Open Access
Add to BibSonomy
Abstract
In this work, we present the development of a femtosecond tunable middle infrared (mid-IR) radiation source for the realization of a hybrid concept compact broadband high-resolution sum-frequency generation (SFG) spectroscopy system. For the realization of the new concept, we used a two-channel picosecond fiber laser as a seed for narrowband (∼1.5 cm-1) and broadband ultrafast radiation sources operating at 1 kHz repetition rate. In order to achieve >500 cm-1 bandwidth widely tunable microjoule level pulses in the mid-IR spectral region (2–10 µm), broadband femtosecond source optimization was performed. Numerical simulations with different nonlinear crystals and optical parametric amplification layouts at given fixed initial conditions paved a way to experimentally realize an optimal scheme for a femtosecond mid-IR channel. Fully operating SFG spectrometer setup was assembled and tested. The developed SFG spectrometer demonstrates a unique combination of parameters: excellent spectral resolution (<3 cm-1) similar to a narrowband scanning picosecond spectrometers and fast simultaneous acquisition of broadband spectra up to more than 850 cm-1.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Sum-frequency generation spectroscopy is a powerful and non-destructive characterization method for surfaces and interfaces [1–4]. The process is based on a resonant nonlinear mixing of tunable infrared (ωIR) and fixed/constant visible (ωVIS) laser radiation in order to produce an output with the sum frequency (ωSFG = ωIR + ωVIS) [5]. Active vibrational modes of the examined molecules at the interface give a resonant contribution to the sum frequency signal. This technology is widely used for many applications [6] such as studies of molecular orientation, surface structure, chemical composition, electrochemistry and epitaxial growth, investigation of interfaces and surfaces of liquids, solids, polymers, biological membranes and other systems. The SFG signal is generated in the visible spectral range so that it can be measured efficiently with sensitive detectors (photomultiplier tube (PMT) or CCD). Nanosecond, picosecond and femtosecond lasers can be used for experimental SFG setups. Fourier-transform-limited pulses with a longer duration allow to achieve better spectral resolution, but very short pulses improve generation efficiency, signal-to-noise ratio and data acquisition speed.
Three types of SFG systems are widely used for research: narrowband, broadband and hybrid. In narrowband SFG systems mid-IR and visible (VIS) laser sources with picosecond pulse durations are used. That provides excellent spectral resolution down to few wavenumbers [7,8]. These are already commercially available systems and extensive research has been done using them. Measurements using these systems are based on wavelength scanning, which gives good spectral resolution and signal-to-noise ratio, but could be a source of systematic errors due to the sample degradation in time [7]. Narrowband SFG requires high energy of picosecond pulses for sufficient SFG signal generation, therefore they can significantly heat up the sample, and some samples are modified or completely damaged until the end of the measurement. Typically, 30 ps pulse duration radiation, IR pulse energy of 0.3 mJ, and 532 nm pulses of energy from a few µJ up to 0.5 mJ are used in that kind of spectrometers [9,10]. Broadband SFG systems use femtosecond laser sources, which allow simultaneous recording of the complete spectrum without laser wavelength scanning and consequently speed up spectroscopic data acquisition. Due to the shorter pulse duration (higher intensity) it requires less energy comparing to narrowband systems, therefore samples are less heated and possible systematic errors are eliminated, nevertheless spectral resolution is limited to ∼10-20 cm-1 [7]. To overcome these limitations spectral narrowing techniques are used to increase resolution down to ∼2-4 cm-1 [11,12]. However, these methods are usually very ineffective and can cause significant energy losses. A slightly different solution to increase resolution was proposed in [13] where Ti:sapphire oscillator seeded femtosecond and picosecond regenerative amplifiers (RA). In this case, the spectrum of ∼1 ps pulses was narrowed with etalon down to the bandwidth of 0.9 cm-1. This approach allowed to reach the SFG resolution down to 1.4 cm-1 although due to only ∼5% transmittance of the etalon the overall efficiency of this system was very low. The nonlinear spectral narrowing technique based on sum-frequency generation with pulses of equal wavelength but opposite chirp showed promising results in [14] regarding the optimization of energy losses, however the setup of this approach is rather complex and not very simple to implement. Hybrid SFG systems with femtosecond and narrowband picosecond laser sources can be used as an alternative [15]. An improved spectral resolution of 0.6 cm-1 was demonstrated using two separate picosecond and femtosecond lasers [16]. The authors indicated and demonstrated, that high spectral resolution and high signal-to-noise ratio are essential for accurate lineshape registration of the SFG vibrational response. Resonant response often contains signatures of fine molecular interactions and heterogeneity which are in order of only a few wavenumbers [17,18]. However, this approach required precise electronic synchronization between the two lasers. Due to the complexity and cost systems of such design are not widely used.
In this work we present a novel concept of compact broadband high resolution all optically synchronized SFG spectrometer, which retains all advantages of the standard hybrid SFG system and overcomes the above-described shortcomings. The system was based on a multiple-channel picosecond fiber laser [19,20] which served as a seed for narrowband (∼1.5 cm-1) picosecond and broadband femtosecond channels. No spectral narrowing was required for picosecond channel because the linewidth was determined by an active medium used in the RA.
In order to develop a hybrid SFG system, first of all, we focused our efforts on a broadband femtosecond mid-IR spectral range channel optimization. As there are several ways to realize this scheme, we started with the modeling of different mid-IR wavelength generation and amplification layouts looking for the most efficient and the most broadband. The broad-bandwidth SFG method was first presented by L. J. Richter et al. in [21] where the term “broad-band bandwidth SFG” referred to spectra obtained over a bandwidth of >200 cm-1. Though it is frequently enough to cover the entire C-H stretch vibrational spectrum, our goal was to cover up to 500 cm-1 range in order to be able to measure the whole SFG spectrum of O-H stretch in water. The aim was to achieve widely tunable (from 2 to 10 µm) microjoule level pulses (>2 µJ) which are required for efficient SFG spectrometer operation. An optimal solution based on modeling results of the broadband mid-IR channel was experimentally realized. Spectrometer resolution and performance were tested using several common specimens in mid-IR vibrational spectroscopy.
2. Numeric simulations of mid-IR stage
The SFG spectrometer design requires to have two laser channels [2]: one narrowband in the visible spectral range, which determines the SFG system spectral resolution, another tunable and/or broadband, covering the spectral range of molecular vibrational oscillations, i.e., middle infrared spectral range laying in 2 –20 µm. The simulations aimed to choose the best suiting scheme for mid-IR channel realization in the broadband high-resolution all optically synchronized SFG system. The next aspect of the task was to select optimal nonlinear crystal along with its length and amplifier parameters to achieve the desired output radiation parameters. The optical characteristics of the pump laser were chosen according to the real experimental values: 250 µJ pulse energy at 1064 nm central wavelength, 400 fs pulse duration. The whole pump laser setup will be described later in the experimental results section.
In order to generate widely tunable mid-IR radiation broadband or widely tunable seed is required. One of the very common methods to produce it is via SC radiation generation. OPA stage seeded with this broadband signal produces idler wave in the desired mid-IR wavelength region [22]. We started with the same approach, however, we were restricted by the pump power and challenging desired output parameters. Nonlinear crystals suitable for broadband parametric amplification in mid-IR wavelength range were selected for simulations: Gallium selenide – GaSe (GS), Silver Thiogallate - AgGaS2 (AGS), Lithium Gallium Sulfide - LiGaS2 (LGS), Lithium gallium selenide - LiGaSe2 (LGSE). Parameters of the investigated crystals at different idler wavelengths are presented in Table 1 where deff - effective nonlinear coefficient, Δν - full width at half maximum (FWHM) phase-matching bandwidth (calculated according to Eq. (24) in [23]), LIDT – laser-induced damage threshold.
Table 1. List of crystals used for the OPA stage modeling and their parameters.
Our model was based on three-wave resonant interaction equations. We used a full 3D grid in time and lateral dimensions of the beam x and y, meanwhile z being the propagation direction. We were able to define beam parameters (wavelength, bandwidth, diameter, M2, ellipticity, shape), pulse duration, shape and dispersion. In multiple OPA stages case, the output of previous stage was used as input for the next stage.
To determine single or multiple OPA stage is optimal, simulations with 1 nJ and 1 µJ seed energies were performed for a set of nonlinear crystals. The pump intensities were chosen according to the LIDT values for different crystals. The simulation strategy was to optimize the crystal length so that the bandwidth of the amplified idler wave would be larger than 500 cm-1 along with the best conversion efficiency. Simulation results for both seeds energies 1 nJ and 1 µJ are presented in Table 2 for comparison. Results at 1 µJ seed energy manifested that it is required minimum to achieve desired conversion efficiencies in the second nonlinear amplification stage.
Table 2. Comparison of the OPA stage pump conversion efficiencies to the idler wave for different nonlinear crystals at 4 µm and 2.5 µm idler wavelength with the 1 nJ and 1 µJ seed energy.
Despite that GS crystal performance was the best at the low seed case, we have to mention that it has some major drawbacks. One cannot cut it at any desired angle, and only the z-cut option is possible. This drawback results in the need for a large aperture crystal due to high phase-matching angle for the 1064 nm pump. Another important drawback is that no AR coating is available for this crystal. Those issues and a quite large phase-matching angle in the crystal lead to tremendous reflection losses. For the cases s(o)+i(e) = p(e) and s(e)+i(o) = p(e) interactions losses approach to 100% and for the s(o)+i(o) = p(e) case losses exceed 80% at 2.5 µm and 30% at 10 µm wavelengths therefore only the last one was analyzed. Here we used the following notations; s-signal; i-idler; p-pump; o-ordinary wave; e-extraordinary wave.
To understand the tendency of conversion efficiency dependence on seed pulse energy simulations at various signal energy levels from 1 nJ to 20 µJ were performed for AGS crystal at 2.5 µm idler wavelength. A crystal length of 0.36 mm, pump pulse duration of 400 fs and pump intensity of 100 GW/cm2 were used in simulations. Two duration cases - transform-limited pulses of 29 fs duration and chirped with 1400 fs2 additional dispersion to 206 fs duration, were compared (Fig. 1(a)). The efficiency of conversion increased rapidly when the signal energy was increased from 1 nJ to few µJ, but saturated at higher seed pulse energies. The saturation level can be increased by applying additional dispersion to the signal pulse. As presented in Fig. 1(a) case ∼7 times increase of duration lead to ∼ 2.5 times increase of the saturation level.
Fig. 1. a) Conversion efficiency dependence on the seed energy level of the AGS OPA stage at 4 µm idler wavelength. b) Conversion efficiency for different idler wavelengths of the BBO OPA stage. c) Conversion efficiency for the different seed wavelength of the AGS OPA stage.
Since the SC energy is at a level of few nJ and additional dispersion influence is noticeable only at energy levels of >0.5 µJ, the best expected conversion efficiency of the first OPA stage could be ∼0.16% at 2.5 µm and ∼0.045% at 4 µm only. In the case of two nonlinear amplification stages pump energy has to be distributed between stages. As the pump laser energy is limited the first stage efficiency is too low to achieve 1 µJ output energy even under the whole pump. Without sufficient seed energy the second OPA stage will be inefficient, therefore more effective first stage is required.
As an alternative to the 1064 nm pumped first OPA stage, the second harmonics (SH) pumped BBO OPA was analyzed theoretically too. The BBO crystal has not only a good transmission in the 1190-1850nm spectral range but also possesses a large nonlinear coefficient, broad amplification bandwidth and finally, significantly higher LIDT [27,28] than all infrared crystals of interest. A similar scheme was implemented by M. Bradler et al. in [29] as the first stage in LiNbO3-based broadband mid-IR OPA.
In the model, the first BBO OPA stage was seeded with the same 1 nJ seed energy pulse with the wavelength in the spectral range from 746 nm to 962 nm. Corresponding idler wavelengths were tuned from 1.19 µm to 1.85 µm. BBO crystal has a higher damage threshold therefore 200 GW/cm2 pump intensity was set. The optimized crystal length was set to 2 mm. The final results of the simulation are presented in Fig. 1(b). Conversion efficiency was calculated relative to the 1064 nm pump pulse energy therefore 66% SH generation efficiency was taken into account. For comparison of schemes, simulations using AGS crystal were performed in the same wavelength range (Fig. 1(c)). In that case, the SC wavelengths from 1.19 µm to 1.85 µm were amplified. Pump intensity was set to 100 GW/cm2 and the crystal length was set to 0.36 mm which ensured broadband amplification. The results manifested, that conversion efficiency to the idler in the BBO crystal OPA nearly 3 orders of magnitude higher comparing to conversion efficiency to the signal in the AGS crystal. Higher conversion efficiency in the BBO crystal OPA allows to increase the efficiency of the second OPA stage and achieve ∼1 µJ energy seed pulses using only the fraction of the pump laser pulse energy.
A slightly non-collinear configuration for BBO OPA is more suitable due to the easier seed and idler beams separation. The drawback of the non-collinear configuration is the angular dispersion of the idler wave as it is used as a seed for the subsequent OPA stage. In order to avoid the angular dispersion, we decided to use a seed of longer than 1 µm wavelength instead of shorter one, in this way whole angular dispersion was transferred to the visible spectrum meanwhile the infrared spectrum was not impacted [30]. Simulations in this configuration manifested comparable conversion efficiencies.
The final setup of mid-IR OPA is presented in Fig. 2. To get more realistic simulation results the first OPA stage amplification was remodeled with higher input energy - 5 nJ seed instead of 1 nJ. Optimization of seed and pump beam diameters along with temporal overlap of the pump and seed pulses were performed. Calculated conversion efficiencies of the 1064 nm pump pulse energy to idler pulses reached 4.5-5.4% for idler wavelengths in the 1852-1190 nm spectral range respectively. Radiation with parameters achieved after the first OPA stage was used as the input signal to the second OPA stage. The pump pulse energy was split to 10 µJ reserved for SC generation, 60 µJ for the first OPA stage and the rest of the pump energy for the second stage. Calculated signal energy after the first stage was 2.7-3.27 µJ in the 1852-1190 nm spectral range. The simulations and optimization of the second stage OPA were performed using several nonlinear crystals. During simulations the optimization of crystal length, the delay between pump and seed and seed dispersion was performed. The best results for several crystals are presented in Fig. 3. For the modeled crystals the lowest conversion in the second OPA stage was 0.6-2% in longer wavelength range at which increased up to 2.7-5.5% at shorter wavelength side. Calculated idler energy after the second stage reached 1-3.4 µJ at 10 µm wavelength and up to 4.5-9.3 µJ at 2.5 µm idler wavelength. The idler pulse bandwidth at the 30% level of maximum was broader than 500 cm-1 in almost full mid-IR tuning range, for all investigated nonlinear crystals. Though intuitively it looks that direct conversion of IR pump pulses to the mid-IR spectral range can be more efficient, modelling revealed, that generation of seed using visible light pumped OPA and subsequent mid-IR radiation amplification using IR pump is much more efficient. The reason is the superior efficiency of the BBO crystal-based OPA stage produced high-energy seed for the mid-IR nonlinear conversion stage. The modeling allowed us to select the layout which enabled us to generate enough mid-IR power for building an SFG spectrometer.
Fig. 3. Comparison of mid-IR OPA stage pump conversion efficiencies to idler and idler bandwidth at the 30% level with different nonlinear crystals for 2.5-10 µm idler wavelength.
3. Measurements and experimental results
Simulation results were taken as a guideline for building the experimental setup therefore, two-stage OPA for mid-IR radiation amplification was implemented in the SFG spectrometer scheme.
The final setup of the SFG spectrometer is presented in Fig. 4. Our hybrid approach of the system front-end part is described in detail in [20]. All-in-fiber 45 MHz repetition rate SESAM mode-locked oscillator operating at 1064 nm wavelength had 3 optically synchronized outputs. One of them was used to seed a picosecond Nd:YVO4 RA, meanwhile another one was a fusion spliced to fiber pigtailed InGaAs photodetector from which the RA controls were synchronized. The 1064 nm wavelength pulses in RA were amplified up to 6.1 mJ at a 1 kHz repetition rate, of which 0.1 mJ was used as narrowband SFG channel seed and 6 mJ as a broadband channel pump. The pump pulse duration after RA was 100 ps. The third output provided seed to the Yb-doped fiber amplifier and pulse stretcher module. After the amplification up to a few nJ energy level, the pulse spectrum was broadened up to 14 nm via self-phase modulation in single-mode fiber and then pulses were temporally stretched up to 115 ps using a chirped fiber Bragg grating. These broadband pulses subsequently were amplified in the picosecond BBO OPA.
3.1 Broadband SFG spectrometer channel
Temporally and spectrally stretched broadband 1064 nm wavelength pulses were coupled out of the fiber via free-space collimator and amplified in two picosecond OPA stages up to 560 µJ. The output of Nd:YVO4 RA was frequency doubled in 9 mm LBO crystal with ∼67% conversion efficiency and was used as a pump of the ps BBO OPA in the broadband channel front-end. Type I phase-matched BBO crystals of 9 mm and 4 mm length were used for the first and second ps OPA stage respectively. The ratio of the pump energy for each stage was optimized in a way to achieve the best overall conversion efficiency with minimal spatial beam distortions. The optimal pump energy distribution ratio in our system was 1 mJ for the first stage and 3 mJ for the second one what resulted in an overall pump to signal conversion efficiency of 14%.
After the cascade of ps OPA’s amplified broadband pulses were compressed by the transmission diffraction grating compressor using 1600 groves/mm gratings operating at Littrow configuration. The autocorrelation trace of the compressed pulse was measured by the SHG autocorrelator (APE pulseCheck) (Fig. 5(a)). The measured pulse duration was equal to 367 fs according to the Gaussian fit with the contrast of ∼96-97%. The efficiency of the compressor was equal to 75%, which resulted in the final pulse energy at the output of the broadband front-end around 420 µJ.
Fig. 5. a) The measured autocorrelation trace of the compressed femtosecond front-end output pulses (black) and Gaussian fit (red). b) The supercontinuum spectrum produced in a 20 mm YAG crystal. The pump radiation was filtered using the long-pass filter (1100 nm).
A small fraction of the compressed femtosecond pulse was focused into the 20 mm length YAG crystal in order to generate the SC radiation. Focusing conditions and pump energy (∼1.6 µJ) were optimized to produce smooth, stable and sufficiently broad especially in the longer wavelengths SC wing (>2 µm) spectrum (Fig. 5(b)).
The remainder was used as a pump for the femtosecond noncollinear OPA (fs NOPA). The pump pulse was split into three parts. The first two of 50 µJ and 120 µJ pulse energies were frequency doubled in 1.5 mm BBO crystals with conversion efficiencies of 56% and 63% respectively and then used as a pump for two stages of fs NOPA where 1150-2200 nm band of supercontinuum was amplified. Noncollinear geometry configuration allowed us to separate idler wave without using broadband dichroic mirrors. The Poynting-vector walk-off compensation (PVWC) geometry was used in both stages instead of tangential phase-matching (TPM) to reduce parasitic signals. For example, the BBO phase-matching angle for SH generation from 1350 nm wave is ∼20.4°, which is very close to the phase-matching angle of the fs NOPA (∼20.66°) pumped by 532 nm at ∼2.5° noncollinearity angle, meanwhile, in PVWC geometry this angle differs much more being around 25.66°. Due to the similar phase-matching angles of SH and NOPA process, the larger part of the signal energy was converted to SH as well as to other wavelengths via four-wave mixing at TPM geometry (Fig. 6). To enhance the signal and pump temporal overlap ratio and achieve better conversion efficiencies the dispersion of signal pulses was controlled using a plate of ZnSe placed between the stages. Tuning the phase-matching angles of BBO crystals and the delay between signal and pump SC radiation was amplified up to ∼10 µJ at 1150 nm wavelength and up to ∼2.5 µJ at 2200 nm (Fig. 7).
Fig. 6. Illustration of multiparametric interactions in the OPA output beams on a white screen after the BBO crystal at TPM and PVWC amplification geometries.
The last part of the remaining pump (∼200 µJ) radiation at 1064 nm wavelength was directed to the mid-IR fs OPA stage collinearly with the broadband signal amplified in fs NOPA. Signal radiation in the 1.15-2.2 µm wavelength range was used to generate 2-13 µm idler wave in AGS, LGS and GS crystals. Ag coated flat and spherical mirrors were used to guide and shape the signal beam therefore around 30% signal energy was lost in the optical path to the last mid-IR fs OPA stage. A comparison of the results achieved with different crystals in mid-IR fs OPA is presented in Fig. 8(a). Pulse spectra were characterized using a scanning monochromator and HgCdTe (MCT) detector. We were able to measure up to 13 µm of idler wavelength after mid-IR fs OPA using 0.5 mm length GS crystal, however, tuning range was limited to ∼5 µm due to large phase-matching angles and limited aperture of used crystal. The energy of the idler wave decreased from ∼0.5 µJ at 5 µm down to ∼0.1 µJ at 13 µm meanwhile, the spectral bandwidth changed from ∼222 cm-1 down to ∼120 cm-1 respectively. The highest idler wave energy of 0.4-1.8 µJ in the 2-5 µm spectral range was achieved with AGS crystal of 0.4 mm length. However, at longer wavelengths (5-10.5 µm) it produced considerably lower energies of 0.1-0.25 µJ. The broadest spectra were achieved with LGS crystal of 1 mm length. The tuning range of the idler wave with this crystal was 2-9.6 µm, pulse energies of 0.1-1.35 µJ and bandwidths of 205-550 cm-1 (Fig. 8(b)).
Fig. 8. a) The comparison of AGS, LGS and GS mid-IR fs OPA idler wave parameters; the output spectral bandwidth at a 30% level of maximum and pulse energies in the wavelength tuning range. b) The LGS mid-IR fs OPA idler wave output spectra and pulse energies in the wavelength tuning range of 2-9.6 µm.
Experimental conversion efficiencies differed from the simulated. It could be due to a few reasons. One of them was the difference in the pump intensity. We used the ∼100 GW/cm2 value in the simulations, however, during the check of the experimental setup we noticed some light-induced modifications of the AGS crystal therefore the pump intensity was reduced down to ∼60 GW/cm2. Other reasons were the two-photon absorption and reflection losses. AGS and LGS crystals had reflection loses of ∼5-10%, meanwhile GS ∼30-35%. Moreover, AGS and GS crystals manifested the transmission dependency on the pump intensity: at ∼60 GW/cm2 total transmission of the pump in the AGS crystal decreased down to ∼85% and for the GS crystal down to 50%. No decrease in transmission was observed in the LGS crystal only. The third reason might have been due to a different beam quality and profile used in simulation and experimental setup. The pump beam in experimental setup was not perfectly Gaussian. Finally, the Sellmeier equations used in simulations might have been inaccurate for the experimentally used crystals.
To increase idler energy 1 mm length AGS crystal was tested. With this crystal, the measured idler energy was from 2 to 4 times higher in 2-5 µm wavelength range (2.4-5.1 µJ) and nearly 2 times higher in 5-10 µm range (0.2-1 µJ). Although the higher energy was reached with the longer crystal, the spectral bandwidth of the pulses was inevitably narrower- 200-300 cm-1 over the tuning range. To increase the bandwidth of the measurement spectrum a fast crystal scanning method was tested in the SFG spectrometer setup. For this purpose, the AGS crystal was mounted on a motorized rotating stage and scanned around the set wavelength position in a range of a few degrees. Since the signal bandwidth was quite broad the crystal angle scanning ensured phase-matching for a whole spectrum. By integrating the SFG signal of many idler pulses over the whole scanning range broad effective spectrum was registered. Due to signal to pump durations ratio, not all of the signal wavelengths were amplified even in applying scanning technique, which caused that the measured idler bandwidth was narrower than a bandwidth of the seed. A comparison of idler wave spectral bandwidths achieved with static AGS crystal of 0.4 mm length and static and scanned 1 mm length AGS crystal is presented in Fig. 9. In the wavelength range of 5-10 µm bandwidth of the amplified idler wave by the scanned AGS crystal of 1 mm length was similar or even slightly broader comparing with the 0.4 mm length crystal case and at 2.5-5 µm spectral bandwidths were up to 500-750 cm-1.
Fig. 9. Comparison of the idler pulse bandwidths at a 30% level of maximum for the static case using the AGS crystal of 0.4 mm length, static and scanned cases using 1 mm length AGS crystal over the tuning range.
After amplification, the mid-IR fs OPA idler wave was compressed using a germanium plate of experimentally selected thickness and directed to the SFG spectrometer.
3.2 Narrowband SFG spectrometer channel
A small part of chirped 100 ps pulses from the regenerative amplifier (100 µJ) was amplified in a Nd:YVO4 single-pass amplifier up to 1 mJ energy. Then the pulses were compressed down to 15 ps with an efficiency of 84% by using the reflection diffraction gratings compressor where 1818 groves/mm gratings were used. Finally, pulses of narrower than 2 cm-1 bandwidth were frequency doubled with 77% conversion efficiency in 6.5 mm length LBO crystal resulting in 650 µJ pulse energy radiation at 532 nm. The narrowband visible radiation was directed to the SFG spectrometer.
3.3 Sum-frequency generation and detection
The narrowband VIS and the broadband mid-IR pulses were combined on the sample in the SFG spectrometer, simplified drawing of which is presented in Fig. 10. The SFG spectrometer setup used in the experiments was similar to the one available commercially from Ekspla. Mid-IR and VIS beams were arranged in a vertical plane that allows performing measurements of liquid samples as well as at the liquid-air interface. Narrowband channel radiation was focused by a 500 mm focal length lens positioning the beam waist on the sample with the spot diameter of 0.3 mm at the 60° angle of incidence. For the broadband channel radiation parabolic off-axis mirror of 100 mm focal length and 45° off-axis angle was used to avoid pulse broadening by dispersion of traditional transmission optical elements.
Fig. 10. The setup of SFG spectrometer. The abbreviations stand for: PM – parabolic mirror, M1-4 – mirror, L1-4 – lens, A1-2 – aperture, GP – Glan prism, NF – notch filter, MC – monochromator, CCD – charge-coupled device camera.
VIS and mid-IR beams overlap on a sample was adjusted by changing sample holder height and by adjusting the parabolic mirror. The sample holder had a shift, tilt and height adjustment. It was possible to change the states of polarization of both VIS and IR beams. The generated sum-frequency signal was collimated by fused silica lens then spatially filtered from scattered and reflected VIS radiation (L2 and A1). After that beam passed through a polarization analyzing Glan prism, it was focused into a monochromator slit. Additional aperture A2 was used to reduce background and scattered radiation and a notch filter to suppress residual VIS radiation. SFG spectrum was registered by the cooled charge-coupled device (CCD) camera. We used Andor Newton 971 EM CCD camera with 1600 × 400 pixel sensor cooled down to -80°C and 328 mm Andor Kymera 328i monochromator with 1200 groves/mm density grating.
For SFG spectrometer characterization several test samples were investigated. The measured spectrum of monoolein (C21H40O4) is presented in Fig. 11(a). Narrow dip in the SFG resonance of monoolein C = O stretching vibration at 1738cm-1 is the result of absorption of water vapor in the air which features a narrow line width. This graph is indicative for the evaluation of the resolution of the created spectrometer. The SFG spectrometer setup efficiently exploited the narrowband VIS beam to reach a spectral resolution down to <3 cm-1. At the same time, the measurement spectral range covered by tuning of mid-IR fs OPA span stretches from 4000 cm-1 to 950 cm-1.
Fig. 11. a) Illustration of the SFG spectrometer resolution and the single measurement spectral range (SFG spectrum of monoolein). Inset: magnified spectrum around the dip in 1740cm-1 line. b) The SFG spectrum of the ethanol/air interface obtained at different exposure times (0.1 s and 30 s).
The presented SFG spectrometer allowed to perform almost real-time spectra measurement of the air/liquid interface of liquid samples such as ethanol/air interface, which illustrates quite good sensitivity. An exposure time of only 0.1 s was enough to observe the spectrum with a quite good signal-to-noise ratio (Fig. 11(b)). Widely used for calibration of SFG spectrometers z-cut quartz generated sum-frequency so efficiently that the signal was visible by naked eye on a white screen.
To evaluate the bandwidth of the mid-IR channel gold/air interface was selected. Due to the non-resonant SFG signal, it corresponds to the laser pulse spectrum. The bandwidth of a broadband mid-IR channel allowed us to record up to 300 cm-1 spectral range at a single measurement (Fig. 12). Applying the crystal scanning technique (extended bandwidth in Fig. 12) mentioned before, we were able to increase the spectral range up to more than 850 cm-1. The crystal angular positions were scanned in less than a second while measurement of SFG spectrum can take tens of seconds or even tens of minutes. Thus, the scanning technique has minor effect on the SFG spectrum measurement. Moreover, it even makes spectrum normalization easier thanks to a broader spectral range of signal acquisition and smoothening of the mid-IR radiation spectrum.
Fig. 12. The experimentally measured SFG spectrum of the gold/air interface using the static and the scanning method of acquisition.
4. Conclusions
We have designed a hybrid SFG spectrometer featuring a tunable from 2 to 10 µm mid-IR channel featuring nearly 500 cm-1 bandwidth of simultaneous spectral data acquisition while assuring less than 3 cm-1 spectral resolution. For this system realization, we assembled a two-channel laser system that generated optically synchronized narrowband (∼1.5 cm-1) visible and broadband mid-IR pulses. Before mid-IR channel construction, the detailed energy conversion to mid-IR efficiency in line with the generation of femtosecond pulses of as broad as possible spectral width was modeled.
Three layouts of mid-IR OPA were analyzed targeting for optimal conversion efficiency, as well as pulse spectral width. The modeling revealed that the mid-IR OPA stage seeded by 1 nJ of SC radiation in 1190 nm-1850nm wavelength range is very ineffective manifesting conversion efficiencies lower than 0.001% using the available on the market nonlinear crystals. The exception was the GaSe crystal. The modeling showed that GaSe crystal in comparison to the rest of the tested crystals has about 100 times higher efficiency, but still too low for the realization of SFG spectrometer. It was found that to achieve acceptable output energies the input signal energy in this kind of OPA should be of microjoule-level. Following the modeling results, a two-stage mid-IR OPA scheme was chosen. Modeling manifested unsatisfying output parameters due to the low efficiency of the first mid-IR OPA stage. The alternative approach exploiting the signal wave of BBO OPA seeded with long wavelength wing (1150 nm-1850nm) of SC radiation and pumped by the second harmonic of pump laser at 532 nm demonstrated acceptable performance of energetic seed generation. This approach demonstrated conversion efficiencies from ∼0.6 to ∼5.5% and several µJ levels of pulse energies amplified in mid-IR OPA, what is 2-3 orders of magnitude higher than produced by mid-IR OPA layouts pumped by fundamental 1064 nm radiation by the same pump energy. In this case, as modeled, the idler pulse energy after the second stage reaches 1-3.4 µJ at 5-10 µm idler wavelength and 4.5-9.3 µJ at 2.5 µm idler wavelength with bandwidth broader than 500 cm-1, which fulfills requirements for the broadband SFG system realization.
Based on the simulation results experimental setup of the SFG spectrometer system was designed, assembled and tested. The 1 mm length AGS, 0.5 mm length GS and 1 mm length LGS crystals were tested for the mid-IR fs OPA stage. We were able to achieve tunability from 5 µm up to 13 µm idler wavelength using the GS crystal and from 2 µm up to 10 µm using the AGS and the LGS crystals. Energies up to 0.4-1.8 µJ of idler wave at 2-5 µm wavelengths were achieved using the AGS crystal and 0.2-0.5 µJ at 5-10 µm wavelengths using the LGS and the GS crystals. The broadest bandwidth over the tuning range of 205-550 cm-1 was achieved using the LGS crystal and 230-470 cm-1 using the AGS crystal. The experimental results differed from the simulated probably due to factors as two-photon absorption in some crystals, lower pump intensity and reflection losses were not taken into account in the simulation.
In order to increase the idler energy use of 1 mm length AGS crystal with angular scanning was analyzed. The increase in the mid-IR energy was from 2 to 4 times in 2-5 µm wavelength range (up to 2.4-5.1 µJ) and ∼2 times in 5-10 µm range (up to 0.2-1 µJ), meanwhile, the bandwidth in 5-10 µm spectral range was similar to the case of 0.4 mm length AGS crystal, around 250-350 cm-1 while in 2-5 µm wavelength range broader effective spectra of around 500-750 cm-1 were achieved.
A fully functional SFG system was tested. The spectral resolution of <3 cm-1 and spectral range of measurement stretching from 4000 cm-1 to 950 cm-1 were achieved. Energy of mid-IR pulses was sufficient to perform real-time measurements of various liquid and solid samples. As an example, at the 0.1 s exposure time ethanol spectrum was obtained with an acceptable signal-to-noise ratio. The crystal scanning technique allowed achieving more than 850 cm-1 bandwidth at 2900 cm-1 having a minor influence on spectra acquisition time.
Acknowledgements
We would like to thank Prof. Dr. Gediminas Niaura and his group from the Center for Physical Sciences and Technology for the collaboration, helpful suggestions and sharing of samples used for SFG experiments.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. X. D. Zhu, H. Suhr, and Y. R. Shen, “Surface vibrational spectroscopy by infrared-visible sum frequency generation,” Phys. Rev. B 35(6), 3047–3050 (1987). [CrossRef]
2. Y. R. Shen, “Surface properties probed by second-harmonic and sum-frequency generation,” Nature 337(6207), 519–525 (1989). [CrossRef]
3. K. B. Eisenthal, “Liquid interfaces probed by second-harmonic and sum-frequency spectroscopy,” Chem. Rev. 96(4), 1343–1360 (1996). [CrossRef]
4. H. F. Wang, W. Gan, R. Lu, Y. Rao, and B. H. Wu, “Quantitative spectral and orientational analysis in surface sum frequency generation vibrational spectroscopy (SFG-VS),” Int. Rev. Phys. Chem. 24(2), 191–256 (2005). [CrossRef]
5. A. G. Lambert, P. B. Davies, and D. J. Neivandt, “Implementing the theory of sum frequency generation vibrational spectroscopy: A tutorial review,” Appl. Spectrosc. Rev. 40(2), 103–145 (2005). [CrossRef]
6. C. S. Tian and Y. R. Shen, “Recent progress on sum-frequency spectroscopy,” Surf. Sci. Rep. 69(2-3), 105–131 (2014). [CrossRef]
7. J. P. Smith and V. Hinson-Smith, “Product Review: SFG coming of age,” Anal. Chem. 76(15), 287A–290A (2004). [CrossRef]
8. B. C. Chow, T. T. Ehler, and T. E. Furtak, “High-resolution sum-frequency generation from alkylsiloxane monolayers,” Appl. Phys. B: Lasers Opt. 74(4-5), 395–399 (2002). [CrossRef]
9. G. Niaura, Z. Kuprionis, I. Ignatjev, M. Kažemékaité, G. Valincius, Z. Talaikyte, V. Razumas, and A. Svendsen, “Probing of lipase activity at air/water interface by sum-frequency generation spectroscopy,” J. Phys. Chem. B 112(13), 4094–4101 (2008). [CrossRef]
10. S. Strazdaite, E. Navakauskas, J. Kirschner, T. Sneideris, and G. Niaura, “Structure Determination of Hen Egg-White Lysozyme Aggregates Adsorbed to Lipid/Water and Air/Water Interfaces,” Langmuir 36(17), 4766–4775 (2020). [CrossRef]
11. A. Lagutchev, A. Lozano, P. Mukherjee, S. A. Hambir, and D. D. Dlott, “Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy Compact broadband vibrational sum-frequency generation spectrometer with nonresonant suppression,” Spectrochim. Acta, Part A 75(4), 1289–1296 (2010). [CrossRef]
12. O. Isaienko and E. Borguet, “Ultra-broadband sum-frequency spectrometer of aqueous interfaces based on a non-collinear optical parametric amplifier,” Opt. Express 20(1), 547 (2012). [CrossRef]
13. J. C. Phys, “Flexible high-resolution broadband sum-frequency generation vibrational spectroscopy for intrinsic spectral line widths Flexible high-resolution broadband sum-frequency generation vibrational spectroscopy for intrinsic spectral line widths,” J. Chem. Phys. 150(7), 074702 (2019). [CrossRef]
14. Z. Heiner, V. Petrov, and M. Mero, “Compact, high-repetition-rate source for broadband sum-frequency generation spectroscopy,” APL Photonics 2(6), 066102 (2017). [CrossRef]
15. L. Velarde, X. Y. Zhang, Z. Lu, A. G. Joly, Z. Wang, and H. F. Wang, “Communication: Spectroscopic phase and lineshapes in high-resolution broadband sum frequency vibrational spectroscopy: Resolving interfacial inhomogeneities of identical molecular groups,” J. Chem. Phys. 135(24), 241102 (2011). [CrossRef]
16. R. Feng, L. Lin, Y. Li, M. Liu, Y. Guo, and Z. Zhang, “Effect of Ca 2+ to Sphingomyelin Investigated by Sum Frequency Generation Vibrational Spectroscopy,” Biophys. J. 112(10), 2173–2183 (2017). [CrossRef]
17. L. Velarde and H. F. Wang, “Unified treatment and measurement of the spectral resolution and temporal effects in frequency-resolved sum-frequency generation vibrational spectroscopy (SFG-VS),” Phys. Chem. Chem. Phys. 15(46), 19970–19984 (2013). [CrossRef]
18. L. Velarde and H. F. Wang, “Capturing inhomogeneous broadening of the -CN stretch vibration in a Langmuir monolayer with high-resolution spectra and ultrafast vibrational dynamics in sum-frequency generation vibrational spectroscopy (SFG-VS),” J. Chem. Phys. 139(8), 084204 (2013). [CrossRef]
19. K. Michailovas, A. Baltuska, A. Pugzlys, V. Smilgevicius, A. Michailovas, A. Zaukevicius, R. Danilevicius, S. Frankinas, and N. Rusteika, “Combined Yb/Nd driver for optical parametric chirped pulse amplifiers,” Opt. Express 24(19), 22261 (2016). [CrossRef]
20. R. Danilevičius, A. Zaukevičius, R. Budriūnas, A. Michailovas, and N. Rusteika, “Femtosecond wavelength-tunable OPCPA system based on picosecond fiber laser seed and picosecond DPSS laser pump,” Opt. Express 24(15), 17532 (2016). [CrossRef]
21. L. J. Richter, T. P. Petralli-Mallow, and J. C. Stephenson, “Vibrationally resolved sum-frequency generation with broad-bandwidth infrared pulses,” Opt. Lett. 23(20), 1594 (1998). [CrossRef]
22. Z. Heiner, L. Wang, V. Petrov, and M. Mero, “Broadband vibrational sum-frequency generation spectrometer at 100 kHz in the 950-1750cm−1 spectral range utilizing a LiGaS2 optical parametric amplifier,” Opt. Express 27(11), 15289 (2019). [CrossRef]
23. G. Cerullo and S. De Silvestri, “Ultrafast optical parametric amplifiers,” Rev. Sci. Instrum. 74(1), 1–18 (2003). [CrossRef]
24. G. B. Abdullaev, K. R. Allakhverdiev, L. A. Kulevskii, A. M. Prokhorov, ÉY Salaev, A. D. Savel’ev, and V. V. Smirnov, “Parametric conversion of infrared radiation in a GaSe crystal,” Sov. J. Quantum Electron. 5(6), 665–668 (1975). [CrossRef]
25. Y. X. Fan, R. C. Eckardt, R. L. Byer, R. K. Route, and R. S. Feigelson, “AgGaS2 infrared parametric oscillator,” Appl. Phys. Lett. 45(4), 313–315 (1984). [CrossRef]
26. L. I. Isaenko and A. P. Yelisseyev, “Recent studies of nonlinear chalcogenide crystals for the mid-IR,” Semicond. Sci. Technol. 31(12), 123001 (2016). [CrossRef]
27. D. N. Nikogosyan, “Beta barium borate (BBO),” Appl. Phys. A 52(6), 359–368 (1991). [CrossRef]
28. H. Nakatani, W. R. Bosenberg, L. K. Cheng, and C. L. Tang, “Laser-induced damage in beta-barium metaborate,” Appl. Phys. Lett. 53(26), 2587–2589 (1988). [CrossRef]
29. M. Bradler, C. Homann, and E. Riedle, “Mid-IR femtosecond pulse generation on the microjoule level up to 5 µm at high repetition rates,” Opt. Lett. 36(21), 4212 (2011). [CrossRef]
30. B.-H. Chen, E. Wittmann, Y. Morimoto, P. Baum, and E. Riedle, “Octave-spanning single-cycle middle-infrared generation through optical parametric amplification in LiGaS2,” Opt. Express 27(15), 21306 (2019). [CrossRef]
