1. Introduction

Motivated by the potential applications in imaging and microscopy, coherent anti-Stokes Raman scattering (CARS) has been extensively researched in the past decades [1,2]. Since CARS microscopy is label-free, has high sensitivity and three-dimensional sectioning capabilities [1], CARS is attractive for applications in medicine and biology [3–5]. CARS is a four-wave mixing process for vibrational spectroscopy, in which a pump field with frequency ω_p, a Stokes field with frequency ω_s, and a probe field with frequency ω_pr interact with a sample and generate an anti-Stokes field at a higher frequency given by ${\omega }_{as}={\omega }_p-{\omega }_s+{\omega }_{pr}$. The anti-Stokes signal can be easily detected, as it is blue-shifted from the probe frequency by the characteristic molecular vibrational frequency W_vib. Thus, using a short-pass filter the CARS signal can be readily separated from the excitation light. CARS spectroscopy can be performed both with narrowband sources, as previously described [1], or with a single broadband source that supplies all the necessary frequencies, ω_p, ω_s, and ω_pr. This method is termed as single-pulse CARS, and uses femtosecond pulses to impulsively excite the Raman level [2].

To perform single-pulse CARS, some form of spectral shaping must be employed in order to regain spectral selectivity. Conventionally, a pulse shaper is used to apply a spectral phase which either selectively excites a chosen level [2], or creates narrow phase features that restore spectral resolution [6,7]. A compact and readily implementable method for performing single-pulse CARS spectroscopy and microscopy has recently been demonstrated in our group [8], where spectral selectivity is achieved with minimal shaping, using a narrow notch filter (NF). The filter is used to eliminate one spectral component from the pulse spectrum, effectively creating a narrowband probe field with much longer duration, as shown in Fig. 1(a). The Raman spectrum is resolved from small narrow features created in the broadband CARS signal, at frequencies $\omega ={\omega }_{NF}+{\Omega }_{vib}$ where ${\omega }_{NF}$ is the frequency eliminated by the notch filter, as shown in Fig. 1(b). That signal, however, is dominated by a large non-resonant (NR) background that is not sensitive to the Raman levels [9]. The NR background is often much stronger than the resonant one, and may distort or even overwhelm the resonant signal of interest [2]. The strong background signal usually necessitates lock-in detection, thereby slowing the imaging process. While narrowband CARS microscopy effectively suppresses the NR background [10], the technique is not easily combined with other forms of nonlinear microscopy that require short pulses. Single-pulse CARS microscopy is compatible with a variety of nonlinear processes such as second- and third-harmonic generation (SHG, THG) and four-wave mixing (FWM) that benefit from the high peak intensity of the femtosecond excitation [11, 12]. It is therefore highly desirable to develop a single-pulse CARS microscopy method that does not require lock-in detection.

 

Fig. 1 Time-frequency plots of single-notch CARS, C-CARS and double-notch C-CARS processes. (a) Single-notch CARS excited by a transform-limited (TL) pulse. Shown are the spectrum of the pulse (ellipse with a hole), the long probe pulse (green ellipse) induced by the notch filter, and the band of vibrational frequencies (blue) that can be excited. (b) The single-notch CARS signal with two dip features (shown are two vibrational levels R₁ and R₂ spaced by Δω_R) induced by a single-notch shaped TL pulse. In this process, the contrast is poor because the whole excitation spectrum contributes to both resonant and NR signals. (c, d) Single-notch C-CARS induced by a chirped pulse. The Raman span is narrower as compared with case a, yet it leads to a higher contrast. (e) Double-notch C-CARS process produced by a chirped pulse shaped with two notches (1 and 2). Here, the two notches are spaced by Δω_n. (f) Double-notch C-CARS signal with highest contrast based on coherent addition when Δω_R = Δω_n.

Download Full Size | PDF

Here, we demonstrate high-contrast single-pulse chirped CARS (C-CARS) microscopy without lock-in detection by applying notch-filtering to a pulse with a relatively small chirp applied to its excitation spectrum. The effect of a small chirp is to elongate the broadband pump field in time, while hardly affecting the narrowband probe field, as seen in Fig. 1(c). The elongation serves to significantly reduce the NR background while causing only a minor reduction of the resonant signal, as shown in Fig. 1(d). An optimal chirp exists for which the ratio of resonant to NR signal is the largest, enabling CARS microscopy with high contrast. This lock-in-free imaging method could find potential applications in live cells and other fast imaging tasks [13]. We note that one disadvantage of this method is the somewhat reduced maximal Raman frequency that can be excited by the chirped pulse. We also note that while chirped pulses are employed in CARS methods based on spectral focusing [14–16], these use much larger chirp values that reduce the peak power significantly, and are therefore incompatible with most other nonlinear modalities.

To additionally increase the contrast and enhance the resonant signal level, multiple-notch probes are combined to achieve coherent enhancement [17]. When using a double-notch probe, for example, enhancement occurs when the spacing of two notches $\Delta {\omega }_n={\omega }_{N1}–{\omega }_{N2}$ matches that of two Raman lines $\Delta {\omega }_R={\Omega }_{vib1}–{\Omega }_{vib2}$_, as shown in Figs. 1(b) and 1(e). In this case, the CARS feature R₂ generated by Notch 1 at $\omega ={\omega }_{N1}+{\Omega }_{vib2}$will constructively interfere with the R₁ CARS feature created by Notch 2 at the same frequency $\omega ={\omega }_{N2}+{\Omega }_{vib1}$. This constructive interference that specifically enhances a signal from a medium with a particular two-line structure results in a feature with a larger amplitude and thus an improved signal-to-background ratio, as shown in Fig. 1(f). We show below that combining chirped CARS with coherent addition enables CARS microscopy that is compatible with other multimodal nonlinear imaging modalities without lock-in detection.

2. Theoretical description

Consider a linearly chirped Gaussian pulse can be written as

Applying a small chirp to the excitation pulse affects the lineshape of the resonant CARS polarization only slightly. The magnitude of the NR polarization, on the other hand, is reduced dramatically as the chirp is increased according to numerical simulation with Eq. (6).

In order to estimate the contrast enhancement, we calculate the ratio between the magnitude of the CARS feature and the NR background for the excitation of a 685cm⁻¹ line (a perfluorodecalin line) and its dependence on the chirp parameter. This ratio can be expressed as $(I-I_{NR})/I_{NR}$, in which the NR background is $I_{NR}\propto {\left|P_{NR}^{(3)}(\omega )\right|}^2$. In our numerical simulation, we assume that FWHM of the notch is 1.3nm, the notch location is fixed at 763nm, and that the central wavelength of the laser is at 800nm. We choose ${\Omega }_{vib}=685c{m}^{-1}$ and ${\Gamma }_{vib}=2c{m}^{-1}$ . $\left|k/{\chi }_{NR}^{(3)}\right|=2.5c{m}^{-1}$was determined by fitting the experimental results. We also assume that the temporal FWHM of the unchirped pulse is $\Delta \tau =15fs$ and that its spectral width ${\Delta }_{TL}$ satisfies ${\Delta }_{TL}\cdot \Delta \tau =0.44$. The pulse width of the chirped pulses is longer than the TL pulses by a factor of ${(1+{\beta }^2)}^{1/2}$, while the instantaneous linewidth is narrower, ${\Delta }_{TL}/{(1+{\beta }^2)}^{1/2}$ . Since the instantaneous bandwidth of the chirped pulses should be larger than the vibrational energy, this limits the chirp value to $\beta \le {[{(4{\Delta }_{TL}\ln 2/{\Omega }_{vib})}^2-1]}^{1/2}$.

Figure 2 presents simulation results, which show that the ratio between the CARS feature and NR background increases at first and then decreases with the chirp parameter. We see that the optimal chirp for perfluorodecalin at 685cm⁻¹ is obtained for β = 3.2. As the NR background decreases, the peak-dip feature [8] becomes a pure peak as is shown schematically in Fig. 1(d). The optimal chirp depends on the vibrational energy and the bandwidth of the TL laser pulses. We note that the optimal value of the chirp parameter decreases with Raman energy.

 

Fig. 2 Simulation results: the ratio of single-notch C-CARS feature’s magnitude at 685 cm⁻¹ of perfluorodecalin to the magnitude of the NR background at the same wavelength as a function of chirp parameter.

Download Full Size | PDF

As mentioned above, coherent enhancement can be achieved by two notches with proper frequency spacing which induce two sets of CARS features that overlap at a particular frequency. This enhancement occurs due to constructive interference between the two resonant CARS features. Coherent enhancement is possible when there are two vibrations that can be excited, and then an effective resonant third-order susceptibility ${\chi }_{R,eff}^{(3)}(\Omega )$ can be modeled as

The merit of this method is that it produces an enhanced feature only for the specific molecule for which it was designed. Therefore, double-notch CARS serves as ‘matched filter’ [18], providing a quantitative measure of the amount of a given substance at a specific point in the sample.

3. Experimental setup

The experimental setup is shown in Fig. 3(a). The laser source is a home-built Ti:Sapphire (Ti:S) oscillator with a bandwidth of 60nm FWHM, central wavelength of 800nm, and 80 MHz pulse-repetition rate. The dispersion is controlled by a pulse compressor composed of a grating (Milton Roy Company, blaze at 750nm, 1200 grooves/mm), a curved mirror, and a planar mirror. The pulses can be tuned to be transform limited with pulse width of approximately 15 femtoseconds (at the sample), and chirp is introduced by moving the distance d between the curved mirror and the grating. The chirp parameter β_m, is given by [19]

 

Fig. 3 Experimental setup for single-notch and double-notch C-CARS spectroscopy and microscopy. (a) Diagram of the experimental setup. A femtosecond laser after compressed by a pulse compressor goes through the first notch filter (NF1) once (i) or twice (ii) and then goes to the microscope. CARS and SHG are divided into corresponding PMT (Photomultiplier) for respective imaging. FM: flip mirror. G: gratings, CM: curved mirror, LPF: long-pass filters, SPF: short-pass filters, BPF: band-pass filters, DM: dichroic mirrors, S: sample, C: condenser lens. (b) Notch shaped spectrum of laser pulses and corresponding CARS signals. (i) Single-notch case. (ii) Double-notch case.

Download Full Size | PDF

4. Results and discussion

4.1 Single-notch C-CARS

As predicted above, the contrast between the resonant CARS signals and NR background is enhanced when the chirp parameter, which is controlled by tuning the distance between the grating and the curved mirror shown in Fig. 3(a), is increased. This can be seen by comparing the measured spectra with the notch filter (solid lines in Fig. 4) and the spectra without the notch filter (dashed lines in Fig. 4) for several chirp values. These measurements were performed both for the 670cm⁻¹ Raman line of chloroform [Fig. 4(a)], and for the 685 cm⁻¹ Raman line of perfluorodecalin (Sigma-Aldrich P9900) [Fig. 4(b)]. The inset figures are the corresponding Raman lines resolved by subtracting the CARS spectrum without the NF, from the spectrum with the NF, for the optimal β_m values (green). Aside from increasing the contrast, adding dispersion obviously reduces the absolute magnitude of the resonant signal, and also changes the shape of the resonant feature. For example, for β_m = 0 in chloroform [black solid line in Fig. 4(a)] the resonant feature appears as a dip in the spectrum, whereas for β_m = 2.2 [pink solid line in Fig. 4(a)] the feature appears as a peak. The contrast reaches maximum at β_m = 3.3 for the 670cm⁻¹ Raman line [Fig. 4(a)] and β_m = 3.2 for the 685 cm⁻¹ Raman line of perfluorodecalin [Fig. 4(b)]. These results agree well with the theoretical prediction above.

 

Fig. 4 Experimental results of single-notch C-CARS: enhancement of the ratio between resonant CARS feature and smooth background. (a) 670cm⁻¹ Raman line of chloroform. (b) 685cm⁻¹ Raman line of perfluorodecalin. The solid curves are measurements of the CARS spectrum with the notch filter, NF1, and the dashed curves are background measurements, without NF1. Insets: the Raman lines, resolved by subtracting the measurements without NF1 from the measurements with NF1. The measurements were performed with 15mW average laser power and 1s integration time.

Download Full Size | PDF

4.2 Double-notch C-CARS spectroscopy

Here we consider the CARS spectrum of toluene, which has Raman lines at 791cm⁻¹, 1003cm⁻¹, and 1212cm⁻¹. With the laser beam passing through the notch filter twice, double-notch shaped pulses are produced as shown in Fig. 5(a). Figure 5(b) shows the raw measured CARS spectrum as obtained using transform-limited, double-notch shaped laser pulses. Although double-notch CARS with TL pulses enhances the signal significantly, it still necessitates lock-in detection [17]. This is usually obtained by dithering the notch filters and retrieving the changing signal by a lock-in amplifier, as seen in Fig. 5(b).

 

Fig. 5 Double-notch shaped excitation pulses and the corresponding CARS- and C-CARS spectra of toluene. (a) The measured spectrum of the double-notch shaped excitation pulses. The two notches are located at 759.5nm and 772.5nm, respectively. (b) Two slightly shifted CARS spectra excited by transform-limited pulses shaped with a double-notch. (c) The measured C-CARS output spectrum, for excitation light shaped with a single-notch (blue dashed curve) and a double-notch (red solid curve). The black dotted curve is the NR background obtained by measuring the spectrum without any notch. The spectra were detected by the spectrometer without any background reduction. Note the significant contrast enhancement in comparison with (b).

Download Full Size | PDF

To avoid lock-in detection, we add a chirp to perform double-notch C-CARS. The resulting signal is shown as the red solid line in Fig. 5(c), which also shows for comparison single-notch C-CARS (blue dashed line) and the NR background (black dotted line). We used a chirp value of 2.7, which is optimal for the 1003cm⁻¹ line. This chirp value is not optimal for the 791cm⁻¹ line, yet the combined feature, which appears around 719nm, shows significant enhancement. We note that the line around 707.5nm is also enhanced – this could be the effect of the relatively weak 1212cm⁻¹ line which shifts to combine with the 1003cm⁻¹ line. The fact that the three lines are nearly evenly spaced in this sample is of course fortuitous. The contrast around 719 nm has been improved by at least an order of magnitude. This high contrast enables direct detection of toluene without any background reduction, a key step for direct imaging applications. We note that the lines in Fig. 5(c) are significantly broader than regular CARS resolved lines. This is caused by the relative phase between the resonant CARS and the NR background due to the chirp. While a disadvantage for spectroscopy, this is much less important in imaging applications, which would benefit from the enhancement of signal-to-background.

4.3 Double-notch C-CARS microscopy

As an example of the use of double-notch C-CARS in multimodal imaging of biological tissues, we present a result of high-contrast multimodal double-notch C-CARS microscopy of ~100μm thick bone sample, shown in Fig. 6. The sample and its preparation were described previously [9]. Single-pulse multimodal CARS microscopy in bone has already been demonstrated but with lower contrast and relatively low imaging speed [9]. Bones are characterized by several Raman lines for phosphate (960cm⁻¹), Phenylalanine (1003cm⁻¹ C-C stretch), $P{O}_4^{3-}$ (1035cm⁻¹ overlaps with proline C-C stretch), Proteoglycan (1058cm⁻¹, overlaps with lipids, collagen), $C{O}_3^{2-}$ (1067cm⁻¹, overlaps with $P{O}_4^{3-}$), and carbonate (1080cm⁻¹) [20], which we can indeed observe using single-notch C-CARS spectroscopy, as shown in Fig. 6(a).

 

Fig. 6 CARS-resolved Raman spectrum and multimodal imaging of bone. (a) The single-notch C-CARS resolved Raman spectrum of bone tissue. (b) The double-notch C-CARS (combined signal of the 960cm⁻¹ and 1080cm⁻¹ lines) image of bone tissue. (c) Simultaneously obtained SHG image. (d) The overlaid image with both of CARS and SHG. The scanned area is 250 μm × 250 μm with 1 μm pixel size and 1ms pixel dwell time. The total imaging time is about 1 minute, and the laser power is <10mW. Scale bar: 50μm.

Download Full Size | PDF

Figure 6(b) shows the result of double-notch C-CARS imaging of the bone sample. The image was extracted directly from the CARS signal without lock-in or any other background reduction technique, using a 1ms pixel dwell time. Two notches were positioned such that their separation was 120cm⁻¹, corresponding to the difference between 960cm⁻¹ phosphate line and the 1080cm⁻¹ carbonate line which combine coherently, giving rise to the enhanced contrast. The SHG contrast arises mainly from the collagen fibrils around an osteon, as shown in Fig. 6(c). The dark structures in Figs. 6(b)-6(d) are attributed to cartilage lacunae which are cavities containing cartilage cells and surrounded by lamellae of calcified matrix.

5. Summary

In summary, we have demonstrated contrast enhancement in C-CARS spectra of different liquids and a biological sample. Optimal contrast was obtained by adding small amounts of dispersion to the excitation pulses. Combining coherent addition with C-CARS by adding a second notch results in a significant contrast enhancement, compared with single-notch C-CARS. Chirped notch CARS can be used to generate high-contrast CARS images with a simple, compact, lock-in-free setup, and is attractive for imaging applications that require fast signal acquisition.