Astrid
Waldner
a,
Luca
Artiglia
a,
Xiangrui
Kong
a,
Fabrizio
Orlando
a,
Thomas
Huthwelker
b,
Markus
Ammann
a and
Thorsten
Bartels-Rausch
*a
aLaboratory of Environmental Chemistry, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland. E-mail: thorsten.bartels-rausch@psi.ch
bSwiss Light Source (SLS), Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland
First published on 4th September 2018
Interactions between trace gases and ice are important in environmental chemistry and for Earth's climate. In particular, the adsorption of trace gases to ice surfaces at temperatures approaching the melting point has raised interest in the past, because of the prevailing pre-melting. Here, we present Near Edge X-ray Absorption Fine Structure (NEXAFS) spectroscopy data at ambient partial pressure of water to better define the onset temperature of pre-melting at the interfacial region of ice. Further, this study directly compares the interaction between an organic acid common in the atmosphere, formic acid, and that of an aliphatic carbon with ice at 253 K. It makes use of X-ray Photoelectron Spectroscopy (XPS) with its inherent narrow probing depth covering both the surface and near-surface bulk region when detecting electrons. We use the tender X-ray range for excitation to locate the organic species within the interfacial region with an extended probing depth compared to published XPS work. Electron kinetic energy dependent C1s photoemission data indicate that, at low coverage of a few 1014 molecules cm−2, the presence of formic acid is restricted to the upper ice layers of the interfacial region. Increasing the dosage, formic acid penetrates 6–7 nm into the air–ice interface. The presence of the more hydrophobic aliphatic carbon is restricted to the upper ice monolayers. This direct comparison of an organic acid with an aliphatic compound confirms the emerging picture where solutes enter the interfacial region of ice at a depth related to their specific tendency to form solvation shells.
Short chain organics are present in the air above ice, and only limited studies of their interactions with ice are available.12 Formic acid (HCOOH), one of the strongest short chain organic acids abundant in the atmosphere, occurs in mixing ratios of around ∼2 parts per billion in boundary layer air.13 The primary sources of formic acid are photochemical oxidation of volatile organic compounds and biogenic sources such as direct forest emissions.12,14–16 Main formic acid sinks include photochemical oxidation17 as well as dry and wet deposition.12,16,18,19 With an atmospheric lifetime in the atmosphere of 2–4 days,20 formic acid represents a substantial sink for in-cloud OH21 and has significant impact on key chemical budgets such as that of nitrogen oxides. This makes the loss processes of formic acid important for atmospheric chemistry.12 Theoretical and experimental work at temperatures below ∼230 K showed that the partitioning of formic acid to ice is significantly smaller than that of inorganic acids such as HCl, but still larger compared to other volatile organics.22–25 Thus, in presence of ice, partitioning of formic acid has the ability to influence the OH budget and the oxidation capacity of the atmosphere.
We have recently reported on the interaction of formic acid with ice at the molecular level using Near Ambient Pressure X-ray Photoelectron Spectroscopy (NAP-XPS), which probes the chemical composition with high surface sensitivity under equilibrium thermodynamic conditions.26 Depth profiles at 251 K have revealed that the presence of formic acid is restricted to the surface of the interfacial region at 3 × 1014 molecules cm−2.26 Increasing the dosage, formic acid enters 3–6 nm into the interfacial region of ice at 251 K. The presence of formic acid deeper in the interfacial region went along with an increase in the liquid character of the interfacial hydrogen-bonding network.26 We have suggested that the driving force for solutes to enter into the interfacial region and to modify the hydrogen-bonding structure is their demand for water molecules to form energetically favourable solvation shells.26,27 This picture of very localized changes to the interfacial structure nicely explains the increase in disorder in presence of solutes, as shown in the pioneering work by McNeill et al.,10 while suggesting structural differences in the pre-melting of neat ice and impurity induced solvation shells. Interestingly, recent large scale molecular simulations also linked the increase of liquid-like character at the ice interface in presence of solutes with the formation of solvation shells.28 In particular, the increase in liquid-like character was found to be less in presence of an oxygenated organic solute than in presence of ions and this difference was linked to the specific energetic attraction for water molecules. Therefore, we compare the depth profiles of formic acid with those of a more hydrophobic carbon species in this work to investigate the link between the solute's ability to modify the hydrogen-bonding network and its solubility.26,27,29,30 The hydrophobic species is adventitious carbon, a ubiquitous carbon species in many photoemission experiments.31,32 A well-known source are radical reactions of carbonaceous precursors that are initiated by secondary electrons emitted from the ice sample upon X-ray irradiation in the gas phase close to the sample spot.33
Last but not least, the results by Bartels-Rausch et al. have been derived at the SIM beamline (Swiss Light Source, Paul Scherrer Institute) with an upper limit of excitation energy of 1600 eV, corresponding to an electron escape depth† of ∼5 nm.26 This poses the question, whether the observed limit of formic acid to the upper 3–6 nm is given by the constraints to the probing depth of those experiments. Therefore, we probe the presence of formic acid in the interfacial region using tender X-rays (2000–7000 eV) with an upper electron escape depth in the ice of 12 nm in this work.
In addition, we measured partial Auger electron-yield NEXAFS spectra at the O K-edge using a kinetic energy window at the background of the Auger lines (450–470 eV). For these NEXAFS measurements, a pass energy of 20 eV and dwell times of 100 ms were used. The beam polarization vector was rotated 54.7° from the analyser detection axis, which corresponds to the so-called magic angle, to ensure a signal intensity independent of the molecular orientation. In the O K-edge region of most interest, 529 to 542 eV, we used a photon energy step size of 0.2 eV instead of 1 eV. The NEXAFS spectra were normalized to background measurements (I0) to account for variations in beamline flux with photon energy as derived from in situ flux measurements at the refocusing mirror of the beamline, to drifts in the pre-edge region using a linear fit, and to their integrated area between 534 and 545 eV photon energy to facilitate comparison.
![]() | ||
Fig. 1 Oxygen K-edge NEXAFS spectra of ice acquired at the SIM beamline at SLS at a temperature of 251 K (light blue line) and 233 K (dark blue line). Also shown is the oxygen K-edge NEXAFS of water at 269 K (grey dotted line) that was not acquired in this work.1 |
The origin of the changes in the NEXAFS spectra between that of ice and that of water can be linked to the coordination at the molecular level as follows.9,11 In tetrahedral ice both OH groups of the water molecules are coordinated with neighbouring water molecules, whereas in liquid water this coordination is weaker. The post-edge results from strongly H-bonded OH, thus the post-edge to main edge peak ratio is well pronounced in ice. The O K-edge NEXAFS spectrum of liquid water exhibits a shift towards the main edge with a maximum around 538 eV. A second difference in the NEXAFS spectra of ice and of water is evident in the pre-edge region. The intensity of this pre-edge peak increases with increasing disorder of the water molecules as we go from solid ice to liquid water. It corresponds to a transition to empty states similar to the 4a1 lowest unoccupied molecular orbital of gas-phase water molecules. Because of the dipole selection rule, the intensity of this transition depends on the degree of s- or p-character of the molecular energy state. In ice, oxygen is tetrahedrally coordinated thus symmetry arguments indicate predominant s-symmetry. In liquid, this symmetry is broken, thus the p-character is more probable and the peak intensity increases. Features at 532.5 eV in the spectra, can be assigned to CO or N
O bonds indicating slight contamination of the ice.
At a closer look, a small shift in the slope of the main edge region in the O K-edge NEXAFS of ice is observed towards that seen in liquid water when increasing the temperature from 233 K to 251 K. Reproducibility of NEXAFS spectra acquired one after the other during the same beamtime at constant temperature was excellent, i.e. the spectra were indistinguishable over the whole excitation energy range. Comparing spectra from samples prepared at different days revealed larger fluctuations in particular in the main edge to post-edge intensity ratio. We attribute this to differences in the stability of the ice sample during the NEXAFS acquisition of 30 min duration: minute changes in the ice thickness directly relate to changes in the distance between the electron analyser's sample orifice and the sample surface. This working distance is a key factor of absolute sensitivity in electron spectroscopy. Nevertheless, the slope of the main edge turned out to be a robust feature to differentiate the NEXAFS at 233 K and 251 K. This minor alteration in the NEXAFS indicates that there are only very little changes in the hydrogen-bonding network associated with a temperature increase to 251 K. Based on analogue Auger yield NEXAFS measurements, Bluhm et al. concluded that quasi-liquid is absent at the air–ice interface at temperatures below 248 K.9 Our data presented here expand this data set to higher temperatures and give experimental evidence that the interfacial liquid character does not significantly increase up to 251 K. Clearly, the onset of surface disorder, defined at the temperature where the liquid-like character is found beyond the outermost crystal layer and as seen by NEXAFS, occurs at temperatures above 251 K.
In summary, the interfacial region of ice is at the dawn of the disorder being significantly present within the interfacial region at 251 K. In the following, we will analyse how trace gases adsorb at the interfacial region of ice just prior to the onset of the quasi-liquid layer.
The origin of the spectroscopic feature at 288 eV in the C1s PE spectrum is less certain. We have found earlier, that the neutral and the dissociated form of acids can be differentiated at the ice interfacial region upon adsorption by means of XPS.27,29 Density functional theory calculations and XPS data taken at liquid water samples have revealed the electron binding energy of deprotonated formic acid to be shifted by −1.3 eV as compared to the binding energy of the neutral acid.36 However, as we cannot exclude that adventitious carbon also contains carbonyl functionalities with typical binding energies of 288 eV, we refrain from analysing this feature further. In summary, the C1s PE spectrum at the interfacial region of ice reveals the presence of adventitious carbon with a dominant aliphatic component and of formic acid. Both spectroscopic features respond to increasing partial pressures of formic acid with an intensity increase reflecting a rise in surface concentration. In the following, we will discuss the distribution of these chemical species with depth at the air–ice interface and their surface concentration as directly seen by XPS. We’d like to note that the precision of the partial pressures given in this work to reflect the gas-phase composition at the sample spot might be rather low. First, capacitance manometers do not work reliable with sticky gases as formic acid. Secondly, formic acid may form dimers at high concentration as used in this work. Third, loss of formic acid at the walls of the flow-through cell during gas-phase transport likely results in a substantially lower partial pressure of formic acid at the ice sample compared to that measured upstream of the sample. Nevertheless, based on these upper partial pressure limits, we conclude that the work was done in the ice stability domain in absence of melting. Freezing point depression data and Henry's law coefficients indicate that melting does not occur below 0.09 mbar and 253 K.26,37 Please note, that traces of formic acid have been observed in the RGA, but were not used for quantification because formic acid fragments strongly in RGA and we have thus not quantified this signal further.
ED = IMFP × cos(θ), | (1) |
![]() | ||
Fig. 3 (A) Calibrated PE ratio of the integrated C 1s carboxylic peak area and O 1s peak area as a function of electron kinetic energy at 253 K. The upper panel shows results for experiments with 0.04 mbar formic acid partial pressure (yellow circles) and the lower one gives the data at 0.004 mbar formic acid (orange circles). The coloured lines represent fits to eqn (2.1) (solid line) and of eqn (3.1) (dashed line). (B) Calculated mean electron escape depth (grey dots) in ice versus electron kinetic energy.2 The grey line shows the fit and extrapolation based on the quantitative prediction.6 |
![]() | ||
Fig. 4 Calibrated PE intensity ratio of the integrated C 1s adventitious carbon peak area and O 1s peak area as a function of electron kinetic energy. The upper panel shows results (yellow triangles) for experiments with 0.04 mbar formic acid and the lower with 0.004 mbar formic acid partial pressure (orange triangles), both at 253 K. The coloured lines represent fits to eqn (2.2) (solid line) and to eqn (3.2) (dashed line). |
A species with homogeneous concentration along the entire sampling depth at the interfacial region would show a steady increase of the calibrated IC1s(FA) and IO1s with increasing photon energy, and consequently a constant IC1s(FA)/IO1s in Fig. 3A and 4. The other extreme, a species that is purely surface adsorbed would result in a constant IC1s(FA) regardless of the electron kinetic energy, an increase of IO1s as the number of water molecules grows with probing volume, and thus a strongly decreasing IC1s(FA)/IO1s. Therefore, the steep profile of the integrated signal for 0.004 mbar formic acid in Fig. 3A indicates that the formic acid is predominantly located at the surface and its availability decreases strongly with depth, whereas the steady profile of Fig. 3A (upper panel) indicates significant presence of formic acid also deeper in the ice. This is fully consistent with earlier findings26 and indicates a substantial increase in the penetration depth into the air–ice interface of the formic acid with increasing partial pressure.
Adventitious carbon shows a distinctly different behaviour compared to formic acid with persistent steep profiles at both surface coverages. The IC1s/IO1s of adventitious carbon (Adv.C) decreases to 75% over the first 1000 eV in the experiments with 0.004 mbar formic acid and to 82% in the experiments at 0.04 mbar formic acid. The decline of the depth profile at low dosage agrees perfectly well with the observations of formic acid at the interface where a decline to 76% was observed. These results are in-line with adventitious carbon not entering the interface at considerable depth, but rather covering the ice sample surface.
In summary, this qualitative analysis of the data gives clear experimental evidence of formic acid entering the interfacial region deeper at higher surface coverage. Further, the more hydrophobic adventitious carbon is found more at the surface compared to formic acid at high coverage. For a more in-depth analysis of the concentration of formic acid and adventitious carbon in the interfacial region, a 3-layer model was used to interpret the profiles quantitatively. The model is similar to that presented by Křepelová et al., but to account for the adventitious carbon an additional surface layer was introduced.27 In brief, a homogeneous layer of thickness d(1) ranging from the air–ice interface at 0 nm to a depth of z1 nm contains adventitious carbon with an atomic concentration of nAdv.C(1) and nO(1) oxygen atoms per unit cell. This layer sits on top of a homogeneous 2nd layer with a thickness of d(2) (depth z1 to z2) containing nFA(2) formic acid and nO(2) oxygen atoms per unit cell. The third layer below z2 is infinitely thick consisting of neat ice with an atomic concentration of nO(3) oxygen per unit cell. Now, the measured IC1s(FA)/IO and IC1s(Adv.C)/IO1s can be described by eqn (2.1) and (2.2), respectively. Iterative fit results, minimizing the residuals of IC1s(FA)/IO1s, IC1s(Adv.C)/IO1s, and of IC1s(FA)/IC1s(Adv.C) simultaneously applying the 3-layer model presented above are displayed in Fig. 3A and 4 (solid lines). Tables 1 and 2 also give the constrains to the fits that were used taken that the model system (eqn (2.1 and 2.2)) holding 4 unknown fitting parameters is underdetermined.
![]() | (2.1) |
![]() | (2.2) |
![]() | (3.1) |
![]() | (3.2) |
Model | Constrains | d(1) [nm] | d(2) [nm] | [FA]surf [molec. cm−2] | X FA(2) [%] | FA(1)/FA(2) [%] | r 2 |
---|---|---|---|---|---|---|---|
Eqn (2.1 and 2.2) | d(1) > 0.3 nm; d(2) > 0.3 nm; X(FA) < 100% | 0.3 | 0.3 | 1.2 × 10 +14 | 16 | — | 0.9 |
Eqn (2.1 and 2.2) | d(1) > 0.3 nm; d(2) > 0.3 nm; 25% < X(FA) < 100% | 0.4 | 0.3 | 1.7 × 10+14 | 25 | — | 0.9 |
Eqn (2.1 and 2.2) | 0.3 < d(1) < 1 nm; 1 < d(2) < 4 nm; X(FA) < 100% | 0.3 | 2.3 | 2.0 × 10+15 | 52 | — | 0 |
Eqn (2.1 and 2.2) | 1 < d(1) < 4 nm; 1 < d(2) < 4 nm; X(FA) < 100% | 1.0 | 0.4 | 4.4 × 10+14 | 100 | — | 0.9 |
Eqn (3.1 and 3.2) | d(1) > 0.3 nm; d(2) > 0.3 nm; X(FA) < 100% | 0.3 | 4.6 | 9.3 × 10+13 | 0.6 | 15 | 0.6 |
Eqn (3.1 and 3.2) | 0.3 < d(1) < 0.5 nm; 0.3 < d(2) < 0.5 nm; X(FA) < 100% | 0.3 | 0.5 | 2.1 × 10 +13 | 1.4 | 22 | 0.2 |
Eqn (3.1 and 3.2) | 1 < d(1) < 4 nm; 0.3 < d(2) < 4 nm; X(FA) < 100% | 1.0 | 4.0 | 1.6 × 10+14 | 1.4 | 34 | 0.3 |
Eqn (3.1 and 3.2) | 1 < d(1) < 4 nm; 0.3 < d(2) < 0.5 nm; X(FA) < 100% | 1.0 | 0.5 | 2.4 × 10+13 | 1.6 | 56 | 0 |
Eqn (3.1 and 3.2) | 0.3 < d(1) < 4 nm; 0.3 < d(2) < 4 nm; 11 < X(FA) < 100% | 4.0 | 1.2 | 3.6 × 10+14 | 11 | 1 | 0 |
Model | Constrains | d(1) [nm] | d(2) [nm] | [FA]surf [molec. cm−2] | X FA(2) [%] | FA(1)/FA(2) [%] | r 2 |
---|---|---|---|---|---|---|---|
Eqn (2.1 and 2.2) | d(1) > 0.3 nm; d(2) > 0.3 nm; X(FA) < 100% | 4.9 | 4.9 | 5.9 × 10+15 | 100 | — | 0.8 |
Eqn (2.1 and 2.2) | 0.3 < d(1) < 2 nm; d(2) > 0.3 nm; X(FA) < 100% | 2.0 | 6.6 | 1.9 × 10 +15 | 11 | — | 0.8 |
Eqn (2.1 and 2.2) | d(1) > 0.3 nm; d(2) > 0.3 nm; X(FA) < 25% | 3.2 | 5.8 | 3.2 × 10+15 | 25 | — | 0.8 |
Eqn (3.1 and 3.2) | 0.3 < d(1) < 2 nm; d(2) > 0.3 nm; X(FA) < 100% | 0.3 | 4.2 | 1.6 × 10 +14 | 1 | 12 | 0.8 |
Eqn (3.1 and 3.2) | 0.3 < d(1) < 2 nm; d(2) > 0.3 nm; 11 < X(FA) < 100% | 2.0 | 2.1 | 8.8 × 10+14 | 18 | 6 | 0 |
Eqn (3.1 and 3.2) | 0.3 < d(1) < 5 nm; d(2) > 0.3 nm; 11 < X(FA) < 100% | 4.7 | 10.0 | 4.2 × 10+15 | 18 | 18 | 0 |
This finding supports the conclusion that, at such high concentrations, formic acid induces disorder to the hydrogen-bonding network at the interfacial region resulting in greater depth at which formic acid is found compared to the data at lower formic acid surface coverage. That the depth of formic acid is limited to about 7 nm is an important finding unambiguously showing that the surface disorder is limited to a nanometre scale region even in presence of large amounts of solute. While being in agreement to our previous analysis, this is the first photoemission study to probe the upper 12 nm of the air–ice interface. Our previous work, showing a depth of 3–6 nm at similar formic acid partial pressure reached the beam-lines probing depth limits of 5 nm. These data at high surface coverage of formic acid indicate that also the adventitious carbon might enter the quasi-liquid to some extend as the first layer of the model (d(1)) with a thickness of 2 nm extents further than the upper monolayer of the ice. Therefore, we tested whether a modified model where formic acid is also present in the layer d(1) might represent the data as well. Model eqn (3.1) and (3.2) holds both formic acid and adventitious carbon in their first layer. The interfacial region of ice is thus described by three layers where the upper one holds formic acid and adventitious carbon, the second one formic acid, and the lowest one no solutes. Introducing formic acid into layer 1 results in poorer fits both to the data at low formic acid partial pressure (Table 1 and Fig. 3A) and at high formic acid concentration (Table 2 and Fig. 3A). For example, when constraining the thickness of layer d(1) and of d(2) to less than 0.5 nm, corresponding to the fitting results of eqn (2.1) and (2.2), the fit does not capture the trend in the data (Fig. 3A). The performance of the fit is improved when allowing a thicker layer with d(2) = 4.6 nm, but overall fits to eqn (2.1) and (2.2) were found to reproduce the data better. The impact of the presence of formic acid in layer d(1) on the fitted profiles of formic acid is significantly more evident than that on the fitted profiles of adventitious carbon (Fig. 4). Despite the uncertainty in these underdetermined systems, these model runs support the conclusion that the upper layer holds no formic acid.
The adsorption of formic acid to ice at temperatures below ∼230 K and at low dosage can be well described by Langmuir isotherms with a saturation coverage of 2–3 × 1014 molecules cm−2.23–25 The XPS data presented here show that formic acid remains at the outermost edge of the interfacial region upon partitioning to roughly the saturation coverage. This XPS result gives thus direct evidence of surface adsorption à la Langmuir. Support comes from molecular dynamics simulations41 and from grand canonical Monte Carlo calculations23 indicating that formic acid forms two exceptionally strong hydrogen bonds with rigid ice surfaces up to a Langmuir saturation coverage.
Increasing the partial pressure of formic acid and thus the apparent surface concentration to 2–3 × 1015 molecules cm−2, we can state that formic acid penetrates deeper into the ice, while the adventitious carbon remains at the upper ice surface. We interpret this as adventitious carbon covering the ice sample surface with a quasi-liquid layer below, which solvates the formic acid. Quantification based on a homogenous quasi-liquid layer gives a thickness of 7 nm at maximum. The adventitious carbon is mainly composed of aliphatic C–H groups (Fig. 2) and thus rather hydrophobic. Hydrophobic compounds have been found not to interact strongly with ice surfaces: laser induced fluorescence work and molecular dynamics calculations revealed that the spectra of polycyclic aromatic hydrocarbons at ice surfaces resemble that of the neat crystal and that these compounds favour self-association rather than forming strong hydrogen bonds with the ice.42,43 That formic acid enters deeper into the interfacial region once the Langmuir saturation coverage is reached is at odds with number density profiles derived with grand canonical Monte Carlo calculations that showed multilayer adsorption on rigid ice.23 Molecular dynamics calculations in presence of pre-melting have revealed a high tendency of neutral formic acid molecules to enter and dissolve in the interfacial region.41 It appears therefore likely that the increased flexibility of the hydrogen bonding network at 253 K facilitates the formation of two strong hydrogen bonds41 as solvation shells are formed within the interfacial region. Formic acid–formic acid bonds and hydrogen bonds with the ice surface beyond the Langmuir monolayer coverage are weak.23 Semantically, we would refer to the interfacial region where the concentration of solutes exceeds that of a solid-solution in crystalline ice as QLL. A similar QLL thickness was observed in presence of HCl. Nota bene, a thickness of the QLL of 8 nm was derived based on XPS data at a surface concentration below a formal monolayer.29 At such low coverage, XPS data reveals that formic acid and acetic acid remain at the upper surface only in this and in earlier work.26,27 Apparently, strong acids attract water molecules from the ice more easily to solvate than weaker acids. This striking ability of strong acids to modify the hydrogen bonding network at the interfacial region of ice is also evident from NEXAFS data in presence of HNO344 and of HCl,29 and from Raman data in presence of HCl and of HNO3.45 The interpretation is that the overall energetics of the system favour strong acids forming solvation shells deeper in the interfacial region where the hydrogen bonding network becomes increasingly ice-like. The overall ice-like structure of the QLL is supported by recent sum frequency generation data comparing spectral features of supercooled water, ice and the QLL40 and of NEXAFS data in presence of solutes revealing a dominating ice-like feature in the spectra of the interfacial region.26,29,44
Footnotes |
† The electron escape depth (ED) is a direct measure of the sampling depth in XPS. |
‡ XPST is an add-on to IgorPro to analyse X-ray Photoelectron Spectroscopy (XPS) data. Written by Dr Martin Schmid, Philipps University Marburg. http://www.igorexchange.com/project/XPStools |
This journal is © the Owner Societies 2018 |