiment, one has to scan one frequency with respect to another to obtain the whole vibrational spectrum. Not only it is time consuming, but it also requires significant number of tweaking through the whole optical system, making those measurements complicated and tricky. However, if one of the waves, for example, ω ₂, is substituted with a broadband wave, and a multichannel detector is used in conjunction with a spectrometer for detection, the whole CARS spectrum can be recorded simultaneously [20–22]. This experimental arrangement greatly simplifies the CARS apparatus, making CARS experiments essentially hands-off measurements. Recent advances in laser engineering have led us to the experimental set-up shown in Fig. 2 [22]. In brief, picosecond high-energy oscillator generates transform-limited pulses at the repetition rate of 6 MHz and an average power of 1.8 W with typical pulse duration of 3–4 ps, centered at 1064 nm. This fundamental wavelength serves as ω ₁ for the CARS process. To produce a broadband continuum wave temporarily synchronized with the fundamental wave, we send about 600 mW of the fundamental radiation into a single mode GeO₂, where a red-shifted ultra-broadband white-light continuum is generated via consecutive stimulated Raman scattering processes. The infrared part of this radiation is then collinearly combined with the temporally delayed remaining fundamental radiation and is used for CARS experiments. All the measurements are done in transmission, in a 1-mm-thick glass cell. Both pump waves are focused with an achromatic lens into the central part of the cell, and the CARS radiation is collected with another achromatic lens and directed into fiber input of a 1/3-meter spectrometer (Jobin-Yvon, Inc.). The combination of 200-μm fiber output diameter and spectrometer dispersion results in about 15 cm^-1 spectral resolution. For a given focusing geometry, we estimated the interaction length to be about 10 μm. However, since the spore’s size is about 1-μm in diameter and on average there is less than one spore in the focal volume of the laser beams, the effective interaction length with the spore’s material is only 1-μm, and phase-matching requirements for CARS signal are not that important. With all the advantages of CARS spectroscopy, it suffers from a so-called non-resonant background. This background arises from both the molecules under study and surrounding molecules (typically, water). There are several well-established methods to suppress non-resonant backgrounds, including polarization spectroscopy [23], coherence-controlled CARS spectroscopy [1, 24], differential spectroscopy [25], and epi-detection CARS [26]. In our laboratory, we are using polarization suppression scheme, which has proved to be a reliable method in obtaining resonant CARS spectra. We set the angle between polarizations of two pump beams to 60°, so that the polarization vector of non-resonant CARS component is directed along the bisector of this angle. The analyzer is then oriented perpendicular to this direction, so that only the resonant CARS signal will be detected in this direction [22, 26]. In reality, depolarization effects in solution and imperfections in polarizing optics result in some leakage of non-resonant CARS signal into the detection system defining the ultimate limit of this technique. By careful choosing of all the optical elements in our system and by pre-compensating for depolarization on optical components using the combination of quarter- and half- waveplates [27], we were able to achieve more than 4.5 orders of magnitude suppression of non-resonant signal. This can be done for any solution of interest in the spectral region from 1800 to 2000 cm^-1 region, where no resonances exist. The recorded signal, which in our experiments exceeded the leakage signal by at least an order of magnitude, was spectrally dispersed and recorded by a liquid-nitrogen-cooled CCD. This spectrum was than normalized by the non-polarized CARS signal recorded in the same geometry for any non-absorbing inorganic material, which typically lacks vibrational resonances in our spectral region of interest. This procedure allowed us to take into account the spectrometer transmission, the CCD spectral response, the spectral modulations of a broadband continuum, and all the wavelength-dependent aspects of our optical set-up [22].

 

Fig. 2. Schematic diagram of the experimental CARS set-up [22].

Download Full Size | PDF

4. Experimental results

The recorded spectrum was rather strong, so that we were able to get all the major spectral features within 1-s exposure time on the CCD; however, it had a significant temporal modulation, which we attributed to the small excitation volume (~10 μm³) and fluctuations of the number of spores, which were present in this volume during the exposure time. That is why we have recorded a series of spectra with an exposure time of 100-s, which were taken for different points inside the cell in order to avoid any possible inhomogeneities of spores distribution inside the cell. While we found very small variations in the signal intensity, and essentially no variations in relative spectral intensity of different lines, the resultant spectrum presented in Fig. 3 (b) and (d) is the average of 5 different spectra. Since a single pixel on the CCD detector corresponds to the spectral width significantly larger than the spectral resolution estimated from the dispersive properties of our spectrometer, all the spectra were smoothened using a standard adjacent averaging procedure (Origin 7.5, Microcal, Inc.). The Raman spectrum from the same spores is shown in Fig. 3 (a) and (c) for comparison. It is clear that, while the relative ratio of different spectral components has changed, the position of the major vibrational lines and bands remains essentially the same. The variations can be easily understood from several key differences. Firstly, the excitation wavelengths for CARS and spontaneous Raman spectra are dramatically different. Secondly, the polarized CARS signal is proportional to |χ ^(3),R|², while spontaneous Raman spectrum - to (IM(χ ^(3),R))². Finally, one has to keep in mind that spontaneous Raman signal is proportional to the concentration of molecular species, while the CARS signal is proportional to the square of their concentration. Since DPA makes up about 15% and has the highest concentration among all the other organic molecules in B. subtilis, we should expect that the relative contribution of DPA to the CARS signal should increase dramatically with respect to its contribution to the Raman signal. Nevertheless, despite of all these minor differences two characteristic isolated peaks at around 825 and 1010 cm^-1 are clearly seen in both spectra signifying the presence of bacterial spores in solution.

 

Fig. 3. The experimentally measured CARS spectrum of bacterial spores in water solution (10mg/mL) - (b) and (d), compared with the experimentally measured spontaneous Raman spectrum of bacterial spores in a powder form (a) and (c).

Download Full Size | PDF

At present, there are many possible ways to improve the sensitivity limit in our detection system. We are using just the simplest background suppression scheme without optimizing the overall signal. It is well known that, for example, heterodyne detection of CARS signal, which was originally demonstrated in 1979 [23] and recently applied for microscopic imaging [28], can significantly amplify the signal, while providing a measure of (Im(χ ^(3),R))², i.e. being directly comparable with spontaneous Raman signal measured at the same excitation wavelength. Another alternative is to use the non-resonant nature of background to numerically subtract it and fit non-polarized CARS spectra with the known spectral shapes [29]. By doing this, we hope to improve the discrimination against the other types of similar spores, as it was done earlier in the case of Raman experiments [3, 17].

5. Conclusion

In summary, we have clearly demonstrated, for the first time, the applicability of CARS spectroscopy to the detection of bacterial spores in aqueous solution. It has been demonstrated that CARS spectroscopy is favorably compared to the spectroscopy of spontaneous Raman scattering and can be potentially used for remote sensing of biohazards.