Experimental Confirmation of H₂O₂ Adsorption at the Water–Air Interface

Abstract

Recent work has reported that hydrogen peroxide is formed at the air–water interface. Given the reduced solvation environment there, this process could give rise to enhanced production of OH from H₂O₂ photolysis at the interface. These considerations give some importance to understanding the adsorption thermochemistry of hydrogen peroxide. Although there are two molecular dynamics studies that provide the adsorption free energy, to date there is no experimental verification that H₂O₂ adsorbs at the air–water interface. Here we use glancing-angle Raman spectroscopy to follow the surface adsorption behavior of this molecule. Using standard states of 1 mol L^–1 for each of the bulk and surface phases yields a ΔG° of −5 kJ mol^–1 at 293 K, comparable to that obtained for DMSO.

Introduction

Hydrogen peroxide plays an important role in atmospheric chemistry, in both gas and aqueous phases.¹ In the aqueous phase, H₂O₂ may act directly as an oxidizing agent (e.g., for S(IV) oxidation);² in both aqueous and gas phases, it acts as a photochemical source for the highly reactive hydroxyl radical, OH. Whereas in gas phase, the quantum yield for the latter process is 2, in aqueous solution it is reduced to about unity³ through the solvent cage effect. If hydrogen peroxide is either formed or accumulates at the air–water interface, the expected smaller cage effect there may result in an enhanced OH production, giving rise to a higher local oxidative potential.

Recent reports from the Zare laboratory⁴ have highlighted the novel possibility of the spontaneous production of hydrogen peroxide, H₂O₂ at the air–aqueous interface. Those authors concluded that the strong electric field present at that interface can promote electrochemical reactions, including the oxidation of hydroxide anions to OH, which may then recombine to yield H₂O₂. A different mechanism has been suggested by Gallo et al.,⁵ in which the uptake to solution and subsequent hydrolysis of trace amounts of gas phase ozone is responsible for H₂O₂ production. Regardless of the specifics of the production mechanism, there is strong experimental evidence^4,5 that under typical tropospheric conditions H₂O₂ will be formed in or at the surface of aqueous aerosol droplets.

To date, there have been no experimental studies of the potential for H₂O₂ to be surface active at the air–aqueous interface. The similar structures and hydrogen bonding abilities of water and hydrogen peroxide suggest that any surface propensity would be weak at best. Indeed, a very recent study by Bredt et al.,⁶ concludes that there is “seamless integration” of hydrogen peroxide into the water structure. Nevertheless, results of Molecular Dynamics (MD) simulations reported in 2004 by Vacha et al.⁷ suggested an adsorption free energy of about −4 kJ mol^–1 for H₂O₂ at the water surface. In a 2010 paper, Nissenson et al.⁸ used experimental measurements coupled with kinetic modeling to determine the possible effect of a reduced surficial solvent cage effect on aqueous OH photoproduction from a number of precursor compounds. They concluded that, although there is no enhancement of the quantum yield for OH production from H₂O₂ photolysis at the aqueous surface, the enhancement of that molecule’s concentration there implied by the MD results⁷ gives rise to a significant increase in the OH production rate. More recently, more powerful QM/MM simulations⁹ have predicted a somewhat higher adsorption free energy, of about −6.6 kJ mol^–1, for H₂O₂ at the water surface. By coupling those simulations with high-quality quantum chemical calculations of the H₂O₂ absorption spectrum at the water interface, Ruiz-Lopez et al.¹⁰ concluded that the photolysis rate of hydrogen peroxide at the surface is similar to that in the gas phase, but that the enhanced water surface concentration of H₂O₂ increases the OH production rate by orders of magnitude there.

The present work aims to establish experimentally whether hydrogen peroxide shows any partitioning behavior toward the air–water interface. The chemical and spectroscopic similarity of H₂O₂ and H₂O make this determination experimentally challenging. Here we use glancing-angle Raman spectroscopy (GARS) to track the intensity of the peroxide O–O stretching mode in the water surface region and in the bulk, as a function of peroxide aqueous concentration. We show that H₂O₂ does indeed show weak surface activity and compare its adsorption to that of dimethyl sulfoxide (DMSO), whose surface adsorption behavior has previously been reported.

Experimental Section

The glancing-angle Raman setup has been fully described in earlier works.^11,12 In the present study, the horizontally polarized 532 nm output of a pulsed, frequency-doubled Nd:YAG laser operating at 10 Hz impinged on the surface of a liquid sample at a glancing angle (∼85° to surface normal). Typical pulse energies were 2.4 mJ in each approximately 5 ns pulse. Scattered light was collected using a 7 mm diameter liquid light guide, held ∼5 mm above the liquid surface at the point where the laser impinged, and directed through a 532 nm notch filter to the entrance slit of a 1/4 m monochromator. Signal was detected using a photomultiplier tube, whose output was monitored using a digital oscilloscope. The integrated intensity in a 30 ns time slice that captured the scattered signal was averaged over 16 laser shots and sent to a lab computer that averaged 10 such inputs for each output point. Spectra were recorded by manually tuning the monochromator over steps of 0.2–0.5 nm and collecting 3–10 output intensity points at each wavelength. Because the spectral resolution was instrument-limited, peak intensities were determined by averaging the signal from 3 to 5 such output points around the maximum of the Raman peak and subtracting the signal from a corresponding number of points from a nearby region with no scattering.

Samples were made up volumetrically, using neat DMSO (Sigma-Aldrich ≥99.0%) and 30% w/w H₂O₂ (Sigma-Aldrich), diluted in ultrapure deionized water. Glancing angle experiments used a ∼ 30 mL Pyrex dish to hold each sample; bulk measurements were made using vertically polarized laser light passing through a 1 cm path length quartz cuvette. In those experiments, scattered light was collected by the liquid light guide oriented at 90° to the laser axis. Adsorption isotherms for DMSO and H₂O₂ in room temperature aqueous solution were obtained by plotting Raman intensities—as measured above—as a function of bulk solution concentration.

Results and Discussion

As outlined in our earlier work,¹² the Raman spectrum of DMSO shows a feature near 700 cm^–1, assigned to the C–S–C bending mode. To measure H₂O₂, we use the O–O stretching mode_, appearing near 860 cm^–1, as displayed for the bulk and the surface region in Figure 1. The intensities of the C–S–C bend and O–O stretching bands were used to construct surface adsorption isotherms for DMSO and H₂O₂, respectively.

Figure 1.

Raman spectrum of the O–O stretch band of an aqueous 30% w/w solution of H₂O₂, measured in a cuvette (left frame) and at the aqueous surface (right frame) as described in the text. The lines are meant only to guide the eye; data points are included to display the difference in S/N between bulk and surface spectra.

In Figure 3 of Wren and Donaldson,¹² we compared bulk and glancing-angle Raman intensities of the C–S–C bending mode of DMSO, as a function of DMSO mole fraction up to X_DMSO = 0.23. Here, those measurements are repeated and extended to include higher DMSO concentrations, up to the neat liquid. The present glancing-angle measurements are shown together with our earlier ones as a function of bulk molar concentration in Figure 2. DMSO is known to be surface-active;¹³⁻¹⁵ the approach to saturation in the Raman intensity with increasing concentration is consistent with adsorption of DMSO at the air–water interface.

Figure 2.

Raman intensity vs bulk solution concentration of DMSO, measured in the surface region using the glancing-angle Raman setup. Data from 4 separate experiments (black symbols) are combined with data from Wren and Donaldson²¹ (red symbols), all normalized to the intensity near 7 mol L^–1. The line shows a Langmuir isotherm fit.

The dependence of the H₂O₂ Raman intensity on molar concentration is plotted in Figure 3 for the bulk (black circles) and glancing-angle (red triangles) experiments, normalized to the intensity of each at 1.9 mol L^–1. Error bars indicate the standard deviation of the average obtained over 2 (bulk) and 4–6 (surface region) measurements. It is evident that the bulk results show excellent linearity of intensity with concentration, as has been reported previously.¹⁶ By contrast, the glancing-angle result also shows a clear approach to saturation of the Raman intensity as the bulk concentration increases; as with DMSO, this is suggestive of partitioning of H₂O₂ to the surface region. The dashed red line in Figure 3 shows a fit to a hyperbolic function,

1

where A represents the value of I at saturation and B is related to how quickly saturation is achieved. The dashed line in Figure 2 shows a fit of the DMSO data to this function. Such a function describes a Langmuir adsorption isotherm.

Figure 3.

Raman intensity vs bulk solution concentration of the O–O stretching mode in aqueous solutions of H₂O₂, for bulk and surface regions. The bulk and surface results are normalized to their values at 1.9 mol L^–1. Error bars indicate the standard deviation of 4–6 measurements for the surface region and 2 measurements for the bulk.

Thermochemical Analysis

It is common to fit water surface adsorption results to a Langmuir isotherm;¹⁷ some germane examples include Dabkowski et al.¹³ and Allen et al.,¹⁵ who each fitted DMSO surface excess to a Langmuir (or modified Langmuir) isotherm, and Karpovitch and Ray,¹⁴ who did the same with surface second harmonic derived results. Similarly, we have also reported such adsorption isotherms, derived from surface tension,¹⁸ surface second harmonic generation,¹⁹ glancing-angle laser-induced fluorescence,²⁰ and glancing angle Raman.¹²

If we assume that the fits of our GARS intensities to a Langmuir isotherm (which is defined kinetically) describes adsorption of DMSO or H₂O₂ to the water surface, then A in eq 1 is related to the saturated surface concentration and B represents the ratio of rate constants for surface desorption to adsorption, B = k_des/k_ads. This is often interpreted as the inverse of a phenomenological “kinetic equilibrium constant”, K_Lang = 1/B, and allows for a comparison with other, similarly determined values. However, a problem arises in extracting thermochemical parameters by equating B to a thermodynamic equilibrium constant (i.e., setting B = 1/K_ads), since K_ads is a thermodynamic quantity

2

where γ_i is the activity coefficient of the adsorbing species at the surface or in the bulk and ΔG°_ads relates the free energy change between the 2 standard states represented by c°_i. The connection between the thermodynamic equilibrium constant and a ratio of forward to reverse rate coefficients is straightforward for a homogeneous reaction (all in a single phase), where all activities relate to the same standard state (often implicit!). In order to make the connection between the kinetically defined Langmuir adsorption fit parameters and thermochemical quantities, these standard states must be well-defined.

In Donaldson et al.,²¹ we discuss this issue at length, and we show how the connection can be made. In that work, for the case of gas–surface adsorption the appropriate relationship between the “kinetic” Langmuir adsorption constant and the true thermochemical adsorption equilibrium constant is derived to be

3

Here p° and π⁰ represent the standard state pressures of the gas phase and surface, respectively, and N^max gives the maximum surface coverage (a “monolayer” of adsorbate). Although the choice of standard states is, in principle, arbitrary, in order to relate derived thermochemical parameters to published values, p° is often taken to be 1 atm (or 1 bar) and the surface standard state of Kemball and Rideal²² is chosen. The Kemball and Rideal standard state imagines a surface layer of gas having 6 Å thickness, with a density equivalent to 1 atm pressure. Numerically, it is π° = 0.06084 dyn cm^–1, equivalent to a surface coverage of (4.41 × 10¹⁴/T) molecules cm^–2.

In the present case, we are interested in understanding the surface partitioning of a soluble solute from solution. Although again the choice of standard states is arbitrary, it makes sense to choose an ideal solution of 1 mol L^–1 for the bulk solution, and by analogy to Kemball and Rideal,²² a surface standard state corresponding to an ideal 1 mol L^–1 solution in a surface layer of 5 Å thickness. This choice corresponds to a surface density of 3.01 × 10¹³ molecules cm^–2. For this, eq 3 becomes

4

from which ΔG°_ads = −RT ln K_ads may be obtained. Surface activity coefficients are not readily available, but it seems reasonable to assume that their concentration dependence at the interface is similar to that in the bulk. In the following, we assume the ratio of activity coefficients to be unity. The value of N^max (related to A in eq 1) is sometimes available experimentally, from fits to measured surface excess vs concentration plots. In the absence of this value, in the present case we will assume that the value of A from a Langmuir fit is directly proportional to the surface density of the pure liquid DMSO or H₂O₂.

An alternative (and equivalent) approach when activity coefficients and N^max are not experimentally available involves using the analysis presented in Donaldson²³ to obtain an “ideal solution” value for ΔG_ads, then applying a standard state correction to obtain ΔG°_ads. From ref (23),

5

is determined graphically to obtain ΔG_ads, and then a correction term of −RT ln{N^max/c°_surf} is added, where {I_surfN^max/A} = c°_surf, and N^max is assumed to be the surface density of the pure liquid solute. As stressed in Donaldson et al.,²¹ regardless of how it is obtained, the value of ΔG°_ads is highly dependent on the specific choices of the surface and solution standard states. This is illustrated below.

Thermochemistry

a. DMSO

Because there are no previous experimental reports of H₂O₂ adsorption thermochemistry, we begin by applying the tools outlined above to the DMSO GARS results, in order to determine how well the latter reproduce previously reported thermochemistry. There are three previous experimental studies that have reported adsorption free energies for DMSO at the air–water interface, so we shall use these to test whether isotherms derived from glancing-angle Raman results can yield comparable values. Dabkowski et al.¹³ report a value of K_Lang (although they do not identify it thus) of 1 L mol^–1, with N^max = 1.3 × 10¹⁴ molecules cm^–2, based on measurements of surface tension in DMSO–water solutions up to 3 mol L^–1. Applying the transformation in eq 4, this corresponds to a value of ΔG°_ads of −3.5 kJ mol^–1 for the 1 mol L^–1 ideal solution choice of standard states given above.

Although Allen et al.¹⁵ show an SFG-derived adsorption isotherm, they rely on data from surface tension measurements to extract an adsorption free energy. Those authors calculated activity coefficients and applied these to the bulk (but not surface!) concentrations to convert mole fractions to activities and reported ΔG°_ads – 19.8 kJ mol^–1. These considerations make a direct comparison to other work (that did not use bulk phase activities) difficult. As an estimate, we replot their adsorption isotherm (Figure 4 of ref (15)), converting the bulk concentration from activity to molar units, and fit the data up to ∼10 mol L^–1 to eq 1 From this fit, we obtain K_Lang ≃ 1, the same value as Dabkowski et al.,¹³ thus also giving a ΔG°_ads ≃ −3.5 kJ mol^–1.

The scatter in the glancing-angle results for DMSO solutions displayed in Figure 2 makes a direct fit to a Langmuir function poorly constrained. Consequently, we fit the data to a linearized form of eq 1:

1a

Here B and A have the same meaning as in eq 1. This is shown in Figure 4a. From the linear fit to these data, a value of K_Lang = (0.30 ± 0.15) L mol^–1 is obtained from 1/B. Because the glancing-angle Raman intensity data includes neat DMSO as the asymptote, we take N^max = 4.2 × 10¹⁴ molecules cm^–2, corresponding to the density of neat DMSO. With these choices, application of eq 3 at a temperature of 293 K yields a value for ΔG°_ads of – (3.5 ± 1) kJ mol^–1, equivalent to that arising from surface tension measurements.¹³ Using the alternative approach,²³Figure 4b displays a fit of the glancing-angle Raman data to eq 4. The intercept of this plot gives ΔG_ads = +(3.7 ± 0.4) kJ mol^–1; after adding the correction term, we obtain ΔG°_ads = −(2.7 ± 0.3) kJ mol^–1. These results are presented together in Table 1.

Figure 4.

Analysis of DMSO results from GARS. (a) Glancing-angle data plotted and fit using a linearized form of the Langmuir isotherm equation, eq 1a. From the slope and intercept, the kinetic equilibrium constant is obtained. (b) Plot of the data, fit using eq 4. The intercept yields a free energy of adsorption, which may then be corrected to obtain the standard free energy. See text for details.

Table 1. DMSO Thermochemistry Results with 1 mol L^–1 Standard State.

	ΔG° (kJ/mol)at 293 K	notes
DMSO	–3.55	Dabkowski et al.,13 surface tension derived isotherm
	– 3.5	Allen et al.,15 surface tension derived isotherm (reanalyzed from their plot)
	– 7.4	Karpovitch and Ray,14 SHG-derived isotherm (reanalyzed from their plot)
	–(3.5 ± 1.5)	present work, eq 1a
	–(2.7 ± 0.4)	present work, eq 4

It is clear that with the “ideal 1 mol L^–1” choice of standard states for both surface and solution, the extracted value of ΔG°_ads for DMSO to the water–air interface is reasonably consistent between glancing-angle Raman and surface tension derived isotherms. In the case of fully soluble and miscible solutes, both techniques explore probe depths below the uppermost 1–2 surface molecular layers.^11,17 Techniques that are truly surface specific, such as surface second harmonic generation (SHG) and sum-frequency generation (SFG—see, e.g., ref (24) and citations therein), have also been used to quantify the partitioning of DMSO to the water–air interface. Karpovitch and Ray¹⁴ fit their SHG-derived results to a Langmuir isotherm; these results are expressed in terms of mole fraction of DMSO and are plotted up to x_DMSO = 0.05. Again, we transform to molar units and replot the isotherm displayed in their Figure 1,¹⁴ and we fit the data to eq 1 to obtain B. Application of eq 4 gives ΔG°_ads = −7.4 kJ mol^–1 if N^max is assumed to be 1.3 × 10¹⁴ molecules cm^–2. This value is presented as well in Table 1.

Notably, the glancing-angle Raman and surface tension derived results are somewhat smaller than the SHG derived result; this is most probably a reflection of the different depths probed by the two methods. Where SHG strictly is sensitive only to the upper 1–2 surface layers, glancing-angle Raman probes many tens of surface layers, as discussed in our earlier works.^11,25 This deeper probing will “dilute” the signal arising from the more surfactant-enriched surface layers, giving an apparent weaker adsorption free energy.

b. H₂O₂

Comparing the standard adsorption free energy of DMSO extracted from the present glancing-angle Raman results to those of earlier reports (with the same choice of standard states) gives some confidence that GARS can provide some meaningful thermochemical data. The adsorption free energy of hydrogen peroxide to the water surface from aqueous solution has been only reported twice, in both instances using molecular dynamics simulation techniques. Although mixtures of H₂O₂ with water have been the subject of much study (e.g., ref (6) and citations therein), there have been no experimental reports of the surface partitioning behavior of H₂O₂. The present glancing-angle Raman study represents the first such result.

The results illustrated in Figure 3 are strongly suggestive that H₂O₂ is at least somewhat surface active. Application of eqs 3 and 4 to the H₂O₂ results is straightforward. Since we do not have Raman intensity data for neat H₂O₂, we take the value of A from fits to eq 1 to reflect N^max, and we assume that this is given by the surface density of the neat liquid. As with DMSO, we further assume that the ratio of surface and bulk activity coefficients for H₂O₂ is unity through the concentration range studied here. Parts a and b of Figure 5 show the same analyses as in Figure 4, applied to H₂O₂. The resulting 293 K ΔG°_ads values are presented in Table 2.

Figure 5.

Analysis of H₂O₂ results. (a) Glancing-angle data plotted and fit using a linearized form of the Langmuir isotherm equation, eq 1a. From the slope and intercept, the kinetic equilibrium constant is obtained. (b) Plot of the data, fit using eq 4. The intercept yields a free energy of adsorption, which may then be corrected to obtain the standard free energy. See text for details.

Table 2. H₂O₂ Thermochemistry Results.

	standard state	ΔG°(kJ/mol)at 293 K	notes
H2O2	c°HOOH = 1 mol L–1	–(5 ± 1)	present work, eq 1a
		–(4.7 ± 0.4)	present work. eq 4
	single molecule at surface	–4.2	Vacha et al.,7 MD simulation
		–6.7	Martins-Costa et al.,9 MD simulation
	c°HOOH = 1 mol L–1	–1.9	Vacha et al.,7 MD simulation
		–2.8	Martins-Costa et al.,9 MD simulation

The glancing-angle experimental results are very close to the values reported from the MD simulations.^7,9 It is not clear how to rigorously apply our standard state choices to the MD results, as the bulk water densities in the simulations are both quite different from that of liquid water (2.65 mol L^–1 in the case of the Vacha et al. work,⁷ and 10.5 mol L^–1 for the Martins-Costa et al. simulations,⁹ vs 55.5 mol L^–1 for liquid water). However, if we ignore this (which may well not be appropriate), we can estimate ΔG° from the equation:

6

For c_surf, we have 1 H₂O₂ molecule adsorbed on a surface area as given in each paper; for c_sol, we have 1 H₂O₂ molecule in the corresponding simulation volume. Both of these quantities differ between the two simulations. With these choices and using a 1 mol L^–1 standard state, the ΔG°_ads values from the two simulations are both somewhat smaller in magnitude than the GARS result. They are also given in Table 2.

The MD simulations follow the trajectory of a single solute molecule, calculating the average free energy difference between its environment in the bulk and at the interface. In this way, they may be best compared to truly surface-sensitive experimental probes, such as SFG or SHG. Unfortunately, results from such probes are not currently available. As discussed above, the GARS probe depth is expected to be considerably deeper than the upper 5–10 A in the surface region, most likely exploring the same depth profile as surface tension measurements. The “seamless integration” of H₂O₂ into the water structure,⁶ due to its very similar hydrogen bonding ability, makes extraction of hydrogen peroxide surface excess from surface tension measurements problematic, however. Nevertheless, given the results presented here, we may say with some confidence that H₂O₂ is enriched in the surface region of aqueous solutions.

Notwithstanding the previous MD results,^7,9 it was not a priori obvious to this author that H₂O₂ should partition to the aqueous–air interface. Hydrogen bonding between water and hydrogen peroxide is stronger than that between either molecule on its own,^6,26 resulting in a somewhat depressed solution vapor pressure²⁷ compared to an ideal solution, as well as a somewhat negative excess enthalpy of mixing.²⁸ Nevertheless, studies of H₂O₂ – H₂O solutions have generally concluded that these solutions are close to ideal, at least in the liquid state.^6,28 The modest value of hydrogen peroxide’s standard adsorption free energy is certainly consistent with these other properties.

Conclusions

The experimental results and analysis presented above demonstrate that glancing-angle Raman spectroscopy can be used to determine adsorption free energies for soluble species at the air–water interface. GARS measurements of DMSO solutions yield a standard free energy of adsorption in good accord with that determined using surface tension, and somewhat smaller than that inferred from truly surface sensitive techniques. Applying GARS-derived isotherms to the adsorption of H₂O₂ to the water surface yields a standard free energy of adsorption for H₂O₂ (using a 1 mol L^–1 standard state) of about −5 kJ mol^–1, similar to the corresponding value determined for DMSO.

Given the potential importance of an elevated air–aqueous surface concentration of hydrogen peroxide in aqueous solutions,^4,8,10 the present experimental verification of its adsorption is significant.

The author declares no competing financial interest.

Special Issue

Published as part of The Journal of Physical Chemistry virtual special issue “Advances in Atmospheric Chemical and Physical Processes”.