We have used quasielastic and inelastic neutron scattering to investigate the structure, dynamics, and phase transitions of water interacting with superhydrophilic CuO surfaces that not only possess a strong affinity for water but also a “grass-like” topography that is rough on both micro- and nanoscales. Here, we report quasielastic neutron scattering (QENS) measurements on two samples differing in water content at five temperatures below 280 K. The QENS spectra show water undergoing two different types of diffusive motion near the CuO surfaces: a “slow” translational diffusion occurring on a nanosecond time scale and a faster rotational motion. Further from the surfaces, there is “fast” translational diffusion comparable in rate to that of bulk supercooled water and the rotational motion occurring in the interfacial water. Analysis of the QENS spectra supports wetting of water to the CuO blades as seen in electron microscopy images. In addition, we observe an anomalous temperature dependence of the QENS spectra on cooling from 270 to 230 K with features consistent with a liquid–liquid phase transition. We suggest that the solvent-like properties of the coexisting bulk-like water in our CuO samples are a significant factor in determining the temperature dependence of the interfacial water’s dynamics. Our results are compared with those obtained from two well-studied substrate classes: (1) silicas that contain ordered cylindrical nanopores but have weaker hydrophilicity and (2) nanoparticles of other transition-metal oxides, such as TiO2, which share the strong hydrophilicity of our samples but lack their porosity.

Over the last two decades, there has been growing interest in the so-called “super” hydrophilic materials that combine a high affinity for water molecules with a rough or porous surface morphology.1–5 Previous research has been concerned primarily with the macroscopic characterization of superhydrophilic surfaces (e.g., by water contact angle measurements),2–5 the role of both surface chemistry and surface roughness in enhancing hydrophilicity,2–5 and the fabrication of superhydrophilic materials.1–5 Technologically, these materials have been used in self-cleaning, antifogging, antibacterial, and heat transfer applications.1–5 Notwithstanding these promising applications, there has been less attention paid to fundamental properties such as the molecular-scale structure and dynamics of water near superhydrophilic surfaces. In comparison, these properties have been more thoroughly studied for water confined to more uniform materials characterized by cylindrical pores but with weaker hydrophilicity, such as mesoporous silicas6–8 as well as water near atomically smooth, planar metal-oxide surfaces with stronger hydrophilicity.9–12 

In the case of mesoporous6–8 and nanoporous silica-like materials,13 both the structure of water and its freezing/melting behavior when confined within hydrophilic pores have been investigated. There have also been studies of the dynamics of supercooled water confined in nanoporous silica materials8,14,15 and mesoporous carbon.16 However, due to their relatively well-ordered pore structure and uniform diameter of their pores, these materials lack the combination of rough surface morphology and high affinity for water that characterize superhydrophilicity.

Recently, we have used neutron scattering techniques to probe the structure and phase transitions of water in proximity to superhydrophilic CuO surfaces, which have a qualitatively different “grass-like” topography (see Fig. 1) characterized by micrometer-size CuO blades and nanostructures surrounding their base.17 

FIG. 1.

(a) SEM image of the CuO surface; (b) a magnified view of the grass-like morphology, showing the micrometer-size CuO blades and the nanostructures surrounding their base.

FIG. 1.

(a) SEM image of the CuO surface; (b) a magnified view of the grass-like morphology, showing the micrometer-size CuO blades and the nanostructures surrounding their base.

Close modal

Our original motivation for this study was technological in nature. It has been found that the superhydrophilic CuO coatings could improve the thermal performance of heat transfer devices known as oscillating heat pipes.18 However, our results proved to be of fundamental interest as well. As will be reviewed in Sec. III, our elastic neutron scattering measurements showed that coating a bare-Cu surface with hydrophilic CuO nanostructures caused the abrupt freezing transition of water at 265 K near a bare-copper surface to be spread into a continuous transition extending down to a temperature of ∼200 K.17 Our neutron diffraction measurements also indicated that water closest to the CuO nanostructures freezes into a noncrystalline phase.17 Evidence of disordered water had been found adjacent to the walls of cylindrical pores in silicas,6,7 but this structure differs qualitatively from that of water near more strongly adsorbing planar metal-oxide surfaces, which can be described in terms of three distinct layers.12 

These findings on the structure and phase transitions of water near the superhydrophilic CuO surfaces have raised questions concerning the interfacial dynamics of their associated water at the molecular level. From our elastic neutron scattering measurements, we inferred the presence of two distinct water populations located at different distances from the CuO nanostructures and differing in their freezing behavior.17 In this paper, we address the questions of whether there are two water populations that differ in their molecular diffusive motions and how their dynamics might change with temperature in the extended freezing transition observed in these superhydrophilic samples. For this purpose, we apply quasielastic neutron scattering (QENS) in much the same way as in our previous studies of water in proximity to single-supported bilayer lipid membranes.19,20 In addition, we use vibrational spectra, measured by inelastic neutron scattering, to corroborate the presence of disordered water near the CuO nanostructures as we had inferred from our elastic neutron diffraction measurements.17 To assess the combined effects of hydrophilicity and porosity on the molecular dynamics of water in a superhydrophilic environment, we compare our results both with water confined in porous materials of lower hydrophilicity, such as nanoporous silica, and in proximity to planar surfaces of comparable hydrophilicity, such as those found on nanoparticles of transition metal oxides.

In Sec. II, we describe our sample preparation method and the techniques of quasielastic and inelastic neutron scattering that we have used to probe the dynamics of water in our CuO samples over a wide range of length and time scales. Section III presents the results of our quasielastic and inelastic scattering measurements. Included in the discussion are the decomposition of the quasielastic neutron scattering (QENS) spectra into broad and narrow Lorentzian components, their anomalous temperature dependence, and their combination with electron microscopy images to infer a growth mode of the water films on the grass-like CuO surface. Section IV summarizes our main results. We also consider possible explanations for differences and similarities between the water dynamics in our superhydrophilic CuO samples with that in the two well-studied substrate classes described earlier: (1) silicas that contain ordered cylindrical nanoscale pores but have weaker hydrophilicity and (2) nanoparticles of TiO2, which share the strong hydrophilicity of our samples but lack their porosity.

The process of growing CuO nanostructures on a copper foil has been described previously.17,21 Briefly, we removed the native oxide from copper foils (All Foils, USA51) of thickness 12.7 µm in a 2.0M HCl solution and then immersed them for 10 min at a temperature of 368 K in a pH 14 solution consisting of NaClO2, NaOH, Na3PO4·12H2O, and deionized-water in the ratio 3.75:5:10:100 by weight. The foils were then rinsed in deionized water and air-dried. This treatment resulted in the formation of a grass-like coating of CuO nanostructures as shown in the scanning electron microscope (SEM) images in Fig. 1. From a survey of SEM images, the typical dimensions of the triangular blades are 2 µm tall, 0.5 µm wide at the base, and ∼20 nm thick. We confirmed the hydrophilicity of our CuO coatings on a macroscopic scale by a zero static water contact angle17 in agreement with previous measurements.21 

The hydrophilicity of our samples was also characterized at the microscale by imaging in situ the condensation and evaporation of water within the CuO nanostructures using an SEM operated in its environmental mode (FEI Quanta 600F51). The SEM chamber was backfilled with water vapor creating a pressure of ∼500 Pa around the sample of size ∼1 × 1 cm2. It was mounted on an oxygen-free high-conductivity copper stub at an angle of 15° to the incident beam direction and cooled with a Peltier cold stage. Following the procedure used by others,22 we initiated condensation by cooling the sample slightly below the dew point (∼270 K) and then recorded a video until the water thickness exceeded the escape length of the scattered electrons.

We found the wetting behavior to differ qualitatively from that which occurs in nanoporous silicas. As will be discussed in Sec. III B 4, our environmental SEM measurements reveal webs of water condensing between CuO blades, a mesoscale wetting phenomenon that does not occur within the cylindrical pores of the silicas.

Our neutron scattering samples were the same as those used previously17 and consisted of 100 copper foil disks, each 5 cm in diameter and coated with CuO nanostructures as described earlier. The large diameter and number of the copper foils were necessary to increase the contribution to the neutron scattering intensity from water near the CuO surfaces, while the foil thickness (∼12.7 µm) was chosen to minimize the incoherent scattering from copper. The foils were stacked in a cylindrical aluminum can (5 cm in diameter and 4 cm tall), heated in air at ∼328 K for at least 24 h to remove excess water, and sealed under a helium atmosphere with an indium O-ring. Sample hydration was established by a droplet of water of known volume (10 µl or 60 µl) added to each cell via micropipette prior to sealing. The sample assembly is illustrated in Fig. 2. After sealing, the assembly was heated again at ∼328 K for 1–2 h to evaporate water from the droplet and produce a humid atmosphere inside the cell.

FIG. 2.

Schematic diagram of the sample cell used for neutron scattering measurements showing the neutron beam of cross section 30 × 30 mm2 incident on a stack of 100 copper foil disks. To hydrate a sample, a droplet of H2O of known volume is placed at the bottom of the cell outside of the neutron beam. The neutron wave vector Q is parallel to the plane of the copper foil disks. Not drawn to scale.

FIG. 2.

Schematic diagram of the sample cell used for neutron scattering measurements showing the neutron beam of cross section 30 × 30 mm2 incident on a stack of 100 copper foil disks. To hydrate a sample, a droplet of H2O of known volume is placed at the bottom of the cell outside of the neutron beam. The neutron wave vector Q is parallel to the plane of the copper foil disks. Not drawn to scale.

Close modal

To estimate the surface area of the CuO coating, N2 adsorption measurements were conducted at 77 K on a Quantachrome Autosorb51 instrument after degassing the sample at ∼515 K for 20 h. Brunauer–Emmett–Teller (BET) analysis yielded surface areas of ∼0.6 and ∼0.1 m2/g of our CuO-coated and untreated copper foil, respectively (see the supplementary material, Fig. S1). These measurements imply that a 22 g sample of CuO-coated copper foils used for our neutron scattering measurements (Fig. 2) would have a total surface area of ∼13 m2 (compared to the planar surface area of ∼0.4 m2). We note that the specific surface area of our samples is two to three orders of magnitude smaller than the mesoporous6–8 and nanoporous14,15 silicas and the metal oxide powders,12 which have been investigated previously. Unfortunately, the relatively small surface area of our samples renders differential scanning calorimetry (DSC) measurements unreliable using commercial instruments.17 In the case of the silicas, DSC measurements have been useful in interpreting the phase changes occurring in the interfacial water.7,14,15 We will also see that the small surface area of our samples necessitates longer counting times in our QENS measurements compared to the silicas, limiting the number of samples and temperatures which can be investigated.

The QENS measurements were performed on the backscattering spectrometer BASIS (backscattering silicon spectrometer)23 at the Spallation Neutron Source, Oak Ridge National Laboratory (ORNL). BASIS has an energy resolution width at zero energy transfer of ΔE ∼ 3.5 µeV, allowing motions occurring on a time scale as long as ∼1 ns to be probed. The largest energy transfer accessible (dynamic range) is ±120 µeV, corresponding to a time scale as short as ∼10 ps. Our measurements spanned a wave vector transfer (Q) range of 0.3–1.7 Å−1.

Due to the large incoherent cross section of hydrogen, the QENS spectra of our samples are dominated by the contribution from the H atoms in the water molecules. As used in previous studies of interfacial water,12,19,20 we found that our QENS spectra could be modeled well by assuming a dynamic structure factor S(Q,ω) consisting of a sum of three terms: an elastic component described by a delta function plus two Lorentzian functions, representing the quasielastic scattering,

SQ,ω=AQδω+BQLNQ,ω+CQLBQ,ω,
(1)

where A(Q), B(Q), and C(Q) are free parameters, Q=kikf is the neutron wave vector transfer, ω=EiEf is the energy transfer to the neutron, and the Lorentzian functions are given by

LiQ=1πΓiQΓiQ2+ω2.
(2)

Here, Γi is the half-width-half-maximum (HWHM) of the Lorentzian and the subscript i is either N or B, denoting “narrow” and “broad,” respectively. A one-Lorentzian fit consistently resulted in residuals near zero energy transfer (see Fig. S2 in the supplementary material), indicating the presence of a narrow Lorentzian component. Attempts to fit spectra with more than two Lorentzians gave either unphysical results or fits that depended on the initial parameter set assumed. In the beginning, we imposed no model-dependent constraints on the Q-dependence of the parameters A(Q), B(Q), and C(Q). As will be described later, we show that, unlike the narrow Lorentzian intensity B(Q), the temperature dependence of the broad Lorentzian intensity C(Q) at low Q (<1 Å−1) differs from that at high Q (>1 Å−1).

Using the DAVE (Data Analysis and Visualization Environment) software,24 we fit the observed quasielastic spectra to the intensity function I(Q,ω) obtained by convoluting Eq. (1) with the instrument resolution function R(Q, ω),

IQ,ω=S(Q,ω)RQ,ω+DQ,ω,
(3)

where DQ,ω=a(Q)ω+b(Q) is a linear background term. R(Q, ω) is measured at a sample temperature of 100 K, which is sufficiently low in temperature that the motion of the H atoms in the water molecules can be neglected. The decomposition of a typical spectrum measured on BASIS is shown in Fig. 3.

FIG. 3.

The QENS spectrum of the wet (60 µl of H2O added) CuO-coated sample measured on BASIS at Q = 1.1 Å−1 after cooling to a temperature T = 260 K. The data points (open circles) have been fitted by folding the instrumental resolution function (dotted spectrum at T = 100 K) with a scattering law composed of three terms: a delta function (black dotted curve) representing elastic scattering and two Lorentzians, a broad Lorentzian (magenta solid curve) plus a narrow Lorentzian (blue solid curve), representing the quasielastic scattering [see Eq. (1)]. The best fit to the spectrum (red solid curve) is obtained after adding a linear background term (dashed red line).

FIG. 3.

The QENS spectrum of the wet (60 µl of H2O added) CuO-coated sample measured on BASIS at Q = 1.1 Å−1 after cooling to a temperature T = 260 K. The data points (open circles) have been fitted by folding the instrumental resolution function (dotted spectrum at T = 100 K) with a scattering law composed of three terms: a delta function (black dotted curve) representing elastic scattering and two Lorentzians, a broad Lorentzian (magenta solid curve) plus a narrow Lorentzian (blue solid curve), representing the quasielastic scattering [see Eq. (1)]. The best fit to the spectrum (red solid curve) is obtained after adding a linear background term (dashed red line).

Close modal

To select temperatures at which to record QENS spectra, it was helpful to consider our previous measurements of the intensity of neutrons scattered elastically from the same samples as a function of temperature.17 These measurements were performed on the high-flux back scattering spectrometer (HFBS) at the NIST Center for Neutron Research,26 using a stationary monochromator in a so-called “fixed-window scan” hereafter denoted FWS. As noted earlier, the incoherent scattering from the H atoms in the water molecules in our samples dominates the elastic signal. The ∼1 µeV energy resolution of the HFBS implies that an increase in the elastic intensity is proportional to an increase in the number of H atoms that are moving on a time scale longer than ∼4 ns. Therefore, the elastic intensity increase on cooling provides a sensitive measure of the number of water molecules that become immobilized or “frozen” as the temperature decreases.

We used the Vibrational Spectrometer (VISION)25 at the ORNL Spallation Neutron Source to obtain inelastic neutron spectra from the same two samples of CuO-coated Cu foils containing different amounts of H2O whose QENS spectra were measured on BASIS (see sample cell in Fig. 2). Samples were quenched in liquid nitrogen to ∼80 K and further cooled to 5 K. VISION is optimized to characterize molecular vibrations in an energy range of 5–600 meV with an energy resolution of ΔE/E < 1.5% for E > 2 meV.

Our aim is to investigate the dynamics of water molecules interacting with the superhydrophilic CuO nano and blade-like structures that coat the surface of the Cu foils. In particular, we wish to determine how the interfacial water dynamics differs from that of water confined in nanoporous silicas and near planar surfaces of metal-oxide particles as well as bulk supercooled water. To increase sensitivity to the interfacial water, it is desirable to work with as low water content as is possible. In practice, we found that a volume of 10 µl of water added to the sample (see Fig. 2) yielded the smallest quasielastic intensity that could be analyzed in spectra measured on BASIS. As it turned out (see Sec. III B 4), the interfacial water comprised about half of the total amount of mobile water in both the dry (10 µl water) and the wet (60 µl water) samples.

To facilitate discussion, we reproduce in Fig. 4(a) the FWS from Ref. 17 for three different samples: CuO-coated copper foils with 60 µl of H2O added (termed “wet”); a similar sample with 10 µl of H2O (termed “dry”); and, to serve as a control, bare copper foils with 60 µl of H2O. In contrast to the abrupt freezing of the water in the sample of bare Cu foils at 265 K, the water in the two CuO-coated samples freezes continuously on cooling from a temperature of 280 down to 237 K. The elastic intensity of the dry sample increases linearly over this entire temperature range, and the intensity of the wet sample has the same linear dependence down to ∼257 K [red points in Fig. 4(a)]. At lower temperatures, the intensity of the wet sample still increases linearly but with a larger slope [green points in Fig. 4(a)]. As discussed in Ref. 17, we interpreted the initial linear temperature dependence of the elastic intensity (red points) as indicating a population of strongly bound water common to both samples and located near the foot of the CuO nanostructures where the density of CuO blades is highest. This interpretation is consistent with electron microscopy images showing that water condensed first within this region. In addition, we proposed that the wet sample contained a second population of water located further from the CuO blades that froze at lower temperatures as indicated by the second linear term in its elastic intensity (green points). We shall see from the presence of two Lorentzian components at all temperatures that the QENS spectra support the presence of two water populations, one located close to the CuO blades and another more distant from them with bulk-like dynamics. However, contrary to our previous interpretation,17 water with bulk-like dynamics is present in both the wet and dry samples and contributes to both linear terms in the FWS of the wet sample.

FIG. 4.

(a) Incoherent elastically scattered neutron intensity vs sample temperature (fixed-window scans denoted FWS) measured on the HFBS spectrometer on cooling from Ref. 17. Data are shown for CuO samples hydrated with 10 µl of H2O (top plot) and 60 µl (middle plot) on cooling at a rate of 0.08 K/min. The intensity is summed over all detectors (0.36 Å−1 < Q < 1.75 Å−1). Vertical dashed lines indicate inflection points in the elastic intensity of the wet sample containing 60 µl H2O. The slope of the red data points for both CuO-coated samples is the same. The bottom plot shows an abrupt freezing transition observed for a sample of bare Cu foils containing 60 µl of water. (b) Temperature dependence of A(Q), the intensity of the delta-function term in the dynamic structure factor defined in Eq. (1) and obtained in fits to the BASIS spectra of both the wet and dry CuO-coated samples. The intensity has been summed over all detectors as in (a). Dashed curves are lines of best fit over all temperatures. In (a) and (b), error bars representing a statistical uncertainty of one standard deviation are smaller than the size of the data points.

FIG. 4.

(a) Incoherent elastically scattered neutron intensity vs sample temperature (fixed-window scans denoted FWS) measured on the HFBS spectrometer on cooling from Ref. 17. Data are shown for CuO samples hydrated with 10 µl of H2O (top plot) and 60 µl (middle plot) on cooling at a rate of 0.08 K/min. The intensity is summed over all detectors (0.36 Å−1 < Q < 1.75 Å−1). Vertical dashed lines indicate inflection points in the elastic intensity of the wet sample containing 60 µl H2O. The slope of the red data points for both CuO-coated samples is the same. The bottom plot shows an abrupt freezing transition observed for a sample of bare Cu foils containing 60 µl of water. (b) Temperature dependence of A(Q), the intensity of the delta-function term in the dynamic structure factor defined in Eq. (1) and obtained in fits to the BASIS spectra of both the wet and dry CuO-coated samples. The intensity has been summed over all detectors as in (a). Dashed curves are lines of best fit over all temperatures. In (a) and (b), error bars representing a statistical uncertainty of one standard deviation are smaller than the size of the data points.

Close modal

Due to the time required to obtain a QENS spectrum, the FWS were helpful in suggesting temperatures on which to focus our attention. We chose to measure QENS spectra at 270 and 260 K where the two samples share common freezing behavior, at 250 and 240 K where the water in the wet sample freezes more rapidly on cooling, and at 230 K where most of the water has been immobilized in both samples. At all temperatures and Q (0.3 Å−1 < Q < 1.7 Å−1), the quasielastic scattering can be modeled by a delta function and two Lorentzians [see Eq. (1)] whose HWHM (Γ) are well separated in energy: a “narrow” Lorentzian (1 ≲ Γ ≲ 5 µeV) and a “broad” Lorentzian (10 ≲ Γ ≲ 60 µeV). The delta-function component represents neutrons scattered elastically from water molecules, i.e., with an energy transfer ≲1 µeV as well as the residual elastic scattering from the Cu and CuO. An increase in the delta-function intensity measures the number of water molecules that become immobile on the time scale of the instrument. The “narrow” Lorentzian represents molecules undergoing a motion on a time scale of 100–400 ps, and the “broad” Lorentzian represents molecules undergoing faster motion (rotational and translational) on a time scale of 4–50 ps.

In Fig. 5, we show the temperature dependence of the intensity of the two Lorentzians obtained in fits to spectra from both the wet and the dry samples. The intensity has been summed over all detectors. To account for differences in the scattering from the foils in the two samples, the spectra of the wet sample have been scaled so that the intensity of the delta function component summed over all Q at 270 K agrees with that of the dry sample (at 270 K, the delta function intensity is dominated by elastic scattering of the foils). At temperatures T < 270 K, we see that the narrow Lorentzian intensity of each sample exceeds that of its broad Lorentzian. Furthermore, the intensity of both the narrow and the broad Lorentzian increases with the water content of the sample. That is, both samples contain more water diffusing “slowly” rather than “rapidly,” and the amount of water moving on each time scale increases as the hydration level of the sample increases.

FIG. 5.

Temperature dependence of the intensity of the narrow and broad Lorentzian components obtained in fits to the BASIS spectra of both the wet and dry CuO-coated samples. The intensity has been summed over all detectors (0.3 Å−1 < Q < 1.7 Å−1) as in the case of the FWSs in Fig. 4. Error bars representing a statistical uncertainty of one standard deviation are smaller than the size of the data points.

FIG. 5.

Temperature dependence of the intensity of the narrow and broad Lorentzian components obtained in fits to the BASIS spectra of both the wet and dry CuO-coated samples. The intensity has been summed over all detectors (0.3 Å−1 < Q < 1.7 Å−1) as in the case of the FWSs in Fig. 4. Error bars representing a statistical uncertainty of one standard deviation are smaller than the size of the data points.

Close modal

The temperature dependence of the delta-function intensity A(Q) [see Eq. (1)] inferred from the fits to the BASIS spectra of the wet and dry samples provides an important reproducibility check by comparing it with their respective FWS measured on the HFBS. In Fig. 4(b), we see that in the range 240 K < T < 260 K, the intensity of both samples increases linearly on cooling as does their FWS. The slope of A(Q) of the wet sample is ∼2.8 times larger than that for the dry sample compared to a ratio of ∼2.7 observed in the FWS in the temperature range of 257–237 K. Due to the sparsity of temperatures and the somewhat lower energy resolution of BASIS, the inflection points at ∼257 and ∼237 K seen in FWS of the wet sample [Fig. 4(a)] cannot be resolved in its delta-function intensity.

1. The broad Lorentzian component

a. Q-dependence of the broad Lorentzian width.

To probe the type of motion (e.g., translational or rotational) occurring on the two different time scales, we examined the Q dependence of the HWHM of the two Lorentzians. In Fig. 6, the HWHM of the broad Lorentzian ΓB obtained in the fits to the QENS spectra of (a) the wet and (b) the dry CuO-coated sample are plotted as a function of Q2 at several temperatures. Also plotted in Fig. 6 are data for the translational motion of bulk supercooled water measured at 253 K from Ref. 27. For clarity, data at 240 and 260 K for our samples are not shown but appear in Fig. S3 in the supplementary material. The data of the two samples are quantitatively similar with about the same widths ΓB in the temperature range of 240–270 K. To within the uncertainty of our measurement, the ΓB of the wet and dry samples follows that of bulk supercooled water at low Q. This behavior of bulk water in which the quasielastic width has a linear dependence on Q2 at low Q (Fick’s law) followed by a leveling off has been well described by models of jump diffusion.27 Also noteworthy is the change in the Q-dependence and magnitude of the HWHM on cooling below 250 K—a feature present in both samples. This observation will be discussed in more detail below.

FIG. 6.

HWHM of the broad Lorentzian ΓB as a function of Q2 for the two CuO-coated samples, (a) wet (60 µl H2O) and (b) dry (10 µl H2O), at selected temperatures as measured on BASIS. Solid curves are drawn as guides to the eye. The horizontal dashed line is the best fit to the data points of each sample at T = 230 K for Q2 ≥ 0.7 Å−2. Although not shown for clarity (see the supplementary material), values of ΓB for both samples at temperatures of 240 and 260 K closely match those at 250 K. Also shown is the HWHM for the translational motion of bulk supercooled water at 253 K from Ref. 27 (black squares). The fall-off of the HWHM at high Q for bulk supercooled water is believed to be due to an imperfect deconvolution of the translational and rotational diffusive motion.27 Error bars represent a statistical uncertainty of one standard deviation. The smaller error bars for the dry sample at 230 K reflect a longer counting time of ∼9 h compared to ∼4 h for the wet sample.

FIG. 6.

HWHM of the broad Lorentzian ΓB as a function of Q2 for the two CuO-coated samples, (a) wet (60 µl H2O) and (b) dry (10 µl H2O), at selected temperatures as measured on BASIS. Solid curves are drawn as guides to the eye. The horizontal dashed line is the best fit to the data points of each sample at T = 230 K for Q2 ≥ 0.7 Å−2. Although not shown for clarity (see the supplementary material), values of ΓB for both samples at temperatures of 240 and 260 K closely match those at 250 K. Also shown is the HWHM for the translational motion of bulk supercooled water at 253 K from Ref. 27 (black squares). The fall-off of the HWHM at high Q for bulk supercooled water is believed to be due to an imperfect deconvolution of the translational and rotational diffusive motion.27 Error bars represent a statistical uncertainty of one standard deviation. The smaller error bars for the dry sample at 230 K reflect a longer counting time of ∼9 h compared to ∼4 h for the wet sample.

Close modal

The close tracking of ΓB of the wet and dry samples to that of bulk supercooled water at Q < 1 Å−1 in the temperature range from 270 down to 240 K suggests that bulk-like water is present in both samples. Furthermore, the fact that this agreement occurs over this entire temperature range allows us to infer that the freezing of bulk-like water is contributing to both linear terms in the FWS of the wet sample (red and green points in Fig. 4)—a conclusion that we could not reach from the FWS alone.17 Presumably, this bulk-like water is sufficiently distant from the CuO nanostructures to have only a weak interaction with them.

A qualitative change in the Q-dependence of ΓB occurs on reducing the temperature from 240 to 230 K. The magnitude of ΓB increases and becomes nearly Q-independent at 230 K as can be seen in Fig. 6 (see Fig. S3 for data at 240 K). This behavior of ΓB at 230 K is inconsistent with the presence of bulk supercooled water undergoing translational diffusion.27 First, at low Q, ΓB no longer tracks the Q-dependence observed for bulk water; and, second, we expect ΓB to decrease when the bulk-like water has completely frozen out rather than increase at 230 K.27 We interpret this result as indicating that there are actually two types of diffusive motions occurring at a comparable rate that are unresolved and contributing to the broad Lorentzian at higher temperatures: (1) translational diffusion of bulk-like water and (2) water molecules undergoing a rotational diffusive motion that contributes a component with a Q-independent width to the quasielastic scattering.28 As the temperature is reduced below 240 K, the translational motion of the bulk-like water freezes out so that by 230 K only a rotational motion contributes to the broad Lorentzian. We note that this abrupt change in ΓB corresponds to an upward inflection in the elastic intensity on cooling below ∼237 K for both the wet and dry CuO-coated samples as shown in the fixed window scans in Fig. 4(a).

b. Temperature dependence of the broad Lorentzian intensity.

In order to substantiate contributions from both translational and rotational diffusive motion to the broad Lorentzian intensity at temperatures above 230 K, it is of interest to compare the temperature dependence of these two components. Although we have been unable to resolve their separate contributions to the QENS spectra in the form of a third Lorentzian, we can probe their temperature dependence more indirectly. The intensity of quasielastic scattering from the rotational motion of a water molecule, either about its center of mass or a symmetry axis, will be a minimum at Q = 0 and increase monotonically in the Q-range of our measurements, whereas, for pure translational diffusion, the quasielastic intensity should be independent of Q.28–30 In Fig. 7(a), we have summed the intensity of the broad Lorentzian for both wet and dry samples over two different Q ranges: low Q (0.3 Å−1 < Q < 0.9 Å−1) and high Q (1.1 Å−1 < Q < 1.7 Å−1). As explained above the sums over the low-Q and high-Q ranges will weight the broad Lorentzian intensity toward the translational and rotational diffusion components, respectively.29,30 For both the wet and dry samples, we see in Fig. 7(a) that the intensity of the broad Lorentzian summed over the high-Q range has a temperature dependence that is concave downward compared to a concave upward dependence when summed over low Q. This difference supports our interpretation of two contributions to the broad Lorentzian intensity—a bulk-like translational motion that on cooling freezes out more quickly as represented by the low-Q summed intensity (concave upward) and rotational diffusion represented by the high-Q summed intensity (concave downward). In addition, the rotational component displays a weaker temperature dependence than the translational component above 250 K, particularly for the dry sample.

FIG. 7.

Temperature dependence of the intensity of (a) the broad and (b) the narrow Lorentzian components in the BASIS spectra of both the wet (60 µl H2O) and dry (10 µl H2O) CuO-coated samples. Unlike in Fig. 5, the intensity of each Lorentzian component has been summed over two different Q ranges: low Q (0.3 Å−1 < Q < 0.9 Å−1) and high Q (1.1 Å−1 < Q < 1.7 Å−1). As in Fig. 5, the Lorentzian intensities have been scaled to allow for a different number of foils in the two samples.

FIG. 7.

Temperature dependence of the intensity of (a) the broad and (b) the narrow Lorentzian components in the BASIS spectra of both the wet (60 µl H2O) and dry (10 µl H2O) CuO-coated samples. Unlike in Fig. 5, the intensity of each Lorentzian component has been summed over two different Q ranges: low Q (0.3 Å−1 < Q < 0.9 Å−1) and high Q (1.1 Å−1 < Q < 1.7 Å−1). As in Fig. 5, the Lorentzian intensities have been scaled to allow for a different number of foils in the two samples.

Close modal
c. Location of the water undergoing rotational diffusive motion.

We next consider where the water undergoing rotational diffusive motion at 230 K is located with respect to the CuO surfaces. The freezing out of the bulk-like translational motion in the temperature range of 230–240 K is in reasonable agreement with the temperature of ∼238 K at which the NMR signals for translational31 and rotational32 motion of bulk supercooled water are lost. At 230 K, the characteristic relaxation time of a rotating H2O molecule is ∼145 ps obtained by extrapolating molecular dynamics simulations of bulk water from 233 down to 230 K [see Fig. 1 in Ref. 32]. This value for bulk water is about 30 times longer than the relaxation time τR = ℏ/(3ΓB) ∼ 4.6 ps for rotational motion12 that we infer at 230 K from the horizontal dashed lines at ΓB ∼ 48 µeV in Fig. 6. Thus, we propose that the faster rotational motion that we are observing in our samples is occurring in the water interacting strongly with the nanoscale and blade-like CuO structures on the CuO surface and located closer to them than the bulk-like water.

The question remains as to why the half-width ΓB increases as the temperature decreases from 240 to 230 K, indicating an increase in the diffusivity on cooling in this temperature range. From Fig. 7(a), we see that the high-Q-weighted broad Lorentzian intensity of both the wet and the dry samples (red squares and green diamonds, respectively) representing the contribution of rotational motion to the QENS spectra exceeds that of the low-Q-weighted intensity representing the translational diffusive motion. Therefore, we can infer that the increase in ΓB between 240 and 230 K must be dominated by an increase in the rotational rate of the water molecules close to the CuO surfaces rather than an increase in the translational diffusion rate of the more distant bulk-like water. We shall return to this point in the discussion of the narrow Lorentzian component in Sec. III B 3.

d. Comparison with other systems.

A variety of other systems containing confined water have shown evidence of rotational diffusive motion at low temperatures. These studies include supercooled water in Vycor29 (porous silica glass), nanoporous silica,14 and water at protein interfaces.30 As with our CuO-coated samples,17 Zanotti et al.29 did not observe Bragg peaks characteristic of crystallization for a water (D2O) monolayer adsorbed in Vycor.29 Furthermore, they recorded FWSs with 1 µeV energy resolution summed over a low-Q and a high-Q range as we have done for the Lorentzian intensities in Fig. 7. The different temperature dependence of the elastic intensity in the two FWSs revealed the presence of translational and rotational motion at low-Q and high-Q, respectively.29 Inflection points in the temperature dependence of the mean-square displacements occurred at 240 K (translational motion) and 220 K (rotational motion). These results appear consistent with our observation of a “fast” translational motion freezing out between 240 and 230 K with the rotational motion persisting to a lower temperature.

In the case of mesoporous silicas, the structure, freezing, and dynamics of their confined water have been investigated in some detail.6,7,15,33 Neutron diffraction measurements have provided evidence of water in a glassy state lining the pore walls in a layer three molecules thick at 240 K.7 However, unlike the water in our samples,17 this state converts reversibly to a defective crystalline (cubic) state at lower temperatures. NMR relaxation studies below 260 K on ice in several different forms of mesoporous silica (86 Å pore diameter) have shown that the proton mobility in the disordered interfacial water layer in the temperature range 180–230 K is intermediate between that of bulk water and that of crystalline ice and that this motion is predominantly rotational rather than translational.6,7 The higher mobility than in the crystalline phase was ascribed to changes in the rotational freedom arising from the creation of a more hydrogen-bonded local environment.7 Thus, the disordered interfacial water was termed a “plastic crystal”6,7 in analogy with other crystals that have well-defined lattice structures but in which the molecules are orientationally disordered.34 

Earlier QENS measurements on water confined in nanoporous silica (22–25 Å pore diameter) that were analyzed with a relaxing cage model have found the average rotational relaxation time to increase by orders of magnitude at 220 K.15 No mention is made of an enhanced rotational motion or a plastic phase at low temperatures. Interestingly, their data show a drop in the rotational relaxation time of almost two orders of magnitude at 230 K, but it is not discussed.

2. The narrow Lorentzian component

We next turn to the narrow Lorentzian component in the QENS spectra, which is contributed by water molecules that are diffusing on a time scale an order of magnitude more slowly than those contributing to the broad Lorentzian.

a. Q-dependence of the narrow Lorentzian.

In Fig. 8, the HWHM, ΓN, of the narrow Lorentzian is plotted as a function of Q2 at several temperatures for our two CuO-coated samples. As was the case for the broad Lorentzian, the wet and dry samples yield similar results. For Q2 > 1.5 Å−2, ΓN of both samples tends to level off for temperatures T ≥ 240 K. As we have noted in the case of the broad Lorentzian, this behavior is characteristic of jump diffusion models, which have been used to interpret the dynamics of bulk supercooled water.27 In these models, the time between particle jumps or residence time is τ0, where ℏ/τ0 is the asymptotic HWHM at large Q. For the curves in Fig. 8, we have τ0 ∼ 150 ps at ∼250 K, i.e., a time which is about a factor of 6 longer than for bulk supercooled water.27 

FIG. 8.

HWHM of the narrow Lorentzian ΓN as a function of Q2 at selected temperatures as measured on BASIS for (a) the wet (60 µl H2O) and (b) the dry (10 µl H2O) CuO-coated samples. Solid curves are guides to the eye. The smaller error bars for the dry sample at 230 K reflect a longer counting time of ∼9 h compared to ∼4 h for the wet sample.

FIG. 8.

HWHM of the narrow Lorentzian ΓN as a function of Q2 at selected temperatures as measured on BASIS for (a) the wet (60 µl H2O) and (b) the dry (10 µl H2O) CuO-coated samples. Solid curves are guides to the eye. The smaller error bars for the dry sample at 230 K reflect a longer counting time of ∼9 h compared to ∼4 h for the wet sample.

Close modal

In addition to the longer residence times for T > 240 K, the results in Fig. 8 differ qualitatively from bulk supercooled water in two respects. First, for both samples, the asymptotic value of ΓN at high Q increases monotonically on cooling from 270 to 240 K from ∼4 to ∼5 µeV rather than decreasing.27 Second, at the lowest temperature of 230 K, ΓN decreases particularly at low Q. To see this behavior more clearly, we have plotted in Fig. 9(a) the parameter ΓNasym, defined to be the HWHM of the narrow Lorentzian averaged over the three data points with Q > 1.3 Å−1; ΓNasym approximates the asymptotic value of the HWHM at large Q. We see that ΓNasymp increases monotonically on cooling from 270 to 240 K as shown by the best fit lines (dashed) in Fig. 9(a). At 230 K, however, ΓNasymp of both the wet and dry samples falls about 30% below the dashed lines. To facilitate comparison with the results for bulk supercooled water, our data are replotted in Fig. 9(b) expressed as the average residence time τNasymp = ℏ/ΓNasymp. As reported in Ref. 27, the residence time τ0 for bulk supercooled water increases monotonically from 4.66 ps at 268 K to 22.7 ps at 253 K, whereas τNasymp in Fig. 9(b)decreases from ∼170 ps at 270 K to 140 ps at 250 K before abruptly rising at 230 K. Although we expect the time scale of the interfacial water dynamics to be longer than for bulk, the decrease in τNasymp on cooling is anomalous. Yet it occurs in both our wet and dry samples. We have investigated whether it might be an artifact of our two-Lorentzian fits. However, as shown in Fig. S2 of the supplementary material, there are no detectable residuals in our two-Lorentzian fits to the spectra summed over Q > 1.1 Å−1 in this temperature range.

FIG. 9.

Plotted for the dry sample (10 µl) and the wet sample (60 µl) is the temperature dependence of (a) the average HWHM of the narrow Lorentzian evaluated for Q > 1.3 Å−1, ΓNasymp. (b) The residence time τNasymp = ℏ/ΓNasymp. The dashed lines are best fits for temperatures T ≥ 240 K.

FIG. 9.

Plotted for the dry sample (10 µl) and the wet sample (60 µl) is the temperature dependence of (a) the average HWHM of the narrow Lorentzian evaluated for Q > 1.3 Å−1, ΓNasymp. (b) The residence time τNasymp = ℏ/ΓNasymp. The dashed lines are best fits for temperatures T ≥ 240 K.

Close modal
b. Type of molecular motion represented by the narrow Lorentzian.

With this evidence of its reproducibility, we turn to the question of the type of molecular motion the narrow Lorentzian component represents. Its much slower time scale compared to that of bulk supercooled water suggests that the narrow Lorentzian is contributed by water interacting strongly with and, hence, close to the CuO nanostructures. In Fig. 8, we see that ΓN, like ΓB of the broad Lorentzian, generally increases with Q at low Q. However, again, the quality of these data does not allow a determination of whether ΓN has a Q2 dependence. Nevertheless, the presence of this low-Q dependence followed by a plateau in ΓN is sufficient to exclude specific models, such as the diffusion of a particle in the interior of a sphere with an impermeable surface.35 At low Q, this model predicts a Q-independent quasielastic width resembling that of rotational motion; while, at larger Q, it reduces to the Q2 dependence predicted for translational diffusion in an infinite medium. Such behavior is clearly inconsistent with the Q-dependence of ΓN for both the wet and dry samples in Fig. 8, although it has been seen for water fully hydrating a protein.36 

As was the case for the broad Lorentzian, it is useful to compare the temperature dependence of the narrow Lorentzian intensity at low and high Q. In Fig. 7(b), we see that, for both the wet and dry samples, the narrow-Lorentzian intensity has a similar temperature dependence in the range of 270–240 K at both low and high Q. This Q-independence differs from the behavior of the broad Lorentzian [Fig. 7(a)] for which a qualitative difference in the shape of the curves (concave downward vs upward) was explained by a contribution to the QENS spectra from the rotational diffusive motion of the interfacial water molecules (see Refs. 29 and 30). As we have already identified the rotational motion of the interfacial water as contributing to the broad Lorentzian component, the absence of Q-dependence in the narrow Lorentzian intensity is consistent with a relatively slow diffusive motion that is purely translational.

c. Comparison with other systems.

To our knowledge, the translational diffusive motion of water with a relaxation time of ∼170 ps at a temperature of 270 K [see Fig. 9(b)] has not been observed for water confined to cylindrical nanopores in silicas.15,37 QENS measurements on these systems15 show a smaller translational relaxation time of ∼11 ps at 270 K. On the other hand, Mamontov et al.12 have observed a longer relaxation time in QENS measurements conducted on BASIS for water interacting with nanoparticles of the transition-metal oxides TiO2 and SnO2. Analyzing their spectra with a two-Lorentzian model similar to ours, they find for TiO2 that the slower component yields a translational relaxation time that increases from 209 ps at 265 K to 660 ps at 225 K.12 This relatively slow translational motion is consistent with a stronger hydrophilic interaction of water with the planar surfaces of the TiO2 nanoparticles than with the surfaces of silica nanopores. At 265 K, the time scale of this motion is comparable to that of the water represented by the narrow Lorentzian component in our wet and dry CuO-coated samples. However, on the planar TiO2 surfaces, the translational relaxation time increases by about a factor of three on cooling to 225 K whereas we observe an ∼25% decrease in τNasymp on cooling to 240 K followed by an abrupt increase to about the value at 270 K [see Fig. 9(b)]. We discuss this discrepancy further in Sec. IV and consider a possible explanation for it.

To interpret their measurements, Mamontov et al.12 performed classical ab initio molecular dynamics (MD) simulations on water multilayers adsorbed on (110) planes of TiO2 and SnO2, the predominant facet on the surfaces of the nanoparticles used in their QENS measurements. Their simulations indicate that water adsorbs in three structurally distinct molecular layers denoted L1, L2, and L3. The L1 species are either intact water molecules or dissociated hydroxyl groups chemisorbed on the oxide surface with a density exceeding that of bulk water. For an instrument with the energy resolution of BASIS, the MD simulations predict that the L1 species would contribute only to the elastic (delta function) component of the spectra. The second-layer L2, like L1, has a higher density than bulk and contains water molecules that are hydrogen-bonded to those in L1 as well as to oxygen atoms in the top oxide layer. The relatively slow translational motion on a nanosecond time scale observed on both the TiO2 and SnO2 surfaces is attributed to water molecules in layer L2 that make direct contact with the immobile species (H2O or OH) in layer L1. A faster diffusion, comprised of both rotational and translational motions, occurs on a time scale of picoseconds and is attributed to molecules in all three hydration layers that require breaking fewer hydrogen bonds compared to translational jumps.12 

To assess whether a similar model may apply to the hydration water in our CuO system, we compare thermogravimetric analysis (TGA) measurements on the TiO2 and SnO2 nanoparticles12 with those on CuO nanoparticles.38 All three materials show desorption peaks at two characteristic temperatures: TiO2 (334 and 471 K), SnO2 (331 and 574 K), and CuO (351 and 538 K). In the case of TiO2 and SnO2, the upper- and lower-temperature peaks can be attributed primarily to the desorption of L1 and L2 water molecules, respectively.12 This behavior contrasts with that of water desorption from nanoporous silica (MCM-41) that has been calcined above 773 K.7,15,33,39 It retains a moderate degree of hydroxylation7 and, hence, some hydrophilicity but shows no peak structure in TGA.39 

From the previous TGA measurements,38 we conclude that the water hydrating the CuO surfaces in our samples likely has a two-stage desorption process similar to that found for nanoparticles of TiO2 and SnO2. The higher-temperature stage would correspond to desorption of a layer L1 composed of chemisorbed water molecules and OH groups that are bound to the predominant (111) facets of the CuO blades with a strength comparable to that of water binding to the (110) surfaces of TiO2 and SnO2. As discussed in Ref. 17, our fixed window scans and diffraction patterns have shown evidence of chemisorbed species in their background levels. As in the case of TiO2, they would contribute only to the elastic component in our BASIS spectra. Following Ref. 12, we suggest that it is the more strongly hydrogen-bonded L2 water molecules that contribute to the relatively slow translational motion (time scale ∼100–200 ps) represented by the narrow Lorentzian component in the QENS spectra of our CuO-coated samples.

3. Anomalous QENS behavior between 240 and 230 K

In addition to observing water diffusion on a nanosecond time scale in their TiO2 sample, Mamontov et al.12 found evidence of a “fragile-to-strong” or “liquid–liquid” phase transition between 220 and 210 K in the L2 water undergoing slow translational diffusion. Previously, evidence of a liquid–liquid transition had been found for water confined in nanoporous silica40 and subsequent studies.33,41–43 There are now a number of systems containing confined water for which such a transition has been reported: Vycor,29 DNA,44 carbon nanotubes,45 and lysozyme.30,46

Although we have QENS spectra at only one temperature in the range (≲230 K) where evidence of a liquid–liquid transition has been reported in these other systems, we have observed anomalous temperature dependence of our QENS spectra from both our wet and dry CuO-coated samples on cooling from 240 to 230 K. Several anomalous features appear to manifest a common origin, which may be indicative of a phase transition.

To begin with, we consider model-independent evidence that does not depend on the fitting of the QENS spectra. In Fig. 10, we have plotted the QENS spectra of the dry sample measured on BASIS after averaging over Q > 1.1 Å−1. Here, we have selected the dry sample due to its longer measurement time (∼9 h). Its spectra are shown at the same temperatures as the curves of ΓN vs Q2 in Fig. 8(b). Focusing on the energy transfer range of ∼4–5 µeV, we see that the spectral intensity shows a slight decrease in each 10 K step from 270 to 240 K. However, between 240 and 230 K, the intensity drops by an amount comparable to that which occurred in the previous 30 K decrement. The intensity at 230 K is still well-separated from that of the resolution function, the spectrum taken at 100 K (see Fig. 3). This relatively abrupt reduction in intensity is suggestive of a phase transition. Further evidence comes from the FWSs, which also do not involve the fitting of the QENS spectra. As can be seen in Fig. 4(a), both CuO-treated samples show an upturn in their elastic intensity just below ∼237 K before leveling off at lower temperatures. Thus, without recourse to the fitting of the QENS spectra, we have observed both a relatively abrupt loss in quasielastic intensity and a corresponding increase in elastic intensity between 240 and 230 K.

FIG. 10.

Scattered neutron spectra averaged over Q > 1.1 Å−1 at various temperatures for the dry sample. The inset shows the full spectra. The spectrum at 100 K provides a measurement of the resolution function of the spectrometer.

FIG. 10.

Scattered neutron spectra averaged over Q > 1.1 Å−1 at various temperatures for the dry sample. The inset shows the full spectra. The spectrum at 100 K provides a measurement of the resolution function of the spectrometer.

Close modal

Next, we discuss evidence of a phase transition from the fitting of our QENS spectra. In Sec. III B 1, we have seen that the width of the broad Lorentzian ΓB becomes Q-independent and increases by ≳50% on cooling from 240 to 230 K (see Fig. 6). Because this Q-independence of ΓB is consistent with rotational motion28 and the translational motion of the bulk-like water has frozen out by 230 K [see broad Lorentzian intensity at low Q in Fig. 7(a)], we have attributed the increase in ΓB to a faster rotation of the water molecules. This temperature dependence is anomalous in that one usually expects the rate of diffusive motion to decrease on cooling. We suggest that the faster rotational motion could be caused by a transition to a low-temperature liquid phase of lower density as occurs in other systems of confined water.30,33,40–46

The temperature dependence of the narrow-Lorentzian component in our QENS spectra also suggests that a phase transition is occurring. In Sec. III B 2, we noted an abrupt decrease between 240 and 230 K in ΓNasym, the asymptotic width of the narrow Lorentzian spectral component at high Q [Fig. 9(a)]. This drop in ΓNaymp appears to preempt a further increase in ΓNaymp anticipated from the trend established from 270 to 240 K. Recall that we have interpreted ΓNasymp as measuring the translational jump rate of water confined near the surface of the CuO blades. On cooling from 240 to 230 K, the translational relaxation time τNasym = ℏ/ΓNasym increases from ∼135 ps to a value of ∼165 ps [Fig. 9(b)]. The time scale of this motion of the interfacial water molecules at 230 K together with the absence of Bragg peaks in neutron diffraction measurements (see Sec. III B 4) are consistent with a transition to a liquid phase.

Thus, from the QENS spectra, we can conclude that there is a relatively abrupt change in the intensity and width of the quasielastic scattering between 240 and 230 K on the two different time scales represented by the broad and narrow Lorentzian components. The FWSs of our two samples provide a finer grid of temperatures. The upturn in elastic intensity that occurs below the linear increase in each sample indicates that the anomalous features occur in the temperature range from 237 to 225 K. These results motivate further QENS measurements on CuO/water samples below 230 K to investigate the possibility of a liquid–liquid transition.

4. Wetting of water to the CuO blades

We can use the intensities of the Lorentzian components of the QENS spectra to compare the amount of the water types in our wet and dry samples and to elucidate the wetting of water to the CuO surface. To begin, we return to the temperature dependence of the intensity of the narrow and broad Lorentzian components of both samples plotted in Fig. 5. At 270 K, we can assume that all of the mobile water in each sample is fluid and at a temperature high enough to neglect the elastic component in the scattering [see Eq. (1)] but low enough that both Lorentzian components lie within the dynamic range of BASIS. Under these conditions, we take the sum of the narrow and broad Lorentzian intensities at 270 K as proportional to the total amount of mobile water within a sample’s scattering volume. Then, from Fig. 5, we find that the total amount of mobile water in the wet sample is ∼2.1 times that in the dry sample. This value agrees well with an estimate of a factor of ∼2.0 made from the FWS scan in Fig. 4, where the increase in elastic intensity between 270 and 205 K is proportional to the amount of frozen water in the sample.17 

Similarly, we can use Fig. 7 to estimate the increase in the bulk-like and interfacial water on going from the dry to the wet sample. As discussed in Sec. III B 1, we take the amount of bulk-like water undergoing translational motion in each sample to be proportional to the low-Q-weighted broad Lorentzian intensity at 270 K. We take the narrow Lorentzian intensity at 270 K as proportional to the amount of interfacial water undergoing slow translational motion. Due to the absence of Q dependence in the narrow Lorentzian intensity, we use the average of the low-Q and high-Q-weighted intensities in Fig. 7(b). In this way, we find the amount of bulk-like water in the wet sample to be ∼1.4 times greater than in the dry sample and the amount of interfacial water to be ∼2.8 times greater in the wet sample.

It is more difficult to determine the ratio of bulk-like to interfacial water in the same sample. While only the translational motion of the interfacial water contributes to the narrow Lorentzian intensity, we have identified contributions to the broad Lorentzian intensity from both the translational motion of the bulk-like water and the rotational motion of the interfacial water. In addition, there may be a contribution from rotational motion of the bulk-like water to the broad Lorentzian intensity. Although the low-Q weighting of the broad Lorentzian intensity [see Fig. 7(a)] reduces these contributions from rotational motion, it does not eliminate them entirely; and it is difficult to quantify them. Likewise, there is contribution from translational motion of the bulk-like water to the broad Lorentzian intensity in the high-Q range that is not included in the low-Q-weighted broad Lorentzian intensity. Using the analysis given in Ref. 29, we infer that the effect of this contribution is dominant so that it results in our low-Q-weighted broad Lorentzian intensity underestimating the amount of bulk-like water. Consequently, the low-Q-weighted broad Lorentzian intensity provides only an approximate measure of the amount of mobile bulk-like water in our samples but one that is sufficient for our purposes in Sec. IV. Taking its ratio with the narrow Lorentzian intensity at 270 K, we can make a rough estimate of the fraction of the mobile water that is bulk-like water in each sample. We find that about half the water is bulk-like in both samples. In Ref. 17, we estimated a total of 80 and 40 µl of water in the scattering volumes of the wet and dry samples, respectively. These amounts include residual water and other H-containing species that were present in the cell prior to loading and any immobilized molecules from the droplets added. Thus, they represent an upper bound on the number of mobile molecules in the samples’ scattering volume.

The fact that the increase in interfacial water is about twice than that of the bulk-like water on going from the dry to the wet sample has implications for the wetting of water to the CuO surfaces. If water were to interact only weakly with the CuO blades, we would expect the increase in the amount of bulk-like water to be greater than that of the interfacial water as the water content of a sample increases. Because the opposite is occurring, we propose that water wets along the length of the strongly hydrophilic CuO blades rather than form a film of uniform thickness covering a foil as considered in Ref. 17. Based on our discussion in Sec. III B 2, we propose that the film coating the blades consists of water chemisorbed in the first layer with mobile water in the next two layers, L2 and L3. This mode of film growth continues until the blades are maximally coated with water at which point the amount of bulk-like water should increase more rapidly. This picture is consistent with electron microscopy images showing water initially condensing at the base of the nanostructures17 followed by webs of water forming between the edges of neighboring CuO blades as shown in Fig. 11 [an image of this region with comparable resolution in the absence of water is shown in Fig. 1(b)]. It is also consistent with the water structure inferred from our neutron diffraction results.17 At 240 K, we did not observe Bragg peaks indicative of hexagonal D2O ice at twice the water content (120 µl) of our wet sample reported here. Only when the water content doubled again (240 µl) did broadened Bragg peaks appear. Combining our QENS and neutron diffraction results, we suggest that the strongly bound water adsorbing along the length of the CuO blades freezes into a disordered structure and that crystalline ice results from the freezing of bulk-like water that condenses after the blades are maximally covered. Apparently, at the coverage of our wet and dry QENS samples, the water exhibiting bulk-like dynamics (see Fig. 6) does not possess sufficient long-range order to crystallize at low temperatures.

FIG. 11.

High-magnification environmental SEM images of CuO nanostructures (a) before and (b) and (c) during water condensation. The images in the second row are identical to those in the first row except for the blue shading of the water to increase contrast with CuO blades. Drawings (a)–(c) in the bottom row illustrate qualitatively the water webs that form between blades. The scale bar in (a) is 0.5 µm.

FIG. 11.

High-magnification environmental SEM images of CuO nanostructures (a) before and (b) and (c) during water condensation. The images in the second row are identical to those in the first row except for the blue shading of the water to increase contrast with CuO blades. Drawings (a)–(c) in the bottom row illustrate qualitatively the water webs that form between blades. The scale bar in (a) is 0.5 µm.

Close modal

The adsorption of a molecularly thin water film along an appreciable length of a micrometer-sized CuO blade in our wet sample leads us to consider whether the film dynamics could change with height along the blade. For example, we would expect different behavior near the foot of the CuO blades where their density is the highest and the complicated nanoporous environment could inhibit freezing. To address this question, we focus on the narrow Lorentzian component in the QENS spectra, which we have identified as representing the relatively slow translational diffusion of the interfacial water. The narrow-Lorentzian intensity of the dry sample is essentially temperature independent from 270 down to 240 K and, in contrast, that of the wet sample decreases steeply from 270 to 260 K and then more slowly down to 240 K (see Fig. 5). Because we expect the wet sample to contain water further up the CuO blades, the loss of its narrow Lorentzian intensity on cooling could be explained by interfacial water in the more open space near the top of a blade beginning to freeze before water nearer its base.

A problem with this argument is that the FWS of the wet and dry samples share the same linear dependence on cooling between 270 and 257 K [see Fig. 4(a)]. Reconciliation of this apparent inconsistency could lie in the FWS measurements taken on heating17 where the two samples exhibit a different hysteresis. The dry sample shows a nearly linear decrease in elastic intensity on heating so that its hysteresis loop fails to close by 280 K. On the other hand, the hysteresis loop of the wet sample is larger in the area but closes by ∼273 K with the steepest decline in elastic intensity occurring between 260 and 270 K. The fact that a similar behavior does not occur in the dry sample on heating is consistent with melting of the interfacial water in the wet sample occurring at a height on the blades not reached by the water in the dry sample. Possibly, the absence of a pronounced freezing effect between 270 and 260 K in the FWS of the wet sample on the initial cooling was due to the poor thermal conductivity of the CuO blades and too fast of a temperature ramp (0.08 K/min). Such an effect would not occur in the QENS data taken on BASIS where the sample was allowed to equilibrate for ∼3 h at each temperature before spectra were recorded.

At this point, we should emphasize that on cooling below 257 K and in the subsequent heating scan (just described), there is no apparent inconsistency between the FWSs taken on the HFBS and the QENS spectra measured on BASIS for the wet and dry samples. As shown in Fig. 4, both samples show a linear dependence of the elastic intensity on cooling below 257 K in their FWS and in A(Q), the intensity of the elastic term in the dynamic structure factor [Eq. (1)]. We have seen that the ratio of the slope of the linear term in the wet sample to that for the dry sample determined from the FWSs [Fig. 4(a)] agrees well with that found using A(Q) [Fig. 4(b)]. Also, we note that the single linear term in the FWS of the dry sample [Fig. 4(a)] is determined mainly by the freezing out of the translational diffusive motion of the bulk-like water proportional to the low-Q-weighted broad Lorentzian intensity [see Fig. 7(a)]. In the case of the wet sample, the steeper linear term below 257 K in the FWS is due to its larger amount of bulk-like water (factor 1.4) and of interfacial water (factor 2.8), the latter being proportional to the narrow Lorentzian intensity.

As reported in Ref. 17, we have used neutron diffraction to investigate the structure of the frozen water in proximity to the CuO nanostructures. Although we had sufficient sensitivity, we did not observe Bragg peaks characteristic of crystallinity in a sample having twice the hydration level of our wet QENS sample reported here. On this basis, we concluded that the water in our CuO-coated samples freezes continuously into a noncrystalline state. Nevertheless, it would be desirable to have positive evidence for this structure rather than only a null result from diffraction measurements.

For this reason, we have conducted inelastic neutron scattering measurements of the vibrational spectra of the frozen water in both our wet and dry samples using the VISION spectrometer.25 We also obtained a vibrational spectrum for a “bare” sample of CuO-coated copper foils, which were dried in vacuum at room temperature for 24 h. Although this dehydrated sample (0 µl of H2O added) could still contain a small amount of tightly bound residual water, subtracting its spectrum from that of our wet and dry samples removes the spectral features due to CuO.

In Fig. 12(a), we show the vibrational spectra at a temperature of 5 K of the same wet and dry CuO-coated samples as measured on BASIS with the background spectrum of our dehydrated CuO-coated sample subtracted. A temperature of 5 K was chosen for these measurements in order to reduce the broadening of the low-energy phonon modes. All spectra were normalized to the intensity of the copper phonon at ∼3 meV (see Fig. S4). For comparison, we also show the spectrum of a sample of bulk crystalline (hexagonal) ice at 5 K also taken on VISION. The inelastic neutron vibrational spectrum of bulk ice has been analyzed previously47 and includes an acoustic mode at 7.1 meV, two optical modes (H-bond stretching) at 28.4 and 37.9 meV, and a broad librational band that spans energy transfers 67–130 meV. The acoustic and optical modes are collective excitations; and the librations include hindered rotations (wagging, twisting, and rocking) of the water molecule about its three principal axes.

For the dry CuO-coated sample (10 µl H2O added), we see [Fig. 12(a)] that the sharp librational edge at ∼67 meV, characteristic of bulk crystalline ice,47 has been significantly broadened at 5 K, extending over the range of 50–100 meV. In the case of the wet sample, the broadening is limited to the leading edge of the librational band. Such broadening has been observed for water confined in nanoporous silica41,43 and near the surface of CuO nanoparticles48 where it has been interpreted as indicating a softening and distortion of the H-bond networks. In addition, the weaker and broader acoustic and optical modes below ∼30 meV, observed in both the dry and wet samples, support the loss of long-range translational order.

These differences between the vibrational spectra of the CuO-coated samples and bulk crystalline ice at a temperature of 5 K are consistent with the presence of disordered ice located close to the CuO nanostructures. Here, we are assuming that there are no surface vibratory modes of water molecules adsorbed on the CuO surfaces that could broaden the librational band. Additional measurements (see Fig. S5) show that there is little temperature dependence of the librational band of a wet sample up to a temperature of 200 K. Thus, our measured vibrational spectra are in accord with the presence of disordered water at temperatures at which Bragg peaks indicative of crystalline ice are absent.17 

FIG. 12.

(a) Vibrational spectra measured on the VISION spectrometer at a temperature of 5 K for CuO-coated samples hydrated with 60 µl (blue) and 10 µl (magenta) of H2O. Spectra were normalized to the intensity of the copper phonon at ∼3 meV. For both samples, the spectrum of a dehydrated CuO-coated sample has been subtracted. A reference spectrum is shown for bulk ice at 5 K (black) whose elastic intensity has been scaled to that of the wet CuO sample over the energy range of ±1 meV. (b) The vibrational density of states for the wet sample in (a) (blue) is compared with that of water confined in nanoporous silica at 170 K and kilobar pressures (red) (from Ref. 43).

FIG. 12.

(a) Vibrational spectra measured on the VISION spectrometer at a temperature of 5 K for CuO-coated samples hydrated with 60 µl (blue) and 10 µl (magenta) of H2O. Spectra were normalized to the intensity of the copper phonon at ∼3 meV. For both samples, the spectrum of a dehydrated CuO-coated sample has been subtracted. A reference spectrum is shown for bulk ice at 5 K (black) whose elastic intensity has been scaled to that of the wet CuO sample over the energy range of ±1 meV. (b) The vibrational density of states for the wet sample in (a) (blue) is compared with that of water confined in nanoporous silica at 170 K and kilobar pressures (red) (from Ref. 43).

Close modal

Above ∼67 meV, the wet sample's spectrum has a shape similar to that of bulk crystalline ice but with smearing of the smaller peak/dip features. The maximum intensity of the wet sample's librational band at ∼67 meV is about a factor of three greater than for the dry sample. Such a large intensity difference might be explained by the signal orginating primarily in the interfacial water of the two samples. As discussed in Sec. III B 4, we estimated from the ratio of their narrow Lorentzian intensities at 270 K that the wet sample had a factor of ∼2.8 more interfacial water than the dry sample.

Of particular interest is the quantitative agreement of the vibrational density of states of our wet sample with that determined for water confined in nanoporous silica at kilobar pressures and a temperature of 170 K [red curve in Fig. 12(b)] as reported in Ref. 43. This agreement suggests that the water confined in the silica host (MCM-41 with a 15 Å pore diameter) has a short-range order similar to that in the interfacial ice of our wet CuO-coated sample. Wang et al.43 argued that the pressure and temperature dependence of both the boson peak in the energy range ∼2–10 meV and the librational band of their sample are consistent with a transition at a temperature of 170 K from a low-density liquid phase at 3 kbar (LDL) to a high-density phase at 4 kbar (HDL). Their conclusion is consistent with the phase diagram that they determined from density measurements on confined heavy water.42 They point out that the broad leading edge of the librational band in Fig. 12(b) compared to the step-like feature that they observe for bulk low-density (LDA) and high-density amorphous ice (HDA) is consistent with the confined H2O being in a liquid phase [see Fig. 5(b1) in Ref. 43].

Analysis of our QENS spectra supports a model of two distinct mobile water populations: a disordered water phase that wets the CuO blades and strongly interacts with them (population 1), and water located further from the nanostructures whose dynamics is “bulk-like” (population 2). Water in population 2 at ∼250 K performs translational diffusive motion at a rate close to that of bulk supercooled water; but, as indicated by our neutron diffraction measurements,17 does not crystallize on cooling to 230 K at the coverages of our wet and dry samples. Population 2 contributes to the broad Lorentzian component in our QENS spectra along with the water in population 1 performing the rotational motion.

We have proposed that water in population 1 contributes to the narrow Lorentzian component in our QENS spectra. The dependence of its width and intensity on temperature and Q is consistent with a slow translational motion confined to regions near the CuO blades. In addition, we have presented evidence that water molecules in population 1 are undergoing rotational motion, which contributes to the broad Lorentzian intensity in our spectra. At a temperature of 230 K, this rotational motion is about 30 times faster than that found in MD simulations of bulk supercooled water.

The diffusive translational motion that we have inferred for population 1 together with the absence of Bragg peaks in our neutron diffraction measurements17 are consistent with population 1 being in a liquid phase. This conclusion is further supported by the librational band that we observe in our inelastic spectra (Fig. 12 and Fig. S5). Its broad leading edge differs from the sharp edges observed for amorphous ice as well as hexagonal ice [Fig. 12(a)]. For these reasons, we conclude that population 1 water remains a liquid rather than transforming to a crystalline or amorphous solid at temperatures down to 230 K.

By combining our QENS results with our previous measurements using electron microscopy and neutron diffraction,17 we have proposed a film growth mode in which the water in population 1 first condenses at the foot of the “grass-like” CuO blades before wetting along their length to form a film a few molecular layers thick. Due to dispersion in the blade heights, bulk-like water presumably appears in the dry sample after the shorter blades are coated. As the longer blades approach maximal coverage, webs of water form between them (see Fig. 11). It is likely that additional bulk-like water fills the space between them until the blades are completely covered. We posit that the crystallization of bulk (hexagonal) ice occurs at this point. In practice, we observed bulk ice by neutron diffraction on cooling a sample with four times the water content of our wet sample.17 

We have compared the interfacial water dynamics inferred from our QENS measurements on the CuO-coated samples with that of water confined in nanoporous and mesoporous silica hosts and with water in proximity to TiO2 nanoparticles. The silicas represent a class of adsorbents that has a relatively well-ordered (cylindrical) pore structure but is less hydrophilic than our CuO coatings whereas the TiO2 nanoparticles are nonporous but have comparable hydrophilicity to our samples. At atmospheric pressure and a temperature of 270 K, the translational diffusion of water molecules in population 1 of our samples as measured by the relaxation parameter τNasymp is about a factor of 15 slower than the average Q-independent translational relaxation time in 22 Å-diameter cylindrical pores in silica (MCM-48-S).15 However, τNasymp depends only weakly on temperature so that on cooling to 230 K the translational diffusion is about a factor of 10 faster than for water confined in MCM-48-S.15 Furthermore, at 230 K, the rate of rotational motion within water population 1 is orders of magnitude faster than for water confined in MCM-48-S.15 Based on the TGA data discussed in Sec. III B 2, we have attributed these differences to the significantly greater hydrophilicity of the CuO blades compared to the surfaces of the cylindrical pores in the silicas.

In contrast to water confined in the silicas, water interacting with TiO2 nanoparticles exhibits a slow component in its QENS spectra with a relaxation time of 209 ps at 265 K.12 At this temperature, the time scale of this motion is comparable to that in population 1 of our CuO samples as characterized by the value of τNasymp ∼ 170 ps. However, the TiO2 and CuO systems differ in the temperature dependence of this slow motion of their interfacial water. As expected, the relaxation time in the TiO2 system increases monotonically on cooling whereas we have found our CuO samples to show an anomalous decrease in τNasymp down to 240 K [Fig. 9(b)].

From TGA data, we have seen that it is reasonable to assume that the interfacial water of the TiO2 and CuO systems has a similar layered structure in the temperature range 270–240 K with chemisorbed water in the first layer and mobile water above it. Therefore, it appears unlikely that the disparity in the temperature dependence of the relaxation time characterizing the slow translational diffusion in this temperature range is due to a qualitatively different structure of the interfacial water. Instead, we believe it essential to consider the difference in the water content of the two systems. Due to their lower hydration level, the TiO2 samples do not contain a component in their BASIS spectra that can be identified as “bulk-like.”12 Although the TiO2 BASIS spectra were modeled using a two-Lorentzian fit as we have done, the broad Lorentzian or “fast” component represents a slower diffusive motion than the bulk-like water (population 2) in our CuO samples. It is attributed to localized motions that take place in all three hydration layers and that require breaking fewer hydrogen bonds compared to translational jumps.12 

In Sec. III B 4, we estimated that roughly half of the mobile water in both our wet and dry CuO samples was bulk-like. We now consider the effect that it might have on the coexisting interfacial water. We suggest that the bulk-like water may act like a solvent that induces a restructuring within the interfacial water layers that differ both structurally and dynamically from itself. Such a restructuring is known to occur in physisorbed hydrocarbon films where the presence of a bulk solvent of homologous heptane molecules compresses a polycrystalline monolayer of a longer alkane chain adsorbed on a graphene surface and thereby alters its molecular orientations.49 In particular, it was shown that a molecular monolayer equivalent of heptane solvent was sufficient to drive a transition from an incommensurate to a commensurate monolayer of re-oriented long-alkane molecules. Upon withdrawal of all of the heptane solvent, the long-alkane monolayer reverted to its incommensurate structure.49 

Based on our estimate of approximately the same amount of bulk-like and interfacial water in both the wet and dry CuO samples, we suggest that the presence of the bulk-like water may stabilize the interfacial water (population 1) in a relatively high-density liquid phase at 270 K. Although we could not withdraw the bulk-like water from our samples, on cooling, it is continuously freezing to a solid state that is both structurally and dynamically mismatched with the denser, liquid interfacial water. Furthermore, due to its long relaxation times, the solid bulk-like water does not contribute to the QENS signal. That is, the frozen bulk-like water is not only invisible in our QENS spectra, it is incapable of restructuring the interfacial water. Thus, we expect a compressive effect of the bulk-like water on population 1 to diminish as the bulk-like water freezes. Under these conditions, the density of the chemisorbed water in the first layer would likely be unaffected and would remain higher than that of the remaining bulk-like liquid. However, a decrease in the compressive effect of the bulk-like liquid could lower the density of the more weakly bound second and third water layers, allowing a more developed hydrogen bond network and faster translational diffusion to occur. In support of this scenario, we see that on cooling from 270 to 240 K the decrease in the amount of liquid bulk-like water as measured by the low-Q-weighted broad Lorentzian intensity in Fig. 7(a) (see Sec. III B 1) tracks the decrease in the residence time τNasymp = ℏ/ΓNasymp as shown in Fig. 9(b) and Fig. S6.

Despite the differences between the dynamics of the water confined in nanoporous amorphous silica and water interacting with crystalline TiO2 nanoparticles, both systems are believed to exhibit a liquid–liquid transition.12,43 In discussing our results on CuO samples, we have noted some features of their water dynamics that appear consistent with the presence of a liquid–liquid transition. In the case of the nanoporous silicas, the transition occurs at kilobar pressures where the shape of the vibrational density of states around the librational band is remarkably similar to that which we have found for our wet CuO sample at 1 bar and comparable temperatures (see Fig. 12 and Fig. S5). This agreement suggests a similarity in their interfacial water structure on a length scale of a nearest-neighbor shell of water molecules.

A compressive effect of the bulk-like solvent in our CuO samples might explain how a liquid–liquid transition could occur at 1 bar, while kilobar pressures are required for water confined in the 15 Å-diameter pores of silica samples in which no bulk-like water is present.33,43,50 We note that QENS measurements on these silica samples at temperatures below 220 K have shown the average translational relaxation time in the high-density liquid phase at 4 kbar to be smaller than in the low-density phase at 1 bar.50 This result is in qualitative agreement with our measurements on CuO samples at 1 bar. As shown in Fig. 9(b), the relaxation time τNasymp that we observe at a temperature of 240 K (interfacial water compressed by bulk-like liquid) is smaller than at 230 K (all bulk-like water frozen).

Our results suggest the possibility of realizing a liquid–liquid transition at 1 bar in silica samples with larger diameter pores. The slowing of the translational diffusive motion accompanied by an increase in the rate of rotational diffusion that we observe for our CuO samples on cooling from 240 to 230 K is reminiscent of the “plastic phase” characterized by enhanced rotational motion that has been proposed from NMR measurements at 1 bar in a temperature range of 180–230 K.6,7 These experiments used silica with 85 Å-diameter pores that are large enough to accommodate bulk-like water along the pore centerline, which could compress the interfacial water into the plastic phase. Unfortunately, the NMR measurements did not yield the time scale of the rotational motion.

In conclusion, the weak scattered intensities imposed by the small surface area of our samples have prevented further measurements in the current study. We have QENS spectra at only one temperature (230 K) in the range where a liquid–liquid transition has been reported in other systems. It would be desirable to have data at additional temperatures and with improved statistics to corroborate the results reported here. In addition, molecular dynamics simulations of water layers interacting with CuO(111) surfaces, the dominant facet in our samples, could provide valuable insight into the interfacial water dynamics. It would be particularly useful to have simulations at a sufficiently high level of hydration to investigate the effect of a coexisting bulk-like water phase.

See the supplementary material for a nitrogen adsorption isotherm on a CuO-coated Cu foil sample (Fig. S1), the Q-dependence of the width of the broad Lorentzian component for both the wet and the dry CuO-coated samples at temperatures of 240 and 260 K (Fig. S2), additional vibrational spectra from both wet and dry samples (Figs. S3 and S4), and comparison of the temperature dependence of the broad Lorentzian intensity at low Q for both the wet and dry CuO samples with that of τNasymp = ℏ/ΓNasymp (Fig. S6).

This work was supported by the U.S. National Science Foundation under Grant No. DGE-1069091 and the University of Missouri Research Reactor. Access to the HFBS was provided by the Center for High Resolution Neutron Scattering, a partnership between NIST and the NSF under Agreement No. DMR-2010792. J.R.T. was partially supported by a GO! Internship funded by Oak Ridge National Laboratory (ORNL). A portion of this research used resources at the Spallation Neutron Source, a DOE Office of Science User Facility operated by the Oak Ridge National Laboratory. Part of the electron microscopy work was supported by the University of Missouri Electron Microscopy Core’s Excellence in Microscopy award. We thank R. A. Winholtz, A. I. Kolesnikov, T. White, and H. B. Ma for helpful discussions.

The authors have no conflicts to disclose.

The data that support the findings of this study are available within the article and its supplementary material.

1.
E.-P.
Ng
and
S.
Mintova
,
Microporous Mesoporous Mater.
114
,
1
(
2008
).
2.
J.
Drelich
,
E.
Chibowski
,
D. D.
Meng
, and
K.
Terpilowski
,
Soft Matter
7
,
9804
(
2011
).
3.
B.
Su
,
Y.
Tian
, and
L.
Jiang
,
J. Am. Chem. Soc.
138
,
1727
(
2016
).
4.
T. A.
Otitoju
,
A. L.
Ahmad
, and
B. S.
Ooi
,
J. Ind. Eng. Chem.
47
,
19
(
2017
).
5.
L.
Zhang
,
N.
Zhao
, and
J.
Xu
,
J. Adhes. Sci. Technol.
28
,
769
(
2014
).
6.
J. B. W.
Webber
,
J. C.
Dore
,
J. H.
Strange
,
R.
Anderson
, and
B.
Tohidi
,
J. Phys.: Condens. Matter
19
,
415117
(
2007
).
7.
J.
Jelassi
,
H. L.
Castricum
,
M.-C.
Bellissent-Funel
,
J.
Dore
,
J. B. W.
Webber
, and
R.
Sridi-Dorbez
,
Phys. Chem. Chem. Phys.
12
,
2838
(
2010
).
8.
T.
Takamuku
,
M.
Yamagami
,
H.
Wakita
,
Y.
Masuda
, and
T.
Yamaguchi
,
J. Phys. Chem. B
101
,
5730
(
1997
).
9.
A.
Verdaguer
,
G. M.
Sacha
,
H.
Bluhm
, and
M.
Salmeron
,
Chem. Rev.
106
,
1478
(
2006
).
10.
Q.
Li
,
J.
Song
,
F.
Besenbacher
, and
M.
Dong
,
Acc. Chem. Res.
48
,
119
(
2015
).
11.
O.
Björneholm
,
M. H.
Hansen
,
A.
Hodgson
,
L.-M.
Liu
,
D. T.
Limmer
,
A.
Michaelides
,
P.
Pedevilla
,
J.
Rossmeisl
,
H.
Shen
,
G.
Tocci
,
E.
Tyrode
,
M.-M.
Walz
,
J.
Werner
, and
H.
Bluhm
,
Chem. Rev.
116
,
7698
(
2016
).
12.
E.
Mamontov
,
L.
Vlcek
,
D. J.
Wesolowski
,
P. T.
Cummings
,
W.
Wang
,
L. M.
Anovitz
,
J.
Rosenqvist
,
C. M.
Brown
, and
V. G.
Sakai
,
J. Phys. Chem. C
111
,
4328
(
2007
);
E.
Mamontov
,
D. J.
Wesolowski
,
L.
Vlcek
,
P. T.
Cummings
,
J.
Rosenqvist
,
W.
Wang
, and
D. R.
Cole
,
ibid.
112
,
12334
(
2008
).
13.
F. G.
Alabarse
,
J.
Haines
,
O.
Cambon
,
C.
Levelut
,
D.
Bourgogne
,
A.
Haidoux
,
D.
Granier
, and
B.
Coasne
,
Phys. Rev. Lett.
109
,
035701
(
2012
).
14.
S.
Takahara
,
M.
Nakano
,
S.
Kittaka
,
Y.
Kuroda
,
T.
Mori
,
H.
Hamano
, and
T.
Yamaguchi
,
J. Phys. Chem. B
103
,
5814
5819
(
1999
).
15.
L.
Liu
,
A.
Faraone
,
C.-Y.
Mou
,
C.-W.
Yen
, and
S.-H.
Chen
,
J. Phys.: Condens. Matter
16
,
S5403
(
2004
).
16.
C. G.
Wiener
,
Z.
Qiang
,
Y.
Xia
,
M.
Tyagi
, and
B. D.
Vogt
,
Phys. Chem. Chem. Phys.
20
,
28019
(
2018
).
17.
J.
Torres
,
Z. N.
Buck
,
H.
Kaiser
,
X.
He
,
T.
White
,
M.
Tyagi
,
R. A.
Winholtz
,
F. Y.
Hansen
,
K. W.
Herwig
, and
H.
Taub
,
J. Appl. Phys.
125
,
025302
(
2019
).
18.
F. Z.
Zhang
,
R. A.
Winholtz
,
W. J.
Black
,
M. R.
Wilson
,
H.
Taub
, and
H. B.
Ma
,
J. Heat Transfer
138
,
062901
(
2016
).
19.
A.
Miskowiec
,
Z. N.
Buck
,
F. Y.
Hansen
,
H.
Kaiser
,
H.
Taub
,
M.
Tyagi
,
S. O.
Diallo
,
E.
Mamontov
, and
K. W.
Herwig
,
J. Chem. Phys.
146
,
125102
(
2017
).
20.
Z. N.
Buck
,
J.
Torres
,
A.
Miskowiec
,
E.
Mamontov
,
H.
Kaiser
,
F. Y.
Hansen
,
H.
Taub
,
M.
Tyagi
,
L.
Collins
, and
K. W.
Herwig
,
Europhys. Lett.
123
(
1
),
18002
(
2018
).
21.
Y.
Nam
and
Y. S.
Ju
,
J. Adhes. Sci. Technol.
27
,
2163
(
2013
).
22.
X.
Chen
,
J.
Shu
, and
Q.
Chen
,
Sci. Rep.
7
,
46680
(
2017
).
23.
E.
Mamontov
and
K. W.
Herwig
,
Rev. Sci. Instrum.
82
,
085109
(
2011
).
24.
R. T.
Azuah
,
L. R.
Kneller
,
Y.
Qiu
,
P. L. W.
Tregenna-Piggott
,
C. M.
Brown
,
J. R. D.
Copley
, and
R. M.
Dimeo
, “
DAVE: A comprehensive software suite for the reduction, visualization, and analysis of low energy neutron spectroscopic data
,”
J. Res. Natl. Inst. Stand. Technol.
114
,
341
(
2009
).
25.
Z.-L.
Xue
,
A. J.
Ramirez-Cuesta
,
C. M.
Brown
,
S.
Calder
,
H.
Cao
,
B. C.
Chakoumakos
,
L. L.
Daemen
,
A.
Huq
,
A. I.
Kolesnikov
,
E.
Mamontov
,
A. A.
Podlesnyak
, and
X.
Wang
,
Eur. J. Inorg. Chem.
2019
,
1065
.
26.
A.
Meyer
,
R. M.
Dimeo
,
P. M.
Gehring
, and
D. A.
Neumann
,
Rev. Sci. Instrum.
74
,
2759
(
2003
).
27.
J.
Teixeira
,
M.-C.
Bellissent-Funel
,
S. H.
Chen
, and
A. J.
Dianoux
,
Phys. Rev. A
31
,
1913
(
1985
).
28.
M.
Bée
,
Quasielastic Neutron Scattering: Principles and Applications in Solid State Chemistry, Biology, and Materials Science
(
Adam Hilger
,
Philadelphia, PA
,
1988
), pp.
150
153
,
180
183
.
29.
J.-M.
Zanotti
,
M.-C.
Bellissent-Funel
, and
S.-H.
Chen
,
Europhys. Lett.
71
,
91
(
2005
).
30.
J.-M.
Zanotti
,
G.
Gibrat
, and
M.-C.
Bellissent-Funel
,
Phys. Chem. Chem. Phys.
10
,
4865
(
2008
).
31.
W. S.
Price
,
H.
Ide
, and
Y.
Arata
,
J. Phys. Chem. A
103
,
448
(
1999
).
32.
J.
Qvist
,
C.
Mattea
,
E. P.
Sunde
, and
B.
Halle
,
J. Chem. Phys.
136
,
204505
(
2012
).
33.
K.
Yoshida
,
T.
Yamaguchi
,
S.
Kittaka
,
M.-C.
Bellissent-Funel
, and
P.
Fouquet
,
J. Chem. Phys.
129
,
054702
(
2008
).
34.
S.-K.
Wang
,
E.
Mamontov
,
M.
Bai
,
F. Y.
Hansen
,
H.
Taub
,
J. R. D.
Copley
,
V. G.
Sakai
,
G.
Gasparovic
,
T.
Jenkins
,
M.
Tyagi
,
K. W.
Herwig
,
D. A.
Neumann
,
W.
Montfrooij
, and
U. G.
Volkmann
,
Europhys. Lett.
91
,
66007
(
2010
).
35.
F.
Volino
and
A. J.
Dianoux
,
Mol. Phys.
41
,
271
279
(
1980
).
36.
M.-C.
Bellissent-Funel
,
J.
Teixeira
,
K. F.
Bradley
, and
S. H.
Chen
,
J. Phys. I
2
,
995
1001
(
1992
).
37.
A.
Jani
,
M.
Busch
,
J. B.
Mietner
,
J.
Ollivier
,
M.
Appel
,
B.
Frick
,
J.-M.
Zanotti
,
A.
Ghoufi
,
P.
Huber
,
M.
Fröba
, and
D.
Morineau
,
J. Chem. Phys.
154
,
094505
(
2021
).
38.
D. A.
Svintsitskiy
,
A. P.
Chupakhin
,
E. M.
Slavinskaya
,
O. A.
Stonkus
,
A. I.
Stadnichenko
,
S. V.
Koscheev
, and
A. I.
Boronin
,
J. Mol. Catal. A: Chem.
368-369
,
95
106
(
2013
).
39.
S. A.
Kozlova
and
S. D.
Kirik
,
Microporous Mesoporous Mater.
133
,
124
133
(
2010
).
40.
A.
Faraone
,
L.
Liu
,
C.-Y.
Mou
,
C.-W.
Yen
, and
S.-H.
Chen
,
J. Chem. Phys.
121
,
10843
(
2004
).
41.
S.
Kittaka
,
S.
Takahara
,
H.
Matsumoto
,
Y.
Wada
,
T. J.
Satoh
, and
T.
Yamaguchi
,
J. Chem. Phys.
138
,
204714
(
2013
).
42.
Z.
Wang
,
K.
Ito
,
J. B.
Leão
,
L.
Harriger
,
Y.
Liu
, and
S.-H.
Chen
,
J. Phys. Chem. Lett.
6
,
2009
(
2015
).
43.
Z.
Wang
,
A. I.
Kolesnikov
,
K.
Ito
,
A.
Podlesnyak
, and
S.-H.
Chen
,
Phys. Rev. Lett.
115
,
235701
(
2015
).
44.
S.-H.
Chen
,
L.
Liu
,
X.
Chu
,
Y.
Zhang
,
E.
Fratini
,
P.
Baglioni
,
A.
Faraone
, and
E.
Mamontov
,
J. Chem. Phys.
125
,
171103
(
2006
).
45.
E.
Mamontov
,
C. J.
Burnham
,
S.-H.
Chen
,
A. P.
Moravsky
,
C.-K.
Loong
,
N. R.
de Souza
, and
A. I.
Kolesnikov
,
J. Chem. Phys.
124
,
194703
(
2006
).
46.
S.-H.
Chen
,
L.
Liu
,
E.
Fratini
,
P.
Baglioni
,
A.
Faraone
, and
E.
Mamontov
,
Proc. Natl. Acad. Sci. U. S. A.
103
,
9012
(
2006
).
47.
J.
Li
,
J. Chem. Phys.
105
,
6733
(
1996
).
48.
E. C.
Spencer
,
N. L.
Ross
,
S. F.
Parker
,
R. E.
Olsen
, and
B. F.
Woodfield
,
Chem. Phys.
427
,
66
(
2013
).
49.
K. W.
Herwig
,
B.
Matthies
, and
H.
Taub
,
Phys. Rev. Lett.
75
,
3154
(
1995
).
50.
Z.
Wang
,
P.
Le
,
K.
Ito
,
J. B.
Leão
,
M.
Tyagi
, and
S.-H.
Chen
,
J. Chem. Phys.
143
,
114508
(
2015
).
51.

Certain commercial equipment, instruments, or materials (or suppliers) are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identified are necessarily the best available for this purpose.

Supplementary Material