Orientational ordering in heteroepitaxial water ice on metal surfaces

Toshiki Sugimoto *ab and Yoshiyasu Matsumoto *c
aDepartment of Materials Molecular Science, Institute for Molecular Science, Myodaiji, Okazaki, Aichi 444-8585, Japan. E-mail: toshiki-sugimoto@ims.ac.jp
bPrecursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency (JST), Saitama 332-0012, Japan
cToyota Physical and Chemical Research Institute, 41-1 Yokomichi, Nagakute, Aichi 480-1192, Japan. E-mail: matsumoto@toyotariken.jp

Received 1st April 2020 , Accepted 11th June 2020

First published on 12th June 2020

Heteroepitaxial growth of crystalline ice thin films of water on metal substrates under ultrahigh vacuum provides an excellent opportunity to investigate the interior and surface structures of crystalline ice that are closely related to their physicochemical properties. Here we present the spectroscopic studies of the orientational ordering and the surface relaxation of crystalline ice films grown on two representative metal surfaces: Pt(111) and Rh(111). A versatile tool for exploring these structures is sum frequency generation (SFG) vibrational spectroscopy; homodyne detection of SFG signals serves as a good measure of orientational ordering in the interior of crystalline ice films while heterodyne detection enables us to determine the direction of water molecules at the interface with metal substrates, in the interior of crystalline ice films, and at their surfaces. Water molecules on the wetting layer of Pt(111) are preferentially oriented in H-down configuration, and the configuration is passed along into the interior of crystalline ice films. In contrast, water molecules on Rh(111) are adsorbed in a mixture of H-down and H-up configurations, leading to orientationally disordered crystalline ice films. The inter-layer distance at the top of the surface is modulated alternately in accordance with the orientation of molecules hydrogen bonded to the bilayer underneath. Therefore, the molecular orientation also plays an important role in their surface relaxation.

image file: d0cp01763a-p1.tif

Toshiki Sugimoto

Toshiki Sugimoto is an associate professor (principal investigator) at the Institute for Molecular Science. He received his BS degree in 2007 from Kyoto University and PhD degree in 2011 from the University of Tokyo. After few months research as postdoctoral fellow of Japan Society at the Promotion of Science at the University of Tokyo, he worked at Kyoto University as an assistant professor until 2018. He works on state-of-the-art molecular spectroscopy and nanoscopy of interfacial water and hydrogen systems. His main research interest is focused on unveiling unique structures, physicochemical properties and quantum dynamics of interfacial hydrogen bonds.

image file: d0cp01763a-p2.tif

Yoshiyasu Matsumoto

Yoshiyasu Matsumoto is an emeritus Professor of Kyoto University and a fellow of Toyota Physical and Chemical Research Institute. After graduated from Kyoto University, he completed his PhD at the University of Tokyo. Following appointments in the University of Pittsburgh, RIKEN institute, Institute for Molecular Science, and the Graduate University for Advanced Studies, he had been a Professor in Kyoto University till 2018. He received the Awards from the Chemical Society of Japan and the Japan Society for Molecular Science. His research interests are molecular spectroscopy and dynamics, ultrafast photoinduced processes at surfaces and interfaces, and the mechanism of photocatalysis.

1. Introduction

The impressive hexagonal symmetry of snowflakes is widely known, and the structure of ice crystals has attracted scientists for a long time. By neutron-diffraction crystallography various phases of ice crystals have been found in wide temperature and pressure ranges.1,2 Although all the phases of ice are characterised using their distinctive periodic structures, they are based on the tetrahedral coordination around a water molecule in which each molecule donates protons to two neighbours and accepts two from the others (Fig. 1B). Among those various phases, ice Ih is the most familiar one in our daily life. In this phase, bilayer planes of six-membered puckering rings are periodically stacked, being connected through inter-plane hydrogen bonds (stitching hydrogen bonds) as shown in Fig. 1A.
image file: d0cp01763a-f1.tif
Fig. 1 (A) Schematic illustration of an ideal full-bilayer termination structure of crystalline ice-Ih (0001) with one of the many possible random orientational alignments of water molecules under the ice rules. Oxygen of four-coordinated water molecules is coloured by red while that of three-coordinated water molecules at the topmost surface layer is coloured by blue. (B) The local tetrahedral coordination of water molecules in ice.

Although ice Ih is usually referred to as one form of crystalline ice, strictly speaking, it is not a truly crystalline material because the positions of hydrogen atoms are disordered; in other words water molecules in ice Ih display orientational disorder. This orientational disorder is the origin of the residual entropy in cryogenic temperature clarified by Pauling.3 The arrangements of hydrogen atoms and hence the orientation of water molecules in crystalline ice are dictated by the simple ice rules:4 (1) two of the hydrogen atoms are positioned closer to the oxygen atom, and (2) each of the other two hydrogen atoms is away from it toward a neighbouring oxygen atom. These rules impose restrictions on the orientation of water molecules in crystalline ice, resulting in constrained disorder in the global topological structure. Restrictions caused by these kinds of local rules have been found in many other materials such as spin glasses, spin ices, etc.5 In the case of ice Ih, the reorientation of a water molecule creates two kinds of defects: D and L defects with two hydrogen atoms and no hydrogen atoms between neighbouring oxygen atoms, respectively, while the proton transfer between neighbouring water molecules creates ionic defects (H3O+ and OH); thus, they violate the ice rules. Consequently, converting the constrained disordered phase to an ordered one requires cooperative reorientation and/or proton transfer in the entire crystalline ice without violating the ice rules; this imposes a substantial kinetic barrier, and overcoming the sluggish kinetics is a formidable task. In fact, partial conversion of ice Ih to ice XI, the counter part of the orientation-ordered phase of Ih, has been accomplished with catalysts, such as alkali hydroxides.6

Heteroepitaxial growth of crystalline ice on a metal substrate provides a good opportunity to investigate the structure of ice.7–10 The growth of crystalline ice on a metal substrate under ultrahigh vacuum minimises the effects of impurities on the structure of bulk ice and its surface. Because water molecules in the first layer directly interact with the metal, the adsorbate–substrate interactions play a critical role in the heteroepitaxial growth mode and the structure of crystalline ice grown. Moreover, if molecules in the first layer have a definite orientational ordering, the structure of the first layer may influence that in bulk crystalline ice grown on it. Because of these advantages the structure of crystalline ice and its growth mode have been intensively studied by the surface science community, and excellent reviews on this subject appeared in the past.11–13

Various surface science techniques have been utilised for clarifying the structure of monolayer water adsorbates and crystalline ice on metal substrates, including low electron energy diffraction (LEED), scanning tunnelling microscopy (STM), atomic force microscopy (AFM), X-ray photoelectron spectroscopy (XPS), infrared reflection absorption spectroscopy (IRAS), etc. Although these methods provide useful information on the structure, morphology, and the nature of hydrogen bonds in crystalline ice, they do not provide direct information of the molecular orientation and the degree of ordering. In this respect, infrared-visible sum frequency generation (SFG) spectroscopy deserves to be a versatile special tool for obtaining these pieces of information. Although SFG is often emphasised as a rare spectroscopic tool sensitive to the surface of materials, this is also very useful for probing the ordering of constituent atoms and molecules in the bulk.14,15 Moreover, heterodyne detection of SFG signals, which is described in more detail later, makes it possible to determine the direction of molecular orientation.16

This perspective describes the molecular orientation and its ordering in crystalline ice thin films grown on Pt(111) and Rh(111). This perspective is organised as follows. In Section 2 we summarise the structure and growth mechanisms of these systems investigated with conventional surface science techniques and density functional theory (DFT) in the past. In Section 3 we present the basics of SFG spectroscopy, in Sections 4 and 5 we describe the major findings obtained with SFG spectroscopy on the structures of hydrogen-bonded networks of ice films grown on these surfaces, focusing on the orientational ordering of water molecules. In the case of Pt(111), the heterodyne detection of SFG provides direct information on the preferential orientation of water molecules in the first wetting layer, and homodyne detection is invaluable to know how this directional ordering passes along into the bulk crystalline thin film. In the case of Rh(111), molecular orientation in the crystalline ice is disordered because of mixed configurations of molecules in the first layer. This orientational disorder of the film enables us to determine the structure of water molecules at the ice/vacuum interface by using the heterodyne detection of SFG. Finally, in Section 6 we conclude our findings and discuss the interplay of metal–water and intermolecular interactions that determine the structure and the molecular ordering in crystalline ice films grown on metal surfaces.

2. Structure and growth mechanism of crystalline ice films on Pt(111) and Rh(111) investigated with conventional surface science techniques

2.1. Monolayer adsorption

The structure and the growth mechanism of the crystalline ice film on Pt(111) have been extensively studied experimentally and theoretically, and the previous studies conducted with conventional surface science techniques (LEED, STM, AFM, XPS, and IRAS) and DFT calculations are summarised in the review of Hodgson.13 Water adsorbs intact on Pt(111) at low temperatures. At the early stage of crystal growth at a substrate temperature around 140 K, the first layer of water is completed; water wets the surface homogeneously. Although a simple commensurate structure of image file: d0cp01763a-t1.tif R30° was postulated in a previous study,17 He scattering18 and LEED measurements19–21 uncovered the complex extended periodic patterns: image file: d0cp01763a-t2.tif R25.3°, which is further compressed to image file: d0cp01763a-t3.tif R16.1° at the saturation of the wetting layer (Fig. 2A and B).
image file: d0cp01763a-f2.tif
Fig. 2 (A) STM image of the image file: d0cp01763a-t4.tif R16.1° superstructure of the saturated first layer grown on Pt(111) at 140 K.22 (B) Schematic illustration of the image file: d0cp01763a-t5.tif R16.1° superstructure.23 (C) STM image of a 2-dimensional water island on Rh(111) at 120 K in a submonolayer coverage.24 (D) Schematic illustration of the image file: d0cp01763a-t6.tif R30° structure.25

The H-down orientation, in which one of the hydrogen atoms of a water molecule points down toward the surface, was suggested to be favoured with X-ray absorption spectroscopy26 under the assumption of the image file: d0cp01763a-t7.tif R30° structure although this assumption is not consistent with the more extended and dense image file: d0cp01763a-t8.tif structures. IRAS measurements19,27–29 showed that the wetting layer has few (if any) dangling OH pointing to the vacuum, suggesting that the H-up orientation is rare in the wetting layer.

The DFT calculations by Feibelman and co-workers23,30,31 predicted the optimised image file: d0cp01763a-t9.tif structures. In these structures, H-down water molecules face directly the surface at one end of hydrogen and are laterally connected with hydrogen bonds through another end of hydrogen; thus, there are no dangling OH bonds in the wetting layer, which is consistent with the IRAS measurements.19,27–29 Thus, both experiments and theories suggest that the H-down configuration is favoured in the first layer. But, the direct experimental proof for the H-down orientation in the wetting layer needed to await heterodyne detection measurements of SFG, which is described in Section 4.

In contrast to the extensive studies of the first-layer water molecules on Pt(111), a few studies have been reported for elucidating the structure and the growth mechanism of crystalline ice films on Rh(111). Pioneering studies on the adsorption and the wetting behaviour of water molecules on Rh(111) were conducted with various techniques including IRAS, STM, LEED and temperature-programmed desorption (TPD) of D2O instead of H2O.24,25 There are couple of differences in the structure and desorption kinetics in comparison with those of Pt(111). First, the commensurate image file: d0cp01763a-t10.tif R30° structure is formed in the first layer on Rh(111) (Fig. 2C and D). Second, molecules in the first layer on Rh(111) desorb with the quasi-first-order (fractional-order) kinetics,25,32 whereas those on Pt(111) desorb with the quasi-zero-order kinetics (Fig. 3).19 Third, the first-layer water molecules adsorb on Rh(111) in mixed configurations of D-up and D-down, which was speculated on the basis of the IRAS measurements.24,25

image file: d0cp01763a-f3.tif
Fig. 3 (A) TPD traces of H2O from Pt(111) for 0.1, 0.45, 1.0, and 1.7 BL coverages at a heating rate of 0.65 K s−1.19 (B) TPD traces of D2O from Rh(111) for 0.47, 0.57, 0.68, 0.85, 1.00 and 2.92 BL coverages at a heating rate of 1.3 K s−1.25 The D2O-TPD traces were shifted from the original data by −4 K (isotope effect) for straightforward comparison with H2O-TPD traces.

2.2. Thick crystalline ice film

AFM and STM investigations clarified the growth mechanism of thick crystalline ice films on Pt(111).33 The adsorption of water on the wetting layer creates 3-dimensional (3D) clusters (Stranski–Krastanov mode). The clusters coalesce to make domain boundaries, and are interconnected by forming spirals. As the film thickness increases by more than 100 layers, the double spirals dominate to form a perfect hexagonal ice. Thus, the thick crystalline ice Ih film grows with a pseudo layer-by-layer mode.

In the case of Rh(111), STM images show that the 2-dimensional (2D) islands of water molecules (Fig. 2C) are connected at the saturation coverage of the first layer, and 3D ice crystallites grow on the first water layer with further adsorption, i.e., the typical Stranski–Krastanov growth. Xe TPD shows that the desorption of Xe from the first layer decreases with increasing D2O layers but a few percent of the first layer still remains uncovered with multilayers. Thus, it takes a larger amount of water adsorbates to make 3D islands to collapse completely with each other.

The hydrogen-bonded OH stretching band is a good indicator for examining the structure of hydrogen-bond networks in ice crystallites formed. Fig. 4A shows the IRAS spectra of the OH stretching band of the H-diluted (∼10%) HDO ice film with a similar thickness (∼25 BL) grown on Pt(111)34 and Rh(111) at ∼140 K. Both the spectra show that the peak of the stretching band is located at ∼3280 cm−1; this peak wavenumber is smaller than those of amorphous ice crystals and rather in good agreement with that of bulk HDO ice Ih, indicating that the films grown under these conditions are crystalline ice, and not amorphous ice. On the basis of the correlation between the wavenumber of the band peak of the isotope diluted OH oscillator and the intermolecular O–O distance RO–O in the crystalline ice (Fig. 4B),35–37 we can estimate the RO–O of the crystalline ice films grown on the metal surfaces to be 2.76 ± 0.02 Å; this is identical to that of bulk ice Ih and XI.38

image file: d0cp01763a-f4.tif
Fig. 4 (A) IRAS spectra of the hydrogen-bonded OH stretching band of H-diluted D2O ice films (25 BL thick, hydrogen mole fraction ∼0.10) grown on Pt(111) (red)34 and Rh(111) (blue) at ∼140 K. The peak positions of the band of bulk crystalline ice Ih and amorphous ices (LDA: low density amorphous ice, HDA: high density amorphous ice) are indicated in figure. (B) The correlation between the frequency of the OH stretching mode and the average intermolecular O–O distance.37

3. SFG spectroscopy for probing orientational ordering

3.1. Second order nonlinear susceptibility

The interaction of light with an ensemble of dipoles creates a macroscopic polarisation,
Pχ(1)E + χ(2)EE + χ(3)EEE + …, (1)
and the macroscopic polarisation acts as a new light source. Here χ(1) is the linear susceptibility and χ(n) is the n-th nonlinear optical susceptibility. As the light field amplitude |E| increases, various nonlinear optical processes originating from the nonlinear terms become prominent. In this perspective we focus on SFG, one of the second order processes governed by χ(2). The second-order term of eqn (1) generates the macroscopic polarisation oscillating the angular frequency of the sum of the frequencies of two incoming light pulses, P(ω1 + ω2), resulting in the generation of outgoing SFG signals with the intensity proportional to |P(ω1 + ω2)|2.

SFG serves as a versatile tool for the vibrational spectroscopy of adsorbates when one of the incoming infrared (IR) light is in the molecular finger print region.39–45 The second-order nonlinear susceptibility χ(2) includes a resonant term composed of molecular hyper-polarisability β represented as46

image file: d0cp01763a-t11.tif(2)
where ω is the frequency of incident IR light; ωq and Γq are the resonant vibrational frequency and the damping constant of the q-th vibrational mode, respectively; Aq is composed of the infrared transition matrix element and the derivatives of Raman polarisability with respect to the nuclear coordinate of the q-th mode. Thus, this term and hence SFG intensity are resonantly enhanced when ω matches with ωq. In other words, infrared-visible SFG can be interpreted as an IR transition accompanied with anti-Stokes Raman scattering.47–50 Thus, a vibrational mode of interest is SFG active if this mode is both infrared and Raman active.

In the case of crystals with inversion symmetry or molecular aggregates with complete random molecular orientation, the SFG signal generated at each site in the crystal or molecule in the aggregate is destructively interfered. Namely, χ(2) = 0 for these materials, implying that SFG signals do not emerge from the interior of materials. However, SFG signals are generated at the surface of the materials because the inversion symmetry is lost at the surface of the crystal and the molecules at the surface of randomly oriented materials tend to align one way or the other to some extent. This is why SFG is usually mentioned as a surface sensitive spectroscopy.

In addition to this useful feature of SFG spectroscopy, there is another advantage of using SFG. Let us consider molecular crystals. If molecules in a crystal are oriented in one direction, the SFG signals from the molecules in the crystal are constructively interfered because of the loss of inversion symmetry, resulting in strong SFG signals coming out of the crystal. In this case, the SFG intensity from the bulk crystal outweighs that from the surface. In other words, SFG can be used for monitoring the ordering of molecular orientation in the molecular crystal.

3.2. Homodyne and heterodyne detection of SFG

Optical spectroscopy usually detects intensity I ∝ |E|2 rather than the electric field E itself of the signal light. Note that the molecular hyper-polarisability and hence χ(2) provide information on the direction of a molecule because they are proportional to the transition dipole fixed in the molecular frame. Because conventional SFG (homodyne detection) measures the intensity of SFG light proportional to |χ(2)|2, the information on the direction of a molecule is lost. Moreover, because χ(2) is a complex number containing a real and an imaginary part, Reχ(2) and Imχ(2), respectively, the spectral line shape is complicated and frequently distorted because both the parts contribute to SFG intensity, i.e., ISFG ∝ |Reχ(2) + Imχ(2)|2. In particular, the distortion could be significant if nonresonant terms in χ(2) are not negligibly small in comparison with the resonant term.

How can we retrieve information on the direction of a molecule by SFG spectroscopy? A detection method called phase sensitive detection or heterodyne detection is suitable for this purpose (hereafter referred to as heterodyne detection).15,16,34,51–53 In this detection scheme, we measure the sum of SFG signals from a sample and an additional source called a local oscillator both generated by the same infrared and visible pulses. Let Es and El be the electric fields of SFG signals from the sample and the local oscillator, respectively. The SFG intensity measured by this method is given as,

ISFG ∝ |Es + El|2 = |Es|2 + |El|2 + 2Re(EsEl*). (3)
The observed signal intensity consists of the sum of the signal intensities from the sample and the local oscillator (the first and second terms, respectively) and the interference term of the two optical fields (the third term). Note that this interference term is composed of the electric fields of the signal from the sample and the local oscillator and the phase difference between the two. Thus, given the phase difference between Es and El, one can obtain the amplitude and phase of the signal field Es. Because the electric field of the signal light is proportional to χ(2), the phase and the amplitude of χ(2), and thereby Imχ(2) that is proportional to the transition dipole, are obtained. Consequently, this method using the optical interference allows us to retrieve the information on the direction of a molecule.

The heterodyne detection of SFG has another advantageous feature when it is applied to samples consisting of different species. Because the signal intensity with the homodyne detection of SFG is proportional to |χ(2)|2, it always has contributions from Reχ(2) in addition to Imχ(2); this makes spectral features complex and makes it difficult to decompose the spectra of mixed samples to each component. In contrast, the imaginary part of the spectrum of each species is free from the interference with the real part and the spectral intensity of each species is simply additive. Therefore, the spectra obtained with the heterodyne detection can be analysed similarly as in conventional infrared absorption and Raman spectroscopy.

3.3. Molecular orientation probed by SFG

Here we describe in more detail how to obtain the direction of molecules of the crystalline ice films grown on metal surfaces. SFG spectra are recorded with various combinations of polarisation of light involved in the SFG process. The polarisation of light is referred with respect to the incident plane: the light is in p-polarisation when the electric vector of light is parallel to the plane while that is in s-polarisation when the electric vector is vertical to the plane. The combination of polarisation is referred by using three indices i, j, and k: i for the polarisation of sum frequency; j for visible; and k for infrared light. Among the possible eight polarisation combinations, this perspective focuses on the ppp combination because this combination gives the most intense signal light in the case of adsorbates on metal surfaces as a result of optical interference between the incident and reflected light waves.54

Let the laboratory-fixed coordinates and the molecule-fixed ones be X, Y, and Z and a, b, and c, respectively (Fig. 5A). Because the Fresnel coefficients greatly reduce the field components parallel to the metal surface,48,49,54 the second order nonlinear susceptibility for ppp-SFG on the metal substrate is dominated by the ZZZ component of χ(2) in the laboratory fixed coordinate, AZZZ,q.54

image file: d0cp01763a-f5.tif
Fig. 5 (A) Relationship between the laboratory-fixed (X, Y, and Z) and the molecule-fixed (a, b, and c) coordinate systems for an OH oscillator of water adsorbed on the metal surface. (B) Classification of OH oscillators in the n-th bilayer (BL) of the hexagonal ice film with a thickness of Θ BLs. Only OH oscillators relevant to discussion are shown. The inter-bilayer OH oscillators in parallel to the Z-axis and point toward the ice surface and substrate are labelled as Bn-up (green spheres) and Bn-down (blue spheres), respectively, while intra-bilayer OH oscillators nearly perpendicular to the Z-axis and lean toward the ice surface and substrate are labelled as Bnp-up (cyan spheres) and Bnp-down (magenta spheres), respectively.

The component AZZZ,q is related to the properties of the molecule itself including the polarisability α and the dipole moment μ with respect to the normal coordinate Q of the q-th vibrational mode, and the directional cosines of the l-th molecule with respect to the surface normal,

image file: d0cp01763a-t12.tif(4)
The relationships: (â·)l,m = sin[thin space (1/6-em)]θl,msin[thin space (1/6-em)]ψl,m, ([b with combining circumflex]·)l,m = sin[thin space (1/6-em)]θl,mcos[thin space (1/6-em)]ψl,m, (ĉ·)l,m = cos[thin space (1/6-em)]θl,m, ∂αaa/∂Q = ∂αbb/∂Q,49 and (∂αaa/∂Q)/(∂αcc/∂Q) ∼ 0.3255–57 lead to the approximate form of AZZZ,q as,
image file: d0cp01763a-t13.tif(5)
With regard to the OH stretching band of water, there are four types of OH oscillators in the n-th layer (Fig. 5B): inter-bilayer H-up (Bn-up) and H-down (Bn-down) OH; intra-bilayer H-up (Bnp-up) and H-down (Bnp-down) OH. The inter-bilayer OH oscillators are directed along the Z-axis while the intra-bilayer ones are almost perpendicular to the Z-axis. If the crystalline ice thin film has net orientational ordering, AZZZ,q is simplified as34
image file: d0cp01763a-t14.tif(6)
where N is the number of water molecules in a bilayer of hexagonal ice: N ∼ 1.1 × 1019 m−2 and the brackets labelled Bn denote the ensemble average over inter-bilayer OH in the n-th bilayer. Note that the contributions of intra-bilayer OHs to AZZZ,q are much smaller than those of inter-bilayer OHs because cos[thin space (1/6-em)]θ is +1/3 (Bnp-up) and −1/3 (Bnp-down) while +1 (Bn-up) and −1 (Bn-down); thus, the contributions of intra-bilayer OHs to AZZZ,q can be neglected. Eqn (6) indicates that, if the crystalline ice film grows with net preferential orientational ordering, the SFG amplitude of the OH-stretch band increases with the film thickness. In contrast, if the ice film with random orientation grows, the amplitude is zero. Therefore, the SFG signal intensity can be used as a probe of orientational ordering in the ice film. Moreover, the sign of Imχ(2) of the OH-stretch band obtained by the heterodyne detection directly provides the direction of net orientational ordering: positive and negative signs for net-H-up and net-H-down OH oscillators, respectively.

4. Water orientation and orientation ordering in crystalline ice films on Pt(111) unveiled with SFG spectroscopy

4.1. Growth of orientationally ordered ice

Fig. 6A shows the SFG spectra (|χ(2)|2) of the OH stretching vibration band of HDO crystalline ice films deposited on Pt(111) at 140 K as a function of nominal film thickness Φ. The choice of HDO has been made because HDO ice has the following advantages over neat H2O ice: because both the intramolecular coupling between OH and OD vibrations and the intermolecular coupling among OH groups are small, the band profile of OH stretching band is very simple. As shown in Fig. 6B, the intensity of the OH band (|χ(2)|2) monotonically increases with Φ and does not saturate.15 Almost the same Φ dependence was observed during the growth of H2O and D2O crystalline ice films.34 These findings indicate that some ordering of water orientation exists in the crystalline ice films.
image file: d0cp01763a-f6.tif
Fig. 6 (A) Homodyne SFG (|χ(2)|2) spectra of the hydrogen-bonded OH stretch band of the HDO crystalline ice film deposited on Pt(111) at 140 K as a function of film thickness. (B) Adsorbed amount Φ dependence of |χ(2)|2 (circle). (Inset) Φ dependence of |χ(2)| and the result of curve fitting with the function: image file: d0cp01763a-t15.tif.

To examine the orientation of water molecules at the metal interface in more detail, we performed heterodyne-detection measurements to obtain Imχ(2) spectra at the early stage of epitaxial growth of the crystalline ice film on Pt(111) (Fig. 7). The water molecules in the first layer interacting directly with the platinum surface have a negative peak at 3370 cm−1 and its intensity is saturated at Φ ∼ 1 BL, indicating that the hydrogen atom of one OH group of water molecules in the wetting layer is directed to the platinum substrate: the H-down configuration. With an increase in the layer of water, another band with a peak at 3260 cm−1 grows, which is characteristic to the hydrogen bonded OH stretching band in the bulk. Note that the sign of the band is also negative, indicating that the water molecules adsorbed in the second layer reflect the orientation of the water molecules in the first layer. This is because the water molecules in the second layer interact through hydrogen bonds with the water molecules in the first layer, so that their orientation is restricted by the molecules in the first layer. In other words, the directional orientation passes along from the first layer into the crystalline ice thin film.

image file: d0cp01763a-f7.tif
Fig. 7 SFG (Imχ(2)) spectra of HDO crystalline ice films grown on Pt(111) obtained with heterodyne-detection of SFG as a function of Φ. Small non-resonant background was subtracted from these spectra.

The H-down orientation of water molecules at Pt(111) is in accordance with the prediction of the first principles calculation23 and is not inconsistent with the IRAS results19,27–29 showing a negligibly small free OH band of the wetting layer. The DFT calculation further suggested that the adsorption of water molecules in the second layer flips the orientation of some fraction of water molecules in the first layer to H-up orientation.23 Considering the precision of DFT calculations for estimating the energetics of subtle changes, one should be cautious about the prediction of the adsorption-induced flip-flop in the first layer. In contrast, the heterodyne-detected SFG measurements give a more decisive conclusion on this process. The constant intensity of |Imχ(2)| of the peak of the wetting layer (∼3370 cm−1) observed during the multilayer adsorption indicates that the adsorption-induced flip-flop does not happen or the fraction of water molecules flip-flopped is too small to detect, if any. This implies that the flip-flop motion is kinetically prohibited at 140 K or the molecules in the first layer is strongly pinned due to the interactions with the platinum substrate and the inter-adsorbate interactions in the hydrogen bond network of the first layer.

4.2. Orientation ordering across the film

Although a large surface voltage would emerge due to the orientational ordering in the crystalline ice film, such a surface voltage has not been observed in the measurements of work function and surface voltage during multilayer growth of crystalline ice films.13,20,58 Thus, the orientational ordering observed in the SFG measurement apparently contradicts with the previous measurements. How can we reconcile this contradiction?

Moreover, note that the ordering in the crystalline ice films is not necessarily perfect. Fitting the Φ dependence in Fig. 6B with the function,

|χ(2)|2Φm, (7)
results in m ∼ 2 for Φ < ∼30 BL and m = 1 for Φ > ∼30 BL. Let pn be the net-orientational ordering of the n-th layer: pn = 1 for complete order; pn = 0 complete disorder. In the case of layer-by-layer growth, m = 2 if pn is perfectly transmitted to the (n + 1)-th layer (pn = pn+1). Thus, the observed Φ-dependence of m appears to indicate some drastic changes in the transmission of ordering information around Φ = 30 BL. However, note that the growth mode of the ice film on Pt(111) changes from the SK mode to the pseudo-layer-by-layer mode with increasing Φ as described in Section 2. One can show that the observed Φ dependence of m is reproduced by assuming that the transition of the growth mode changes at Φ = 30 BL and image file: d0cp01763a-t16.tif holds irrespective of Φ. Therefore, the net-orientational order is not lost abruptly but gradually deteriorates throughout the overlayer of crystalline ice.15

There must be some mechanism to prevent from perfect ordering during the film growth. One plausible mechanism is the screening of the depolarisation field.59–61 Ferroelectric ordering of polar molecules in a thin film creates a huge depolarisation field across the film.62 If both sides of the film are attached to a metal, the metal electric field compensates well the depolarisation field to stabilise the film. Because the crystalline ice film in the current study is grown on the metal substrate in vacuum, the interface of the film with vacuum is unscreened. Thus, the H-down polarisation in the film causes a negative polarisation-charge density. Because this polarisation-charge density induces upward band bending of the valence band of ice, this crosses the Fermi level near the platinum substrate, resulting in electron injection from the crystalline ice film to the metal.15 The leftover holes are stabilised as hydronium cations through subtraction of H atoms from the neighbouring water molecules as H2O+ + H2O → H3O+ + OH.63,64 The mobile hydronium cations (protons)65 would be redistributed in the film to shield the negative polarisation charge, resulting in the formation of the stabilised crystalline ice film.

The distribution of hydronium cations across the crystalline ice film was estimated to be image file: d0cp01763a-t17.tif,15 where n is the number of bilayers in the crystalline ice film. The introduced hydronium cations form defects that allow the violation of the ice rules to randomise orientational ordering during the crystalline ice film growth. Therefore, this randomisation is the origin of the little work-function change reported in the Kelvin probe measurements.20,58

4.3. Thermal stability of orientational ordering

Transition from bulk ice XI to Ih is believed to be the first-order type if the ice rules strictly govern the configurations of water molecules in the hydrogen-bonded network. As described in the previous section, the crystalline ice films grown on Pt(111) likely have ionic defects as well as Bjerrum L and D defects that violate the ice rules. Thus, it is interesting to examine the thermal stability of orientational ordering in the crystalline ice films grown on Pt(111). For this purpose it is useful to measure homodyne SFG signals, the indicator for orientational ordering, as a function of temperature. Because ice starts to sublime around 145 K, it is necessary to perform TPD measurements simultaneously for evaluating the loss of water molecules out of the film.

Fig. 8 shows the temperature dependence of |χ(2)| of the hydrogen-bonded OH band of HDO derived from homodyne SFG signal intensity together with the desorption rate measured simultaneously at a heating rate of 0.10 K s−1. The intensity of |χ(2)| decreases with increasing temperature even in the temperature range below 145 K where sublimation of ice is negligible (Fig. 8A). As depicted with a red thin solid line, the decrease in the SFG intensity is well accounted for by the temperature dependence of Raman hyper-polarisability β. By taking into account this temperature dependence of β and the loss of ice layer due to sublimation, we plot the order parameter image file: d0cp01763a-t18.tif in Fig. 8B. The steep drop of η(T) starting at around T = 150 K is caused by the randomisation of orientational ordering in the crystalline ice film.

image file: d0cp01763a-f8.tif
Fig. 8 (A) Simultaneous measurements of the temperature dependence of |χ(2)| intensity (red open circle) and the desorption rate of HDO molecules (blue dot) for the crystalline ice film (330 BL) grown on Pt(111) at 135 K. The film was pre-cooled to 110 K for measurement after the film growth. Three horizontal black arrows indicate the temperature ranges of heating (+0.10 K s−1), cooling (−0.05 K s−1) and reheating (+0.10 K s−1) processes. The thick black line is the adsorbed amount of the film estimated from the TPD curve and the thin red line is the decay of |χ(2)| due to the temperature dependence of molecular hyper-polarisability. (inset) Enlarged view of the temperature dependence of |χ(2)| below 160 K. (B) Order parameter η normalised at ∼110 K (red circle). The red solid line is the thermal randomisation curve calculated by a statistical theory taking account of the ice rules and their partial violation with a defect formation energy of ELD = 92 ± 5 meV and an orientational anisotropic energy of E0 = 11 ± 1 meV. The blue dotted line is a result of curve fitting with an empirical function of (TcT)γ with Tc = 174 K and γ = 0.10.

H2O as well as D2O molecules in the crystalline ice thin film grown on Pt(111) also show the H(D)-down configuration at the interface with the metal substrate similar to HDO crystalline ice films. And, a similar curve of η(T) has been observed in both H2O and D2O crystalline ice films grown on Pt(111).34

The slopes of the decay of η(T) of these films are not much steeper than that of bulk XI ice,6 indicating that the order–disorder transition of molecular orientation is close to the second order transition. However, fitting η(T) with an empirical function for the second order-transition η(T) = (TcT)γ failed, implying that the transition is not purely second-order. The statistical model66 is more appropriate to describe the thermal stability of the orientational ordering of crystalline ice films. This model assumes that the reorientation of water molecules takes place under the short-range constraint of the ice rules that is partially violated through the formation of Bjerrum L and D defects. Thus, molecular reorientation is characterised with two terms: the energy difference between H-up and H-down orientation E0 and the formation energy of an L- and D-defect pair ELD.

Fitting the order parameter of HDO crystalline ice films (330 BL thick) epitaxially grown on Pt(111) with this model gives ELD = 92 ± 5 and E0 = 11 ± 1 meV.15 The value of ELD is much smaller than that of bulk ice Ih (ELD ∼ 500 meV). This is a manifestation of the second order type transition of the crystalline ice films; the statistical model predicts that the transition from the orientation-ordered to the disordered phase follows the first order transition if ELDE0.

The small ELD of the crystalline ice films probably stems from the stabilisation of L and D defects in the films. Because L and D defects are negatively and positively charged, respectively, hydronium cations spontaneously introduced into the film during the deposition process would efficiently split these defect pairs by trapping L defects preferentially. The stabilisation of L defects and the resultant local structure distortions would lower effectively the apparent defect formation energy.

Almost the same ELD and E0 were observed for H2O and D2O crystalline ice films on Pt(111), and the small isotope effect on Tc was also confirmed in the thermal stability of orientational ordering of these two systems: D2O crystalline ice shows a shift of +4 K from Tc of H2O crystalline ice.34 The small isotope shift of Tc of crystalline ice is in stark contrast to that of other high-Tc hydrogen-bonded ferroelectric materials.67–70

5. Structures of hydrogen-bond networks at the surfaces and interfaces of crystalline ice on Rh(111) unveiled with SFG spectroscopy

In contrast to the crystalline ice film on Pt(111), the homodyne SFG measurement of crystalline ice on Rh(111) shows that |χ(2)|2 does not change in the range of film thickness from 100 to 180 BL. This suggests that the orientations of molecules are completely disordered inside the crystalline ice film.

The heterodyne SFG measurement displays more details of the orientational disorder in the film. Fig. 9 shows the Imχ(2) spectrum of the HDO crystalline ice film (1 BL thick) grown on Rh(111). It reveals a positive OD dangling band and a negative OH stretching band, indicating that water molecules in the first BL are a mixture of H(D)-up and H(D)-down configurations.71 This is consistent with the previous IRAS results.24,25

image file: d0cp01763a-f9.tif
Fig. 9 Imχ(2) spectrum of dangling OD and hydrogen bonded OH stretch bands for 1 BL isotope diluted HOD ice films on Rh(111) measured at 95 K. The dangling OD band at 2718 cm−1 is normalised by the isotopic concentration ratio of the sample.

The broad negative peak at 3310 cm−1 changes to the ‘bipolar’ spectra composed of a positive and a negative peak even at 1.5 BL (Fig. 10A). The intensity of the bipolar spectrum increases with the coverage of HDO and is saturated at around 120 BL. The Xe TPD indicates that the total terrace area of the ice islands facing vacuum increases with the amount of adsorbed water molecules, and it reaches some constant value when most of the islands collapsed on each other (Fig. 10C).24 As shown in Fig. 10B, the coverage dependence of |χ(2)| mimics the growth of the total terrace area estimated with the result of Xe TPD. This suggests that the SFG signal originates from the top of ice islands. The origin of the SFG signal was confirmed by titrating D2O on the HDO crystalline ice film: the SFG intensity of the OH-stretching band was steeply decreased by deposition of a few bilayers of D2O, indicating that the deposition of additional D2O ice bilayers converts the surface region of the original ice film into bulk.54 Thus, this makes it clear experimentally that the SFG signal from the orientationally disordered ice crystalline films is contributed by only a few layers at the vacuum/ice interface.

image file: d0cp01763a-f10.tif
Fig. 10 (A) Heterodyne SFG (Imχ(2)) spectra in the hydrogen-bonded OH stretch region of the HOD crystalline ice films on Rh(111) as a function of coverage. The spectra were recorded at 95 K. The spectra for coverages higher than 1 BL are multiplied by factors indicated in the figure and each spectrum is arbitrarily shifted vertically. (B) Coverage dependence of the peak area of |χ(2)| (red squares), and the surface area of ice islands estimated from the Xe TPD data in ref. 24 (blue triangles). (C) The schematic images of the S–K growth of crystalline ice films on Rh(111) where red lines indicate the terrace of the ice crystallites.

In addition to the major contributions to the SFG signals from only a few layers at the vacuum/ice interface, we note that the SFG signals of inter-bilayer OH oscillators (Bn-up, Bn-down) dominate over those of intra-bilayer OH oscillators (Bnp-up, Bnp-down) as described in Section 3.3. Thus, the crystalline ice grown on Rh(111) provides a good opportunity to probe the stitching hydrogen bonds at the vacuum/ice interface by using SFG spectroscopy.

To understand the peculiar dispersion-like feature of the spectra in Fig. 10A, we simulated the Imχ(2) spectra with QM/MM calculations combined with classical molecular dynamics (MD) for ice Ih(0001).54 Fig. 11A and B show the contributions of each bilayer to the SFG intensity and the sum of the contributions, respectively. The largest contribution comes from the first bilayer (B1) at the top of the ice film. The first and the second bilayers are connected by stitching hydrogen bonds of B1-down and B2-up OH. The calculations show that B1-down and B2-up OH contribute a negative and a positive peak to Imχ(2), respectively, and the peak of B2-up is redshifted from the peak of B1-down. Thus, the contributions from the two stitching bonds result in the dispersion-like feature. The deeper into the crystal, the narrower the splitting between the two opposite peaks and the smaller the difference in the amplitude of both signals. This is caused by the recovery of inversion symmetry in the interior of crystalline ice.

image file: d0cp01763a-f11.tif
Fig. 11 The ZZZ-component of resonant Imχ(2) spectra derived from the QM/MM calculation of the hydrogen-bonded OH-stretching band of isotope-diluted ice Ih(0001). (A) Spectra of Bn-down OH (blue), B(n + 1)-up OH (green) and their sum (red, Bn-down OH + B(n+ 1)-up OH) for n = 1 and 2. (B) Spectra of B1-down OH + B2-up OH (dashed), B2-down OH + B3-up OH (dotted) and their sum (solid).

The splitting of the two peaks with opposite signs stems from the hydrogen-bond strength of the stitching bonds. The MD calculations reveal that B1-down OH experiences larger librational fluctuation (δϕ ∼ 9.3°) than B2-up OH (δϕ ∼ 8.9°) as shown in Fig. 12B. This indicates that the stitching bond of B1-down OH is more flexible in libration motion than that of B2-up OH. The difference in flexibility causes inequivalent weakening and elongation between the stitching hydrogen bond of B1-down OH and B2-up OH: RO–O of B1-down OH is longer than that of B2-up OH. This is the origin of the splitting of the two peaks. Moreover, RO–O is the largest at B1-down OH and gradually decreases as going deeper into bulk. Thus, the difference in hydrogen-bond in the first two layers mainly contributes to the splitting.

image file: d0cp01763a-f12.tif
Fig. 12 (A) Schematic illustration of the inter-bilayer H-bond network connected with the B1-down OH bond and B2-up OH bond. The first solvation shells of water molecules with B1-down and B2-up OH bonds are marked with blue and green shadow, respectively. The values of ΔRO–O and δϕ for inter- and intra-bilayer H-bonds are also shown. Hydrogen atoms in the intra-bilayer hydrogen-bond network are not shown and the number attached to the oxygen atom (red sphere) is the coordination number of H-bond for each molecule. (B) Structural parameters ΔRO–O (filled square) and δϕ (open circle) for the inter-bilayer hydrogen bond calculated from the MD simulations. The black dashed line is guide to the eye.

The question arises: why is the hydrogen-bond strength of B1-down OH weaker than that of B2-up OH? Considering the environment around the stitching bond of B1-down OH and B2-up OH, we note that the coordination number of water molecules is different: while the water molecules in the B1 layer are three-coordinated, those in the B2 layer are four-coordinated (Fig. 12A). This difference makes the water molecules in the B1 bilayer more flexible than those in the B2 bilayer. In fact, the libration motion of B1-down OH is more excited according to the calculations. Thus, the inter-bilayer surface hydrogen bond formed by B1-down OH is weaker than that formed by B2-up OH.

6. Conclusion and remarks

In this perspective, we have described that the orientation of water molecules is influenced by the metal substrate on which crystalline ice thin films are grown. All results are obtained by utilisation of infrared-visible sum frequency generation spectroscopy. This spectroscopy serves as a versatile method in multi-fold ways: (1) the SFG signal intensity with homodyne detection is a good measure for the ordering of molecular orientation; (2) the sign of the SFG signal derived from heterodyne detection provides the direction of molecular orientation; (3) SFG is surface sensitive in the case of materials with inversion symmetry or completely random structures.

We have focused on two metal surfaces: Pt(111) and Rh(111). A series of SFG measurements15,34 have clearly demonstrated that water molecules in the first layer at the interface with Pt(111) have a preferential orientation: H-down configuration. This orientation in the first layer continues to propagate into the overlayer of the crystalline ice film although the orientational ordering is gradually deteriorated as going away from the metal substrate. In contrast, the crystalline ice film grown on Rh(111) does not have a unanimous orientation configuration; water molecules at the metal surface have both H-down and H-up configurations, leading to the growth of the orientationally disordered multilayer crystalline ice films.

An interesting question is why the net orientational order of water molecules emerges on Pt(111) but not on Rh(111). One might argue that this substrate-dependent orientational ordering in the first layer originates due to the difference in water–metal interactions. In analogy with the structures of interfacial water molecules at biomolecular sites72 the differences in the structure of water molecules on the two metal surfaces could be a sign of difference in hydrophobicity at the interface.73,74 However, this concept cannot be applied to the current cases because both surfaces are almost equally hydrophilic, which was clearly confirmed from the TPD measurements (Fig. 3):19,25,71 the water molecules in the multilayer start to desorb around ∼140 K that is much lower than the desorption temperature of water in the first layer directly interacting with Pt(111) and Rh(111) surfaces. This indicates that water–metal interactions are stronger than water–water interactions; thus, both the metal surfaces are hydrophilic.

As noted in Section 2.1, there are some other important differences between the two systems, which are relevant to the question raised. First, water molecules in the saturated first layer on Pt(111) form a dense image file: d0cp01763a-t19.tif R16.1° super-structure whereas those on Rh(111) form a simple image file: d0cp01763a-t20.tif R30° structure (Fig. 2). Second, the desorption behaviours of water molecules are quite different; the first-layer water molecules on Pt(111) desorb in the quasi-zero-order kinetics while those on Rh(111) desorb in the quasi-first-order kinetics (Fig. 3). These stringent differences between the two substrates imply that not only water–metal interactions but also the lateral water–water interactions come into play. Thus, a delicate balance between the two interactions would ultimately determine the orientation in the first layer and thereby the orientational ordering of overlays in crystalline ice films. The larger lattice mismatch of Pt(111) (6.3%) than that of Rh(111) (3.1%) in comparison with the basal plane of ice Ih could be one factor tuning the balance between the two interactions: the water–metal interaction is stronger than the intermolecular one on Rh(111) whereas they are well balanced on Pt(111). But, the detailed understanding of the mechanism of ordering cannot be gained only by the comparison between the two systems. It is necessary to extend measurements over various metal substrates and to perform sophisticated first-principles calculations with high precision for clarifying the correlation between the interactions and the orientational ordering.

The structure of water ice crystallites at the other end, i.e., at the interface with vacuum, has also been elucidated. Because the ice crystallite is truncated so that the periodicity along the c-axis is lost, the structure of the ice surface is relaxed. With the aid of quantum chemical and molecular dynamic calculations, we have shown that the ice surface has a ripple structure and the surface relaxation persists for a couple of surface layers. The structural relaxation at the surface of the ice crystal is quite relevant to melting75,76 and reactions.10,77 Therefore, further SFG studies are necessary in combination with various microscopy techniques and large-scale computational simulations for understanding not only the static structure but also the dynamic features of the ice surface at the molecular level.

Conflicts of interest

There are no conflicts to declare.


This work was supported by MEXT KAKENHI: Grant-in-Aid for Scientific Research on Innovative Areas, No. 16H00937; JSPS KAKENHI Grant-in-Aid for Specially Promoted Research, No. 17H06087; Grant-in-Aid for Scientific Research (A), No. 19H00865; Grant-in-Aid for Scientific Research (B), No. 19H02681; Grant-in-Aid for JSPS Research Fellow, No. 17J08362. We thank Kazuya Watanabe for discussions and contributions in the early stage of the project. We are also grateful to Yuji Otsuki, Norihiro Aiga, Tatsuya Ishiyama, and Akihiro Morita for their invaluable contributions.

Notes and references

  1. T. Bartels-Rausch, V. Bergeron, J. H. E. Cartwright, R. Escribano, J. L. Finney, H. Grothe, P. J. Gutiérrez, J. Haapala, W. F. Kuhs, J. B. C. Pettersson, S. D. Price, C. I. Sainz-Díaz, D. J. Stokes, G. Strazzulla, E. S. Thomson, H. Trinks and N. Uras-Aytemiz, Ice structures, patterns, and processes: A view across the icefields, Rev. Mod. Phys., 2012, 84, 885–944 CrossRef CAS.
  2. C. G. Salzmann, Advances in the experimental exploration of water's phase diagram, J. Chem. Phys., 2019, 150, 060901 CrossRef PubMed.
  3. L. Pauling, The structure and entropy of ice and of other crystals with some randomness of atomic arrangement, J. Am. Chem. Soc., 1935, 57, 2680–2684 CrossRef CAS.
  4. J. D. Bernal and R. H. Fowler, A theory of water and ionic solution, with particular reference to hydrogen and hydroxyl ions, J. Chem. Phys., 1933, 1, 515–548 CrossRef CAS.
  5. A. Ortiz-Ambriz, C. Nisoli, C. Reichhardt, C. J. O. Reichhardt and P. Tierno, Colloquium: Ice rule and emergent frustration in particle ice and beyond, Rev. Mod. Phys., 2019, 91, 041003 CrossRef CAS.
  6. Y. Tajima, T. Matsuo and H. Suga, Phase transition in KOH-doped hexagonal ice, Nature, 1982, 299, 810–812 CrossRef CAS.
  7. N. Materer, U. Starke, A. Barbieri, M. A. Van Hove, G. A. Somorjai, G.-J. Kroes and C. Minot, Molecular surface structure of ice(0001): Dynamical low-energy electron diffraction, total-energy calculations and molecular dynamics simulations, Surf. Sci., 1997, 381, 190–210 CrossRef CAS.
  8. H. Bluhm, D. F. Ogletree, C. S. Fadley, Z. Hussain and M. Salmeron, The premelting of ice studied with photoelectron spectroscopy, J. Phys.: Condens. Matter, 2002, 14, L227–L233 CrossRef CAS.
  9. N. Avidor and W. Allison, Helium diffraction as a probe of structure and proton order on model ice surfaces, J. Phys. Chem. Lett., 2016, 7, 4520–4523 CrossRef CAS PubMed.
  10. F. Kato, T. Sugimoto, K. Harada, K. Watanabe and Y. Matsumoto, Unveiling two deuteration effects on hydrogen-bond breaking process of water isotopomers, Phys. Rev. Mater., 2019, 3, 112001 CrossRef CAS.
  11. P. A. Thiel and T. E. Madey, The interaction of water with solid surfaces: Fundamental aspects, Surf. Sci. Rep., 1987, 7, 211–385 CrossRef CAS.
  12. M. A. Henderson, The interaction of water with solid surfaces: Fundamental aspects revisited, Surf. Sci. Rep., 2002, 46, 1–308 CrossRef CAS.
  13. A. Hodgson and S. Haq, Water adsorption and the wetting of metal surfaces, Surf. Sci. Rep., 2009, 64, 381–451 CrossRef CAS.
  14. X. Su, L. Lianos, Y. R. Shen and G. A. Somorjai, Surface-induced ferroelectric ice on Pt(111), Phys. Rev. Lett., 1998, 80, 1533 CrossRef CAS.
  15. T. Sugimoto, N. Aiga, Y. Otsuki, K. Watanabe and Y. Matsumoto, Emergent high-Tc ferroelectric ordering of strongly correlated and frustrated protons in a heteroepitaxial ice film, Nat. Phys., 2016, 12, 1063–1068 Search PubMed.
  16. R. Superfine, J. Y. Huang and Y. R. Shen, Phase measurement for surface infrared-visible sum-frequency generation, Opt. Lett., 1990, 15, 1276–1278 CrossRef CAS PubMed.
  17. L. E. Firment and G. A. Somorjai, The surface structures of vapor-grown ice and naphthalene crystals studied by low-energy electron diffraction, Surf. Sci., 1976, 55, 413–426 CrossRef CAS.
  18. A. Glebov, A. P. Graham, A. Menzel and J. P. Toennies, Orientational ordering of two-dimensional ice on Pt(111), J. Chem. Phys., 1997, 106, 9382–9385 CrossRef CAS.
  19. S. Haq, J. Harnett and A. Hodgson, Growth of thin crystalline ice films on Pt(111), Surf. Sci., 2002, 505, 171–182 CrossRef CAS.
  20. J. Harnett, S. Haq and A. Hodgson, Electron induced restructuring of crystalline ice adsorbed on Pt(111), Surf. Sci., 2003, 528, 15–19 CrossRef CAS.
  21. G. Zimbitas, S. Haq and A. Hodgson, The structure and crystallization of thin water films on Pt(111), J. Chem. Phys., 2005, 123, 174701 CrossRef CAS PubMed.
  22. S. Maier, B. A. J. Lechner, G. A. Somorjai and M. Salmeron, Growth and structure of the first layers of ice on Ru(0001) and Pt(111), J. Am. Chem. Soc., 2016, 138, 3145–3151 CrossRef CAS PubMed.
  23. S. Nie, P. J. Feibelman, N. C. Bartelt and K. Thümer, Pentagons and heptagons in the first water layer on Pt(111), Phys. Rev. Lett., 2010, 105, 026102 CrossRef CAS PubMed.
  24. A. Beniya, Y. Sakaguchi, T. Narushima, K. Mukai, Y. Yamashita, S. Yoshimoto and J. Yoshinobu, The growth process of first water layer and crystalline ice on the Rh(111) surface, J. Chem. Phys., 2009, 130, 034706 CrossRef PubMed.
  25. A. Beniya, S. Yamamoto, K. Mukai, Y. Yamashita and J. Yoshinobu, The first layer of water on Rh(111): Microscopic structure and desorption kinetics, J. Chem. Phys., 2006, 125, 054717 CrossRef PubMed.
  26. H. Ogasawara, B. Brena, D. Nordlund, M. Nyberg, A. Pelmenschikov, L. G. M. Pettersson and A. Nilsson, Structure and bonding of water on Pt(111), Phys. Rev. Lett., 2002, 89, 276102 CrossRef CAS PubMed.
  27. H. Ogasawara, J. Yoshinobu and M. Kawai, Water adsorption on Pt(111): From isolated molecule to three-dimensional cluster, Chem. Phys. Lett., 1994, 231, 188–192 CrossRef CAS.
  28. M. Nakamura, Y. Shingaya and M. Ito, The vibrational spectra of water cluster molecules on Pt(111) surface at 20 K, Chem. Phys. Lett., 1999, 309, 123–128 CrossRef CAS.
  29. H. Ogasawara, J. Yoshinobu and M. Kawai, Clustering behavior of water (D2O) on Pt(111), J. Chem. Phys., 1999, 111, 7003–7009 CrossRef CAS.
  30. P. J. Feibelman, Comment on “vibrational recognition of hydrogen-bonded water networks on a metal surface”, Phys. Rev. Lett., 2003, 91, 059601 CrossRef PubMed.
  31. S. Meng, E. G. Wang and S. Gao, Water adsorption on metal surfaces: A general picture from density functional theory studies, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 69, 195404 CrossRef.
  32. A. Beniya, K. Mukai, Y. Yamashita and J. Yoshinobu, Coverage-dependent sticking probability and desorption kinetics of water molecules on Rh(111), J. Chem. Phys., 2008, 129, 016101 CrossRef PubMed.
  33. K. Thürmer and S. Nie, Formation of hexagonal and cubic ice during low-temperature growth, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 11757–11762 CrossRef PubMed.
  34. N. Aiga, T. Sugimoto, Y. Otsuki, K. Watanabe and Y. Matsumoto, Origins of emergent high-Tc ferroelectric ordering in heteroepitaxial ice films: Sum-frequency generation vibrational spectroscopy of H2O and D2O ice films on Pt(111), Phys. Rev. B: Condens. Matter Mater. Phys., 2018, 97, 075410 CrossRef CAS.
  35. B. Kamb, W. C. Hamilton, S. J. LaPlaca and A. Prakash, Ordered proton configuration in ice II, from single-crystal neutron diffraction, J. Chem. Phys., 1971, 55, 1934–1945 CrossRef CAS.
  36. D. D. Klug and E. Whalley, The uncoupled O–H stretch in ice VII. The infrared frequency and integrated intensity up to 189 kbar, J. Chem. Phys., 1984, 81, 1220–1228 CrossRef CAS.
  37. D. D. Klug, O. Mishima and E. Whalley, High-density amorphous ice. IV. Raman spectrum of the uncoupled O–H and O–D oscillators, J. Chem. Phys., 1987, 86, 5323–5328 CrossRef CAS.
  38. V. F. Petrenko and R. W. Whitworth, Physics of ice, Oxford University Press, 1999 Search PubMed.
  39. Y. R. Shen, Optical second harmonic generation at interfaces, Annu. Rev. Phys. Chem., 1989, 40, 327–350 CrossRef CAS.
  40. Y. R. Shen, Surfaces probed by nonlinear optics, Surf. Sci., 1994, 299–300, 551–562 CrossRef CAS.
  41. K. B. Eisenthal, Liquid interfaces probed by second-harmonic and sum-frequency spectroscopy, Chem. Rev., 1996, 96, 1343–1360 CrossRef CAS PubMed.
  42. G. L. Richmond, Molecular bonding and interactions at aqueous surfaces as probed by vibrational sum frequency spectroscopy, Chem. Rev., 2002, 102, 2693–2724 CrossRef CAS PubMed.
  43. F. Vidal and A. Tadjeddine, Sum-frequency generation spectroscopy of interfaces, Rep. Prog. Phys., 2005, 68, 1095–1127 CrossRef CAS.
  44. Y. R. Shen, Surface nonlinear optics [invited], J. Opt. Soc. Am. B, 2011, 28, A56–A66 CrossRef CAS.
  45. Y. R. Shen, Basic theory of surface sum-frequency generation, J. Phys. Chem. C, 2012, 116, 15505–15509 CrossRef CAS.
  46. X. Wei, P. B. Miranda, C. Zhang and Y. R. Shen, Sum-frequency spectroscopic studies of ice interfaces, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 66, 085401 CrossRef.
  47. K. C. Chou, S. Westerberg, Y. R. Shen, P. N. Ross and G. A. Somorjai, Probing the charge – transfer state of CO on Pt(111) by two - dimensional infrared – visible sumfrequency generation hyperspectroscopy, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 69, 153413 CrossRef.
  48. A. G. Lambert, P. B. Davies and D. J. Neivandt, Implementing the theory of sum frequency generation vibrational spectroscopy: A tutorial review, Appl. Spectrosc. Rev., 2005, 40, 103–145 CrossRef CAS.
  49. H.-F. Wang, W. Gan, R. Lu, Y. Rao and B.-H. Wu, Quantitative spectral and orientational analysis in surface sum frequency generation vibrational spectroscopy (SFG-VS), Int. Rev. Phys. Chem., 2005, 24, 191–256 Search PubMed.
  50. Y. R. Shen, Phase-sensitive sum-frequency spectroscopy, Annu. Rev. Phys. Chem., 2013, 64, 129–150 CrossRef CAS PubMed.
  51. V. Ostroverkhov, G. A. Waychunas and Y. R. Shen, New information on water interfacial structure revealed by phase-sensitive surface spectroscopy, Phys. Rev. Lett., 2005, 94, 046102 CrossRef PubMed.
  52. I. V. Stiopkin, H. D. Jayathilake, A. N. Bordenyuk and A. V. Benderskii, Heterodyne-detected vibrational sum frequency generation spectroscopy, J. Am. Chem. Soc., 2008, 130, 2271–2275 CrossRef CAS PubMed.
  53. S. Yamaguchi and T. Tahara, Heterodyne-detected electronic sum frequency generation: ‘Up’ versus ‘down’ alignment of interfacial molecules, J. Chem. Phys., 2008, 129, 101102 CrossRef PubMed.
  54. Y. Otsuki, T. Sugimoto, T. Ishiyama, A. Morita, K. Watanabe and Y. Matsumoto, Unveiling subsurface hydrogen-bond structure of hexagonal water ice, Phys. Rev. B: Condens. Matter Mater. Phys., 2017, 96, 115405 CrossRef.
  55. W. F. Murphy, The rovibrational Raman spectrum of water vapour ν1 and ν3, Mol. Phys., 1978, 36, 727–732 CrossRef CAS.
  56. X. Wei, P. B. Miranda and Y. R. Shen, Surface vibrational spectroscopic study of surface melting of ice, Phys. Rev. Lett., 2001, 86, 1554–1557 CrossRef CAS PubMed.
  57. Q. Du, R. Superfine, E. Freysz and Y. R. Shen, Vibrational spectroscopy of water at the vapor/water interface, Phys. Rev. Lett., 1993, 70, 2313 CrossRef CAS PubMed.
  58. M. J. Iedema, M. J. Dresser, D. L. Doering, J. B. Rowland, W. P. Hess, A. A. Tsekouras and J. P. Cowin, Ferroelectricity in water ice, J. Phys. Chem. B, 1998, 102, 9203–9214 CrossRef CAS.
  59. J. Junquera and P. Ghosez, Critical thickness for ferroelectricity in perovskite ultrathin films, Nature, 2003, 422, 506–509 CrossRef CAS PubMed.
  60. N. Sai, A. M. Kolpak and A. M. Rappe, Ferroelectricity in ultrathin perovskite films, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 72, 020101 CrossRef.
  61. S. Nie, N. C. Bartelt and K. Thürmer, Evolution of proton order during ice-film growth: An analysis of island shapes, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 84, 035420 CrossRef.
  62. P. Parkkinen, S. Riikonen and L. Halonen, Ice XI: Not that ferroelectric, J. Phys. Chem. C, 2014, 118, 26264–26275 CrossRef CAS.
  63. M. Kunst and J. M. Warman, Nanosecond time-resolved conductivity studies of pulse-ionized ice. 2. The mobility and trapping of protons, J. Phys. Chem., 1983, 87, 4093–4095 CrossRef CAS.
  64. K. Mizuse, J.-L. Kuo and A. Fujii, Structural trends of ionized water networks: Infrared spectroscopy of watercluster radical cations (H2O)n+ (n = 3–11), Chem. Sci., 2011, 2, 868–876 RSC.
  65. C. Kobayashi, S. Saito and I. Ohmine, Mechanism of fast proton transfer in ice: Potential energy surface and reaction coordinate analyses, J. Chem. Phys., 2000, 113, 9090–9100 CrossRef CAS.
  66. G. Shirane and T. Oguchi, On the transition in KH2PO4, J. Phys. Soc. Jpn., 1949, 4, 172–175 CrossRef CAS.
  67. M. Ichikawa, The O–H vs. O⋯O distance correlation, the geometric isotope effect in OHO bonds, and its application to symmetric bonds, Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem., 1978, 34, 2074–2080 CrossRef.
  68. M. Ichikawa, Correlation between two isotope effects in hydrogen-bonded crystals: Transition temperature and separation of two equilibrium sites, Chem. Phys. Lett., 1981, 79, 583–587 CrossRef CAS.
  69. S. Horiuchi and Y. Tokura, Organic ferroelectrics, Nat. Mater., 2008, 7, 357–366 CrossRef CAS PubMed.
  70. K. T. Wikfeldt, Nuclear quantum effects in a 1-D model of hydrogen bonded ferroelectrics, J. Phys.: Conf. Ser., 2014, 571, 012012 CrossRef.
  71. Y. Otsuki, K. Watanabe, T. Sugimoto and Y. Matsumoto, Enhanced structural disorder at a nanocrystalline ice surface, Phys. Chem. Chem. Phys., 2019, 21, 20442–20453 RSC.
  72. D. Russo, J. Teixeira, L. Kneller, J. R. D. Copley, J. Ollivier, S. Perticaroli, E. Pellegrini and M. A. Gonzalez, Vibrational density of states of hydration water at biomolecular sites: Hydrophobicity promotes low density amorphous ice behavior, J. Am. Chem. Soc., 2011, 133, 4882–4888 CrossRef CAS PubMed.
  73. S. J. Cox, S. M. Kathmann, B. Slater and A. Michaelides, Molecular simulations of heterogeneous ice nucleation. I. Controlling ice nucleation through surface hydrophilicity, J. Chem. Phys., 2015, 142, 184704 CrossRef PubMed.
  74. M. Fitzner, G. C. Sosso, S. J. Cox and A. Michaelides, The many faces of heterogeneous ice nucleation: Interplay between surface morphology and hydrophobicity, J. Am. Chem. Soc., 2015, 137, 13658–13669 CrossRef CAS PubMed.
  75. T. Sugimoto, Y. Otsuki, T. Ishiyama, A. Morita, K. Watanabe and Y. Matsumoto, Topologically disordered mesophase at the topmost surface layer of crystalline ice between 120 and 200 K, Phys. Rev. B: Condens. Matter Mate. Phys., 2019, 99, 121402 CrossRef CAS.
  76. F. Tang, T. Ohto, S. Sun, J. R. Rouxel, S. Imoto, E. H. G. Backus, S. Mukamel, M. Bonn and Y. Nagata, Molecular structure and modeling of water–air and ice-air interfaces monitored by sum-frequency generation, Chem. Rev., 2020, 120, 3633–3667 CrossRef CAS PubMed.
  77. F. Kato, T. Sugimoto and Y. Matsumoto, Direct experimental evidence for markedly enhanced surface proton activity inherent to water ice, J. Phys. Chem. Lett., 2020, 11, 2524–2529 CrossRef CAS PubMed.

This journal is © the Owner Societies 2020