1. Introduction

It is well known that molecular vibrations are probed via the Raman scattering process, providing information about the biochemical constituents of substances or biological specimens [1–3]. Implementing Coherent Raman scattering (CRS) micro-spectroscopy/imaging is attractive due to its capacity to perform a label-free visualization and chemical analysis of endogenous bio-samples. CRS's critical advantages over spontaneous Raman scattering are the high detection sensitivity and fast acquisition speed, where coherence is mainly responsible for the inherently stronger signal [1].

CRS imaging/spectroscopic techniques are commonly based on coherent anti-Stokes Raman scattering (CARS) and stimulated Raman scattering (SRS). The key difference between CARS and SRS is the way of retrieving the spectral signal. For single-frequency detection in CARS, the resulting low-level intensity anti-Stokes signal wavelength, compared to the higher intensity pump and Stokes excitation signals, is blue-shifted. Therefore, photomultiplier tubes are commonly used [4]. For single-frequency SRS detection, the vibrational signal can be measured utilizing a fast photodiode either at the pump or the Stokes excitation wavelengths instead. The above via either the stimulated Raman-loss (SRL) or the stimulated Raman-gain (SRG) processes co-occurring over the pump and the Stokes beams, respectively, when their difference frequency matches the vibrational Raman resonance of the probed molecule. For multiplex signals, to obtain the spectra, diffraction gratings (or prisms) are needed to disperse the light onto a CCD detector [5–7]. Indeed, several CRS micro-spectroscopy configurations can be found elsewhere nowadays. For instance: single-source excitation with multiplexed detection [8], single-beam heterodyne [9], single-source excitation with chirp [10] and spectral focusing [3,4], time-delayed probed detection [11], spectral interferometry [12], temporal gate [13] and boxcar averaging method [14], interferometric time-domain Fourier transform [15], dual-comb spectroscopy [16] and scanned-mirror interferometer [17], among many others.

Many experimental methods have also been used to extract the CARS signal, such as the ones based on high signal to noise ratio [6,18,19], on maximum entropy [20], and the subtraction between the data measured and neighboring phase gates [21]. Additionally, some computational methods have also been used for CARS signal extraction based on the fast Fourier transform [22], wavelet prism decomposition analysis [23], Kramers–Kronig transform [24], and fast non-iterative algorithm [25]. Instead, in most SRS experiments, the Raman spectral information is retrieved using a lock-in amplifier system to collect the signal with high accuracy [2,7,26,27].

One particular drawback of obtaining the coherent Raman spectra with many of the modalities mentioned above is the limited spectral range to perform wideband micro-spectroscopy/imaging. Therefore, most works focus their studies either at the CH vibrational frequencies (2800–3100 cm⁻¹) [3,4,28,29] or the fingerprint region (500–1800 cm⁻¹) [30–34], and some few works have spanned the spectral range from fingerprint to CH vibrational frequencies [5,17,33,35,36]. However, organic molecules’ fingerprint region is still more desired to probe since it provides information on vibrational frequencies specific to a molecule's chemical bonds and symmetry.

This work presents a straightforward experimental method to identify coherent Raman spectra utilizing a conventional low-cost spectrophotometer. This method's objective is to be used as a tutorial for those struggling for the first time with complex CRS signal detection while implementing for the first time CRS imaging/spectroscopy, to have a cost-effective methodology to detect and analyze coherent Raman spectra. For this, we perform CRS experiments in a non-optimum configuration using a broadband femtosecond laser to obtain both the excitation pump and Stoke beams, and high group velocity dispersion glass to stretch the obtained beams. Notice that high spectral resolution CRS can be achieved by applying similar linear chirp to both beams [3,4,29,37]. However, a common experimental situation non-CRS experts may face when trying to match the split beams’ chirp is obtaining non-perfectly time-tailored pulses and, therefore, ending up with a so-called temporal slit configuration [5]; yet a nonideal configuration now for multiplex CRS. Despite such a non-optimum configuration, we demonstrate our methodology's feasibility to extract complex to discern coherent Raman spectra. We used dimethyl sulfoxide (DMSO) in our experiments since it is utilized typically for characterizing CRS experiments either in the C-H vibrational frequencies [28,38] or towards the fingerprint region at around 670 cm⁻¹ [30,33,34]. Additionally, we also analyze CRS signals in the fingerprint region obtained from BBO crystal. To our knowledge, such crystals have been studied little using CRS techniques [39–41]. We demonstrate that the Raman spectral range from 300 to 800 cm⁻¹ can be identified with this simple methodology.

2. Theoretical background

The theory of spontaneous Raman and (single-frequency and multiplexed) coherent Raman scattering processes is found elsewhere [1–3,31,32,42]. For the sake of this work, we present only a brief description here with the help of Fig. 1.

 

Fig. 1. Schemes of spontaneous Raman process (a), coherent Raman scattering process (b), and Multiplex CRS processes using broadband sources (c).

Download Full Size | PDF

2.1 Spontaneous Raman and CRS-Raman

The spontaneous Raman scattering process is an inelastic collision of light with the vibrational energy of molecules. When incident light of frequency ${\omega _{inc}}$ interacts with a molecule of vibrational frequency $\mathrm{\Omega }$, the resulting scattered light irradiates at different energy with lower (or higher) frequency than the incident one (see Fig. 1(a)). The energy difference between the incident and the scattered light is equal to the molecule's vibrational energy lost (or energy gained). The lower frequency is called the Stokes component, $\; {\omega _s} = \; {\omega _{inc}} - \mathrm{\Omega }$, while the higher one is called the anti-Stokes component ${\omega _{as}} = {\omega _{inc}} + \mathrm{\Omega }$.

Instead, two incident fields interact with the molecule coinciding spatially and temporally for the coherent Raman Scattering process. If their frequency difference $\mathrm{\Delta }\omega = \; {\omega _p} - {\omega _s}$ matches a molecular vibrational frequency, $\mathrm{\Omega }$, the molecule is driven coherently, resulting in a Raman signal amplification due to the resonant stimulated excitation. Implying that the electron cloud surrounding the chemical bond is vigorously oscillating with the difference frequency $\; \Delta \omega $ and such electron motions alter the sample's optical properties inducing both a periodic modulation of the refractive index of the material and emission of new frequencies. Figure 1(b) depicts the CRS processes, which can be divided into two classes: the first is where the CRS signal is detected at one of the two exciting light fields by measuring the induced intensity modulation either at the pump ${\omega _p}$ or at the Stokes ${\omega _s}$ beams. The second class is where the signal is measured at a newly generated frequency ${\omega _p} + ({\omega _p} - {\omega _s})$, i.e., at a different color. As discussed next, SRS corresponds to the first class, while CARS corresponds to the second class.

In SRS, as the pump and Stokes beams propagate through the sample, their interacting light fields ${\textrm{E}_p}$ and ${\textrm{E}_s}$ are energy coupled via the modulated refractive index. Consequently, the Stokes beam's intensity, ${\textrm{I}_s}$, experiences a slight gain, $\mathrm{\Delta }{\textrm{I}_s}$ (stimulated Raman gain, SRG), while the intensity of the pump beam, ${\textrm{I}_p}$, experiences a proportional loss, $\mathrm{\Delta }{\textrm{I}_p}$ (stimulated Raman loss, SRL), as depicted in Fig. 1(b). For SRL detection, the Stokes beam's intensity is modulated at a specific frequency so that the loss $\mathrm{\Delta }{\textrm{I}_p}$ induced at the same frequency is measured at the pump beam [1,30]. Otherwise, for SRG detection, the pump beam intensity is modulated, and the gain $\mathrm{\Delta }{\textrm{I}_s}$ is detected at the Stokes beam.

In CARS, once the pump and Stokes beams drive the molecules coherently at their frequency difference, a third laser beam can essentially probe the induced molecular motion. Therefore, if a probe beam at a frequency ${\omega _{pr}}$ interacts with the periodically altered medium at a ${\omega _p} - {\omega _s}$ frequency, it will scatter at their sum-frequency ${\omega _{as}} = {\omega _{pr}} + ({{\omega_p} - {\omega_s}} )$, which is the frequency of the resulting anti-Stokes emission. Notice that the probe beam is usually conformed by photons of the pump beam arriving after the first pump/Stokes excitation, i.e., ${\omega _{pr}} = {\omega _p}$. Therefore, the measured CARS frequency is given by

Because SRS and CARS processes co-occur when the difference frequency of the excitation beams are at resonance with a specific molecular Raman vibration, $\mathrm{\Delta }\omega = \mathrm{\Omega }$, one can rewrite Eq. (1) by using the frequency-wavelength relationship ($\omega = 2\pi c/\lambda $), to obtain the following expressions:

2.2 CARS using femtosecond pulses and glass dispersion

Typical Raman resonances have coherence times in the picosecond range; therefore, single frequency CRS excitation is optimized in spectral selectivity using picosecond pulses. However, it is also possible to achieve high spectral selectivity in CARS using femtosecond pulses because the spectral resolution is not determined by the spectrum of the individual exciting pulses (with the electric fields ${E_p}$ and ${\textrm{E}_s}$), but by the spectrum of their temporal interference $\varepsilon (t )= {E_p}E_s^\ast $ which drives the molecular excitation [4,43]. The spectral width of this interference is centered at the instantaneous frequency difference (IFD) given by the Fourier limit of the temporal envelope of these interfering pulses, which can be optimized to match the Raman linewidths to improve the spectral resolution by the pulse-shaping the pulses using a glass of high group velocity dispersion (GVD) [4].

Consider two transform-limited (unchirped) Gaussian pump and Stokes pulses, traveling with a delay time ${t_0}$, with electric fields ${E_p} \propto exp[{ - {t^2}/\tau_{Gp0}^2 + i{\omega_{p0}}t} ]$ and ${E_s} \propto exp[{ - {{({t - {t_0}} )}^2}/\tau_{Gs0}^2 + i{\omega_{s0}}({t - {t_0}} )} ]$, centered frequencies at ${\omega _{p0}}$ and ${\omega _{s0}}$, with duration times ${\tau _{p0}} = {\tau _{Gp0}}\sqrt {2ln2} $ and ${\tau _{Gs0}}\sqrt {2ln2} $ (intensity FWHM). When the beams propagate through a length z of dispersive material, they acquire an instantaneous frequency

Langbein&Borri's group demonstrated that this beam shaping method with high GVD glass results in an efficient and alignment-insensitive way to spectrally focus femtosecond pulses for single-frequency CARS with a high spectral resolution by applying similar linear chirp (${\beta _p} \approx {\beta _s}$) to pump and Stokes pulses yielding a constant IFD [4,29,37,42]. Here we use this beam-shaping technique at the so-called “temporal slit configuration” [5], where both pulses are stretched to the picosecond range, but with longer Stokes pulse's widths (${\beta _p} > {\beta _s}$) to tailor the pump beam energy for multiplex CARS. Notice that being a non-optimum configuration for CRS experiments neither for spectral focusing the excitation beams nor for optimum mixing of a broadband pulse with a narrow band pulse, the configuration used in this work has generally been avoided because that makes it difficult to interpret the spectra obtained. Nevertheless, because this work intends to give new CRS users an alternate option to implement and understand their initial results, we prefer this configuration. Additionally, the CARS signal is proportional to the pump beam intensity's square, which is better suited for non-highly efficient detection.

Figure 2 depicts the effect on the spectral resolution by stretching the pulses with GVD glass. When the pump and Stokes pulses are transform-limited, i.e., unchirped $({\; {\beta_p} = {\beta_s} = 0} )$, their interaction results in multiple simultaneous vibrational frequency excitations and, therefore, a low spectrally resolved multiplexed CARS spectrum is obtained (see Fig. 2(a)). For the case of interacting pulses with the same chirp $({{\beta_p} \approx {\beta_s} \ne 0} )$, a single vibrational frequency is excited efficiently, having the so-called spectral focusing case used for optimum spectral resolution [4,42,44]. Finally, the case for pulse stretching both exciting pulses with longer Stokes pulse's widths, i.e., chirp values $\; {\beta _s} > {\beta _p}$, is shown in Fig. 2(c). Here, when one of the pulses is time delayed over its counterpart, the pump beam acts as a “temporal slit” interacting with its full frequency bandwidth with a fraction of the frequencies contained in the Stokes’ bandwidth (see also Fig. 1(c)).

 

Fig. 2. Interaction between the pump and Stokes pulses a) multiplex spectrum, b) spectral focusing, and c) temporal slit.

Download Full Size | PDF

3. Experimental

3.1 Experimental setup

The optical setup used in our experiments is shown in Fig. 3. It uses a Titanium-Sapphire (Griffin; KMLabs) laser oscillator, which delivers 120 fs width pulses at 80 MHz repetition rate, a wavelength bandwidth of 80nm (760–840 nm), and average output power of 250 mW. Note that the broadband pulse dispersion is not intracavity compensated, and therefore non-transform-limited pulses are delivered. The light beam passes through a polarizing beam splitter cube PBS (PBS102, Thorlabs), which transmits p-polarization and reflect s-polarization light components. The achromatic half-wave plate (AHWP05M-980, Thorlabs), located before the PBS, adjusts the laser power. While the achromatic quarter-wave plate (AQWP05M-980, Thorlabs), placed after the PBS, changes the linear polarization light to circular polarization light. A short-pass dichroic mirror, DBS (DMSP805, Thorlabs), transmits the spectral range from 760 to 800 nm for the pump and reflects the spectral region from 800 to 840 nm for the Stokes beams, respectively. Two extra filters, a long-pass filter (FEL0800, Thorlabs) in the pump arm, and a short-pass filter (FES0800, Thorlabs) in the Stokes's arm, are used to minimize interference around the remaining overlapped spectral region at around 800 nm; this is possible because each beam (pump and Stokes) passes twice through the short- and long-pass filters ensuring enough OD (>4) at the ∼795–805 nm region.

 

Fig. 3. Experimental setup using temporal slit configuration. In the scheme, M: mirrors; PBS: polarizing beam splitter; AHWP: achromatic half-wave plate; AQWP: achromatic quarter-wave plate; DBS: dichroic beam splitter; FES800: short-pass filter at 800 nm; FEL800: long-pass filter at 800 nm; HGVDG: high group velocity dispersion glass; EO: excitation objective and CO: collection objective.

Download Full Size | PDF

A double pass of the Stokes beam through high group velocity dispersion glass (SF57) introduces linear chirp to this part of the spectrum, stretching the pulse temporarily. When both pulses, pump, and Stokes, travel back to the PBS, the quarter-wave plate changes their circular polarization to linear s-polarization. In this way, both pulses are efficiently reflected by the PBS towards the excitation objective, EO (20x, NA:0.40; Mitutoyo), to focus their energy on the sample. Notice that the temporal pulse-width ratio (Stokes/pump) at the rear part of the EO is 4.2 times, with Stokes and pump pulses of around 765 fs and 182 fs, respectively. Therefore, the 182 fs temporal overlap of the pulses at the focal plane can be adjusted using the mirror (M3) mounted on a linear translation stage at the pump beam arm. In this way, the more efficient overlap for coherent Raman excitation and probation can be found experimentally. Finally, the collection objective, CO (10x, NA: 0.25; Olympus), is used to collect the excited Raman signals sent to the spectrophotometer (BLK-C-SR, StellarNet).

3.2 Samples

We used dimethyl sulfoxide (DMSO, Sigma-Aldrich, St. Louis, MO) and a β-phase barium metaborate (BBO) crystal for the experiments. Both samples have discrete and well-defined Raman bands in the region between 300 and 3000 cm⁻¹, useful for characterizing Raman signals. We center our study in the fingerprint region, from 300 to 1000 cm⁻¹, due to the Raman spectral range excited by our pump and Stokes pulses’ spectral widths. In this spectral region, the characteristic frequencies of the Raman molecular vibrations of DMSO are: 306 cm⁻¹, from the C-S-C bend; 334 cm⁻¹, due to the C-S-O antisymmetric bend; 383 cm⁻¹, attributed to the C-S-O symmetric bend; 670 cm⁻¹ associated to C-S symmetric stretch and 700 cm⁻¹ is related to the C-S antisymmetric stretch [45]. While for the BBO crystal, the Raman molecular vibrations bands are at 639 cm⁻¹ and 773 cm⁻¹ due to the intra-ring angle bending and stretching vibrations of intra-ring B-O bonds, respectively [46].

3.3 CRS signals extraction methodology

The CRS extraction method proposed in this paper is briefly summarized as follows: first, the CRS signal spectra (denoted by A in Fig. 4(a)) excited at the sample is collected by the CO objective. Then both pump and Stokes spectrums are collected individually and added to get a new spectrum (denoted by $\textrm{B} = \textrm{P} + \textrm{S}$ in Fig. 4(a)). Finally, the subtraction between the CRS spectrum, A, and the summation of the pump and Stokes spectrum, B, is performed (see Fig. 4(b)), which will be denoted as ($\textrm{A} - \textrm{B}$) CRS spectrum. In this way, wherever the difference is zero ($\textrm{A} - \textrm{B} = 0$), there is no Raman excitation at such wavelengths. In contrast, if the difference is non-zero ($\textrm{A} - \textrm{B} \ne 0$), it implies that such wavelengths may be interacting with another wavelength through a molecular coupling with an oscillation frequency of a molecule. The coupled wavelengths, where the SRG and SRL processes occur, can be found by inspecting the “gain and loss” peaks in Fig. 4(b) and, employing Eq. (2), seeking what pair corresponds to a specific frequency resonance $\Delta \tilde{\nu }$ of the known Raman spectrum of the sample. Then the corresponding CARS signal wavelength is computed using Eq. (3). The SRL wavelengths are identified by the pump spectral region's negative (loss) peaks, while the positive (gain) peaks at the Stokes spectral region identify the SRG wavelengths. As shown in the results, the CARS wavelength will be a positive peak found at shorter wavelengths of the pump spectra, i.e., below the corresponding SRL wavelength, as is expected because the anti-stokes signal is scattered at higher energy than the excitation signals.

 

Fig. 4. Extraction of the CRS spectrum in DMSO. a) CRS signal (curve A) and addition of both pump and Stokes spectrums (denoted by curve B). b) Difference between curve A and B to obtain the (A-B) CRS spectrum.

Download Full Size | PDF

The above methodology is repeated for each step the mirror M3 is moved to find the optimum overlap between the pump and Stokes pulses to create a more efficient excitation and probation of the CARS process; because the CRS spectrum (denoted by A in Fig. 4(a)) changes drastically depending on the exciting beams’ spatio-temporal overlap. This process is depicted in Fig. 5(a), where $\varepsilon (t )$ represents the Gaussian beam envelope of the pump, and Stokes beams interference, represented by ${E_p}(t )$ and $E_s^\ast ({t - {t_0}} )$ with a time delay ${t_0}$. Figure 5(b) shows the (A-B) CRS-spectra changes for 26 different mirror M3 positions when using a block of 4 cm length of HGVD glass (see Stokes arm in Fig. 4(b)). The spectra are taken every 50 µm-steps, the delay ${t_0}$ values are given in picoseconds. Notice that for delay times 0.49, 0.73, 0.98, 1.22, and 1.46 ps, the (A-B) CRS-spectra peaks are higher, indicating that the molecular vibrations are excited more efficiently. As shown in the results, this simple method can identify molecular vibrations in the spectral range from 300 to 800 cm⁻¹. Notice that all the spectra were acquired with an acquisition time of 300 ms.

 

Fig. 5. a) Pulse temporal overlap scheme. b) Difference (A-B) CRS spectrum variation due to the time delay (in picoseconds) between pulses when a block of 4 cm length of HGVD glass is placed in the Stokes's arm.

Download Full Size | PDF

4. Results and discussion

The feasibility to identify CRS at very low vibrational frequencies using our methodology is shown in Fig. 6, where the retrieved (A-B) CRS spectrum of DMSO depicts the different SRL/SRG and CARS excited wavelengths (Fig. 6(a)), corresponding to excited molecules at the spectral range from 250 to 400cm⁻¹ (Fig. 6(b)). Here a block of 4 cm length of HGVD glass in the Stokes's arm to stretch the beam from the fs to ps regime to improve the spectral resolution. Due to a nonideal spectral focusing, the resulting CRS spectrum poses multiple SRL/SRG peaks. Despite that, it is still possible to assign the molecular vibrations (Δν) by inspecting which of the SRL and SRG peaks are coupled with them using Eqs. (3) or (4). For example, the SRL peaks at 780 nm and 781 nm are coupled with the SRG peak at 801 nm, exciting the CARS signal at 761 and 765nm, respectively, which can be verified by computing the CARS signal wavelengths using Eq. (5) for a Stokes wavelength at 801 nm and a pump wavelength range from 770 to 785 nm. The results are shown in Fig. 6(b), which are in good agreement with the experiment. The red, blue, and black solid lines in Fig. 6(b) correspond to the Stokes, pump, and CARS wavelengths. DMSO's spontaneous Raman spectrum is also shown (in solid-gray line) to visualize the corresponding SRL and SRG coupled wavelengths and the excited CARS at each vibration frequency band. Notice that the vertical (dotted) lines show three different wavelengths triads (SRG, SRL, and CARS). The first vertical line (from left to right) corresponds to the vibrational frequency at 308.5 cm⁻¹ of DMSO; this resonant band couples the pump and Stokes pulses at 781.6 nm and 801 nm, exciting the CARS signal at 763.2nm, which is indeed close to the obtained experimentally at 765nm (see Fig. 6(a)). At 335.8 cm⁻¹, corresponding to the second vibrational frequency, the interaction between the pump at 780 nm and Stokes at 801 nm occurs, exciting a CARS signal at 760.5nm (761nm in the experiment). Finally, the vibrational frequency at 384.6 cm⁻¹ results from the interaction between 776 nm and 801 nm for the pump and Stokes pulses, respectively, exciting the CARS signal at 754.4 nm (753 nm experimentally). Notice that the CARS spectrum of Fig. 6(b) is reproduced in the retrieved (A-B) CRS spectrum of Fig. 6(a) (indicated with the black dashed square).

 

Fig. 6. Coherent Raman scattering process in DMSO, placing the mirror M3 in the position corresponding at a time delay between pulses of 0.49 picoseconds. a) Retrieved (A-B) CRS spectrum obtained using the proposed methodology and b) Spontaneous Raman spectrum used for the estimation of SRG, SRL, and CARS signals. The spectrum of Fig. b) is retrieved in Fig. a) indicated by the dashed square.

Download Full Size | PDF

Figure 7(a) shows the retrieved (A-B) CRS spectrum of DMSO at a different spectral range from 340 to 760cm⁻¹. For that, mirror M3 was positioned at a different beams-overlap position (−2.68 ps), where the monitored pump/Stokes spectra presented intense beating with the mirror movement, presumably due to the beam coupling through the excited molecular resonances. A more considerable length of the HGVD glass (8 cm) was used here in the Stokes's arm to improve the spectral resolution against the results shown in Fig. 6. Nevertheless, the resulting (A-B) CRS spectrum also possesses multiple SRL/SRG peaks due to the nonideal spectral focusing achieved, but with more intense SRL/SRG peaks resulting from the spectral resolution. For example, the SRL peak at 770.5 nm is coupled with the SRG peak at 812.5 nm, exciting the CARS signal at 732 nm; which is verified in Fig. 7(b), where the computed Stokes (red line), pump (blue line), and CARS (gray line) wavelengths, laying along the second vertical dotted line (from left to right), at the 670.8 cm⁻¹ frequency of spontaneous Raman spectrum of DMSO, are in excellent agreement with the experiment. Note that while the CARS signal for this frequency is slightly visible at 732 nm (see Fig. 7(a)), the ones at 765nm and 729nm, for the frequencies at 384.6 cm⁻¹ and 701.6 cm⁻¹ (first and third vertical dotted lines in Fig. 7(b)) are not distinguishable at all. Despite that, the excited vibrational frequency at 384.6 cm⁻¹ is still excited, as can be deduced by the appearance of the SRL and SRG peaks at 788 and 812.5 nm, respectively, in Fig. 7(a).

 

Fig. 7. Coherent Raman scattering processes in DMSO using a glass of 8 cm length in the Stokes's arm and placing the mirror M3 in the position corresponding at a time delay between pulses of −2.68 picoseconds. a) (A-B) CRS spectrum obtained with the proposed methodology. b) Spontaneous Raman spectrum of DMSO and estimation of SRG, SRL, and CARS signals.

Download Full Size | PDF

Similarly, the coupled peaks at 768.5 and 812.5 nm indicate that the vibrational frequency at 701.6 cm⁻¹ is also excited by the SRL and SRG processes. We attribute the small CARS signal due to a non-efficient readout of the excited vibration, by the pump pulse, despite a presumably efficient vibration creation via the pump and Stokes pulses’ instantaneous frequency difference, as can be seen by the more intense SRL/SRG peaks. Ideally, the maximum CARS signal occurs for a pump pulse arriving after the Stokes pulse, and the influence of such a time-ordering effect on the spectral response function is negligible only for strongly chirped pulses [4], which is not the case in our experiments.

Finally, Fig. 8(a) shows the (A-B) CRS spectrum obtained with a BBO crystal. The crystal poses multiple SRL/SRG peaks. Here, the SRL peak at 777.5 nm is coupled with the SRG peak at 828 nm, exciting the CARS signal at 733 nm, as is verified by the computed CARS that results in 732.8 nm, indicated with the fourth vertical dotted line (from left to right) in Fig. 8(b). From Fig. 8(b), it can be seen that the other three CARS signal wavelengths (746.2, 748.9, and 743.5 nm) are also in good agreement with the obtained experimentally in the retrieved (A-B) CRS spectrum of Fig. 8(a) (indicated with the black dashed square).

 

Fig. 8. Coherent Raman scattering processes in BBO, placing the glass of 4 cm length in the Stokes's arm and positioning the mirror M3 at a time delay between pulses of −0.24 picoseconds. a) (A-B) CRS spectrum obtained with the proposed methodology. b) Spontaneous Raman spectrum of BBO at 0 degrees and estimation of SRG, SRL, and CARS signals.

Download Full Size | PDF

Note that in all the (A-B) coherent Raman spectra showed in Fig. 6(a), Fig. 7(a), and Fig. 8(a), other gains, and losses peaks are observed at the Stokes beam. We attribute those peaks due to the excitation and probation of molecular vibrations with wavelengths within Stokes pulse's spectral width. For instance, for the BBO crystal, it is known that its Raman spectrum possesses bands between 350 cm⁻¹ and 500 cm⁻¹ (not shown in Fig. 8); at 385 and 480 cm⁻¹ [47]. Therefore, the bands centered at 789.5 nm and 792.5 nm may be CARS signal-induced via the Raman vibration at 480 cm⁻¹. The former is excited by the pair wavelengths 820.8 nm (SRL) and 854.5 nm (SRG) and the latter by 823.6 nm (SRL) and 857.5 nm (SRG), which are close to the peaks indicated by the red and blue arrows accordingly. Additionally, the Raman band at 796.5 nm is associated with the vibrational mode of 380 cm⁻¹, which results from the interaction of 821.3 nm (SRL) and 847.8 nm (SRG).

5. Conclusions

An experimental method to identify coherent Raman spectra using a conventional low-cost spectrophotometer has been presented. The proposed methodology can identify SRG, SRL, and CARS peaks from molecular vibrations range from 300 to 800 cm⁻¹ in both DMSO and BBO crystal samples. This method represents a straightforward tutorial for complex CRS signal detection and analysis for implementing CRS imaging/spectroscopy.