Total reflection infrared spectroscopy of water-ice and frozen aqueous NaCl solutions Rachel L. Walker, Keith Searles, Jesse A. Willard, and Rebecca R. H. Michelsen Citation: The Journal of Chemical Physics 139, 244703 (2013); doi: 10.1063/1.4841835 View online: http://dx.doi.org/10.1063/1.4841835 View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/139/24?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Understanding the role of ions and water molecules in the NaCl dissolution process J. Chem. Phys. 139, 234702 (2013); 10.1063/1.4840675 Molecular events in deliquescence and efflorescence phase transitions of sodium nitrate particles studied by Fourier transform infrared attenuated total reflection spectroscopy J. Chem. Phys. 129, 104509 (2008); 10.1063/1.2973623 Organic monolayers detected by single reflection attenuated total reflection infrared spectroscopy J. Vac. Sci. Technol. A 24, 668 (2006); 10.1116/1.2180270 Structural evolution of aqueous NaCl solutions dissolved in supercritical carbon dioxide under isobaric heating by mid and near infrared spectroscopy J. Chem. Phys. 122, 094505 (2005); 10.1063/1.1858440 Interfacial melting of thin ice films: An infrared study J. Chem. Phys. 116, 4686 (2002); 10.1063/1.1449947
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
THE JOURNAL OF CHEMICAL PHYSICS 139, 244703 (2013)
Total reflection infrared spectroscopy of water-ice and frozen aqueous NaCl solutions Rachel L. Walker,a) Keith Searles,b) Jesse A. Willard,c) and Rebecca R. H. Michelsend) Department of Chemistry, Randolph-Macon College, P.O. Box 5005, Ashland, Virginia 23005, USA
(Received 2 September 2013; accepted 22 November 2013; published online 26 December 2013) Liquid-like and liquid water at and near the surface of water-ice and frozen aqueous sodium chloride films were observed using attenuated total reflection infrared spectroscopy (ATR-IR). The concentration of NaCl ranged from 0.0001 to 0.01 M and the temperature varied from the melting point of water down to 256 K. The amount of liquid brine at the interface of the frozen films with the germanium ATR crystal increased with salt concentration and temperature. Experimental spectra are compared to reflection spectra calculated for a simplified morphology of a uniform liquid layer between the germanium crystal and the frozen film. This morphology allows for the amount of liquid observed in an experimental spectrum to be converted to the thickness of a homogenous layer with an equivalent amount of liquid. These equivalent thickness ranges from a nanometer for water-ice at 260 K to 170 nm for 0.01 M NaCl close to the melting point. The amounts of brine observed are over an order of magnitude less than the total liquid predicted by equilibrium thermodynamic models, implying that the vast majority of the liquid fraction of frozen solutions may be found in internal inclusions, grain boundaries, and the like. Thus, the amount of liquid and the solutes dissolved in them that are available to react with atmospheric gases on the surfaces of snow and ice are not well described by thermodynamic equilibrium models which assume the liquid phase is located entirely at the surface. © 2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4841835] I. INTRODUCTION
It is now widely recognized that the surfaces of ice and snow in Earth’s environment participate in reactions that affect the composition of the local atmosphere. Notable among these reactions are bromine explosion and ozone depletion events.1 It is thought that bromide from sea salt is activated to bromine radicals through photochemistry and reactions on snowpack and sea ice surfaces. The reactivity of ice surfaces proves to be different in many cases from bulk aqueous chemistry in unexpected and as yet unexplained ways. The primary reaction products can be different when ice is involved. For example, in studies of interhalide formation, Br2 Cl− is the favored product from frozen media containing halides and acids, as opposed to Cl2 Br− , which is formed in liquid media containing otherwise identical reactants.2 Additionally, reaction rates on frozen surfaces are sometimes enhanced. For example, photolysis of p–nitroanisole was 40 times faster on ice than in aqueous media.3 Photolysis of harmine is also faster on ice, unless there are salts present in the frozen solution.4 Ozone reacts faster with halides on the surface of frozen solution than on liquid surfaces.5 On the other hand, the reaction of OH radical with benzene is suppressed at the surface of ice relative to liquid.6 One a) Present address: Department of Chemistry, West Virginia University,
Morgantown, West Virginia 26506, USA.
b) Present address: Department of Chemistry, University of Pennsylvania,
Philadelphia, Pennsylvania 19104, USA.
c) Present address: School of Law, University of Richmond, Richmond,
Virginia 23173, USA.
d) Author to whom correspondence should be addressed. Electronic mail:
[email protected].
0021-9606/2013/139(24)/244703/8/$30.00
reason reactions may happen faster at ice surfaces is the freeze concentration effect: solutes are excluded from the ice crystal lattice as it forms resulting in higher concentrations in the remaining liquid at the surface. This concentration enhancement was measured as greater than an order of magnitude for halide ions7 and an enormous three to six orders of magnitude for methylene blue.8 Differences in exclusion from the ice matrix, reaction products, and reaction rates likely result from different molecular-level interactions or mechanisms. Understanding the reactivity of ice surfaces at the molecular level requires a thorough description of the microphysical properties of the ice surface. Pure ice undergoes premelting; that is, the surface is composed of a layer of disordered water molecules9–11 as a result of the interface. The nature of this disordered interface, sometimes called the quasi-liquid layer (QLL), is the subject of much inquiry. Measurements of the thickness of the disordered interface vary greatly depending on the experimental technique used.11, 12 For example, atomic force microscopy measured a liquid-like layer of 90 nm close to the melting point,13 while a photoelectron spectroscopy study found only a couple nanometers of disorder at the surface.14 When there are solutes in the solution that freezes, the result is quite different. In particular, for salt solutions near the melting point of ice, a liquid brine layer forms at the surface.1, 15 This brine is a distinct phase in equilibrium with the ice as described by the phase diagram. The physical and chemical properties of a brine layer are similar to a liquid aqueous environment, whereas those of the disordered interface on water-ice are not necessarily similar. For example, HCl gas will dissociate and acidify the brine layer
139, 244703-1
© 2013 AIP Publishing LLC
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-2
Walker et al.
on frozen saltwater, but frozen freshwater with HCl exposure does not show the same increase in acidity.16 Reaction kinetics in the brine layer tend to be similar to bulk aqueous kinetics, as in the reaction of ozone with frozen salt solutions.17 Changes in acidity in the brine can occur with freezing, as in the case of frozen NaCl solutions resulting in a brine basic enough to catalyze the oxidation of gallic acid.18 Besides appealing to the empirical phase diagram for solute-water mixtures, several attempts have been made to treat the brine layer on a frozen surface as a thermodynamically stable phase in equilibrium with the bulk ice. Cho et al. assume the solutes are dilute and derive an expression for the maximum liquid water fraction.19 Kuo et al. follow a similar treatment, but include deviation from ideality, which is especially important for concentrated brines.20 The thermodynamics of a brine phase is not at issue; however, the idea that all the liquid will reside at the surface may be problematic. Experimental evidence suggests that in the case of nitrate, thermodynamics overpredicts the amount of nitrate at the liquid surface of frozen solutions.21 Without direct measurements or relevant parameterizations, models of snow-atmosphere interactions are currently based on questionable assumptions about how much liquid to incorporate at the surface of snow and what concentrations of solutes to use.22–25 In this work, we compare experimental IR spectra to calculated reflection spectra for liquid/solid water films with a simple two-layer morphology. We present measurements of the amount of liquid detected near the surface of frozen solutions measured via attenuated total reflection infrared spectroscopy (ATR-IR) over ranges of temperature and salt concentration.
II. METHODS A. Attenuated total reflection infrared spectroscopy
Frozen aqueous films were studied via ATR-IR to minimize the interference of water vapor in the spectra. Solutions were introduced to the top of a germanium prism embedded in a copper plate. The plate and germanium crystal were cooled via recirculating ethanol from a Huber ministat cc1 chiller. The ATR plate was positioned on a VeeMax II accessory (Pike Technologies) in the sample cell of a ThermoElectron Nicolet 6700 FT-IR spectrometer with a DTGS detector. A constant 45◦ angle of incidence was used. For water on germanium, the penetration depth of the IR beam is around 200 nm at this angle. (See Eq. (7)) The sample compartment, the ATR plate holder, and the area above the germanium crystal were purged with dry, CO2 -free air. Solutions (0.8–1 ml for water and 1.0 ml for NaCl solutions) were introduced to the Ge surface via a glass syringe. When the Ge crystal was ∼256 K, freezing occurred in less than a minute. No pressure was applied to the frozen films. The temperature of the frozen films was measured by a type K thermocouple (Omega) which was positioned in the film in contact with the Ge/solution interface. The temperature ranged from room temperature down to 256 K and was stable to ±0.1◦ . Background spectra (with a clean Ge crystal) were taken on the same day at approximately 263 K. All spectra are an average of 32 scans and were taken with 4 cm−1 resolution. Sodium chloride (Sigma
J. Chem. Phys. 139, 244703 (2013)
Aldrich, ACS reagent grade) was used without further purification. An initial solution of 1.006 M NaCl (1.026 mol/kg) in reagent grade water (Sigma Aldrich) was mixed, and then serial tenfold dilutions were made down to 10−4 M. In order to follow spectral changes in frozen films as a function of temperature, a solution was frozen at around 256 K. After a spectrum was taken, the temperature of the film was increased about a degree. The film was allowed to equilibrate for 10–20 min and then another spectrum was taken. This procedure was repeated until the entire film melted. Since the Ge crystal was cold, but the air above it was warmer, the films froze from the bottom up. There was no measurable temperature gradient between the interface with the germanium crystal and the top of a stable, frozen film. The air temperature above a frozen film was around 275 K. Thus, while frozen, the uncertainty in temperature is similar to that of the thermocouple (±0.1◦ ). However, some spectra show partially solid films at temperatures slightly above the melting point, e.g., 273.8 K for water-ice. Thus the uncertainty in the temperature may be higher, especially close to the melting point. It is also possible that films that were in the process of melting experienced a temperature gradient, and/or the thermocouple tip moved away from the interface. During melting, the ice was observed to float in the liquid, as expected. B. Spectral calculations
Reflection spectra were calculated from the Fresnel equations26 and the complex indices of refraction of ice and liquid water:27 cos θ − i sin2 θ − n221 r⊥ = , (1) cos θ + i sin2 θ − n221 n221 cos θ − i sin2 θ − n221 r|| = . n221 cos θ + i sin2 θ − n221
(2)
Here, r⊥ and r|| are the amplitude reflection coefficients for which the electric field vector of the light is perpendicular and parallel to the plane of incidence, respectively. The angle of incidence is θ and n21 = n2 /n1 , the ratio of refractive indices of water (n2 = n + ik)27 and germanium (n1 = 4.0275).28 Reflectance, R, is the product of the amplitude reflection coefficient with its complex conjugate. In this study, R = r⊥∗ r⊥ was sufficient to reproduce spectra for our system since r|| is small when the angle of incidence is less than the Brewster angle.26 Extinction, which is analogous to absorption in transmission spectroscopy, is given by E = − log
R , R0
(3)
where R0 = 1 since reflection is total when no absorbing film is in contact with the crystal. In this manner, the reflection spectra of liquid water and ice were modeled, showing excellent agreement to experiment, as discussed below. The complex refractive indices of water are slightly temperature dependent. The experimentally determined values
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-3
Walker et al.
J. Chem. Phys. 139, 244703 (2013)
Since the penetration depth depends on the media present at the reflecting surface, n21 for liquid water was used in Eq. (7). The thickness of the liquid film, t, is the only parameter in the model, and spectra were calculated for liquid layer thicknesses varying from 10 to 200 nm using Mathematica. In our experiments, dp varies from 230 nm at 2800 cm−1 to 163 nm at 3800 cm−1 . III. RESULTS FIG. 1. Simplified sketch of a frozen NaCl (aq) film on Ge. Only the bottom layer of brine and a portion of the ice film contribute significantly to reflection spectra.
from Zasetsky et al. were used: T = 263 K or 273 K for liquid water and T = 235 K for ice.27 Additionally, a non-absorbing solute changes the index of refraction very slightly when the concentration is high. For example, at 589 nm, the index of refraction changes from 1.33 for water to 1.36 for 4.3 mol/kg NaCl solutions.28 Since such changes are small, NaCl was not accounted for in our calculated spectra. In practice, the shape of the calculated water spectrum was slightly different than the shape of the experimental liquid spectra. No polarization filters were used in obtaining the experimental spectra. For the situation in which the absorbing film is liquid at the interface with the germanium prism, but ice at some point above the interface, the following simple model, depicted in Figure 1, was used. It was assumed that the total reflectance, R, for such a film can be reasonably represented as a linear combination of reflectance due to an ice film, Rice , and a liquid film, Rliq , R = cliq Rliq + cice Rice .
ATR-IR spectra were taken of frozen water and aqueous NaCl films as the temperature was increased incrementally until they melted. The concentration in the solutions ranged from 0.0001 to 0.01 M NaCl at room temperature. Higher concentration solutions (0.03–0.1 M NaCl) show distinct behavior, and their spectra are described in Ref. 31. In this study, we focus on the OH stretching region of the spectrum. The frozen films all show an initially intense peak at 3228 cm–1 that gradually decreases in intensity with increasing temperature. Upon complete melting, the spectra show a broad, less intense peak with a maximum around 3350 cm−1 . In Figure 2, two such experiments are shown: the top panel is 0.01 M NaCl and the bottom panel is 0.001 M NaCl. In both panels the spectrum of pure water-ice (∼258 K) is shown as a dashed line. As the temperature increases, the intensity at 3228 cm−1 falls considerably, while a shoulder around 3400 cm−1 increases. The broad peak of the fully melted films centered around 3350 cm−1 closely resembles that of liquid water.
(4)
R is proportional to the square of the electric field amplitude of the evanescent wave that penetrates the film in contact with the reflecting interface. Because this electric field amplitude decreases exponentially,29 the coefficients, cliq and cice , were scaled exponentially relative to the penetration depth. The penetration depth, dp , of the evanescent wave is defined as the distance from the interface at which the electric field amplitude falls to 1/e times its value at the interface.29 The axis perpendicular to the interface is z. Thus, for a liquid film thickness of t and a semi-infinite ice film at z > t, the coefficients may be defined as30 t exp(−2z/dp )dz = 1 − exp(−2t/dp ), cliq = 0∞ (5) exp(−2z/dp )dz 0
∞
exp(−2z/dp )dz cice =
t
= exp(−2t/dp ).
∞
(6)
exp(−2z/dp )dz 0
The penetration depth depends on wavelength, λ, and the indices of refraction, dp =
λ/n1
2π sin2 θ − n21
.
(7)
FIG. 2. Series of ATR-IR spectra showing the melting of a 0.01 M NaCl film (top) and a 0.001 M NaCl film (bottom). The dashed lines are spectra of pure ice at 259 K and 257 K, respectively. All spectra were offset to zero extinction at 4000 cm−1 . In the top panel, 5 out of 12 spectra are shown (for clarity) with temperatures of 257 K, 263 K, 267 K, 270 K, and 271 K. In the bottom panel, 7 out of 17 spectra are shown with temperatures of 257 K, 260 K, 263 K, 266 K, 270 K, 273 K, and 274 K.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-4
Walker et al.
The spectra also show an isobestic point in the vicinity of 3325 cm−1 , where all the spectra cross. This behavior indicates that there are two (or more) species present – the liquid (brine or water) and the solid ice – and that these readily interconvert.32, 33 The spectra taken at intermediate temperatures are combinations of the all-liquid and all-ice spectra. Thus, as the temperature increases, the spectra show that the films become more liquid and less solid in character, within hundreds of nanometers of the interface. In addition, the spectra show that at the same temperature, films for more highly concentrated salt solutions have more liquid. For example, the coldest 0.01 M NaCl film at 257 K in the top panel of Figure 2 has a lower extinction at 3228 cm−1 (0.114 a.u.) than the coldest 0.001 M NaCl film at 257 K in the bottom panel (0.136 a.u.). This trend holds for 0.0001 M NaCl films as well; however, their spectra are essentially indistinguishable from those of pure water-ice films (data not shown).
IV. DISCUSSION A. Morphology of frozen films
The spectral evidence suggests the solid films of water and aqueous NaCl have liquid-like or liquid water present at the interface of the solid film and the germanium prism. For the higher concentration NaCl films (≥0.01 M) at warmer temperatures, this interfacial liquid layer lubricated the solid film, which could easily move on the surface of the prism. Water-ice and more dilute NaCl films (≤0.001 M), however, were frozen in place. These observations suggest that, with the exception of the warmest, most concentrated films, there was not a uniform liquid film between the ice and the Ge. Rather, there must have been some ice in contact with the prism, or any liquid present was not sufficient to lubricate the solid films. On the other hand, the amount of liquid observed increased somewhat over ∼10 min after a solution was frozen and while held at a constant temperature. This observation is likely due to the denser brine flowing down grain boundaries and/or the outer surface due to gravity and suggests that the liquid does not reside solely in isolated pockets. According to the two-component phase diagram for water and sodium chloride,7, 34 when a solution freezes, the NaCl should be excluded to a liquid brine phase of higher concentration while the solid portion is simply ice. Molecular dynamics simulations support the idea that sodium and chloride ions will be almost completely excluded by the ice lattice.35–37 Experimental studies also corroborate the phase diagram predictions of concentrated NaCl brine at the surface of frozen solutions.15, 38, 39 Thus we expect that NaCl will be almost exclusively located in the liquid portions of the films, even though we cannot detect it directly. In order to explore the idea of liquid located at and near the interface further, we calculated reflection spectra as described in Sec. II. We assumed a simple morphology similar to Figure 1, that is, a layer of liquid water with a uniform thickness t between the germanium prism and a semi-infinite solid ice film. The resulting calculated spectra for pure ice, pure water, and three values of t are shown in Figure 3. The calculated spectra show that as the thickness of the interfacial
J. Chem. Phys. 139, 244703 (2013)
FIG. 3. Calculated reflection spectra for liquid water, ice, and layers of varying thickness of liquid water at the interface of ice and Ge.
liquid layer increases from zero to hundreds of nanometers, the intensity of the ice peak decreases and a shoulder to the blue increases. These spectra also show an isobestic point at 3326 cm−1 since they are a linear combination of the spectra of two inter-related species. The similarity between the calculated spectra and the experimental spectra indicate that we are observing a combination of solid and liquid water. The model morphology does not allow the liquid thickness to vary with location at the interface, nor does it take into account liquid at grain boundaries, triple junctions, or inclusions. Because our films were frozen quickly and are polycrystalline, these more complex morphologies almost certainly contribute to our experimental spectra. In fact, when 0.01 M NaCl frozen film was annealed from −17 ◦ C to −5 ◦ C and back to −17 ◦ C, the amount of liquid observed increased dramatically, and persisted even when the film was cooled back down. This result indicates that there is a significant amount of liquid located in the interior of the frozen film when initially frozen. However, annealing a 0.001 M NaCl film shows an increase in liquid that decreases back to the initial amount when cooled back to −17 ◦ C. Since we preferentially probe the region near the interface (∼200 nm), the contribution of liquid from grain boundaries and inclusions is likely to be small and these liquids are not all isolated from the liquid at the interface.
B. Measurement of liquid fraction and equivalent thickness
Using the presence of the isobestic point and our simplified morphology, we can further use our experimental spectra to deduce the amount of liquid or liquid-like water near the interface. For a spectrum that exhibits a combination of both ice and liquid, the extinction at a particular wavelength is the absorptivity (or extinction) of each component times the fraction present of each component. Since there are only two components, their fractions sum to one. Therefore, the fraction of ice, cice , can be determined from the extinction of the spectrum, E, and the extinctions of pure, cold ice, εice , and pure
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-5
Walker et al.
J. Chem. Phys. 139, 244703 (2013)
liquid, εliq , all at a particular wavelength, cice =
E − εliq . εice − εliq
(8)
The extinction of the pure components (which are analogous to molar absorptivity in transmission spectroscopy) at a particular wavelength can be obtained from the spectra of pure ice and pure liquid. For ice films, the coldest ice spectrum and first completely melted water spectrum were used to measure the extinction values. This approach assumes there is no significant liquid-like layer at our coldest temperatures. Other ATR-IR studies of ice have found negligible QLL thickness below −10 ◦ C.40, 41 For aqueous salt films, a cold, pure ice film and a cold but completely melted film were used to find extinction values for ice and the melted film. The ice spectrum, when offset to 0 at 4000 cm−1 , closely matched the isobestic point of the salt film spectra. In this manner, the fraction of liquid (cliq = 1 – cice ) in each spectrum was determined. The liquid fraction of the spectra ranges from 0.01 for the coldest water to 0.81 for the warmest 0.01 M NaCl film. In ATR-IR, the signal from an absorbing species depends exponentially upon its distance from the reflecting interface.29, 30, 41 As a result, cliq is not a mass or volume fraction of liquid. In order to convert the fraction of liquid observed in a spectrum to a physical quantity, the location of the absorbing species relative to the interface must be known. For these experiments, it is not known how much of the liquid observed is adjacent to the interface and how much is located in grain boundaries or inclusions. However, the liquid observed can be referenced to a hypothetical homogenous layer of uniform thickness with an equivalent amount of water. Thus we can express our observations in a physically meaningful way. We refer to this quantity as “equivalent thickness” or “equivalent t” throughout. Once the fraction of ice is calculated from Eq. (8), then the equivalent thickness of the liquid layer, t, can be calculated from Eq. (6). The values of both the observed liquid fraction and the equivalent thickness are reported for water-ice and NaCl solutions in the supplementary material.42 For water-ice films, the ice peak at 3228 cm−1 was usually seen to decrease slightly (5%–10%) from 258 K to 273 K. Occasionally, only small, random differences were seen in the intensity of the ice peak while warming. This behavior is likely due to poor contact between the ice film and the Ge crystal; good contact where the IR beam intersects the interface is required for spectra of consistent intensity. Annealing a frozen film until it was just melted and refreezing usually resulted in better contact and a higher intensity ice peak. For high-intensity ice films showing a decrease with warming, 3228.3 cm−1 was used to calculate equivalent thickness (dp = 204 nm) since the largest change in the spectra was observed at this wavelength. The results are shown as triangles in Figure 4. The data from four separate experiments have been binned into 1◦ increments for clarity. The vertical error bars are the standard deviations for the data points in each bin. The equivalent thicknesses range from 1 nm at 260 K to 18 nm at 274 K. The two data points for solid ice films in the 274 K bin (a temperature above the melting point) may indicate that the film was not thermally equilibrated in
FIG. 4. Measured equivalent thicknesses of liquid-like water between Ge and ice as a function of temperature relative to melting point (supplementary material42 ). Triangles are data from this work for water-ice. Data for frozen D2 O, ×, and frozen D2 O with dissolved protein, +, are from Ref. 41 and were provided by Y. Furukawa.
the time allowed, or that as the film was in the process of melting, the thermocouple moved away from the interface. Since there were no solutes intentionally dissolved in these films, this liquid-like layer at the surface of the ice films may be the disordered interfacial layer, or QLL, commonly referred to in the literature.12 This disordered region arises from the asymmetry at the interface and is not a separate phase, making it distinct from the brines discussed below. Liquidlike water on ice has been observed in three experiments similar to ours. Ochshorn and Cantrell estimated the thickness of disordered water at the interface of ice with silicon to be ∼30 nm at −7 ◦ C.43 In an experiment that observed an entire ice film, including both the ice-substrate interface and the ice-air interface, the sum of liquid-like water ranged from
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
THE JOURNAL OF CHEMICAL PHYSICS 139, 244703 (2013)
Total reflection infrared spectroscopy of water-ice and frozen aqueous NaCl solutions Rachel L. Walker,a) Keith Searles,b) Jesse A. Willard,c) and Rebecca R. H. Michelsend) Department of Chemistry, Randolph-Macon College, P.O. Box 5005, Ashland, Virginia 23005, USA
(Received 2 September 2013; accepted 22 November 2013; published online 26 December 2013) Liquid-like and liquid water at and near the surface of water-ice and frozen aqueous sodium chloride films were observed using attenuated total reflection infrared spectroscopy (ATR-IR). The concentration of NaCl ranged from 0.0001 to 0.01 M and the temperature varied from the melting point of water down to 256 K. The amount of liquid brine at the interface of the frozen films with the germanium ATR crystal increased with salt concentration and temperature. Experimental spectra are compared to reflection spectra calculated for a simplified morphology of a uniform liquid layer between the germanium crystal and the frozen film. This morphology allows for the amount of liquid observed in an experimental spectrum to be converted to the thickness of a homogenous layer with an equivalent amount of liquid. These equivalent thickness ranges from a nanometer for water-ice at 260 K to 170 nm for 0.01 M NaCl close to the melting point. The amounts of brine observed are over an order of magnitude less than the total liquid predicted by equilibrium thermodynamic models, implying that the vast majority of the liquid fraction of frozen solutions may be found in internal inclusions, grain boundaries, and the like. Thus, the amount of liquid and the solutes dissolved in them that are available to react with atmospheric gases on the surfaces of snow and ice are not well described by thermodynamic equilibrium models which assume the liquid phase is located entirely at the surface. © 2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4841835] I. INTRODUCTION
It is now widely recognized that the surfaces of ice and snow in Earth’s environment participate in reactions that affect the composition of the local atmosphere. Notable among these reactions are bromine explosion and ozone depletion events.1 It is thought that bromide from sea salt is activated to bromine radicals through photochemistry and reactions on snowpack and sea ice surfaces. The reactivity of ice surfaces proves to be different in many cases from bulk aqueous chemistry in unexpected and as yet unexplained ways. The primary reaction products can be different when ice is involved. For example, in studies of interhalide formation, Br2 Cl− is the favored product from frozen media containing halides and acids, as opposed to Cl2 Br− , which is formed in liquid media containing otherwise identical reactants.2 Additionally, reaction rates on frozen surfaces are sometimes enhanced. For example, photolysis of p–nitroanisole was 40 times faster on ice than in aqueous media.3 Photolysis of harmine is also faster on ice, unless there are salts present in the frozen solution.4 Ozone reacts faster with halides on the surface of frozen solution than on liquid surfaces.5 On the other hand, the reaction of OH radical with benzene is suppressed at the surface of ice relative to liquid.6 One a) Present address: Department of Chemistry, West Virginia University,
Morgantown, West Virginia 26506, USA.
b) Present address: Department of Chemistry, University of Pennsylvania,
Philadelphia, Pennsylvania 19104, USA.
c) Present address: School of Law, University of Richmond, Richmond,
Virginia 23173, USA.
d) Author to whom correspondence should be addressed. Electronic mail:
[email protected].
0021-9606/2013/139(24)/244703/8/$30.00
reason reactions may happen faster at ice surfaces is the freeze concentration effect: solutes are excluded from the ice crystal lattice as it forms resulting in higher concentrations in the remaining liquid at the surface. This concentration enhancement was measured as greater than an order of magnitude for halide ions7 and an enormous three to six orders of magnitude for methylene blue.8 Differences in exclusion from the ice matrix, reaction products, and reaction rates likely result from different molecular-level interactions or mechanisms. Understanding the reactivity of ice surfaces at the molecular level requires a thorough description of the microphysical properties of the ice surface. Pure ice undergoes premelting; that is, the surface is composed of a layer of disordered water molecules9–11 as a result of the interface. The nature of this disordered interface, sometimes called the quasi-liquid layer (QLL), is the subject of much inquiry. Measurements of the thickness of the disordered interface vary greatly depending on the experimental technique used.11, 12 For example, atomic force microscopy measured a liquid-like layer of 90 nm close to the melting point,13 while a photoelectron spectroscopy study found only a couple nanometers of disorder at the surface.14 When there are solutes in the solution that freezes, the result is quite different. In particular, for salt solutions near the melting point of ice, a liquid brine layer forms at the surface.1, 15 This brine is a distinct phase in equilibrium with the ice as described by the phase diagram. The physical and chemical properties of a brine layer are similar to a liquid aqueous environment, whereas those of the disordered interface on water-ice are not necessarily similar. For example, HCl gas will dissociate and acidify the brine layer
139, 244703-1
© 2013 AIP Publishing LLC
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-2
Walker et al.
on frozen saltwater, but frozen freshwater with HCl exposure does not show the same increase in acidity.16 Reaction kinetics in the brine layer tend to be similar to bulk aqueous kinetics, as in the reaction of ozone with frozen salt solutions.17 Changes in acidity in the brine can occur with freezing, as in the case of frozen NaCl solutions resulting in a brine basic enough to catalyze the oxidation of gallic acid.18 Besides appealing to the empirical phase diagram for solute-water mixtures, several attempts have been made to treat the brine layer on a frozen surface as a thermodynamically stable phase in equilibrium with the bulk ice. Cho et al. assume the solutes are dilute and derive an expression for the maximum liquid water fraction.19 Kuo et al. follow a similar treatment, but include deviation from ideality, which is especially important for concentrated brines.20 The thermodynamics of a brine phase is not at issue; however, the idea that all the liquid will reside at the surface may be problematic. Experimental evidence suggests that in the case of nitrate, thermodynamics overpredicts the amount of nitrate at the liquid surface of frozen solutions.21 Without direct measurements or relevant parameterizations, models of snow-atmosphere interactions are currently based on questionable assumptions about how much liquid to incorporate at the surface of snow and what concentrations of solutes to use.22–25 In this work, we compare experimental IR spectra to calculated reflection spectra for liquid/solid water films with a simple two-layer morphology. We present measurements of the amount of liquid detected near the surface of frozen solutions measured via attenuated total reflection infrared spectroscopy (ATR-IR) over ranges of temperature and salt concentration.
II. METHODS A. Attenuated total reflection infrared spectroscopy
Frozen aqueous films were studied via ATR-IR to minimize the interference of water vapor in the spectra. Solutions were introduced to the top of a germanium prism embedded in a copper plate. The plate and germanium crystal were cooled via recirculating ethanol from a Huber ministat cc1 chiller. The ATR plate was positioned on a VeeMax II accessory (Pike Technologies) in the sample cell of a ThermoElectron Nicolet 6700 FT-IR spectrometer with a DTGS detector. A constant 45◦ angle of incidence was used. For water on germanium, the penetration depth of the IR beam is around 200 nm at this angle. (See Eq. (7)) The sample compartment, the ATR plate holder, and the area above the germanium crystal were purged with dry, CO2 -free air. Solutions (0.8–1 ml for water and 1.0 ml for NaCl solutions) were introduced to the Ge surface via a glass syringe. When the Ge crystal was ∼256 K, freezing occurred in less than a minute. No pressure was applied to the frozen films. The temperature of the frozen films was measured by a type K thermocouple (Omega) which was positioned in the film in contact with the Ge/solution interface. The temperature ranged from room temperature down to 256 K and was stable to ±0.1◦ . Background spectra (with a clean Ge crystal) were taken on the same day at approximately 263 K. All spectra are an average of 32 scans and were taken with 4 cm−1 resolution. Sodium chloride (Sigma
J. Chem. Phys. 139, 244703 (2013)
Aldrich, ACS reagent grade) was used without further purification. An initial solution of 1.006 M NaCl (1.026 mol/kg) in reagent grade water (Sigma Aldrich) was mixed, and then serial tenfold dilutions were made down to 10−4 M. In order to follow spectral changes in frozen films as a function of temperature, a solution was frozen at around 256 K. After a spectrum was taken, the temperature of the film was increased about a degree. The film was allowed to equilibrate for 10–20 min and then another spectrum was taken. This procedure was repeated until the entire film melted. Since the Ge crystal was cold, but the air above it was warmer, the films froze from the bottom up. There was no measurable temperature gradient between the interface with the germanium crystal and the top of a stable, frozen film. The air temperature above a frozen film was around 275 K. Thus, while frozen, the uncertainty in temperature is similar to that of the thermocouple (±0.1◦ ). However, some spectra show partially solid films at temperatures slightly above the melting point, e.g., 273.8 K for water-ice. Thus the uncertainty in the temperature may be higher, especially close to the melting point. It is also possible that films that were in the process of melting experienced a temperature gradient, and/or the thermocouple tip moved away from the interface. During melting, the ice was observed to float in the liquid, as expected. B. Spectral calculations
Reflection spectra were calculated from the Fresnel equations26 and the complex indices of refraction of ice and liquid water:27 cos θ − i sin2 θ − n221 r⊥ = , (1) cos θ + i sin2 θ − n221 n221 cos θ − i sin2 θ − n221 r|| = . n221 cos θ + i sin2 θ − n221
(2)
Here, r⊥ and r|| are the amplitude reflection coefficients for which the electric field vector of the light is perpendicular and parallel to the plane of incidence, respectively. The angle of incidence is θ and n21 = n2 /n1 , the ratio of refractive indices of water (n2 = n + ik)27 and germanium (n1 = 4.0275).28 Reflectance, R, is the product of the amplitude reflection coefficient with its complex conjugate. In this study, R = r⊥∗ r⊥ was sufficient to reproduce spectra for our system since r|| is small when the angle of incidence is less than the Brewster angle.26 Extinction, which is analogous to absorption in transmission spectroscopy, is given by E = − log
R , R0
(3)
where R0 = 1 since reflection is total when no absorbing film is in contact with the crystal. In this manner, the reflection spectra of liquid water and ice were modeled, showing excellent agreement to experiment, as discussed below. The complex refractive indices of water are slightly temperature dependent. The experimentally determined values
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-3
Walker et al.
J. Chem. Phys. 139, 244703 (2013)
Since the penetration depth depends on the media present at the reflecting surface, n21 for liquid water was used in Eq. (7). The thickness of the liquid film, t, is the only parameter in the model, and spectra were calculated for liquid layer thicknesses varying from 10 to 200 nm using Mathematica. In our experiments, dp varies from 230 nm at 2800 cm−1 to 163 nm at 3800 cm−1 . III. RESULTS FIG. 1. Simplified sketch of a frozen NaCl (aq) film on Ge. Only the bottom layer of brine and a portion of the ice film contribute significantly to reflection spectra.
from Zasetsky et al. were used: T = 263 K or 273 K for liquid water and T = 235 K for ice.27 Additionally, a non-absorbing solute changes the index of refraction very slightly when the concentration is high. For example, at 589 nm, the index of refraction changes from 1.33 for water to 1.36 for 4.3 mol/kg NaCl solutions.28 Since such changes are small, NaCl was not accounted for in our calculated spectra. In practice, the shape of the calculated water spectrum was slightly different than the shape of the experimental liquid spectra. No polarization filters were used in obtaining the experimental spectra. For the situation in which the absorbing film is liquid at the interface with the germanium prism, but ice at some point above the interface, the following simple model, depicted in Figure 1, was used. It was assumed that the total reflectance, R, for such a film can be reasonably represented as a linear combination of reflectance due to an ice film, Rice , and a liquid film, Rliq , R = cliq Rliq + cice Rice .
ATR-IR spectra were taken of frozen water and aqueous NaCl films as the temperature was increased incrementally until they melted. The concentration in the solutions ranged from 0.0001 to 0.01 M NaCl at room temperature. Higher concentration solutions (0.03–0.1 M NaCl) show distinct behavior, and their spectra are described in Ref. 31. In this study, we focus on the OH stretching region of the spectrum. The frozen films all show an initially intense peak at 3228 cm–1 that gradually decreases in intensity with increasing temperature. Upon complete melting, the spectra show a broad, less intense peak with a maximum around 3350 cm−1 . In Figure 2, two such experiments are shown: the top panel is 0.01 M NaCl and the bottom panel is 0.001 M NaCl. In both panels the spectrum of pure water-ice (∼258 K) is shown as a dashed line. As the temperature increases, the intensity at 3228 cm−1 falls considerably, while a shoulder around 3400 cm−1 increases. The broad peak of the fully melted films centered around 3350 cm−1 closely resembles that of liquid water.
(4)
R is proportional to the square of the electric field amplitude of the evanescent wave that penetrates the film in contact with the reflecting interface. Because this electric field amplitude decreases exponentially,29 the coefficients, cliq and cice , were scaled exponentially relative to the penetration depth. The penetration depth, dp , of the evanescent wave is defined as the distance from the interface at which the electric field amplitude falls to 1/e times its value at the interface.29 The axis perpendicular to the interface is z. Thus, for a liquid film thickness of t and a semi-infinite ice film at z > t, the coefficients may be defined as30 t exp(−2z/dp )dz = 1 − exp(−2t/dp ), cliq = 0∞ (5) exp(−2z/dp )dz 0
∞
exp(−2z/dp )dz cice =
t
= exp(−2t/dp ).
∞
(6)
exp(−2z/dp )dz 0
The penetration depth depends on wavelength, λ, and the indices of refraction, dp =
λ/n1
2π sin2 θ − n21
.
(7)
FIG. 2. Series of ATR-IR spectra showing the melting of a 0.01 M NaCl film (top) and a 0.001 M NaCl film (bottom). The dashed lines are spectra of pure ice at 259 K and 257 K, respectively. All spectra were offset to zero extinction at 4000 cm−1 . In the top panel, 5 out of 12 spectra are shown (for clarity) with temperatures of 257 K, 263 K, 267 K, 270 K, and 271 K. In the bottom panel, 7 out of 17 spectra are shown with temperatures of 257 K, 260 K, 263 K, 266 K, 270 K, 273 K, and 274 K.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-4
Walker et al.
The spectra also show an isobestic point in the vicinity of 3325 cm−1 , where all the spectra cross. This behavior indicates that there are two (or more) species present – the liquid (brine or water) and the solid ice – and that these readily interconvert.32, 33 The spectra taken at intermediate temperatures are combinations of the all-liquid and all-ice spectra. Thus, as the temperature increases, the spectra show that the films become more liquid and less solid in character, within hundreds of nanometers of the interface. In addition, the spectra show that at the same temperature, films for more highly concentrated salt solutions have more liquid. For example, the coldest 0.01 M NaCl film at 257 K in the top panel of Figure 2 has a lower extinction at 3228 cm−1 (0.114 a.u.) than the coldest 0.001 M NaCl film at 257 K in the bottom panel (0.136 a.u.). This trend holds for 0.0001 M NaCl films as well; however, their spectra are essentially indistinguishable from those of pure water-ice films (data not shown).
IV. DISCUSSION A. Morphology of frozen films
The spectral evidence suggests the solid films of water and aqueous NaCl have liquid-like or liquid water present at the interface of the solid film and the germanium prism. For the higher concentration NaCl films (≥0.01 M) at warmer temperatures, this interfacial liquid layer lubricated the solid film, which could easily move on the surface of the prism. Water-ice and more dilute NaCl films (≤0.001 M), however, were frozen in place. These observations suggest that, with the exception of the warmest, most concentrated films, there was not a uniform liquid film between the ice and the Ge. Rather, there must have been some ice in contact with the prism, or any liquid present was not sufficient to lubricate the solid films. On the other hand, the amount of liquid observed increased somewhat over ∼10 min after a solution was frozen and while held at a constant temperature. This observation is likely due to the denser brine flowing down grain boundaries and/or the outer surface due to gravity and suggests that the liquid does not reside solely in isolated pockets. According to the two-component phase diagram for water and sodium chloride,7, 34 when a solution freezes, the NaCl should be excluded to a liquid brine phase of higher concentration while the solid portion is simply ice. Molecular dynamics simulations support the idea that sodium and chloride ions will be almost completely excluded by the ice lattice.35–37 Experimental studies also corroborate the phase diagram predictions of concentrated NaCl brine at the surface of frozen solutions.15, 38, 39 Thus we expect that NaCl will be almost exclusively located in the liquid portions of the films, even though we cannot detect it directly. In order to explore the idea of liquid located at and near the interface further, we calculated reflection spectra as described in Sec. II. We assumed a simple morphology similar to Figure 1, that is, a layer of liquid water with a uniform thickness t between the germanium prism and a semi-infinite solid ice film. The resulting calculated spectra for pure ice, pure water, and three values of t are shown in Figure 3. The calculated spectra show that as the thickness of the interfacial
J. Chem. Phys. 139, 244703 (2013)
FIG. 3. Calculated reflection spectra for liquid water, ice, and layers of varying thickness of liquid water at the interface of ice and Ge.
liquid layer increases from zero to hundreds of nanometers, the intensity of the ice peak decreases and a shoulder to the blue increases. These spectra also show an isobestic point at 3326 cm−1 since they are a linear combination of the spectra of two inter-related species. The similarity between the calculated spectra and the experimental spectra indicate that we are observing a combination of solid and liquid water. The model morphology does not allow the liquid thickness to vary with location at the interface, nor does it take into account liquid at grain boundaries, triple junctions, or inclusions. Because our films were frozen quickly and are polycrystalline, these more complex morphologies almost certainly contribute to our experimental spectra. In fact, when 0.01 M NaCl frozen film was annealed from −17 ◦ C to −5 ◦ C and back to −17 ◦ C, the amount of liquid observed increased dramatically, and persisted even when the film was cooled back down. This result indicates that there is a significant amount of liquid located in the interior of the frozen film when initially frozen. However, annealing a 0.001 M NaCl film shows an increase in liquid that decreases back to the initial amount when cooled back to −17 ◦ C. Since we preferentially probe the region near the interface (∼200 nm), the contribution of liquid from grain boundaries and inclusions is likely to be small and these liquids are not all isolated from the liquid at the interface.
B. Measurement of liquid fraction and equivalent thickness
Using the presence of the isobestic point and our simplified morphology, we can further use our experimental spectra to deduce the amount of liquid or liquid-like water near the interface. For a spectrum that exhibits a combination of both ice and liquid, the extinction at a particular wavelength is the absorptivity (or extinction) of each component times the fraction present of each component. Since there are only two components, their fractions sum to one. Therefore, the fraction of ice, cice , can be determined from the extinction of the spectrum, E, and the extinctions of pure, cold ice, εice , and pure
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 138.26.31.3 On: Tue, 30 Sep 2014 08:29:46
244703-5
Walker et al.
J. Chem. Phys. 139, 244703 (2013)
liquid, εliq , all at a particular wavelength, cice =
E − εliq . εice − εliq
(8)
The extinction of the pure components (which are analogous to molar absorptivity in transmission spectroscopy) at a particular wavelength can be obtained from the spectra of pure ice and pure liquid. For ice films, the coldest ice spectrum and first completely melted water spectrum were used to measure the extinction values. This approach assumes there is no significant liquid-like layer at our coldest temperatures. Other ATR-IR studies of ice have found negligible QLL thickness below −10 ◦ C.40, 41 For aqueous salt films, a cold, pure ice film and a cold but completely melted film were used to find extinction values for ice and the melted film. The ice spectrum, when offset to 0 at 4000 cm−1 , closely matched the isobestic point of the salt film spectra. In this manner, the fraction of liquid (cliq = 1 – cice ) in each spectrum was determined. The liquid fraction of the spectra ranges from 0.01 for the coldest water to 0.81 for the warmest 0.01 M NaCl film. In ATR-IR, the signal from an absorbing species depends exponentially upon its distance from the reflecting interface.29, 30, 41 As a result, cliq is not a mass or volume fraction of liquid. In order to convert the fraction of liquid observed in a spectrum to a physical quantity, the location of the absorbing species relative to the interface must be known. For these experiments, it is not known how much of the liquid observed is adjacent to the interface and how much is located in grain boundaries or inclusions. However, the liquid observed can be referenced to a hypothetical homogenous layer of uniform thickness with an equivalent amount of water. Thus we can express our observations in a physically meaningful way. We refer to this quantity as “equivalent thickness” or “equivalent t” throughout. Once the fraction of ice is calculated from Eq. (8), then the equivalent thickness of the liquid layer, t, can be calculated from Eq. (6). The values of both the observed liquid fraction and the equivalent thickness are reported for water-ice and NaCl solutions in the supplementary material.42 For water-ice films, the ice peak at 3228 cm−1 was usually seen to decrease slightly (5%–10%) from 258 K to 273 K. Occasionally, only small, random differences were seen in the intensity of the ice peak while warming. This behavior is likely due to poor contact between the ice film and the Ge crystal; good contact where the IR beam intersects the interface is required for spectra of consistent intensity. Annealing a frozen film until it was just melted and refreezing usually resulted in better contact and a higher intensity ice peak. For high-intensity ice films showing a decrease with warming, 3228.3 cm−1 was used to calculate equivalent thickness (dp = 204 nm) since the largest change in the spectra was observed at this wavelength. The results are shown as triangles in Figure 4. The data from four separate experiments have been binned into 1◦ increments for clarity. The vertical error bars are the standard deviations for the data points in each bin. The equivalent thicknesses range from 1 nm at 260 K to 18 nm at 274 K. The two data points for solid ice films in the 274 K bin (a temperature above the melting point) may indicate that the film was not thermally equilibrated in
FIG. 4. Measured equivalent thicknesses of liquid-like water between Ge and ice as a function of temperature relative to melting point (supplementary material42 ). Triangles are data from this work for water-ice. Data for frozen D2 O, ×, and frozen D2 O with dissolved protein, +, are from Ref. 41 and were provided by Y. Furukawa.
the time allowed, or that as the film was in the process of melting, the thermocouple moved away from the interface. Since there were no solutes intentionally dissolved in these films, this liquid-like layer at the surface of the ice films may be the disordered interfacial layer, or QLL, commonly referred to in the literature.12 This disordered region arises from the asymmetry at the interface and is not a separate phase, making it distinct from the brines discussed below. Liquidlike water on ice has been observed in three experiments similar to ours. Ochshorn and Cantrell estimated the thickness of disordered water at the interface of ice with silicon to be ∼30 nm at −7 ◦ C.43 In an experiment that observed an entire ice film, including both the ice-substrate interface and the ice-air interface, the sum of liquid-like water ranged from
