1. Introduction

Narrowband picosecond (ps) pulses are crucial for coherent vibrational techniques such as coherent anti-Stokes Raman (CARS) [1,2], sum frequency generation (SFG) [3,4], and femtosecond stimulated Raman spectroscopy (FSRS) [5,6]. In these spectroscopies, the narrowband ps pulse provides a well-defined virtual state energy resulting in well resolved vibrational spectra. Vibrational spectroscopy is particularly useful for studying chemical reaction dynamics as it probes reactive nuclear evolution with structural precision. In this endeavor, FSRS has three main advantages: simultaneous high time and frequency resolution [7], effective rejection of fluorescence interference, and simple implementation [5]. Thus far, FSRS has been used to elucidate ultrafast structural dynamics in a number of diverse systems including photoswitches [8,9], charge-transfer systems [10,11] and photoactivated proteins [12–14]. However, the convenient production of high power tunable ps pulses necessary to provide the Raman pump pulse in FSRS has been a consistent challenge.

In order to have synchronized broadband femtosecond and narrowband ps pulses, a portion of a femtosecond laser source is typically split and filtered. Common filtering techniques include grating filters and narrow band interference (NBI) filters [5]. However, each of these methods has intrinsic disadvantages: grating filters are inherently inefficient because, as illustrated in Fig. 1, the beam must interact twice with a ruled diffraction grating. Additionally, the grating filter can adversely affect the laser beam’s mode quality when the slit width is smaller than the input beam diameter. NBI filters are not broadly tunable, have large bandwidths and are susceptible to laser damage.

 

Fig. 1 Schematics of the grating filter (top) and the etalon pulse shaper (bottom). Input and output pulse shapes are shown for both filters. The etalon device is optically simpler, more efficient and it provides improved beam quality compared to the grating filter.

Download Full Size | PDF

Fabry-Pérot etalons offer an attractive alternative filtering method for the generation of narrowband pulses and have been applied in SFG [3,4] and double resonance 2D-IR [15,16] spectroscopy. Etalons have not been previously used in FSRS but offer potential advantages; they have higher energy throughput than grating filters and narrower bandwidths than NBI filters while retaining tunability and durability. For example, the etalon described herein has a bandwidth of 2.8 cm⁻¹, an efficiency of 0.6% and a tuning range of 776-814 nm. Etalon filters also offer instrumental advantages; alignment consists of simply placing the optic in the beam path, the center wavelength of the etalon can be tuned by changing the angle of incidence [17] and the beam’s mode quality is preserved regardless of its diameter. However, etalons are more expensive than the alternatives; therefore a critical examination of their utility is needed.

An additional, important, difference between etalon and grating or NBI filters is the resulting time domain pulse shape. Both grating and NBI filters create pulses that are symmetric in the time domain. The exact functional form depends on the nature of the individual filter; for narrow bandwidths, the resulting pulse is generally Gaussian. When the bandwidth of a Gaussian pulse is not sufficiently narrow, the resulting FSRS spectra are significantly broadened and the line-shapes are altered as well. In extreme cases the vibrational lines will take on “ringing” side wings [18,19]. On the other hand, etalons generate time asymmetric pulses (see Fig. 1), a property that has been successfully used in SFG spectroscopy to suppress the non-vibrationally resonant background [3]. These pulses have a steep rise and an exponentially decaying tail meaning that while vibrational line-shapes may be broadened (the amount of which is determined by the bandwidth of the etalon) the vibrational lines will remain Lorentzian. It is also anticipated that the asymmetric etalon pulse profile will be advantageous for FSRS because it can reduce the electronically resonant non-linear background.

With these motivations, we investigated the use of etalon filters to produce the ps Raman pump pulse in FSRS. We accurately characterize the effect of the time delay between Raman pump and probe on signal characteristics for both the grating and etalon filters. A theoretical framework for modeling the etalon pulse and its role in the FSRS process, which quantitatively explains our results, is presented. Finally, we demonstrate the improvement in FSRS baselines by comparing the excited state spectra of flavin adenine dinucleotide (FAD) when probed with the grating pulse versus the etalon pulse.

2. Materials and methods

2.1 Femtosecond stimulated Raman

The laser system has been described in detail previously [5]. It consists of a home built Kerr lens mode-locked Ti:sapphire oscillator (30 fs 5.3 nJ/pulse, 91 MHz) that seeds a Ti:sapphire regenerative amplifier (B.M. Industries, Alpha 1000 US, 991 Hz, 70 fs, 0.91 mJ/pulse, λ_max = 790 nm) pumped by a Q-switched Nd:YLF (B.M. Industries, 621-D). A small portion of the fundamental is focused onto a 3 mm sapphire plate (ThorLabs) generating a broadband continuum which serves as the Stokes probe pulse in the FSRS process [20]. The probe pulse (8 fs, 5 nJ/pulse, λ_max = 883 nm) is subsequently compressed in the 820-945 nm region with a pair of BK7 equilateral dispersing prisms (CVI Melles Griot) [21]. Approximately 200 µJ of the fundamental is used to pump a homebuilt noncolinear optical parametric amplifier (NOPA) [22] which generates the actinic pump pulse (30 fs, 150 nJ/pulse, λ_max = 475 nm). An F2 prism pair (ThorLabs) is used to compress the actinic pump.

The remaining ~690 µJ of the fundamental is used to generate the narrow bandwidth Raman pump. Traditionally, this has been done with a grating filter [23] consisting of a ruled diffraction grating blazed at 26.7° with 1200 groves/mm, a 200 mm fl cylindrical lens, an adjustable slit and a flat mirror. The bandwidth of the grating filter can be tuned by the slit width; however, the power is proportional to the bandwidth. When the bandwidth is set to ~2.8 cm⁻¹ the maximum output pulse energy is ~0.7 µJ. Alternatively, we show here that a Fabry-Pérot etalon (TecOptics, Design #A6) can be used to produce the ps Raman pump pulse. This etalon has been designed with a reflectivity of 98.5 ± 0.5% and a spacing of 18.12 µm giving a free spectral range (FSR) of ~290 cm⁻¹ and a finesse of ~100 resulting in a spectral bandwidth of ~2.8 cm⁻¹. The maximum output pulse energy for the etalon filter is ~2.3 µJ for the same input power. Input powers were measured immediately before the filters, however, the output powers were measured right before the sample. The measured absolute efficiency of the etalon is 0.6% which corresponds to a peak transmission of 44%. For different designs a sacrifice of some bandwidth will increase peak efficiency to ~85%.

All three beams are polarized parallel to the table and are spatially and temporally overlapped in the sample cell after being focused by a 100 mm fl achromatic lens. The instrument response function is measured in the sample cell in the relevant an aqueous buffer using the optical Kerr effect. After the sample, the probe beam is recollimated and directed to a spectrograph (Instruments SA, HR320) which disperses the beam onto a fast CCD (Princeton Instruments, PIXIS 100F). A phase locked chopper (Newport, 3501) blocks every other Raman pump pulse allowing a full Raman spectrum, calculated as ln(Probe_{Raman Pump On}/Probe_{Raman Pump off}) to be collected for every two laser pulses. The delay between the actinic pump and the Raman pump/probe pulse pair is varied by a computer controlled delay stage (Nanomotion II, Melles Griot) while the delay between the Raman pump and probe is controlled with a manual micrometer. Initial data processing and instrument control is performed by a custom LabVIEW program [24].

2.2 Sample preparation

Spectral grade benzene (EMD, >99.7%) was used as received. Measurements were performed in a 1 mm path length cell with 1.2 mm glass windows (Starna Cells, 21-G-1). A Raman pump energy of 0.7 µJ/pulse was used. The Raman shift axis was calibrated using benzene.

FAD (TCI America, >94%) was used as received. FAD was dissolved in a 10 mM Tris, 50 mM NaCl buffer at pH = 8.0 forming a 2.2 mM solution which gives an optical density of 1 per 500 µm at 475 nm. The solution is flowed through a 500 µm path length cell with 200 µm quartz windows (Starna Cells, 48-Q-1-UTWA) from a 40 mL reservoir using a peristaltic pump at a rate sufficient to replenish the sample volume between consecutive laser pulses. A Raman pump energy of 1.5 µJ/pulse was used and the Raman shift was calibrated using cyclohexane.

Results

Figure 2(a) presents contour plots of the stimulated Raman signal of the 992 cm⁻¹ mode of benzene as a function of Raman pump and probe delay for the grating filter (top, dashed) and the etalon filter (bottom, solid). The center traces show the signal amplitude as a function of delay, which mirrors the temporal profile of the pulses. In the case of the grating filter, the vibrational band is initially narrow and weak and then grows and broadens as the delay is increased. Increasing the delay further causes the line-shape to continue to broaden while the intensity decreases. Conversely, the signal resulting from the etalon filter increases sharply in intensity near a delay of zero and then decays exponentially; the width stays remarkably constant. Note that even though both pulses are ~0.7 µJ the maximum signal intensity for the etalon is nearly twice that for the grating filter.

 

Fig. 2 (a) Comparison of the effect of the time delay between the Raman pump and probe on the stimulated Raman signal of the 992 cm⁻¹ mode of benzene. Contour plots of the stimulated Raman spectra are shown for the grating (top, dashed) and for the etalon (bottom, solid). The intensities of the 992 cm⁻¹ peak are shown in the middle. The asymmetry in the grating filter data is due to the presence of a slight baseline. (b) Simulations of the data in (a) using the theory outlined in the text. Important model parameters are: vibrational FID, 2.2 ps; etalon reflectivity, 98.75%; etalon separation, 18.12 μm; FWHM of the electric field of the grating pump, 7 ps.

Download Full Size | PDF

Presented in Fig. 2(b) are simulations of the signals shown in Fig. 2(a) following the theory presented in refs. 20 and 23. To numerically generate the result of the etalon filter we explicitly simulated a transform limited pulse interacting with two highly reflective surfaces using Eq. (1) [3,4].

In order to evaluate the expected improvement in excited state baselines, transient stimulated Raman spectra of FAD were measured. The transient absorption of FAD is known to be strongly affected by the Raman pump when it is tuned to 795 nm [25] making FAD a challenging model system for this investigation. Figure 3 presents the transient stimulated Raman signal of FAD at 7.5 ps delay after excitation at 475 nm using the grating filter (dashed line) and the etalon filter (solid line). These results agree qualitatively with those of Weigel et al. [25] The etalon signal has been scaled by 0.45 so that the Raman features are the same magnitude, emphasizing the reduced baseline. Relative to total pump energy, the signal from the etalon is more than twice as intense as that from the grating filter, in agreement with the results presented in Fig. 2.

 

Fig. 3 A comparison of the excited state Raman spectra of FAD at 7.5 ps delay taken with the grating (dashed) and the etalon (solid). The etalon spectrum has been scaled to that of the grating using the intensities of the FAD peaks in order to emphasize reduced baseline interference with the etalon.

Download Full Size | PDF

Discussion

Optimally shaped narrow band ps pulses can be used to improve femtosecond stimulated Raman spectra. Proper tailoring of the Raman pump pulse results in a two-fold enhancement of the signal and a concomitant reduction in the background. Besides the intrinsic improvement in data quality, etalons are also much more simple to align and tune than grating filters while offering much narrower bandwidths and improved durability compared to NBI filters. Unlike grating filters, etalon filters maintain the laser beam’s mode quality regardless of the beam diameter and have higher energy throughput.

Shaping the ps pulse to be asymmetric in time allows for nearly double the signal intensity per unit pulse energy, as shown in Fig. 2. For asymmetric pulses, almost all of the pump photons contribute to the signal because they arrive at the sample at the same time or later than the stimulating probe pulse. On the other hand, nearly half of the pulse photons in symmetric pulses, such as those produced by grating or NBI filters, ineffectively arrive at the sample before the probe. Furthermore, the exponential shape of the etalon pulse mirrors the natural shape of the vibrational free-induction decay meaning that the resulting line shapes for homogeneously broadened vibrations remain Lorentzian and the width does not change with the time delay.

Another significant improvement, illustrated by the FAD data in Fig. 3, is the reduction of excited state baselines. Because FSRS is a heterodyne technique it is susceptible to both resonant and non-resonant backgrounds. Properly removing baselines from FSRS data can be a challenging part of the data analysis. Although, some innovative experimental approaches have been developed to minimize baselines [8], theses approaches are technically complex and ill-suited for some systems. The largest contribution to the background is the electronically resonant interaction of the Raman pump with the sample. For example, near a ground state resonance the Raman pump can create an excited state population. Alternatively, near an excited resonance the interaction can lead to a non-negligible dumping or promotion of the excited state population. These scenarios commonly result in a different absorption of the probe when the Raman pump is coincident. For a majority of systems the excited state absorption and stimulated emission cover the entire visible wavelength region and this differential absorption leads to broad featureless baselines in the Raman spectra. This effect can result in particularly complicated baselines when the Raman pump transfers a portion of the population to another excited state. Significant perturbation of the excited state population can also lead to unreliable kinetics [25]. Because these effects depend on the Raman pump they are directly correlated with the signal and therefore cannot be removed by conventional methods, such as signal averaging or increasing the excitation power. The time asymmetric pulse generated by the etalon minimizes the interaction of the Raman pump with the sample before the actual probing event thereby mitigating the perturbation of the system and reducing these unwanted background signals.

Finally, etalon filters can be incorporated into broadly tunable systems as long as the etalon’s dielectric coating is sufficiently broadband and the FSR is large enough. Commercial amplifiers coupled with state of the art NOPAs provide sufficient power to produce multi-µJ tunable narrow band pulses with an etalon filter. The technical simplicity and improved power throughput compared with grating and NBI filters should reduce the barriers to implementing FSRS. While etalons are more expensive than conventional filters, the cost is relatively minor when compared to the femtosecond light source and the simplicity is advantageous for both new and experienced stimulated Raman spectroscopist.