1. Introduction

Vibrational sum frequency generation (VSFG) spectroscopy [1–4] has become a major tool for surface science due to its unique interfacial selectivity and sensitivity. Within the electric-dipole approximation, VSFG is symmetry forbidden in isotropic media, and a VSFG signal can therefore only be generated at interfaces for which anisotropy breaks inversion symmetry. VSFG has been used widely to study interfaces of chemical and biological interest. VSFG is a second-order nonlinear process in which a visible beam of frequency ω_vis and an infrared beam of frequency ω_IR are overlapped temporally and spatially at the interface being studied to generate a third, signal beam that has a frequency of ω_sig=ω_vis+ω_IR. As shown in Fig 1, the visible and IR beams can either come from the same side of the surface normal (the co-propagating geometry) or from opposite sides of the surface normal (the counter-propagating geometry). Narrowband VSFG spectroscopy has been performed in both geometries (see [5] for references to examples of both), but broadband VSFG [6] has been performed almost exclusively with the co-propagating geometry. Even in narrowband VSFG, the co-propagating scheme has been widely preferred, particularly due to the minimal dispersion of the signal in this geometry [5]. Moreover, in the co-propagating geometry the signal beam and the visible beam are almost collinear, which makes delivering the signal beam into a detector relatively straightforward. The counter-propagating geometry, on the other hand, is technically more challenging to implement because the signal propagates in a direction that is significantly different from that of both the visible and IR beams. In most cases the signal beam is too weak to be seen by eye, making the delivery of the signal into a detector all the more difficult.

 

Fig. 1. Schematic layouts of the co-propagating geometry for VSFG (a) and the counter-propagating geometry for VSFG (b). Angles β to the left of the surface normal are taken as negative and those to the right of the surface normal are taken as positive.

Download Full Size | PDF

The one prior example of a broadband, counter-propagating VSFG spectrometer was reported by Eliel and coworkers [7]. These researchers used a free-electron laser producing pulses having roughly 50 cm^-1 bandwidth at 1050 cm^-1 to measure the C-O stretching frequency of thiophenol monolayers chemisorbed on a silver substrate. Their work demonstrated the feasibility of acquiring broadband VSFG spectra in a counter-propagating geometry, but they did not fully exploit this geometry’s ability to emphasize specific vibrational features that are often weak and difficult to resolve in co-propagating experiments.

Here, we report a broadband, counter-propagating VSFG spectrometer based on a Ti:sapphire regenerative amplifier system. We demonstrate that despite the experimental challenges associated with the counter-propagating scheme, VSFG experiments carried out with this geometry have many attractive features. The large angular separation of the VSFG signal from the reflected visible beam helps to prevent the input visible light from reaching the detector. This spatial separation of the signal and the reduced non-resonant background for buried interfaces in this geometry make it possible to attain high dynamic range. The counter-propagating geometry also provides more space in the sample area, making adjustment of optics convenient. Due to the comparatively small beam deviation caused by the sample, the signal can be regained rapidly when the sample is changed. In addition, as compared to the co-propagating geometry, different vibrational modes are emphasized under certain polarization conditions, which can be useful in making spectral assignments.

2. Theory

We begin with a brief summary of the VSFG theory relevant to our spectrometer. The VSFG intensity is proportional to the magnitude squared of the second-order nonlinear optical susceptibility [8],

Here, I_i is the intensity at frequency ω_i, c_o is the speed of the light in vacuum, χ ⁽²⁾ _eff is the effective second-order nonlinear susceptibility, n_i is the refractive index at frequency ω_i in the bulk medium in which the beams propagate, and β_sig is the angle that the signal beam makes with the interface normal (Fig. 1). In the frame of reference used here, angles on the same side of the surface normal as the incoming IR beam are taken to be negative and angles on the other side of the surface normal are taken to be positive. The conservation of momentum relation

determines the direction of the signal beam.

χ ⁽²⁾ _eff is a 27-element tensor that can be reduced to four independent nonzero elements if the surfaces are azimuthally symmetric: χ ⁽²⁾ _zzz, χ ⁽²⁾ _xxz=χ ⁽²⁾ _yyz, χ ⁽²⁾ _xzx=χ ⁽²⁾ _yzy and χ ⁽²⁾ _xxz=χ ⁽²⁾ _yyz, where x_y is defined as the plane of the interface and the surface normal lies along z. The polarizations are listed in the order of the VSFG signal, the visible beam and the IR beam, respectively.

The four most commonly used experimental polarization combinations are SSP, SPS, PSS and PPP. The detected χ ⁽²⁾ _eff under each of these four polarization conditions is given by [8]

and

where L_ii (i=x, y, z) is a diagonal element of a 3×3 Fresnel matrix determined by the refractive indices of the two bulk phases and the interfacial layer, as well as by the incident and reflected angles. In SSP spectra, only molecules with a component of the IR transition dipole along the surface normal will appear. In SPS spectra, only molecules with a component of the IR transition dipole in the surface plane will contribute to the signal. Therefore, VSFG spectra can provide information about average molecular orientation at interfaces [9]. Eqs. 3 through 6 illustrate how judicious choice of experimental geometries can enhance or suppress specific elements in the χ ⁽²⁾ tensor. This last consideration is especially relevant for spectra acquired under PPP conditions. In a co-propagating geometry, there is an almost complete cancellation of the χ ⁽²⁾ _xzx and χ ⁽²⁾ _zxx terms under PPP conditions. In a counter-propagating geometry the significantly different Fresnel factors for these two terms tend to prevent strong cancellation. This dependence of the measured susceptibility on geometry was explored in detail by Wang and co-workers using a narrowband, scanning VSFG spectrometer [10].

3. The spectrometer

A schematic of our experimental system is shown in Fig. 2. A Coherent Legend Elite Ti: sapphire regenerative amplifier generates 130-fs, 800-nm pulses at a repetition rate of 1 KHz and an average power of 3 W. 30% of the 800-nm power is used to pump a commercial optical parametric amplifier (TOPAS-C, Light Conversion) with a noncollinear difference-frequency generation module (Light Conversion) to generate IR pulses that are tunable from 2 µm to 12 µm. At a wavelength of 3.5 µm the full-width-at-half-maximum bandwidth of the IR pulses is 100 cm^-1 and their average power is 10 mW as measured immediately before the sample. A second portion of 30% of the 800-nm light is frequency-narrowed with an optical stretcher (Fig. 2). The typical spectral bandwidth of this beam is between 14 and 18 cm^-1. The average power of this spectrally-narrowed beam at the sample is about 45 mW. The angles of the incidence of the 800 nm and IR beams are 64.5° and -54° from the surface normal, respectively, and the signal propagates at an angle of approximately -35° to the surface normal. The VSFG signal is collimated and then is directed onto the entrance slit of a spectrometer (Acton, SP2300i). The VSFG spectrum is detected using a 100×1340 pixel CCD array (Spec-10:100, Roper Science). Data are obtained at a resolution of 0.02 nm/pixel.

 

Fig. 2. Schematic experimental layout for the counter-propagating VSFG spectrometer. OPA=optical parametric amplifier, nDFG=noncollinear difference-frequency generation module, pol=polarizer, HWP=half-wave plate, CCD=charge-coupled device camera.

Download Full Size | PDF

Quartz IR cells (Hellma USA) are used to hold the samples. Silica/vapor and silica/liquid interface measurements are performed in a cell with a 1-mm path length (part number: 404-1-46), while liquid/vapor interfaces are studied in cells with a 1-cm path length (part number: 404-10-46). All cells were cleaned with an oxygen plasma immediately prior to use.

4. Alignment

Obtaining VSFG signal requires both temporal and spatial overlap of the visible and IR pulses at an interface. Attaining this overlap, especially temporally, can be challenging in the counter-propagating geometry. To overcome these difficulties, we have developed the following procedures. First, spatial overlap is achieved by focusing the visible and IR beams onto the same spot on a 2-mm-thick ZnSe substrate. Blue luminescence appears on the ZnSe surface when the IR beam is focused tightly, providing a simple means of ensuring overlap with the 800 nm beam. To find timing, we take advantage of the fact that the IR pulse can change the 800-nm transmission of the ZnSe substrate over a time that is much longer than the duration of either pulse but that is much shorter than the 1 ms laser repetition time. By monitoring the 800-nm transmission, we can therefore determine in a straightforward way whether the 800-nm light arrives at the sample before or after the IR pulse, and thus can find the time at which both pulses overlap. To perform this procedure, the IR beam is chopped synchronously with the firing of the amplifier, but at half of the repetition rate (i.e., 500 Hz). A photodiode is used to monitor the 800 nm transmission of the ZnSe. The photodiode output is fed into a lock-in amplifier that is referenced to the chopper frequency.

Typical IR pump/800-nm probe data are shown in Fig. 3. The red and black lines are the change in 800-nm transmission and the SFG signal for the ZnSe, respectively, plotted as a function of the IR/800-nm delay time. When the IR pulse arrives after the 800-nm pulse, the 800-nm transmission is unaffected by the relative timing of the pulses. When the IR pulse arrives before the 800-nm pulse, the transmission of the 800-nm pulse increases. The ZnSe SFG signal, which comes from both the surface and the bulk of the substrate, should reach its maximum intensity roughly half-way up the rising edge of the IR pump/800-nm probe signal. As shown in Fig. 3, the SFG timing agrees with this prediction.

 

Fig. 3. SFG signal (black) and change in 800-nm transmission (red) for a ZnSe substrate as a function of the delay time between the IR pulse and the 800-nm pulse.

Download Full Size | PDF

The SFG signal from the ZnSe substrate is strong enough to be seen by naked eye on a piece of paper while wearing safety glasses that block the IR and 800-nm light. This SFG signal can therefore be used to align the detection optics. After finding timing and aligning the detection optics with the ZnSe substrate, the sample is changed to a gold plate, for which all of the SFG signal is generated at the surface. The sample height is adjusted to maximize the signal, and the detection optics are optimized once more. As the final step, a sample cell containing dimethyl sulfoxide (DMSO) is used and the IR wavelength is set to be resonant with the symmetric stretch of DMSO (~3.44 µm). A HeNe laser tracer beam is first reflected off of the gold surface and an iris is placed in the beam path approximately 20 cm away from the surface. After changing to the DMSO sample, the height of the sample stage is adjusted until the reflection of the tracer laser from the DMSO liquid/vapor interface (which is identified by tapping on the cell to disturb the liquid surface) is centered on the iris again. At this stage height, the VSFG signal is once again optimized.

One particular advantage of the counter-propagating geometry is that samples can be changed and signal regained rapidly. Two different facets of the counter-propagating geometry make rapid sample changes possible. First, this geometry helps to preserve overlap between the IR and 800-nm laser spots upon changing samples. Each laser beam refracts upon entering the quartz from the air. In the counter-propagating geometry this refraction is readily compensated by adjusting the height of the sample stage. The displacement of the VSFG signal due to changing the sample position is negligible because the refractive effects of the two input beams nearly cancel one another. In a co-propagating geometry, regaining the VSFG signal is usually more difficult. The 800-nm and IR beams experience different refractive indices, and so can bend away from one another. This effect must be compensated by realignment of one of the input beams.

Second, propagation of the signal beam through a sample window leads to beam displacement. The magnitude of the displacement depends on the thickness and refractive index of the window and the frequencies and incident angles of the input beams. As shown in Fig. 4, the displacement from this effect in the counter-propagating geometry is generally much smaller than in the co-propagating geometry, allowing signal to be regained rapidly upon switching samples. For example, using the above 800-nm and IR beam incident angles for a cell with quartz windows, the displacement is less than 5% of the window thickness in the counter-propagating geometry and is about 50% of the window thickness in the co-propagating geometry.

 

Fig. 4. Displacement of the signal beam due to refraction in going from a liquid/vapor or solid/vapor interface (dotted lines, interface not shown) to an interface below a window (solid lines) for the co-propagating VSFG geometry (a) and the counter-propagating VSFG geometry (b). The signal displacement in the latter geometry is considerably smaller than that in the former geometry.

Download Full Size | PDF

To demonstrate the ease of switching samples with our system, in Fig. 5 (Media 1) we show real-time VSFG data in the symmetric methyl stretch spectral region under PPP polarization conditions for the silica/vapor interface of DMSO (Fig. 5a) and acetonitrile (Fig. 5b). Switching the sample cells and regaining signal took less than 17 seconds in this particular instance, which is a typical result for switching between samples in which the interface is defined by a solid surface. In order to regain signal when at least one of the samples has a liquid/vapor interface, additional time is required to find the appropriate sample height. The spectra shown in Fig. 5 were each obtained with an integration time of 1 sec, underscoring the high sensitivity of the spectrometer.

 

Fig. 5. Real-time PPP VSFG signals (with an integration time of 1 second) from the silica/vapor interfaces of (a) DMSO and (b) acetonitrile. The time elapsed in switching from the DMSO sample to the acetonitrile sample and regaining signal was less than 17 seconds (Media 1).

Download Full Size | PDF

5. Results

To demonstrate the sensitivity of our counter-propagating VSFG spectrometer, we present data for the methyl stretching spectral region of DMSO, a well-studied model system [11]. We have used our spectrometer to study the silica/vapor, silica/liquid and liquid/vapor interfaces of neat liquid DMSO. Below we present the SSP, SPS and PPP spectra of DMSO acquired at these interfaces. The infrared frequency is calibrated by measuring the absorption spectrum of polystyrene using the nonresonant SFG signal from a gold substrate. The SFG spectrum of gold is featureless, but when a thin (approximately 60 µm) film of polystryene is inserted into the IR path, the Au spectrum shows sharp attenuation at the polystyrene IR absorption bands. The resulting 5-point calibration correlating polystyrene IR absorbance and detected SFG signal enables us to assign observed VSFG resonances with an accuracy of ± 2 cm^-1 across the C-H stretching region (2800 cm^-1 to 3050 cm^-1). Each broadband spectrum is obtained by summing over individual VSFG spectra acquired at seven consecutive wavelength steps around the center wavelength (3300 nm) with a spectral step size of 50 nm. The summed spectra were smoothed using a 10-point (5 cm^-1) rolling average, which removes some of the high-frequency noise seen in the raw data in Fig. 5.

In the SSP spectra of DMSO at all three interfaces (Fig. 6), only the methyl symmetric stretch appears. This mode appears at 2916, 2908 and 2900 cm^-1 at the silica/vapor, silica/liquid and liquid/vapor interfaces, respectively. The distribution of frequencies for the same vibrational mode of molecules at different interfaces is significant. The higher frequencies observed for DMSO adsorbed to silica (both at the silica/vapor and silica/liquid interfaces) illustrate the inductive effects that hydrogen-bond-donating surface silanol groups can have on the vibrational modes of adjacent, hydrogen-bond-accepting species. This distribution of frequencies is also important to acknowledge if a given VSFG band is used as an internal frequency standard.

 

Fig. 6. VSFG spectra of DMSO under SSP polarization conditions at the silica/vapor (black), silica/liquid (red) and liquid/vapor (green) interfaces. The intensities in this and the following figures are all in the same units and can be compared directly.

Download Full Size | PDF

In the SPS spectra (Fig. 7), we observe the methyl antisymmetric stretch at 2992, 2998 and 2977 cm^-1 at the silica/vapor, silica/liquid and liquid/vapor interfaces, respectively. This is the first time, to our knowledge, that this mode of neat DMSO has been observed in a VSFG spectrum. This observation is made possible by the high dynamic range of our apparatus. Isotopic labeling experiments have shown our experimental apparatus to have a dynamic range of greater than 500. In addition, the SPS spectrum at the liquid/vapor interface was obtained in a closed cell, which minimized the influence of the environment and improved the S/N ratio of our data.

 

Fig. 7. VSFG spectra of DMSO under SPS polarization conditions at the silica/vapor (black), silica/liquid (red) and liquid/vapor (green) interfaces.

Download Full Size | PDF

In the PPP spectra (Fig. 8), both the symmetric and antisymmetric methyl stretches are observed. The methyl symmetric (antisymmetric) stretch shows up at 2921 cm^-1 (2992 cm^-1), 2901 cm^-1 (2981 cm^-1) and 2909 cm^-1 (2980 cm^-1) at the silica/vapor, silica/liquid and liquid/vapor interfaces, respectively. Features observed in both SSP and SPS spectra appear in the PPP spectra, because the PPP spectra contain contributions from χ ⁽²⁾ _xxz, χ ⁽²⁾ _zzz χ ⁽²⁾ _xzx and χ ⁽²⁾ _zxx. In the co-propagating geometry, the last two contributions generally cancel, whereas in the counter-propagating geometry the cancellation is not complete and these contributions do appear. Changes in spectral frequencies at different interfaces indicate that the methyl stretches are sensitive to their local environment. Additionally, in all cases we observe a stronger VSFG signal at the silica/vapor interface than at the other interfaces, an effect that arises primarily from differences in Fresnel factors.

The DMSO SPS signal is considerably weaker than the SSP signal for all of the interfaces studied here. Part of this decrease in intensity is due to differences in the nonlinear Fresnel factors (see Eqs. (3) and (4)) and in the overall spectroscopic strengths of the symmetric and antisymmetric methyl stretches. As has been shown previously [12], methyl group rotation can also lead to a significant decrease in the VSFG signal from an antisymmetric methyl stretch, which in the case of DMSO is the only mode seen in the SPS spectrum.

 

Fig. 8. VSFG spectra of DMSO under PPP polarization conditions at the silica/vapor (black), silica/liquid (red) and liquid/vapor (green) interfaces.

Download Full Size | PDF

We can also glean some information on molecular orientation from the spectra in Figs. 6–8. SSP spectra are sensitive to the projection of IR transition dipoles along the interface normal, whereas SPS spectra are sensitive to the projection parallel to the surface. The symmetric methyl stretch appears in the SSP spectrum but not the SPS spectrum, and the converse is true for the antisymmetric methyl stretch. These observations indicate that the three-fold axis of the methyl groups has a large average projection along the interface normal for all three types of interface. The smaller ratio of the signal for the symmetric methyl stretch compared to the antisymmetric methyl stretch in the PPP spectrum at the silica/vapor interface is indicative of somewhat different structuring in this system as compared to the silica/liquid and liquid/vapor interfaces. We will explore the structure of DMSO at these interfaces in more detail in a future publication.

It is important to note that for all of the VSFG spectra in Figs. 6–8, the nonresonant background is quite low. This feature is another advantage of the counter-propagating geometry. For a primary dielectric interface, the nonresonant background arises predominantly from the electric quadrupole response from the volume within less than a wavelength from the interface. In such a case, the nonresonant background is generally small. However, for a buried dielectric interface nonresonant background is generated in any region of a bulk medium in which the IR and visible beams propagate together. As can be seen from Fig. 4, the overlap of these beams in the bulk medium is considerably smaller in a counter-propagating geometry than in a co-propagating geometry. This reduction in nonresonant background is one of the factors responsible for our ability to observe the anti-symmetric methyl stretch of DMSO. We note that for studies involving liquids and/or vapors, VSFG spectroscopy should generally be performed in a closed system to avoid chemical exposure and contamination of the sample. This goal is most easily accomplished in a sample cell, which requires propagating the fields through a dielectric window to a buried interface.

To demonstrate the magnitude of the difference in the nonresonant background between the two geometries, in Fig. 9 we show SSP data for the DMSO silica/vapor interface obtained with two different broadband VSFG spectrometers. The black trace is from our counter-propagating system, and the red trace from a co-propagating system. The spectrum from the co-propagating system is narrower than the spectrum from the counter-propagating system because 800 nm light with a smaller bandwidth was used in the former case. However, it is clear from these spectra that the non-resonant background in the co-propagating case is considerably larger than that in the counter-propagating case.

 

Fig. 9. Counter-propagating (black) and co-propagating (red) VSFG spectra of DMSO and the silica/vapor interface. The bandwidth of the visible light used for the co-propagating spectrum was narrower than that for the counter-propagating spectrum. The nonresonant background is considerably smaller in the counter-propagating spectrum.

Download Full Size | PDF

6. Discussion

The difficulty in achieving alignment and finding the proper timing between the IR and visible pulses has been an impediment to the use of broadband VSFG spectroscopy in a counter-propagating geometry. The methods presented here simplify the alignment and timing procedures, making the counter-propagating geometry a viable alternative for broadband VSFG.

The co-propagating geometry is comparatively simple to align, and features relatively small dispersion in the signal. In addition, there is nearly complete cancellation of the χ ⁽²⁾ _xzx and χ ⁽²⁾ _zxx contributions under PPP polarization conditions in this geometry, simplifying the PPP spectrum and offering advantages in the determination of molecular orientation [5]. Counterbalancing these advantages is the fact that the signal beam deviation is relatively large in this geometry upon changing samples or interfaces, and the non-resonant signal from buried dielectric interfaces can be large leading to asymmetric lineshapes that can be difficult to fit and interpret.

In the counter-propagating geometry, the large angular separation of the signal from the input beams and their reflection helps to contribute to a high dynamic range. Because the input beams do not co-propagate, the non-resonant background is reduced substantially in this geometry at buried dielectric interfaces. The minimal signal beam deviation in this geometry makes changing samples simpler as well. On the flip side, the counter-propagating geometry is more challenging to align, although the methods introduced here alleviate many of the difficulties. There is also more dispersion in the signal beam than in the co-propagating geometry. Furthermore, because the signal propagates in a significantly different direction from the input beams, in general the χ ⁽²⁾ _xzx and χ ⁽²⁾ _zxx contributions to the PPP spectrum no longer cancel, leading to a more complex spectrum than in the co-propagating geometry. The additional features that appear in PPP spectra using a counter-propagating geometry can provide assistance in assigning transitions and calculating functional group orientations.

The co-propagating and counter-propagating geometries have complementary strengths and weaknesses for broadband VSFG. The geometry that is preferred will depend upon the application, but the ability to implement broadband VSFG spectroscopy in a relatively simple manner provides another tool in the arsenal for studying interfacial structure.

7. Conclusions

In conclusion, we have constructed a broadband, counter-propagating VSFG spectrometer based on a Ti:sapphire regenerative amplifier. The data presented here demonstrate that the counter-propagating geometry has a number of significant advantages for broadband VSFG spectroscopy, such as high dynamic range, more space in the sample area, reduced non-resonant background, and ease of changing samples. We have presented a simple protocol for finding timing and aligning the spectrometer, and have demonstrated the sensitivity of this apparatus on different interfaces of neat DMSO.