1. Introduction

Vibrational sum frequency generation (VSFG) spectroscopy is a surface-specific technique that is important for the characterization of material interfaces [1–11]. In these measurements, an IR beam is spatially and temporally overlapped with a visible or NIR beam at the sample, and an outgoing beam at the sum of their frequencies is analyzed in a phase matched direction [5,9]. The technique derives its surface specificity from selection rules that, within the electric dipole approximation, require macroscopic noncentrosymmetry to give non-zero tensor elements in the second-order susceptibility, χ⁽²⁾ [5,9]. When noncentrosymmetry is present, of the 27 possible elements in this third-rank tensor, only seven are allowed to be nonzero, of which four are redundant [5,9,12]. As a result, χ⁽²⁾ can be fully characterized by measuring VSFG signals using four different input and output beam polarization combinations: ppp, ssp, sps, and pss, where the polarizations are listed in order of decreasing beam energy (e.g. ssp = s – signal, s – visible, and p – IR). Combining this information with the input beam geometries and measurable optical constants and film thicknesses at all relevant frequency ranges enables one to determine a wealth of information, including the average molecular structure and orientation at the material interface.

In practice, obtaining absolute molecular parameters is rarely achieved, in part because the local field corrections due to beam geometries and optical coefficients are not applied because they are not known. In addition, orientation distributions of the transition dipoles can have a profound influence on the determined quantities, sometimes to the point of obscuring them altogether when wide distributions are present at the interface [13]. Furthermore, for absolute orientation to be determined, the signal phase is required, which is only obtained in heterodyne detected experiments [7,14,15]. Nonetheless, VSFG excels at providing semi-quantitative orientation information about interfacial environments and has been used to determine order and orientation in small molecule and polymeric semiconducting organic thin films [11,16–20], at aqueous interfaces [3,7,21,22], and in self assembled monolayers [21,23–25].

Orientational studies require more than one of the nonzero χ⁽²⁾ tensor elements to be sampled. This is typically achieved by performing separate experiments in which different polarization combinations are used. Building on the work of Smits and associates, we previously reported a method of collecting all four VSFG polarizations at once [26–28]. The utility of this approach was highlighted by a study in which we monitored interfacial ordering and orientation in real-time while thermally annealing an organic thin film [16]. A challenge in that work was that it required simultaneous alignment of three pump beams onto the sample and four signal beams onto the detector. Since the visible pump beams were aligned to the sample at different incident angles, the coefficients for reflection and the phase matching directions were slightly different, introducing a source of error in the measurement [26]. The generated signal beams that contained a mixture of polarizations (to be later separated by a birefringent optic) also needed to be independently steered to the entrance slit on the monochromator.

In this work, we present a new method for multiplex polarization data collection. By using a modified 4f pulse shaper, a simple frequency comb is generated that can then be used as the upconverting beam with a broadband IR source in a VSFG experiment. This method has improvements over the previous approach, including collinear alignment and independent tailoring of pulse timing, bandwidth, and polarization within the visible beam frequency comb. This ultimately allows for faster data collection with ease of alignment and improved data fidelity for samples that change on the time scale of a measurement and/or are spatially heterogeneous.

2. Experimental setup and measurement procedure

Broadband IR pulses for VSFG were produced by using half of the output (500 mW) of a regeneratively amplified ultrafast Ti:Sapph laser (Spectra-Physics Spitfire), centered at 800 nm with 30 nm bandwidth (FWHM) at 1 khz, to pump an optical parametric amplifier (Spectra-Physics, OPA-800C). The near-IR signal and idler outputs of the OPA were difference frequency mixed in a 1 mm silver gallium sulfide (AgGaS₂) crystal to produce tunable mid-IR light (150 cm⁻¹ FWHM, ~120 fs). The mid-IR was sent though a delay stage to control its timing relative to the visible pulses, followed by an IR half waveplate (Altechna) and a wiregrid polarizer to control its polarization at the sample.

The visible frequency comb was created by amplitude-shaping 500 mW of the uncompressed regen output (prior to the compressor) in a modified 4f pulse shaper, as shown in Fig. 1. A typical 4f pulse shaper consists of a diffraction grating placed at the focal point of a double convex lens and slit with a mirror behind it at the other focal point [29,30]. The incoming light is dispersed horizontally by the diffraction grating and is imaged to a stripe at the retroreflector. The slit selects a portion of the input bandwidth to be reflected back through the lens, recombined at the diffraction grating, and reflected out of the shaper (typically at a slightly different height). The modified 4f shaper in Fig. 1 introduces a D-shaped broadband dielectric mirror after the refocusing lens, sending approximately half of the dispersed light to a second slit and mirror at the focal point of the lens. Retroreflections from both mirrors in the pulse shaper recombined at the diffraction grating to a single beam that consisted of two narrowband pulses (approximately 1 nm FWHM) that were collinear but separated in wavelength by about 15 nm. The input 800 nm light was uncompressed because it was picked off from the main amplifier output prior to compression. As a result, the two collinear pulses were also dispersed in time by tens of ps. This setup allowed individual control of the bandwidths of the two frequency comb “teeth”. The retroreflector and slit pairs were on translation stages to optimize the selected frequencies and the retroreflector positions relative to the focal plane of the lens. For our purposes the comb was shaped as shown in Fig. 1 with two spectral peaks at 790 and 806 nm. An alternative approach would be to use a double slit at a single mirror rather than dividing the beam into two arms of the shaper. This would simplify alignment but might offer less flexibility in tailoring the bandwidths of the spectral teeth. We also note that the arrangement in Fig. 1 enables one to independently adjust the focal properties of the two beams to give different focal spot sizes.

 

Fig. 1 Right: Diagram of the modified 4f pulseshaper to produce a simple frequency comb showing the relative positions of the diffraction grating, lens, mirrors, and slits. The mirror after the lens directs half of the light to a second slit that is retroreflected back and recombined at the grating with the shaped throughput from the other slit. Left: The input and output spectrum of the modified pulse shaper.

Download Full Size | PDF

For VSFG to occur both the visible and mid-IR pulses need to be overlapped in space and time at the sample. Given the uncompressed nature of the 800 nm broadband light entering the pulse shaper, the two narrowband pulses needed to be retimed to arrive at the sample at the same time. This was achieved with a simple retiming section constructed from two sharp cut-on dichroic beamsplitters that reflected the short wavelength pulse and transmitted the long wavelength pulse (Semrock, 785 nm BrightLine®), as shown in Fig. 2. The dichroic beamsplitters went from 5% to 95% transmission over 5 nm with a 50% transmission at about 798 nm. The two legs of the retiming section enabled adjustment of the arrival times of the two visible pulses at the sample as well as independent beam steering to optimize their collinearity as they continued on to the sample. Before recombining at the second dichroic beamsplitter, the shorter wavelength beam transmitted through a half waveplate and thin film polarizer to independently control its polarization from the longer wavelength beam. Both beams were then recombined at the second beamsplitter and aligned as a single beam to the sample.

 

Fig. 2 Diagram of the retiming portion of the visible beam path. The frequency comb ‘teeth’ are separated by a dichroic beamsplitter for indepent timing, pointing, and polarization control. The 790 nm beam polarization is controlled by a halfwave plate and a thin film polarizer. The 806 nm polarization is cleaned up by a Glan-Thompson polarizer.

Download Full Size | PDF

The visible and IR beams were focused at the sample using a 60° off-axis parabolic reflector and the emitted VSFG signal was collected and collimated by a matching parabolic reflector. The emitted signal was directed through a twisted periscope and through a lowpass filter to remove the visible pump. It was then focused onto the slit of a 150 mm monochromator with a 15 micron slit size followed by detection with a liquid N₂ cooled CCD camera (Princeton Instruments). In a typical alignment procedure, the visible and IR beams were overlapped in space at the sample stage in a ZnS crystal, which produced an extremely large nonresonant VSFG signal for initial alignment and timing. The delay stages for the IR and one of visible delay stages in the retiming section (Fig. 2) were used to maximize signal. After this, these two IR and visible stages were not moved again and the delay stage of the other visible beam was used to maximize its VSFG signal through the monochromator. After finding the correct timing of the pulses, the sample was switched to a 100 nm thin film of ZnO deposited by atomic layer deposition on a Si wafer. This sample gave significant amounts of VSFG due to its lack of centrosymmetry through the bulk of ZnO, but from a thin slab so that fine adjustments could be made to the sample position, beam pointing, and timing to further optimize the signal before switching to the organic thin films of interest. A HeNe laser was used to optimize sample positioning as described previously [31,32].

Thin films of α–sexithiophene (6T, 35 nm) and dioctyl-perylenetetracarboxydiimide (PTCDI, 100 nm) were prepared by vapor deposition onto RCA-cleaned glass and thermal oxide SiO₂ wafers, respectively, using a homebuilt organic vapor desposition chamber [33]. For VSFG measurements on organic thin films, two exposures were collected first with both IR and visible beams and second with the IR blocked to subtract out the background. Five spectra were collected with 5 and 3 minute exposure times for 6T and PTCDI films, respectively, followed by averaging. Cosmic rays were corrected by a program that automatically identified a cosmic ray and compared pixels of the same wavelength in the other exposures. A reference spectrum was collected on ZnO using a ppp polarization combination for both visible beams. Further details of data analysis are described in the discussion below.

3. Results and discussion

Figure 3(a) shows the reference spectrum collected in ppp from ZnO using the frequency comb upconverting beam. Two wide, approximately Gaussian peaks are observed at 706 and at 721 nm, each with 150 cm⁻¹ bandwidth to match the bandwidth of the IR beam. These are nonresonant VSFG signals generated by the 790 and 806 nm visible beams, respectively. The peaks are slightly misshapen by water vapor absorption of the IR beam, which is tuned here to approximately 1450 cm⁻¹. The spectra are similar in shape and intensity. It is believed that subtle differences in intensity are due to errors in coalignment of the two visible pulse in the frequency comb leading to very slight differences in signal phase matching and alignment to the monochromator slit as differences in effeciencies of the gratings and dichroic optics alone does not account for this difference. This is not overly important as the subsequent sample spectra will be divided by this reference spectrum and will inherit the same pointing differences to compensate for any systematic inconsistencies. Similarly, the diffraction grating in the monochromator has small differences in efficiency at these two wavelengths (<5%), but this will also be normalized in the sample spectra when dividing by the reference spectra shown here. The shorter wavelength VSFG peak is marginally broader than the longer wavelength peak. This is because the spectral resolution is lower at shorter wavelengths – a larger energy range falls on each pixel at shorter wavelengths, which will become more apparent in the resonant VSFG spectra discussed below.

 

Fig. 3 a) VSFG spectrum of ZnO thin film and b) a 6T thin film, both collected with the ppp polarization combination. The bottom frequency axis has been created by down converting the summed frequencies with the known center wavelengths of the frequency comb teeth.

Download Full Size | PDF

Figure 3(b) shows the VSFG spectrum generated from a 35 nm thin film of 6T on glass using the ppp polarization combination. The lower axis shows the down-converted frequency ranges using the measured upconverting visible beam spectra, but we emphasize that this figure shows a single continuous spectrum collected from the sample using the same IR beam and the frequency comb beam to upconvert two spectrally separated VSFG regions. The two spectra are identical within our signal-to-noise level, and have relative intensities that match those of the reference spectrum in Fig. 3(a). These demonstrate multiplex VSFG data collection from the same sample spot in a single spectrum using the spectral resolution imparted by a frequency comb rather than vertical displacement into separate spectra.

A more interesting case is shown in Fig. 4 in which we apply this approach to a 100 nm film of PTCDI, a material that we have shown to have a markedly different VSFG response in different polarization combinations [31]. For this experiment, we rotated the polarization of the short wavelength part of the comb to be s-polarized, while setting the IR beam to a mixed polarization. From previous work with this system, we expected that the ssp VSFG spectrum would be over an order of magnitude more intense than the sps combination [31]. Since a single integration time was used to collect the multiplex spectrum, the IR beam was rotated to put 3% of the power into the s-polarization and 97% in the p-polarization so that the signal-to-noise ratio would be the same for the two regions. The spectrum in Fig. 4(a) is a single spectrum with the low and high wavelength portions color coded to indicate that these regions were upconverted with a different part of the comb. The low wavelength region shows the raw ssp VSFG spectrum; the high wavelength region shows the sps VSFG spectrum. A ppp reference spectrum was collected (not shown) with the IR tuned to the 1680 cm⁻¹ region, and was used to normalize the ssp and sps spectra, resulting in the spectra in Fig. 4(c) (with consistent color coding). The normalization process accounts for the differences in intensity across the IR bandwidth, and the spectral features are accordingly reshaped. The resulting spectra are identical to those previously reported but were collected in half the time [31].

 

Fig. 4 a) VSFG spectra of a PTCDI thin film collected in ssp (black) and sps (red), and b) using sps (blue) and sps (green) polarizations. c) ssp and sps (from part a) and d) sps and sps (from part b) after normalization by the ppp reference spectrum.

Download Full Size | PDF

Interestingly, this optical design actually generates all four VSFG polarization combinations at once (ppp, ssp, sps, and pss). Before entering the monochromator the analyzing polarizer removes the ppp and pss signals and transmits the ssp and sps. The ssp and sps combinations give different information and their comparison allows one to determine the orientation of the IR transition dipole for a particular resonance [10]. In a previous report, we demonstrated how film thickness could be used to favor the VSFG signal from the buried interface of a thin film, and we showed that ssp and sps were the only polarization combinations that could be paired together to favor the buried interface for the same beam angles and film thickness [31,34]. In this context, the current results are unique in their ability to provide this particular pair of VSFG signals simultaneously. Previous reports have always differentiated multiplexed signals by the analyzer polarization, forcing pairs of ppp/sps or ssp/pss using mixed polarization IR, or ppp/ssp, sps/pss using mixed polarization visible, or ppp/spp and ssp/psp in chiral SFG [26–28]. Here we use a single analyzer polarization with mixed IR polarization and mixed visible polarization, but separate the signals with the monochromator rather than a birefringent beam displacer after the sample.

To demonstrate the immunity of this beam arrangement to misalignment, Fig. 4(b) shows the VSFG spectrum for the same PTCDI sample but with both visible pulses in the frequency comb rotated to p-polarization and the IR returned to full s-polarization. In this case, the longer wavelength comb was slightly misaligned onto the sample leading to a pointing error in the generated signal. As a result, the relative peak ratios between the short and long wavelength regions are not the same. However, the ppp reference spectrum also shares this misalignment and can be easily used to correct the data assuming no beam adjustments were made between the reference and sample spectra. Figure 4(d) shows that once they are divided by the ZnO ppp reference spectrum the data appear to be nearly identical. We believe that the pointing sensitivity is the result of spatial chirp in the IR beam at the sample, which would cause a difference in the frequencies mixed at the interface that give rise to the VSFG signal. However, the fact that normalizing by the ZnO reference spectrum corrects this shows the considerable resilience of this technique.

4. Conclusions and future directions

In conclusion, we have demonstrated for the first time that a simple frequency comb can be used for VSFG, and that the spectral resolution of the resulting signals opens the door to a new, simpler approach to multiplex polarization measurements. This system allows for collection of the ssp and sps polarization combinations (and, in principle, pss and ppp) from the same spot, at the same time, and with the same incident angles in a reliable and robust way. Generally speaking, this is an improvement over other traditional VSFG approaches as it reduces problems with multiple beam alignments at the sample and at the detector, sample heterogeneity, and changes in laser power, while enabling one to collect data faster. Compared to our previous multiplex detection scheme, this one allows the simultaneous collection of ssp and sps from the same spot with the same incident angle, a combination that shows a greater affinity for the buried interface in layered systems.

In addition to the practical improvements of the frequency comb approach, a number of additional directions could be explored. To begin with, one might design a double (or triple or quadruple) slit using a single retroreflector in the 4f pulse shaper to simplify its alignment. Furthermore, one could envision replacing the dual slit 4f pulse shaper with a single optic such as an etalon in the mode of Lagutchev and associates [35]. For example, a simple calculation shows that an air etalon can be designed for the 800 nm central wavelength using 83% reflective mirrors with a separation of 22.5 microns (6.66 THz free spectral range) to give two narrow Lorentzian peaks at 790 and 804 nm. This single optic could replace the entire shaper portion of our design, but at the expense of control over the spectral width of the teeth of the frequency comb. With enough bandwidth from the amplifier, one might also modify the pulse shaper (or design an etalon) to give three or four peaks in the frequency comb. This would require three dichroic beamsplitter retiming stages in order to obtain the requisite timing control over all four beams, but when combined with mixed polarization IR would give all four VSFG polarization combinations simultaneously in a single spectrum. We reiterate that even without the additional teeth, our current arrangement generates all four polarization combinations, and the other two (pss and ppp) are readily obtained by analyzing for the p-polarized VSFG signal. Furthermore, both the s-polarized and p-polarized sections could be spatially separated at the detector with a displacing optic to obtain all four polarization combinations using only two beam paths [26]. The independent timing of the two legs opens new directions as well. Changing timing allows for the mistiming of one leg to look at data simultaneously with and without nonresonant signal suppression [35]. Also, independent control of the slit of either leg allows for data collection with different spectral resolutions, perhaps in cases where signal may be very low and full resolution is not needed in one polarization. And finally, replacing the half waveplate with a quarter waveplate would enable experiments in which one beam is circularly polarized while the other is linearly polarized.