Two Liquid Phases of Water in the Deeply Supercooled Region and

Apr 28, 2007 - Low-density liquid water is the mother of ice: on the relation between mesostructure, thermodynamics and ice crystallization in solutio...
2 downloads 0 Views 131KB Size
Download PDF
5628

J. Phys. Chem. B 2007, 111, 5628-5634

Two Liquid Phases of Water in the Deeply Supercooled Region and Their Roles in Crystallization and Formation of LiCl Solution Ryutaro Souda* Nanoscale Materials Center, National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan ReceiVed: NoVember 1, 2006; In Final Form: January 12, 2007

The properties of supercooled liquid water and the mechanism of crystallization in it were investigated using time-of-flight secondary ion mass spectrometry and reflection absorption infrared spectroscopy. The selfdiffusion of the water molecules commences at 136 K, and then the liquid-liquid phase transition occurs at 160-165 K. The latter is evidenced not only by the occurrence of fluidity but also by the formation of a LiCl solution. The infrared absorption band also changes drastically above 160 K due to crystallization of water (on the Au film) and the formation of LiCl solution (on the LiCl film). The immediate crystallization and dissolution of LiCl are thought to be characteristic of normal water that is created in a deeply supercooled region, indicating that viscous liquid water (T > 136 K) is transformed into supercooled liquid water at around 160 K. The crystallization kinetics is different between these two phases because the former (latter) involves nuclear growth (spontaneous nucleation). Without nuclei, crystallization is quenched below 160 K in the present experiment. It is suggested that the viscous liquid phase coexists at the surface or grain boundaries of metastable ice Ic.

1. Introduction Structural transformations of glassy and liquid water are attracting growing research attention. Amorphous solid water (ASW), which is formed by slow deposition of water molecules onto cold substrates, is structurally different from hyperquenched glassy water (HGW) prepared by rapid cooling of liquid water. ASW becomes identical to HGW after heating to about 110120 K.1 Another route to glassy water is pressure-induced amorphization of hexagonal ice.2 The high-density amorphous phase is called HDA (high-density amorphous ice), and the lowdensity phase produced by decompressing HDA and heating is called LDA (low-density amorphous ice). The transition between LDA and HDA is reversible and first-order. X-ray and neutron diffraction measurements suggest that LDA is structurally identical to HGW and annealed ASW.3 The glassy water thus prepared appears to become a highly viscous liquid above the glass transition temperature, Tg. Currently, the precise assignment of Tg is a subject of debate.4-9 Tg has been determined as 136 K in calorimetric studies,4,5 but the related endothermic peak is too small to assign a definite glass transition temperature.6-8 From a comparison of the Tg-scaled heat capacity between water and other inorganic glasses, Tg has been reassigned to 165 ( 5 K.6 Because glassy water finally crystallizes to ice Ic above 150 K,4 it remains unresolved whether supercooled liquid water, if any, is a thermodynamic extension of normal water or a distinct phase. Within the framework of polyamorphism,10-12 LDA and HDA should have corresponding liquid forms above Tg, termed low- and high-density liquids (LDL and HDL, respectively). It is hypothesized that the transition between HDL and LDL is first order as well, and terminates at a second critical point (Tc′ ) 220 K, Pc′ ) 100 MPa). The neutron diffraction study has revealed that LDA (HDA) has local order similarities to * E-mail: [email protected].

crystalline ice (liquid water);13 the similarities extend to the second and third coordination shells. Therefore, LDL is assigned to a tetrahedrally structured liquid with strong, straight hydrogen bonds, whereas HDL is characterized by weak, distorted hydrogen bonds. According to this second critical point hypothesis,10-12 normal water consists of microscopic domains of the HDL and LDL phases, whereas deeply supercooled water should be categorized as LDL. Because the domain structure of liquid is expected to become evident in the deeply supercooled regime,14 it is important to clarify the nature of deeply supercooled liquid water. For investigating structural transformations of water, the interaction of water with solute species is useful, because the ability to hydrate solute species relates to the microscopic hydrogen-bond structures of the water molecules. It is known that the nonpolar molecules can be incorporated in the bulk of the ASW film and are released abruptly at around 160 K.15,16 This phenomenon has been regarded as abrupt crystallization of water.15 On the other hand, we have assigned this phenomenon to the glass-liquid transition because the morphology of the thin water film changes at around 165 K following the abrupt release of alkane molecules incorporated in the bulk of the ASW film.16 The dominance of the supercooled liquid phase over crystal ice Ic has been claimed in the diffraction and imaging studies using transmission electron microscopy (TEM).17 The presence of the two characteristic temperatures (136 and 165 K) for the glass-liquid transition has been attributed to a kinetic problem;18 the onset of the self-diffusion of water molecules at Tg ) 136 K is followed by the occurrence of fluidity in the film after a certain aging time. On the other hand, because the calorimetric studies have suggested that homogeneous crystallization occurs above 150 K,4,19 the supercooled liquid phase is either present only in a narrow temperature range of 136-150 K (if Tg ) 136 K) or entirely absent (if Tg ) 165 K). Currently, the occurrence of direct crystallization from glassy water has

10.1021/jp0672050 CCC: $37.00 © 2007 American Chemical Society Published on Web 04/28/2007

Two Water Phases in the Deeply Supercooled Region been identified by calorimetric6,7 and dielectric9 studies, protontransfer measurement,20 and FTIR;21 the kinetics of crystallization has been discussed extensively using thermal desorption22-28 and infrared absorption27-30 spectroscopy. So far, very little has been discussed about the presence of the liquid phase despite the fact that crystallization most likely occurs in supercooled liquid water. In the present paper, we investigated the structural transformations of water in the deeply supercooled regime using temperature-programmed time-of-flight secondary ion mass spectrometry (TOF-SIMS) and reflection absorption infrared spectroscopy (RAIRS). These two techniques are complementary to each other. TOF-SIMS provides information about the composition and hydrogen-bond structure of the water molecules in the surface region, whereas the phase transition in the bulk is investigated by RAIRS. So far, the liquid-like nature of water has been explored by the TOF-SIMS analysis of the selfdiffusion of the water molecules and the morphology change of the thin water films.16,18 On the other hand, crystallization of ASW and HGW has been discussed extensively by IR absorption studies on the basis of the shape change in the O-H stretch band.27-32 More insights into the properties of supercooled liquid water would be gained from its role as a solvent. In this respect, hydration not only of the nonpolar molecules mentioned above but also of electrolytes like LiCl is of interest, because the former is believed to be caged in LDL,16,33 whereas the latter tends to dissolve in liquid water or HDL.34-36 The kinetics of crystallization and formation of aqueous LiCl solutions in a deeply supercooled regime was investigated from the RAIRS analysis in comparison with the TOF-SIMS result. 2. Experimental The TOF-SIMS and RAIRS experiments, respectively, were performed in separate ultrahigh-vacuum (UHV) chambers of base pressure below 1 × 10-8 and 5 × 10-8 Pa. The TOFSIMS measurement was made by bombardment of a sample surface (floated with a bias voltage of +500 V) with a pulsed He+ beam (2.0 keV, ∼10 pA/cm2), and the positive ions, extracted into a field-free TOF tube (∼65 cm) through a grounded stainless steel mesh (placed 4 mm above the sample surface), were detected using a channel electron multiplier. The Ni(111) surface was cleaned in the UHV chamber by heating up to 1200 K by electron bombardment from behind and then cooling to 10 K by means of a closed-cycle helium refrigerator. The polycrystalline LiCl film with a thickness of approximately 500 nm was deposited on it at room temperature. Isotopically labeled water molecules, such as H216O, H218O, and D216O, were admitted into the UHV chamber through high precision leak valves. The ASW film was deposited on the Ni(111) and LiCl surfaces by backfilling the UHV chamber with the water molecules. The coverage of the water molecules was determined from the evolution curves of sputtered ion intensities as a function of exposure. One monolayer (1 ML) of the water molecules was attained by exposure of ∼2.5 Langmuir () 1 × 10-6 Torr s). TOF-SIMS spectra were recorded every 30 s at a ramping speed of 5 K min-1. The infrared absorption measurements were made using a BioRad FTS40A FTIR spectrometer equipped with a liquid nitrogen-cooled mercury cadmium telluride (MCT) detector. The IR beam was reflected from a gold film deposited on a mirrorfinished Ni plate at a grazing angle. The substrate was cooled with liquid nitrogen to 85 K. The LiCl film was deposited in the UHV chamber on the gold substrate at room temperature.

J. Phys. Chem. B, Vol. 111, No. 20, 2007 5629

Figure 1. TOF-SIMS intensities of typical secondary ions sputtered from the water films grown on the Ni(111) surface as a function of temperature. The binary films consisting of (a) 20 ML H216O on 20 ML H218O and (b) 200 ML H216O on 200 ML H218O were formed at 100 K, and then the temperature was increased at a rate of 5 K min-1.

The spectra were taken over the wave number range 400-4000 cm-1 with an 8 cm-1 resolution. The temperature was increased at the same ramping speed as in TOF-SIMS, and 90 spectra were taken continuously in the temperature range 90-290 K. The water films were deposited from the mixture of D2O (5 mol %) in H2O to enable the decoupled OD stretching band of the HOD molecules to be studied.29 For this HOD concentration, a narrow OD stretch band was observed because of the absence of the OD-OD interactions. 3. Experimental Results 3.1. TOF-SIMS. Figure 1a shows TOF-SIMS intensities of typical secondary ions sputtered from a binary film consisting of 20 ML H216O on 20 ML H218O. The film was deposited on the Ni(111) substrate at 100 K. The evolution curve of the H318O+ ion reveals that the H218O molecules embedded underneath the H216O layer diffuse to the surface at temperatures above 140 K. This result shows that the onset of translational molecular diffusion is close to the conventional Tg of 136 K. On increasing temperature above 160-165 K, the Ni+ ion is sputtered abruptly from the substrate due to dewetting of the water film. The mixing of the water molecules is completed after dewetting. The dewetted water film finally evaporates above 180 K, where the H+ and H3O+ intensities decrease. Thus, the diffusion of individual water molecules at around 136 K is followed by a more drastic “phase-transition” at 165 K. The experimental result for a much thicker water film (200 ML H216O on 200 ML H218O) is shown in Figure 1b. In this case, the Ni+ ion is not sputtered until the water film evaporates at 190 K. The H318O+ ion evolves above 165 K, and the H+ intensity exhibits a dip at this temperature. The H+ ion tends to be sputtered from the “free OH group” via the bond-breaking mechanism, whereas the H3O+ ion basically originates from the proton-transfer reactions during collisions of water molecules.16 From a comparison of the results between Figure 1a,b, it is

5630 J. Phys. Chem. B, Vol. 111, No. 20, 2007

Souda

Figure 3. Intensities of typical secondary ions sputtered from the 200 ML H216O molecules deposited on the LiCl film (∼500 nm) as a function of temperature. Figure 2. Intensities of typical secondary ions sputtered from the binary water film of 200 ML H216O on 200 ML D216O formed on the Ni(111) substrate as a function of temperature.

revealed that the changes in morphology of the thicker water films can be monitored from the dip in the H+ intensity. In fact, dewetting occurs in ASW films thinner than ∼80 ML, whereas the dip in the H+ intensity is observed at 165 K in films as thick as 1600 ML. This result indicates that the phase transition at 165 K is a bulk phenomenon and that the top 80 ML of the water molecules take part in the morphology change of the film. In Figure 1b, therefore, the embedded H218O layer should not be exposed after the morphology of the H216O overlayer (200 ML) changes; the embedded H218O molecules need to travel through at least the 120 ML thick H216O layer to appear at the surface, indicating that the water molecules are diffusive in the high-temperature phase (T > 165 K). The diffusive nature of the water molecules was also confirmed in the H/D exchange experiments. Figure 2 shows the TOF-SIMS intensities from the layered water film consisting of 200 ML H216O on 200 ML D216O as a function of temperature. The H2O film is so thick that the embedded D2O layer is not exposed at 165 K, but the H/D exchange occurs efficiently above 165 K. It should be noticed that the intensities of the H+ and H3O+ ions increase markedly at 190 K, just before complete evaporation of the film occurs. Similarly, the D+ and D3O+ ions are emitted after the decay of the H+ and H3O+ ions. These results indicate that the isotope scrambling is not complete. The pure H2O and D2O domains, which avoid H/D exchange, appear to come from crystals. They are thought to be surrounded by the liquid-like phase and appear at the surface only after the isotope-scrambled layer evaporates completely. The same tendency is observed in Figure 1b for mixing between the H216O and H218O molecules, but it is not observed for the 20 ML thickness film in Figure 1a. In order to gain more insights into the properties of liquidlike water above Tg, the interactions between H2O and LiCl were analyzed. The TOF-SIMS intensities of typical secondary ions are shown in Figure 3. The 200 ML water molecules were deposited on the LiCl film at 100 K, and then, the temperature was increased. The Li+ and Li+(H2O) ions are sputtered from the surface above 165 K. The surface of the LiCl film buried

underneath the 200 ML ASW film is not exposed even after the film morphology changes. Therefore, the Li+ ion cannot be sputtered at 165 K unless LiCl mixes with the water film. The LiCl solutions should not be formed at the interface between the crystalline water and LiCl films below the eutectic temperature (∼199 K).37 Therefore, this result should be interpreted as the LiCl dissolving in liquid water or vise versa. The mixing occurs when much thicker water films (up to 1600 ML) are used.35 The Li+(LiCl) ion is sputtered from the crystalline LiCl film and appears only above 185 K. At this temperature, the water film evaporates and the LiCl surface is partly exposed. In the temperature range 165-185 K, therefore, the supercooled aqueous LiCl solution is formed. Above the evaporation temperature of the water film at 185 K, the Li+(LiCl) ion evolves more gradually than the Ni+ ion in Figure 1a; the H+ and (Li+(H2O) ions are sputtered with considerable intensities up to 250-260 K, indicating that the desorption of the water molecules from the aqueous LiCl solution is much slower than that from the pure water films. 3.2. RAIRS. The water molecule in the gas phase displays three modes of vibration, corresponding to symmetric stretch (V1), asymmetric stretch (V3), and bending (V2) of the OH bonds. The IR bands are broadened considerably in the condensed phase as a result of anharmonicity of the vibrations generated by extensive hydrogen bonding. So far, the OH stretching band is used for the discussion of phase transition to ice Ic,17,27,28 but precise assignment of the broadly peaked structures in the band is difficult. We have, in fact, measured the IR absorption spectra of pure H2O films (20-400 ML), but the shape of the stretching band and the red-shift value of the band due to the phase transition depend largely on the thickness and morphology of the film. To avoid these difficulties, we focused on a decoupled OD oscillator in 10 mol % HOD in H2O. In this HOD concentration, OD-OD interactions do not contribute to the bandwidth: the OD stretching band becomes narrow and the spectrum is very much simplified.29,30 The OD stretch band of HOD in the 200 ML H2O film is shown in Figure 4 as a function of temperature. The band from the as-deposited film at 85 K shows a maximum at 2453 cm-1, and the shape of the spectrum is almost unchanged up to 160

Two Water Phases in the Deeply Supercooled Region

Figure 4. RAIR spectra in the OD stretching region of HOD (10 mol %) in H2O as a function of temperature. The 200 ML of water molecules were deposited on the Au film at 85 K and the temperature was ramped at a rate of 5 K min-1.

K. After the film morphology changes, the band shifts abruptly to 2435 cm-1. The red shift and narrowing of the spectra have been explained by crystallization.27-30 Thus, this result shows that spontaneous nucleation and nuclear growth occur above 160 K. The IR spectrum after crystallization is asymmetric: the narrow crystal peak at 2435 cm-1 is overlapped by a broader peak, as revealed by the existence of a hump at around 2450 cm-1 and a tail toward the low-frequency side. The RAIR spectra observed here are quite similar to the FTIR spectra of HGW reported by Hage et al.,29 except that the peak positions are slightly different between these two experiments. They obtained the fraction of crystallized water by deconvolution of the spectra and claimed that water completely crystallizes at 155 K within 20 min. In the present experiment, however, complete crystallization is not observed until the water film evaporates if the sharp, symmetric peak centered at 2435 cm-1 is assigned to the absorption band of crystals, according to the assignment by Hage et al.29 Although the real absorption band of ice Ic is not known, the overlapped broad peak is most likely to be caused by the contribution of the coexisting amorphous (liquid) phase. The coexistence of crystals with liquid-like water is consistent with the TOF-SIMS observations in Figures 1 and 2; liquid-like water at the surface or grain boundaries of crystals is measured preferentially by surface-sensitive techniques like TOF-SIMS. It is thus suggested that ice Ic is metastable and equilibrated with the liquid-like phase. It should be noticed that the abrupt crystallization at 160165 K is apparently inconsistent with the results of the calorimetric study; the crystallization exotherm is observed in a differential scanning calorimetry (DSC) output from HGW over a temperature range 150-175 K.4,19 The DSC scan was made at a faster ramping speed (30 K min-1) than in the present study (5 K min-1), and hence, this discrepancy cannot be ascribed to crystallization kinetics. This point is very important, because most of the previous studies assign the crystallization temperature to 150 K on the basis of the calorimetric data, and it is

J. Phys. Chem. B, Vol. 111, No. 20, 2007 5631

Figure 5. RAIR spectra in the OD stretching region of HOD (10 mol %) in H2O as a function of temperature. The 200 ML of water molecules was deposited on the LiCl film at 85 K. The small structure at around 2350 cm-1 is due to uncompensated CO2.

believed that the supercooled liquid phase is effectively not accessible by heating the glassy water films. This is not the case in the present experiment. The liquid-like phase is dominant at least in the temperature range of 136-160 K, and it remains in the surface regions of ice Ic even above 160 K. Figure 5 shows the IR absorption band during formation of the aqueous LiCl solution for the 200 ML water film deposited on the LiCl film. Above 160 K, a broad peak appears at around 2500 cm-1 in addition to the narrow peak from the crystals. The broad peak is characteristic of the IR band of pure liquid water or water solutions. The crystallization competes with the formation of the LiCl solution, the latter prevails with time, and finally a uniform LiCl solution results. When a thinner water film (40 ML) is formed, the LiCl solution evolves predominantly over crystals as seen in Figure 6. The kinetics of crystallization and aqueous LiCl formation cannot be discussed directly from the present experiment because the former occurs in the bulk whereas the latter evolves at the interface. However, the formation of the LiCl solution apparently prevails over crystallization, as revealed from the results of the thinner water film. For much thicker water films deposited on the LiCl film (800 ML, not shown explicitly), on the other hand, the crystal peak evolves first but it disappears completely below the evaporation temperature of the water films, indicating that the crystals tend to dissolve into the LiCl solution. It appears likely that the liquidlike phase coexisting at the surface plays a role in the dissolution of ice Ic into the LiCl solution. If normal liquid water is formed from the pure ASW film, it should exhibit a broad IR absorption band similar to that of the LiCl solution. However, no such band is observed in Figure 4. The liquid-like water present at the surface of ice Ic above 165 K is a distinct phase, because no blue-shifted component is observed in the IR absorption band. Figure 7 shows the IR absorption band of the supercooled LiCl solution with increasing temperatures above 200 K. They were taken continuously after the spectra shown in Figure 5. The broad absorption band, characteristic of a liquid, persists

5632 J. Phys. Chem. B, Vol. 111, No. 20, 2007

Figure 6. RAIR spectra in the OD stretching region of HOD (10 mol %) in H2O as a function of temperature. The 40 ML of water molecules were deposited on the LiCl film at 85 K.

up to 220 K, after which the OD stretch band narrows suddenly due to crystallization into hydrates of LiCl. The blue shift of the peak position indicates that the number of the coordinated water molecules decreases. Indeed, the crystallization is associated with the reorganization of the water molecules, as revealed from the drop in the H+ and Li+(H2O) intensities at around 220-230 K (see Figure 3). 4. Discussion 4.1. Properties of Supercooled Liquid Water. The TOFSIMS study clearly revealed that the glass-liquid transition is a two-step process: the self-diffusion of water molecules commences at 136 K, and then, the film morphology changes at 160-165 K as a consequence of the evolution of fluidity. These two characteristic temperatures coincide with the conventional and reassigned Tg of water. The slow relaxation of glassy water is thought to be responsible for the presence of these characteristic temperatures. In fact, relaxation into the liquid phase requires a long aging time at lower temperatures (4-20 min for 150 > T > 135 K) as demonstrated by the isothermal TOF-SIMS study.18 The glass-transition temperature should be defined as the onset of the translational diffusion (Tg ) 136 K), so that the following change in the nature of liquid water at around 165 K should be assigned to the liquid-liquid transition. Thus, two liquid-like phases of water exist in the deeply supercooled regime. The second liquid phase links to normal liquid water as inferred from the high solubility of LiCl. Immediate crystallization is also characteristic of normal water that is unstable below 235 K. Therefore, we assign the second liquid phase to “supercooled liquid water”. The first liquid phase emerging just above Tg appears to be a distinct phase characterized by an ultraviscous nature (“ultraviscous water”). Although no crystalline LiCl is dissolved in it, the hydrophobic species, such as hydro- and fluorocarbons, are known to be accommodated in its bulk.16 The molecules incorporated in the D2O film survive at Tg ) 136 K and finally desorb at around 165 K just before the film morphology changes.

Souda

Figure 7. Same as in Figure 5, but evolutions of the spectra in a highertemperature region are shown.

Because the nonpolar solute species are thought to be accommodated in the cages of the water molecules, this result indicates that the hydrogen-bond structures are not changed at 136 K. This is also confirmed from the invariance of the IR absorption band as seen in Figure 4. At this temperature, therefore, glassy water (LDA) transforms into its liquid phase, LDL. LDL (or ultraviscous water) is regarded as a tetrahedrally structured liquid into which the hydrophobic solute species can be incorporated.33 On the other hand, LiCl does not dissolve in LDL because the Li+ and Cl- ions disturb the hydrogen bonds between the water molecules. The hydrogen-bond structures of the water molecules change drastically during the liquid-liquid transition (LDL f LDL + HDL), as evidenced by the occurrence of the dip in the TOF-SIMS H+ intensity. The HDL domain of supercooled liquid water (LDL + HDL) plays a role not only in the fluidity but also in the dissolution of LiCl, thereby forming the supercooled LiCl solution. The HDL domain is almost undetectable in the IR absorption spectra because it transforms immediately into ice Ic without the presence of LiCl. On the other hand, the LDL domain is thought to survive on the surface and grain boundaries of ice Ic even after crystallization; LDL + HDL f LDL + Ic. Because of the coexisting LDL, crystalline water exhibits a liquid-like nature, and is metastable with respect to the supercooled LiCl solution. The liquid-liquid transition can be identified as a small endothermic hump at around 165 K in the DSC output,4 although it is weakened considerably because of the occurrence of crystallization. In the deeply supercooled region, Tg < T < 1.2Tg, the inverse relationship between the translational diffusivity and viscosity (Stokes-Einstein relation) is known to be broken for “fragile” liquids like water.14 The Stokes-Einstein relation is based on the macroscopic hydrodynamics that assumes the liquid to be a continuum, and its breakdown is related to the formation of domain structures. This phenomenon has been explained in terms of the dynamical or spatial heterogeneity.14 The domain

Two Water Phases in the Deeply Supercooled Region structure is expected to occur during the phase transition from ultraviscous water (LDL) to supercooled liquid water (LDL + HDL). Although no experimental data for viscosity and diffusivity of water are available in this temperature range, the computer simulation reveals that most of the water molecules are caged by the hydrogen-bond network, with a small fraction of the molecules breaking out of their cages.12 The caging molecules are assignable to LDL, and the breaking-out molecules appear to be HDL. Thus, the occurrence of the liquidliquid transition following the glass-liquid transition is thought to be responsible for the long relaxation time of glassy water and the presence of the deeply supercooled region. 4.2. Kinetics of Crystallization. In the calorimetric study, the crystallization rate appears to increase above 165 K after the onset at 150 K.19 This behavior has been explained in terms of the two-stage crystallization.38 The crystallization kinetics is diffusion controlled, and hence, crystals should be nucleated and grow in the liquid phase. The two-stage crystallization can be explained as being the result of the different crystal-growth rates between the LDL phase (T < 165 K) and supercooled liquid water (LDL + HDL). However, crystal growth in the former is completely absent in the present experiment. This inconsistency with the previous study may be caused by the difference in the sample quality between ASW and HGW. The HGW sample used in the calorimetric study was prepared by vitrification of aqueous aerosol droplets at an extremely fast cooling rate, but it contains at least 5-30% crystal ice Ic,39 allowing the crystals to grow around the nuclei. This is likely to be responsible for the crystallization in the LDL phase of HGW (T < 165 K). In the present experiment, however, this channel is quenched completely because of the absence of such nuclei; nucleation and nuclear growth occur only after evolution of supercooled liquid water (LDL + HDL). Homogeneously crystallized films are currently believed to occur when water films are heated (or deposited) above 140150 K.20,21 Kay and co-workers discussed the crystallization of water at around 160 K using temperature programmed desorption (TPD).15 They concluded that crystallization occurs on the basis of the abrupt drop in the desorption rate of the water molecules and the explosive desorption of the CCl4 molecules embedded in the thick water films. The former has been explained by the fact that the crystalline ice has a lower desorption rate than the amorphous ice, whereas the latter is ascribed to the formation of cracks in the crystallized water films. We have reinvestigated these phenomena carefully in comparison with the TOF-SIMS results and concluded that these behaviors can be ascribed to the glass-liquid transition.16,36 Although crystallization occurs successively after evolution of supercooled liquid water (LDL + HDL), these two phenomena should be rigorously distinguished to avoid confusion. From the isotope-mixing measurements using TPD, Kay and co-workers claimed that the diffusivity of the water molecules in the range 150-157 K is smoothly connected to the high-temperature data.40 However, the applicability of the assumed one-dimensional transport model is doubtful, because the film morphology changes at 160-165 K. The diffusivity of supercooled liquid water (LDL + HDL) is virtually impossible to determine, because of its short lifetime. The abrupt desorption of the embedded CCl4 molecules may be caused by exposure of the embedded CCl4 layer due to dewetting, but this phenomenon is more likely to be ascribed to the dehydration that occurs during the liquid-liquid phase transition.36 Recently, Kimmel et al. reconfirmed dewetting of the water films formed on a Pt(111) surface by measuring TPD of

J. Phys. Chem. B, Vol. 111, No. 20, 2007 5633 physisorbed Kr.26 In contradiction to our conclusion, however, they explained this phenomenon in terms of nonwetting growth of crystalline ice. This interpretation, however, encounters at least three serious contradictions. First, they attribute the growth of dewetted crystalline ice to hydrophobicity of the water monolayer because the Pt(111) surface has been regarded as an ideal substrate for epitaxial growth of crystalline ice. In this respect, we have compared the dewetting kinetics of the water films formed on the hydrophobic graphite and hydrophilic Ni(111) surfaces;41 dewetting of the moderately thick water films has nothing to do with the substrates and the presence/absence of the template water monolayer because it is caused by the surface tension of the fluidized water films. Therefore, the observed nonwetting growth of the crystalline ice is explicable simply by the occurrence of crystallization in the droplets of supercooled liquid water (LDL + HDL), leaving the ordered water monolayer on the dewetted patches of Pt(111). The difficulty like hydrophobicity of the water monolayer is encountered if the results are forced to be interpreted in terms of the epitaxial growth of the crystalline ice. The crystallinity of the water films should be poor on any substrates because of the coexistence of the amorphous phase. Second, discussions of crystallization by TPD and isothermal desorption rely on the different desorption rates of water molecules between the amorphous and crystalline phases by assuming the zero-order desorption kinetics,22,23 but this most important assumption may not be correct because the surface area changes due to the film dewetting. It is apparent that the crystallization and morphological change of the film cannot be concluded simultaneously from TPD. The desorption rate of the water molecules may be decreased due to the surface tension of liquid-like water. Kimmel et al. envisioned that zero-order desorption kinetics is maintained for the dewetted water films.26 This is reasonable, because the water droplets and the monolayer of water on the Pt(111) substrate are equilibrated due to the presence of the liquid-like phase (LDL). Third, the growth of the crystal domains in the ASW films has been discussed extensively on the basis of the bimodal TPD peaks of the physisorbed nonpolar molecules;24,25,27 the high- (low-) temperature peak is assigned to the molecules desorbing from the crystal (amorphous) domains. Kimmel et al. observed quite similar structures in the TPD spectra of Kr from the D2O film,26 but they explained that the high- (low-) temperature peak comes from the dewetted surface area (the crystal grains), in contradiction to their own previous interpretations24 and those of others.25,27 In this context, it is apparent that what have been observed in TPD and TOF-SIMS are identical to each other. The morphological change of the film is not induced by crystallization: instead, crystallization occurs in the droplets of supercooled liquid water (LDL + HDL). The crystalline ice is thermodynamically most stable, but this appears not to be true for the crystalline ice Ic, because it dissolves into the supercooled LiCl solution. This result is likely to be caused by incomplete crystallization; i.e., the liquid phase (LDL) remains in the surface regions and grain boundaries of ice Ic that are equilibrated. The coexisting liquid-like water may play a role in the progressive transformation of metastable ice Ic to ice Ih. This picture is supported by the diffraction and imaging studies by Jenniskens et al.17 Using TEM, they observed the formation of viscous droplets during crystallization of a 175 ML water film to cubic ice on an amorphous carbon film. Both scattered droplets (at 155 K) and more continuous granular films (183 K) exhibit no crystallographic morphology expected from cubic ice. The crystallization is inhibited when only 30% has transformed, and most of the ice persists as a viscous liquid

5634 J. Phys. Chem. B, Vol. 111, No. 20, 2007 between 140 and 210 K prior to crystallization to ice Ih. This result is basically consistent with the IR absorption results shown in Figure 4. Jenniskens et al. also measured the RAIR spectrum of thin H2O films (70 ML).17 However, their OH stretch vibration band changes very slowly during annealing at 160 K, and no sharpening of the IR band, which is characteristic of crystallization, is observed. Their result is inconsistent with the present work. Without using the decoupled oscillator, at least the red-shift of the IR band associated with crystallization should be observed.28 Backus et al.27 reported a shape change of the O-H stretch band of the 45 ML H2O film within 500 s under the isothermal condition at 139 K. This phenomenon has been explained in terms of crystallization. However, the drastic shape change in the coupled O-H band cannot be attributed solely to the crystallization because it is related to the morphological change of the film.42 5. Conclusion The kinetics of the glass-liquid transition and crystallization of water was investigated by means of TOF-SIMS and RAIRS. In contrast to the conventional belief that crystallization occurs immediately above 150 K, the amorphous phase is dominant, at least up to 160-165 K, as revealed from the absence of the shape change in the IR absorption band. The onset of the translational diffusion at 136 K is assigned to the glass-liquid transition (LDA f LDL), whereas the successive change in the water’s properties at 165 K should be ascribed to the phase transition of LDL into supercooled liquid water (LDL + HDL). The formation of supercooled liquid water is evidenced by the simultaneous occurrence of the morphological change of the fluidized film and the dissolution of LiCl at around 165 K. The immediate crystallization into ice Ic is also characteristic of normal liquid water that is formed in the deeply supercooled regime. A long aging time is required for the structural relaxation because of the slow evolution of the HDL domains in the LDL phase. This behavior leads to two characteristic temperatures in the glass-liquid transition and is responsible for the controversies concerning the precise assignment of calorimetric Tg. The crystallization kinetics is largely different between the LDL phase and supercooled liquid water (LDL + HDL). The former involves only nuclear growth, whereas the latter is characterized by homogeneous nucleation. The crystallization in the LDL phase is quenched completely in the present experiment because of the absence of the crystallization nuclei in ASW, whereas the nuclear growth commences at 150 K in HGW because it includes a small amount of ice Ic. The crystallization of water is not complete, and ice Ic is equilibrated with the viscous liquid phase (probably LDL) present at the surface or grain boundaries, as evidenced by the overlapping of the broad IR band and the occurrence of the self-diffusion of the water molecules observed by TOF-SIMS. The dissolution of ice Ic into the supercooled LiCl solution may also be induced by the coexisting liquid phase.

Souda References and Notes (1) Debenedetti, P. G. J. Phys.: Condens. Matter 2003, 15, R1669. (2) Mishima, O. L.; Calvert, D.; Whalley, E. Nature 1984, 310, 393. (3) Bellissent-Funel, M. C.; Bosio, L.; Hallbrucker, A.; Mayer, E.; SridiDorbez, R. J. Chem. Phys. 1992, 97, 1282. (4) Johari, G. P.; Hallbrucker, A.; Mayer, E. Nature 1987, 330, 552. (5) Johari, G. P. J. Chem. Phys. 2002, 116, 8067. (6) Velikov, V.; Borick, S.; Angell, C. A. Science 2001, 294, 2335. (7) Yue, Y.; Angell, C. A. Nature 2004, 427, 717. (8) Giovambattista, N.; Angell, C. A.; Sciortino, F.; Stanley, H. E. Phys. ReV. Lett. 2004, 93, 047801. (9) Minoguchi, A.; Richert, R.; Angell, C. A. Phys. ReV. Lett. 2004, 93, 215703. (10) Poole, P. H.; Sciortino, F.; Essmann, U.; Stanley, H. E. Nature 1992, 360, 324. (11) Mishima, O.; Stanley, H. E. Nature 1998, 396, 329. (12) Debenedetti, P. G.; Stanley, H. E. Phys. Today 2003, 56, 40. (13) Finney, J. L.; Hallbrucker, A.; Kohl, I.; Soper, A. K.; Bowron, D. T. Phys. ReV. Lett. 2002, 88, 225503. (14) Debenedetti, P. G.; Stillinger, F. H. Nature 2001, 410, 259. (15) Smith, R. S.; Huang, C.; Wong, E. K. L.; Kay, B. D. Phys. ReV. Lett. 1997, 79, 909. (16) Souda, R. J. Chem. Phys. 2004, 121, 8676. (17) Jenniskens, P.; Banham, S. F.; Blake, D. F.; McCoustra, M. R. J. Chem. Phys. 1997, 107, 1232. (18) Souda, R. Chem. Phys. Lett. 2005, 415, 146. (19) Johari, G. P.; Hallbrucker, A.; Mayer, E. J. Chem. Phys. 1990, 92, 6742. (20) Cowin, J. P.; Tsekouras, A. A.; Iedema, M. J.; Wu, K.; Ellison, G. B. Nature 1999, 398, 405. (21) Fisher, M.; Devlin, J. P. J. Phys. Chem. 1995, 99, 11584. (22) Smith, R. S.; Huang, C.; Wong, E. K. L.; Kay, B. D. Surf. Sci. 1996, 267, L13. (23) Lofgren, P.; Ahlstrom, P.; Chakarov, D. V.; Lausmaa, J.; Kasemo, B. Surf. Sci. 1996, 367, L19. (24) Dohnalek, Z.; Kimmel, G. A.; Ciolli, R. L.; Stevenson, K. P.; Smith, R. S.; Kay, B. D. J. Chem. Phys. 2000, 112, 5932). (25) Safarik, D. J.; Meyer, R. J.; Mullins, C. B. J. Chem. Phys. 2003, 118, 4660. (26) Kimmel, G. A.; Petrik, N. G.; Donalek, Z.; Kay, B. D. Phys. ReV. Lett. 2005, 95, 166102. (27) Backus, E. H. G.; Grecea, M. L.; Kleyn, A. W.; Bonn, M. Phys. ReV. Lett. 2004, 92, 236101. (28) Ostblom, M.; Ekeroth, J.; Konradsson, P.; Liedberg, B. J. Phys. Chem. B 2006, 110, 1695. (29) Hage, W.; Hallbrucker, A.; Mayer, E.; Johari, G. P. J. Chem. Phys. 1994, 100, 2743. (30) Hage, W.; Hallbrucker, A.; Mayer, E.; Johari, G. P. J. Chem. Phys. 1995, 103, 545. (31) Delzeit, L.; Devlin, J. P.; Buch, V. J. Chem. Phys. 1997, 107, 3726. (32) Devlin, J. P.; Joyce, C.; Buch, V. J. Phys. Chem. A 2000, 104, 1974. (33) Paschek, D. Phys. ReV. Lett. 2005, 94, 217802. (34) Suzuki, Y.; Mishima, O. J. Chem. Phys. 2002, 117, 1673. (35) Souda, R. J. Phys. Chem. B 2006, 110, 14787. (36) Souda, R. J. Chem. Phys. 2006, 125, 181103. (37) Elarby-Aouizerat, A.; Jal, J.-F.; Chieux, P.; Letoffe, J. M.; Claudy, P.; Dupuy, J. J. Non-Cryst. Solids 1988, 104, 203. (38) Johari, G. P. J. Chem. Phys. 2002, 116, 8067. (39) Hallbrucker, A.; Mayer, E. J. Phys. Chem. 1987, 91, 503. (40) Smith, R. S.; Kay, B. D. Nature 1999, 398, 788. (41) Souda, R. J. Phys. Chem. B 2006, 110, 17524. (42) Souda, R. Unpublished work, 2007.