Heterodyne-detected sum frequency generation of water at surfaces with varying hydrophobicity

Aqueous interfaces are ubiquitously involved in natural and technological processes. The structure and function of biomolecules depend on their interactions with water.^1–6 Technological advancements, such as proton transport in fuel cells and membrane filtration, depend on efficient processes at aqueous interfaces.^7–10 When water comes into contact with a surface, the bulk hydrogen-bonding network is terminated and the interfacial water molecules restructure according to the surface chemistry of the interface. Given the complexity of both natural and technical interfaces, obtaining a detailed understanding of the dependence of the structure of interfacial water on the chemical and physical characteristics of the interface is a first step in elucidating the role of water at these interfaces. However, interfacial water is notoriously difficult to study experimentally since the number of water molecules at the interface is miniscule compared to the bulk.

Sum frequency generation (SFG) is a second order nonlinear vibrational spectroscopy, which is forbidden in centrosymmetric media under the dipole approximation. As a result, only the interfacial water results in a SFG signal since the surface breaks the inversion symmetry that exists in bulk water. In recent years, SFG has been used to study interfacial water in a variety of different systems by probing the OH stretch region, as overviewed in several review papers.^11–16 However, the majority of SFG studies use conventional, homodyne-detected SFG, which measures the norm squared of the second-order nonlinear susceptibility (|χ⁽²⁾|²). Although homodyne-detected SFG provides the vibrational spectrum of the interfacial water, the sign information of the nonlinear susceptibility is lost and the spectrum is distorted by interferences between resonances and with the nonresonant background. Heterodyne-detected SFG (HD-SFG) is an interferometric method that allows the complex χ⁽²⁾ to be measured directly. The imaginary part of χ⁽²⁾ (Im χ⁽²⁾) is purely absorptive and is thus analogous to the bulk vibrational spectrum, while the sign of the spectrum provides information on the orientation of molecular species. The use of HD-SFG has aided the understanding of the structure and orientation of water at the air/water and surfactant (or lipid)/water interfaces. At the air/water interface, HD-SFG showed that the hydrogen-bonded water molecules have the opposite orientation than the free OH.^17–22 For lipid/water interfaces, the comparison of homodyne-detected SFG and HD-SFG showed that the apparent red-shift of the hydrogen-bonded water peak for cationic lipids compared to anionic or zwitterionic lipids was skewed due to the real part of χ⁽²⁾, not from stronger hydrogen-bonded water.²³ The sign information from HD-SFG also helped identify weakly hydrogen-bonded water molecules interacting with the carbonyl of the lipid headgroup.²⁴ However, the use of HD-SFG to probe solid/water interfaces is still limited due to the added experimental complexity.^25–27 We recently presented a new experimental geometry, which is robust and flexible and allows for the solid/water interface to be probed nearly as easily as the exposed surface.²⁵ 

The chemical composition of surfaces can be conveniently tuned using self-assembled monolayers (SAMs). Both thiol-based SAMs on gold and silane-based SAMs on silica have been studied with SFG.^28–37 However, silane monolayers are simpler to implement for studying the interaction of water with SAMs since the interface can be accessed through the substrate. Fused silica substrates obstruct the high-frequency side of the OH stretch frequency due to silanol groups. In order to probe the entire OH stretch frequency, an infrared (IR)-grade fused silica window or calcium fluoride window coated with a thin layer of silica need to be used.^32,38 Silica-coated calcium fluoride offers an additional advantage of being transparent in the OD stretch region as well. Hydrophobic monolayers, namely, octadecyltrichlorosilane (OTS), have been studied by a number of groups. The overall structure of the water is similar to the hydrophobic water/air interface, but the spectrum is dependent on experimental geometry and monolayer order.^29,30 However, only a few studies examining water at a hydrophilic or charged monolayer exist.^33–35 By using different silanes, the surface character can be tuned from hydrophilic to hydrophobic. An ideal hydrophilic monolayer would be OH terminated, but the terminal hydroxyl group renders it incompatible with silane based SAMs. In this study, we use a commercially available silane containing the polyethylene glycol (PEG) repeating unit as the hydrophilic monolayer of interest. Monolayers with intermediate surface character can be made by using a silane with a headgroup of intermediate hydrophobicity, such as methoxy terminated,³⁹ or by creating a mixed monolayer of a hydrophilic and a hydrophobic silane.^40,41 Mixed monolayers are heterogeneous and chemically complex and provide a controlled way to approach the natural heterogeneity of biological molecules. In this study, we employ HD-SFG to probe water in contact with hydrophobic, hydrophilic, and mixed monolayers in order to elucidate how surface chemistry changes the hydrogen-bonding strength and structure at the interface.

Samples were prepared on 1 mm thick calcium fluoride windows (Crystran Ltd.) coated with 10 nm of silica. The silica was deposited using an Oxford Instruments ALD FlexAL atomic layer deposition system (110°C, plasma). After deposition of silica, a 150 nm thick gold reference spot with a 5 nm titanium adhesion layer was sputtered on to the substrates with a magnetron argon sputtering deposition system. The order of the silica and gold reference is important for achieving consistent interference with the local oscillator between the gold and the sample. The substrates were annealed at 800 °C in a high-vacuum furnace before the synthesis of self-assembled monolayers.

Prior to self-assembled monolayer deposition, all glassware was cleaned for at least 30 min in a Nochromix solution prepared by dissolving Nochromix cleaning reagent (Godax Laboratories) in concentrated sulfuric acid (Fisher Scientific, certified ACS Plus) according to directions on the Nochromix packaging. All glassware was washed to neutral pH with ultrapure water (18.2 MΩ·cm resistivity at 25 °C, 5 ppb TOC) generated with a Milli-Q Advantage A10 system (EMD Millipore) and dried in an oven at 150 °C for an hour. The hydrophobic monolayer was synthesized according to a previously published procedure.³² Briefly, 15 µl of octadecyltrichlorosilane (OTS, ≥95%, Sigma-Aldrich) was added to 15 ml of hexanes (HPLC grade, Fisher Scientific). The substrates were submerged for 1.5 h, then washed with hexanes, and sonicated for 5 min in hexanes and 5 min in Milli-Q water. The OTS monolayer was dried with nitrogen and stored dry. A silane containing the PEG repeating unit, 2-[methoxy(polyethyleneoxy)_6-9propyl]trichlorosilane (Gelest), was used for the synthesis of the hydrophilic and mixed monolayers studied. The deposition of the hydrophilic and mixed monolayers was adapted from previously published procedures.^40,41 The hydrophilic monolayer, referred to as PEG, was synthesized by adding 1.5 ml of 2-[methoxy(polyethyleneoxy)_6-9propyl]trichlorosilane to 15 ml of toluene (HPLC grade) and submerging the substrates for 24 h. After 24 h, the substrates were washed with toluene and sonicated for 5 min in toluene and 5 min in Milli-Q water. The PEG monolayer was dried with nitrogen and stored in toluene. The mixed monolayer, referred to as OTS/PEG, was synthesized in a two-step process due to the large difference in deposition time for OTS and PEG. First, the substrate was submerged in a solution of 15 µl of OTS in 15 ml of toluene for 0.5 h. The substrate was rinsed with toluene and placed in a solution of 1.5 ml of PEG in 15 ml of toluene for 24 h. Then, it was washed with toluene and sonicated for 5 min in toluene and 5 min in Milli-Q water. Finally, the OTS/PEG monolayer was dried with nitrogen and stored in toluene. In depositing the mixed monolayer, we aimed to create a monolayer with both OTS and PEG character. In addition to the mixed monolayer conditions used in this study, two other synthesis conditions were tested with contact angle goniometry. If OTS and PEG were deposited in a single step, the contact angles were very close to pure PEG for most spots. In a two-step procedure where OTS was deposited for 1 h instead of half an hour, the average contact angle was only slightly lower than pure OTS. The two-step monolayer deposition process with only a half an hour deposition of OTS resulted in a contact angle that is approximately half way between the pure OTS and pure PEG monolayers and was thus chosen for the SFG studies. However, our experiments do not provide a way of measuring the monolayer composition precisely. Depositing the OTS monolayer from a solution in toluene was also investigated but led to more disordered monolayers. Accordingly, hexanes were used for the synthesis of OTS monolayers instead of maintaining a consistent solvent for all three monolayers.

The macroscopic surface character of the SAMs was determined using a home-built contact angle goniometer. A collimated white light source (tungsten-halogen lamp, ThorLabs QTH10) illuminates a stage, and a microscope objective (5×, Zeiss) focuses the image onto the camera (1280 × 1024 monochrome CMOS camera, ThorLabs DCC1545M). A neutral density filter (ND20, ThorLabs) and a blue dichroic filter (FD1B, ThorLabs) are placed between the microscope objective and camera to reduce the intensity of light into the sensor and increase the image resolution on the monochrome CMOS sensor by reducing the effect of chromatic aberration, respectively. The sample stage was cleaned with acetone before placing a substrate on the stage. Using a 25 µl syringe, 3 µl of Milli-Q water is deposited on the substrate. An image of the droplet is collected using UC480 Camera Manager and the contact angle is analyzed using the DropSnake plugin in ImageJ. Three droplets (six angles total) were analyzed for each monolayer to determine the average contact angle and distribution.

The heterodyne-detected sum-frequency generation spectrometer has been described in detail elsewhere.²⁵ Briefly, a Ti:sapphire amplifier (Coherent Legend Elite Duo) seeded by a Ti:sapphire oscillator (Coherent Micra-5) generates 25 fs, 800 nm, 5 mJ pulses at a 1 kHz repetition rate. The visible, upconversion pulse is generated by filtering 1 mJ of the output with a Fabry–Pérot etalon (TecOptics, Inc.). The resonant IR beam is generated by converting 3 mJ of 800 nm light into tunable broadband IR pulses in a commercial optical parametric amplifier (Coherent OPerA Solo). The full-width-half-maximum bandwidth of the IR pulses generated with the optical parametric amplifier (OPA) is approximately 250 cm⁻¹. Henceforth, four OPA positions were used to cover the entire CH and OH stretch regions. Before the sample, the visible (central wavelength 792.8 nm, bandwidth 0.6 nm, pulse energy 10 µJ) and IR (central wavelength between ∼3000 cm⁻¹ and ∼3600 cm⁻¹ and pulse energy 5 µJ) are focused into a 150 nm thin film of ZnO sputtered onto a 1 mm thick CaF₂ window to generate the local oscillator. The three beams (SFG_LO, visible, and IR) are collimated with a 90° off-axis parabolic mirror (ThorLabs, MPD269-P01), and the SFG_LO is delayed in time with respect to the visible and IR beams by a 2 mm thick CaF₂ window placed in the collimated region. The beams are refocused with a 60° off-axis parabolic mirror (PIKE Technologies, 300-1246-51) onto the sample to generate SFG_reference or SFG_sample. The spot size of the IR beam is ∼450 µm at the sample position, and the visible beam is slightly larger to ensure upconversion of all the excited vibrations. The sample cell consists of a polytetrafluoroethylene (PTFE) coated O-ring sandwiched between the sample and a PTFE back plate with two holes, as illustrated in Fig. 1. All the SFG data presented here are from the backside solid-air and solid-water interfaces. To go from dry to wet monolayers, water was injected into the cell through PTFE tubing stuck in the holes of the PTFE back plate. The SFG signals are focused onto the slit of a polychromator (Princeton Instruments, Acton SP2500) where they are dispersed by a diffraction grating (600 grooves/mm, blazed at 500 nm), and the resulting signal was imaged on a liquid nitrogen cooled CCD camera (Princeton Instruments, model 7509-0001, 1340 × 400 pixels). To collect “homodyne-detected” data, the LO was blocked in the collimated region.

The phasing procedure for our HD-SFG experiments has been previously been described in detail.²⁵ A gold spot is deposited on the silica coated CaF₂ in the same plane as the SAM monolayer and is used as an initial nonresonant reference for phasing our HD-SFG spectra. A 15 nm thin adhesion layer of titanium is used to fasten the gold. Prior to performing the heterodyne analysis, the data from each OPA position were summed together, as seen in Fig. S4. No phase drift between OPA positions was observed. Since the absolute phase of gold can change from sample to sample, the relative phase between the gold reference and the sample was adjusted after the initial phasing by a few tens of degrees such that the imaginary spectrum is zero where no molecular absorbances exist (i.e., the regions below the CH stretches and above the OH stretching frequencies). This procedure has previously been used successfully.^27,42 The method only assumes that the phase from gold is constant across the spectrum but does not rely on knowing the absolute phase of gold and corrects for small differences in the sample and reference position. We estimate our error in our phase to be better than ±20°. Further details on the phasing procedure used are described in the supplementary material.

Contact angle goniometry and homodyne-detected SFG of the CH stretches were performed in order to characterize the chemical surface character and order of the synthesized SAMs. The contact angle is determined by the differences in the surface tension between the water/solid, water/air, and solid/air interfaces.⁴³ A larger contact angle is indicative of a more hydrophobic surface. Contact angles can range from 0° (superhydrophilic, no drop) to 180° (superhydrophobic, the drop only touches the solid at a single point) with 90° being the turning point between hydrophilic and hydrophobic. By measuring multiple water droplets for each sample, the overall hydrophilicity/hydrophobicity and the uniformity of the SAMs using the average contact angle and standard deviation were determined. Representative contact angles for each of the monolayers are displayed in Fig. 2. For the OTS monolayer synthesized in hexanes, the contact angle was 114.7° ± 2.4°, indicating that the surface was hydrophobic and uniform. When the OTS monolayer was deposited in toluene instead, the contact angle was 102.9°. The lower contact angle corresponds to a less hydrophobic monolayer, which is one reason why we focused on the OTS monolayer deposited from hexanes in this study. The pure PEG monolayer was determined to be hydrophilic and uniform with a contact angle of 38.5° ± 2.9°. This PEG monolayer is less hydrophilic than freshly cleaned silica, which has a contact angle near 0°, due to the less hydrophilic nature of the PEG repeating unit compared to the silanol groups. As expected, the mixed OTS/PEG monolayer exhibited a contact angle between the two pure monolayers, 80.5° ± 10.7°, but closer to the pure hydrophobic monolayer. The standard deviation for OTS/PEG is significantly larger than the pure monolayers, indicating that the surface is not uniform, some regions have more OTS character, and some regions have more PEG character. In depositing the mixed monolayer, we aimed to create a monolayer with both OTS and PEG character. In addition to the mixed monolayer conditions used in this study, two other synthesis conditions were tested with contact angle goniometry. If OTS and PEG were deposited in a single step, the contact angles were very close to pure PEG for most spots. In a two-step procedure where OTS was deposited for 1 h instead of half an hour, the average contact angle was only slightly lower than pure OTS. Of the three synthesis conditions tried, we chose the mixed monolayer that was closest to 50/50 as determined by the contact angle. However, our experiments do not provide a way of measuring the monolayer composition precisely.

While the contact angles reflect macroscopic interactions, the SFG spectroscopic measurements probe microscopic interactions. In order to study the microscopic order of the SAMs and how it relates to the macroscopic order measured with the contact angles, we probed the CH stretches of the dry monolayers with SFG. Due to the symmetry requirements of SFG, it is highly sensitive to the order of the monolayers.^30,32 Each of the monolayers exhibits two types of CH groups, the terminal methyl groups and the methylene groups in the chains. For a well ordered monolayer, every other methylene group points in opposite directions and destructively interferes, producing no overall SFG response from the methylene groups. Conversely, the terminal methyl groups point in the same direction and result in a strong signal. However, the introduction of gauche defects in the SAMs disrupts the perfect destructive interference of the methylene groups. Previously, our group studied the order of mixed length hydrophobic monolayers through the appearance of SFG signals from the methylene groups.³² Here, we use the relative intensity of methylene to methyl SFG signal to qualitatively investigate the order of the different SAMs. The dry OTS monolayer exhibits small amounts of methylene character (Fig. 2). The two main peaks result from methyl groups (the symmetric methyl stretch and the symmetric methyl stretch + Fermi resonance), and the shoulders result from methylene groups not canceling perfectly. Accordingly, the dry SFG spectrum of OTS indicates that the monolayer is well ordered but contains a few gauche defects. Unlike OTS, the PEG monolayer has significant methylene character and is consequently significantly more disordered. This is largely due to the variable chain length of the commercial silane (6–9 PEG units), which allows for longer chains to bend over shorter chains. While van der Waals interactions are responsible for keeping the chains ordered and extended, if a long chain is deposited next to a short chain, the end of the long chain is free to bend and introduce methylene character into the SFG spectrum. Even though there is disorder as a result of the variable chain length and the precise ratio of chain lengths is unknown, the SFG spectrum of the dry PEG monolayer is similar at multiple locations on the sample, as displayed in Fig. S1 of the supplementary material, which confirms that the sample is uniform as predicted by contact angle measurements. The gauche defects from the mismatched chain lengths also exposes the oxygen atoms in the PEG chain, improving the hydrophilicity of the monolayer. The mixed OTS/PEG monolayer has the largest methylene peak relative to the methyl peak and is accordingly the most disordered monolayer. This is a result of the inherent disorder of the PEG monolayer with different chain lengths, as well as disorder and defects created by mixing OTS and PEG. As expected from the large standard deviation of contact angles, the mixed OTS/PEG monolayer is not uniform as indicated by different CH stretch spectra at different regions of the sample, as shown in Fig. S2. Our SFG spectrometer does not provide enough spatial resolution to specially resolve the composition and patterning of the sample region. The sample position resulting in the spectrum displayed in Fig. 2 was chosen for studying the interaction of water with the mixed monolayer since it had the largest methyl peaks confirming the presence of hydrophobic chains. Furthermore, the methylene peaks at this spot line up with the methylene peaks of pure OTS and pure PEG, and accordingly, both hydrophobic and hydrophilic moieties exist within the probe area.

Upon the addition of Milli-Q water, we first examined how the CH stretch region changed after interacting with water using homodyne-detected SFG. The hydrophobic, OTS monolayer does not strongly interact with the water, and the CH stretch region retains the two strong peaks resulting from the terminal methyl groups. However, the PEG containing monolayers change in shape because the water can strongly interact with the oxygen atoms within the PEG chains. The PEG containing monolayers also exhibit a relative decrease in intensity, which is most dramatic in the pure PEG monolayer since the entire monolayer can hydrogen bond with the water. However, the mixed OTS/PEG monolayer maintains some of its structure due to the hydrophobic OTS chains. The examination of the CH stretch region is useful in comparing the relative interactions and flexibility of the monolayers; however, the OH stretching region provides a more direct comparison of the strength and orientation of water molecules at the various monolayers.

The OH stretch region in both homodyne-detected and HD-SFG spectroscopy is shown in Fig. 3. The OH region of water in contact with all three monolayers exhibits the double peaked structure, which has been seen in the SFG spectra of water at a variety of interfaces and is caused by the intramolecular and intermolecular coupling between the OH groups.^11–16 However, the relative intensity of the two peaks differs between the samples with the PEG containing monolayers producing a stronger relative water signal. The OTS spectrum additionally contains the free OH at ∼3680 cm⁻¹ that is characteristic of water in contact with the hydrophobic surface. Water in contact with OTS has been studied by a number of groups.^30,44–46 Overall, the present homodyne detected spectrum of OTS (Fig. 3) is comparable to the spectra in the literature. The free OH peak, predominant methyl stretches, and large contact angle confirm the hydrophobicity of our OTS monolayer. Differences in the hydrogen-bonded region of water at OTS likely result from differences in monolayer order and experimental geometry.^28,30 Tyrode and co-workers explored the relation between the intensity of the hydrogen-bonded OH stretch region of water and the monolayer order, and suggested that the hydrogen-bonded OH stretching intensity is dominated by direct silica-water interactions from water in cracks in the monolayer. Previous atomic force microscopy measurements have quantified the amount of cracks in OTS monolayers to be only a few percent of the surface area.^32,46 If water molecules inside defects in the monolayer covering a few percent of the surface area would dominate the hydrogen-bonded water signal, they would have to be highly aligned, which seems unlikely. Furthermore, the fact that we observe spectral differences between the three different monolayers directly shows that the hydrogen-bonded region is not solely due to water molecules in contact with silica. While it is difficult to estimate the contribution of such trapped water molecules, the study by Tyrode and co-workers clearly shows that the hydrogen-bonded region can vary significantly depending on the monolayer order, which presents some uncertainty in quantifying the spectrum of monolayer-water interfaces.

Since homodyne-detected SFG measures the norm squared of χ⁽²⁾, which results the complex mixing of the real and imaginary components, we performed HD-SFG of water at each interface to extract the purely absorptive imaginary component of χ⁽²⁾. Although the general peak structure of the hydrogen-bonded OH stretching region is maintained, there is a red shift of about 50 cm⁻¹ compared to the homodyne-detected spectrum (Fig. 3). This frequency shift is consistent between the three monolayers, which were independently phased, and remains significant within our phase accuracy of ±20° (Fig. S6) confirming that the frequency shift is not an artifact of phasing. This red shift of the OH stretch peak could result from the real part of the χ⁽²⁾ skewing the spectrum or interferences between peaks of opposing signs and was previously observed in HD-SFG spectra of lipid-water interfaces.²³ Since the 50 cm⁻¹ shift is consistent for all three monolayers, while OTS does not have a negative hydrogen-bonding OH stretch peak, the red shift in the double peaked spectral feature results from the real part of χ⁽²⁾. Often, SFG spectra of water are used to describe the hydrogen-bonding strength of interfacial water since the OH stretching frequency correlates with hydrogen-bonding strength. However, due to the real part of χ⁽²⁾ skewing the spectrum, using homodyne-detected SFG can lead to an inaccurate picture of the hydrogen-bonding strength of interfacial water.

By performing HD-SFG, the sign of the Im(χ⁽²⁾) spectrum is also recovered, which provides information on the relative orientation of the dipoles. This additional sign information furthermore makes it easier to identify weak spectral features. For example, in the homodyne-detected SFG spectrum of OTS (Fig. 3), only two main CH stretching peaks are apparent. However, in the HD-SFG spectrum, the Im(χ⁽²⁾) has two positive peaks and two negative peaks in the CH stretching peaks. These peaks have previously been identified as the CH₃ symmetric stretch (r+), the CH₂ asymmetric stretch (d−), the CH₃ Fermi resonance of the bending overtone with r+ (r + FR), and the CH₃ asymmetric stretch (r−).²⁵ The fit analysis of the CH stretching region of OTS (Table I) is in good agreement with previous analyses of the imaginary spectrum of OTS.^25,47,48 The OTS/PEG and PEG monolayers have similar CH stretching peaks with alternating signs. Even though the peaks corresponding to the terminal methyl symmetric stretching vibrations are positive, the methyl group is pointing down since the hyperpolarizability of the CH stretch is negative.⁴⁹ As expected, the free OH and the methyl stretch in OTS have opposite orientations, indicating the free OH points toward the hydrophobic interface. However, the hydrogen-bonded water and the free OH have the same sign and orientation. This finding agrees with previous measurements of the OTS-water interface^45,50 but differs from the water-air interface, which has been shown to have the free OH in the opposite orientation of the hydrogen-bonded water.^17–22 The hydrogen-bonded region in the imaginary spectrum of water in contact with OTS (Fig. 3) differs to some extent from that measured by Shen and co-workers.⁴⁵ However, both the CH peaks and free OH match well. From homodyne SFG, it is known that the monolayer order has a large effect on the hydrogen-bonded OH stretch response,³⁰ which could potentially explain the discrepancy in the imaginary spectra. Additionally, Shen and co-workers phased their spectrum with quartz, which has since been shown to result in a peak not related to a vibrational resonance at the water-air interface around 3000 cm⁻¹,¹⁷ and problematic as a phase reference for the buried surfaces due to difference in propagation media,²⁶ which could be another possible reason for the discrepancy. Roke and co-workers have also presented an imaginary spectrum of OTS that has the same sign and general structure of the imaginary spectrum of OTS in Fig. 3.⁵⁰ By using the sign information from the HD-SFG spectrum of the OTS, a snapshot of the average water orientation at the interface can be built up, as illustrated in Fig. 4.

The SAMs were assembled on a thin layer of silica. All the experiments described here were performed under neat water with approximately pH 5.6 due to adsorbed CO₂. At this pH, the underlying silica layer would be negatively charged. Studies of water at the bare silica surface have shown a large Eisenthal χ⁽³⁾ effect due to the negative charge of silica at neutral and basic pH values.^26,51–58 Accordingly, a negative charge on the silica under the SAMs would create a static electric field that is likely contributing to a net orientation of the hydrogen-bonded water molecules toward the interface. Previous homodyne-detected SFG studies of the OTS-water interface showed a spectral dependence on the pH and ionic strength, confirming an Eisenthal χ⁽³⁾ effect due to the underlying silica.^44,46 However, the spectral differences between the monolayers show that the chemical interactions between the monolayer and the water cause distinct spectral shapes independent of the charge on the silica. By screening the charge from the silica surface with salt, the overall intensity of the water spectrum would be expected to decrease. In order to determine the full extent of the silica’s effect on the spectrum, further studies varying the pH and ionic strength are needed.

As described above, HD-SFG allowed for the identification of additional CH stretching peaks based on differences in sign. Similarly, distinct spectral features in the OH stretch region of water can be identified and fitted in the spectra for PEG containing monolayers based on the spectral sign. For both the mixed OTS/PEG monolayer and the PEG monolayer, there is a broad negative peak between the CH stretching region and the dominant, positive double-humped OH structure (Fig. 3). Since this strongly hydrogen-bonded water only exists for the PEG containing monolayers, it likely results from water hydrogen bonding with the oxygen atoms in the PEG chains, as illustrated in Fig. 4. With the lack of free OH and the presence of the negative strongly hydrogen-bonded peak, the structure of the interfacial water at the mixed OTS/PEG monolayer is dominated by the PEG chains.

A more detailed comparison of the hydrogen-bonded region can be achieved by fitting the imaginary spectra. Typically, SFG data are fit with Lorentzians since they explicitly contain real and imaginary components,^59–62 where a few studies have used Voigt profiles to describe the inhomogeneous broadening while maintaining explicit real and imaginary components.^63–65 Obtaining the imaginary component of the nonlinear susceptibility allows the absorptive spectrum to be fit by peaks with any line shape function, such as Gaussians, to quantify the spectral features. The details of the fitting can be found in the supplementary material. Briefly, the CH peaks were fit with the imaginary component of a Lorentzian line shape, whereas the water resonances were fit with Gaussian line shapes. The fitted peaks and the total fit compared to the experimental spectra for all three SAMs are shown in Fig. 5. The total fits are in good agreement with the data, as illustrated by the bold, black curves (fits) overlapping the bold, red curves (experiments) well across the entire spectrum. The full set of fit parameters can be found in Table I. Each spectrum was fit with a total of eight or nine total peaks. The number of resonances included in the fit was chosen based on the minimum number that still yielded a reasonable fit, as based on the residuals and guided by previous assignments in the literature. The CH region was fit with 5 Lorentzians corresponding to different CH₃ and CH₂ stretching modes, and the OH stretch region was fit with up to 4 Gaussians for each spectrum.

For the pure OTS monolayer, the observed CH resonances are comparable to those obtained in previous experiments.^25,47,48 Lorentzians 1–5 used to fit the CH stretching modes correspond to d+ (CH₂ symmetric), r+ (CH₃ symmetric), d− (CH₂ asymmetric), r + FR (CH₃ symmetric + Fermi Resonance), and r− (CH₃ asymmetric), respectively. To our knowledge, no prior SFG studies exist for similar PEG monolayers and we do not attempt to further assign the CH resonances here. Qualitatively, the CH resonances for the PEG containing monolayers are similar to OTS and were thus fit to the same CH stretching modes. However, the CH resonances for the pure PEG monolayer generally exhibit broader linewidths compared to the pure OTS monolayer reflecting the larger disorder caused by the distribution of chain lengths. The mixed monolayer could exhibit a number of resonances observed in either pure monolayer. For simplicity, we also fit CH region of the mixed monolayer to 5 Lorentzians since this was sufficient to obtain a reasonable fit. For the mixed monolayer, the linewidths lie in between those for the two pure monolayers. We note that the resolution of the heterodyne-detected experiment is not fine enough to resolve separate peaks for the OTS and PEG components for each CH mode.

The OH stretching region was fit with 4 Gaussians for the pure OTS and 3 Gaussians for OTS/PEG and PEG. These Gaussians are numbered 1 through 5 by increasing frequency, with Gaussian 1 only existing for the PEG containing monolayers and Gaussians 4 and 5 only existing for the pure OTS monolayer. For all three monolayers, Gaussians 2 and 3 are the hydrogen-bonded OH stretch modes that make up the double peaked feature typically observed around 3200 and 3400 cm⁻¹ for H₂O surfaces and result from the intermolecular and intramolecular couplings between the OH modes. Gaussian 4 around 3600 cm⁻¹ is only present in the sample of the pure OTS monolayer and was added to improve the fit. It has a similar frequency to a silica peak recently observed with SFG⁵¹ and likely results from the silica coated CaF₂. This peak is not always present but has been observed on multiple samples. Gaussian 5 is the free OH. For the mixed OTS/PEG and pure PEG monolayers, Gaussian 1 of opposite sign as Gaussians 2 and 3 was added to improve the fit. We attribute these features to water molecules that are hydrogen bonded to the PEG chains more strongly than the rest of the water population. Similar peaks with signs opposite to the main hydrogen-bonding peaks have previously been seen at the lipid-water interface.²⁴ Fitting the water resonances with Lorentzians instead of Gaussians led to significantly worse fit, as shown in Fig. S6. The inaccuracy of Lorentzian line shapes reproducing the imaginary spectrum of water highlights that the OH spectrum of water is inhomogeneous broadened and that obtaining the purely absorptive line shape is needed to accurately fit the water spectrum. In order to fit the homodyne-detected spectrum of water, an algorithm to extract the imaginary component, such as the Maximum Entropy Method, or a more complex line shape, such as a Voigt profile, must be used. However, while fitting the imaginary susceptibility allows for quantifying the spectral features, care must be taken in assigning them to specific subpopulations of water given the spectral distortions caused by intermolecular and intramolecular couplings. In order to fully understand the subpopulations of water interacting with monolayers, isotope dilution experiments are needed to remove the intermolecular and intramolecular couplings.

Without assigning the hydrogen-bonded water peaks to specific populations at the interface, the spectral fits allow for quantification of the relative hydrogen-bonding strengths between monolayers by calculating the ratio of the amplitudes of the two peaks that comprise the double-peak structure. The ratios of the stronger hydrogen-bonded peak to the weaker hydrogen-bonded peak are 1.16, 1.5, and 1.67 for OTS, OTS/PEG, and PEG, respectively. This indicates that on average the hydrogen-bonding strength of the interfacial water in contact with PEG is the strongest and that the ratio for the mixed OTS/PEG monolayer is between the two pure monolayers, as expected given its composition, but closer to that of the pure PEG monolayer. Overall, the HD-SFG spectroscopic data show that the molecular hydrogen-bonded structure of water in contact with the mixed monolayer is closer to that of water in contact with the pure PEG monolayer, whereas the macroscopic properties as captured in the contact angle measurement are closer to the pure OTS monolayer.

In this study, we have obtained homodyne-detected and HD-SFG spectra of water in contact with monolayers with tunable chemical functionality. Analyzing the HD-SFG spectra of water in contact with the monolayers revealed a red shift of the hydrogen-bonded water feature for all monolayers, illustrating that the interfacial water is not as strongly hydrogen bonded as previous homodyne SFG spectra have suggested. While the charged silica surface under the monolayers potentially contributed to the net ordering of the interfacial water, the distinct spectral differences between the monolayers indicate that the chemical structure of the SAMs affects the interfacial water orientation and allows for an analysis of the microscopic water structure in contact with the monolayers.

We observe that although the contact angle of the mixed monolayer is closer to the pure hydrophobic monolayer, the HD-SFG spectrum of the mixed monolayer is closer to that of the hydrophilic monolayer. This indicates that the macroscopic chemical properties of the surface are dominated by the hydrophobic parts of the monolayer but that the molecular hydrogen-bonded water structure of the interfacial water is dominated by the hydrophilic parts of the monolayer. The study thus highlights the need for spectroscopic measurements to understand molecular-level structures since these are not easily extracted from macroscopic measurements. However, care must be taken to not assign peaks in the spectra of isotopic pure water to specific subpopulations due to the complicated intermolecular and intramolecular couplings in water, warranting further studies with isotopic dilutions.

See supplementary material for further description of the uniformity of the monolayers, the solvent dependence of the monolayer deposition, the covering of the whole spectral region, comparison of the homodyne and heterodyne data, and the fitting procedure. This material is available free of charge via the Internet.

This work was supported by the National Science Foundation under a NSF CAREER grant (No. CHE-1151079). S.E.S. was supported by a NSF Graduate Research Fellowship. This work made use of the Cornell Center for Materials Research Shared Facilities, which are supported through the NSF MRSEC program (No. DMR-1719875), and the Cornell NanoScale Facility, a member of the National Nanotechnology Coordinated Infrastructure (NNCI) which is supported by the NSF (Grant No. ECCS-1542081).