Introduction

Porous coordination polymers (PCPs) or metal-organic frameworks1 have emerged as a new class of nanoporous materials constructed from organic ligands and metal ions. Compared with other porous materials, PCPs have an excellent ability to control pore structures, sizes, and surface properties, as a result of the wide variety of combinations of organic ligands and metal ions. This capability means that the nanopores of PCPs could be designed to realize an exotic state of adsorbed gas molecules not seen in the bulk phase using confinement effects2. Such states are interesting for molecular physics in nanospaces, and also for potential new applications in facilitating chemical reactions in the nanopores of PCPs.

Gaseous water molecules are significantly affected by the size, geometry, and inner wall properties of the confined space. Previous studies show that confinement effects occur in spaces narrower than 100 nm scale and can be classified into two regimes. The first regime appears in spaces from 2 nm to a few tens of nanometers, where the confined water is conceptually divided into a water core with bulk liquid-like properties and a shell of water existing around the inner walls with properties different from bulk water3,4. The local water around the inner walls and its electrostatic forces affect the collective behavior of bulk liquid-like water, resulting in a higher viscosity and proton mobility than bulk water5,6. The second regime appears in spaces narrower than 2 nm, where all water molecules interact with inner walls7,8,9. In this regime, nanoconfined water is strongly affected by the space restrictions and overlapped potentials of the inner walls, giving rise to exotic properties. Among nanoporous materials, CNTs have been extensively used as a well-defined hydrophobic nanospace8,10,11. Previous studies showed that the H-bonds of water in hydrophobic nanospace are highly restricted, and as a result confined water at ambient temperature and pressure has a number of H-bonds in between gas and liquid water. The intermediate value is a signature for liquid–gas supercritical water8. Furthermore, recent molecular dynamics simulations predict an intriguing possibility of hydrophobic confinement to achieve solid–liquid supercritical water12 when water density becomes comparable to liquid water (e.g. the critical density = 1.33 g/cm3 in nanofilms13). Due to the space limitations, the structural symmetry is not a well-defined order parameter in nanospaces. This allows the existence of a solid liquid critical point, which is generally restricted in the bulk phase by the famous symmetry argument that an isotropic liquid cannot continuously transform into a solid with a discrete symmetry14,15. Above a critical point, water is allowed to have intermediate physical parameters (e.g. number of H-bonds and diffusivity) between the limits of ice and liquid phase.

Considering the previous studies, water molecules confined in the 1-nm-size nanospace with the large density have a possibility to exhibit an exotic state not seen in the bulk phase. In this paper, we realized confined water with a density of 1 g/cm3 in the 1-nm-size hydrophilic nanochannels of PCPs, which provides an attractive potential for water molecules. The potential can dramatically lower the free energy of the adsorbed state due to the large gain of the enthalpy and increase the density of confined water. We investigated the structural and dynamical properties of the confined water in PCPs. By using X-ray diffraction analysis and IR spectroscopy, we revealed that water in PCPs has an ordered structure (characteristic in ices) with a significant number of broken H-bonds (characteristic in liquids). We showed that nanoconfined water in PCP-1 is exotic with no corresponding states in the bulk phase diagram. We found that the structural properties of water in PCP-1 are quite similar to those of solid–liquid supercritical water in CNTs. This result will open up new potential applications of an exotic state of water in PCPs to control chemical reactions but also experimental systems to clarify the existence of solid–liquid critical points.

Results

Isotherm of water adsorbing PCP

We synthesized a water-adsorbing PCP that is composed of copper ions and isophthalates. The water adsorption/desorption isotherm was measured at 298 K, as shown in Fig. 1. At P/P0 = 0.04 where P0 is the saturated water vapor pressure, the data exhibit a very sharp adsorption/desorption of two water molecules per copper ion (referred to as “guest water” hereafter). In the following, the PCP fully occupied by guest water molecules is called PCP-1, while PCP-2 refers to the one without guest water at low relative pressure. Heavy water adsorption/desorption isotherms were also measured (Fig. 1), and we verified that the adsorption/desorption properties exhibit no isotope effects. We also performed thermogravimetric analyses of PCP-1 (see Supplementary Methods). The data revealed another water molecule (referred to as “coordinated water” hereafter) per copper ion that is strongly bound to the PCP framework (Supplementary Figs. 2 and 3). As shown below, this corresponds to the water molecule coordinated to the copper ion.

Fig. 1: Water isotherms.
figure 1

Adsorption/desorption isotherm of PCP adsorbing water at 298 K.

Full size image

Ordered structure of guest water in PCP-1

We performed single-crystal X-ray diffraction (SXRD) analysis of PCP-1 to investigate whether guest water in PCP-1 has an ordered structure as seen in ices or disordered structure as seen in liquids12. Figure 2a shows the unit cell of PCP-1, which is a 2D sheet structure with a square lattice framework (Supplementary Data 1). The 2D sheets are stacked along the c-axis (Fig. 2b) with a separation distance of 6.9 Å, creating two kinds of quasi-1D channels (CH1 and CH2) with approximately 1 nm diameter. We observed all water molecules, i.e., three per copper ion, in CH1, which means CH2 is empty (Fig. 2a). The fact that oxygen atoms of water molecules are observed by SXRD means that confined water molecules form an ordered structure along the channel (Fig. 2b). This is characteristic in ices12, although measured at room temperature. Three types of water molecules (whose oxygen atoms are colored red, blue, and purple) exist in different environments. The red-colored oxygen atoms are in close proximity to the copper ions (green) and strongly bound by a coordination bond (coordinated water). In contrast, the other two oxygen atoms (blue and purple) are located near the center of the channel at different positions along the c-axis (sites I and II in Fig. 2b, respectively). The distances between blue (site I) and red oxygens, and purple (site II) and brown oxygens are 2.75(1) and 3.219(8) Å, respectively (Fig. 2a). This means that these are the guest water molecules weakly bound by H-bonding. The analyses are supported by powder crystal X-ray diffraction (PXRD) and simulation results (Supplementary Fig. 4 and 6). Also, the distance between the four water molecules (2.87(1) Å) at sites I and II suggest the existence of H-bonding network among them (Fig. 2a)16. However, the guest water molecules at sites I and II are separated by a distance of 3.33(2) or 3.58(2) Å, which is comparable to the upper limit of the H-bond length17. This indicates considerably weak H-bond interactions between them (Fig. 2b)16,17. We also confirmed that there are no differences between the framework structures of PCP-1 and PCP-2 (Supplementary Fig. 4). The averaged O–O distance among neighbored water molecules is calculated as (2.75 + 2.87 + 2.87 + 3.33 + 3.58)/5 = 3.08Å. This corresponds to the density of 1.02 g/cm3, and water density in PCP-1 is comparable to liquid water: 0.997 g/cm3 at 298 K, 0.1 MPa17.

Fig. 2: Crystal structure of water adsorbing PCP.
figure 2

a The unit cell of the crystal structure of PCP-1. The bondings are shown, but atoms are colored as follows: Cu, green; C, gray; O of isophtalates, brown; O of coordinated water, red; O of guest water, blue and purple. Hydrogen atoms are omitted. The magnified view shows the confined water structure with distances between oxygen atoms. See also Supplementary Fig. 1. b Side view of the channel CH1 of PCP-1.

Full size image

Significant fraction of broken H-bonds of confined water in PCP-1

SXRD results show that the periodic array of the oxygen atoms of guest and coordinated water molecules are realized in CH1 even at room temperature. However, it is quite difficult to determine the positions of hydrogen atoms by SXRD measurements. To investigate the H-bond structure, we measured infrared (IR) absorption spectra of the OH stretching modes of confined water, which is sensitive to the H-bonds state16,18,19. The blue curve in Fig. 3a shows the IR absorption spectrum of PCP-2 with only coordinated water. It shows two sharp absorption peaks (H1 and H2) owing to the OH stretching modes of coordinated water. This interpretation was confirmed by the isotopic effect observed for the two absorption peaks (see Supplementary Methods). The IR spectrum of PCP-2 with coordinated D2O (red curve in Fig. 3a) shows new absorption peaks (D1 and D2) at frequencies that are 2−1/2 times lower than those of H1 and H2, respectively20. As shown by the black curves in Fig. 3a, each peak is well-defined by the Lorentz function with parameters shown in Table 1, which means that the inhomogeneity of H-bonding structure of coordinated water is negligible. Due to the strong correlation between the resonance frequency of the OH stretching mode and H-bond length (O–O distance between water molecules), the peaks H1 (at 3450 cm−1) and H2 (at 3655 cm−1) can be assigned to H-bonded OH (HB–OH) with a H-bond length of ≈3 Å and non-H-bonded OH (Free-OH)16, respectively. The number of H1 and H2 arms determined from the absorbance area normalized by oscillator strength are almost the same (see Supplementary Methods), which means that every coordinated water molecule has both HB-OH and free-OH arms. The crystallographic data (Data File S1) shows that several oxygen atoms (colored brown in Fig. 3c) are located within ≈3 Å. However, only one atom satisfies the tetrahedral geometry formed by the sp3 hybrid orbital of the oxygen atoms in coordinated water molecules. As shown in Fig. 3c, the HB-OH arm of coordinated water is H-bonded to an oxygen atom of an isophthalate in the adjacent layer, and the other OH arm is directed to the interior of the 1D nanochannel, which plays a role of guest water attractions. Furthermore, based on this knowledge and the SXRD result (Fig. 2a), we can infer the H-bond network structure of the guest water as shown in Fig. 3d. Due to the negligible H-bonds between guest water in site I and II (Fig. 2b), guest water molecules in site I have free or (considerably weak H-bonded) OH-arms.

Fig. 3: Hydrogen bond structure of coordinated and guest water.
figure 3

a The IR absorption spectra of PCP-2 adsorbing only coordinated H2O (blue line) and D2O (red line) at 295K below 4% RH. The small peaks (2800–3100 cm−1) correspond to the CH stretching modes of the framework. The black broken and solid lines are the result of Lorentz curve fittings of absorption peaks due to coordinated H2O and D2O, respectively. b The IR absorption spectra of PCP-1 adsorbing H2O (blue line) and D2O (red line) at 30% RH. Each IR band is well decomposed into three Gaussian functions (black dashed) and one Lorentz function (solid line). Yellow and cyan dashed lines are the sum of four deconvoluted functions of the red and blue line, respectively. c The schematic diagram of H-bond structure of coordinated water in PCP-2. Hydrogen atoms are colored by cyan. d The schematic diagram of H-bond structure of coordinated and guest water in PCP-1 inferred from SXRD and IR experiment results.

Full size image
Table 1 Results of peak positions, FWHMs, and ratio of absorbance area for Lorentz fitting of coordinated H2O and D2O in Fig. 3a and H4 & D4 in Fig. 3b.
Full size table

The blue curve in Fig. 3b shows the IR absorption spectrum of PCP-1 adsorbing guest water. A broad absorption band (H3) is observed in the frequency range of the OH stretching modes. H3 shows an approximately 2−1/2 times isotopic shift by replacing all the guest and coordinated H2O with D2O (red curve in Fig. 3b), indicating that H3 is due to the adsorbed H2O. The broad absorption band H3/D3 is decomposed into three Gaussian functions and one Lorentz function (H4/D4) as shown in Fig. 3b. These are assigned to OH stretching modes with strong (3202 cm−1), intermediate (3390 cm−1), weak (3539 cm−1), and no H-bond (H4, 3630 cm−1)19. The surprising point is the existence of weak H-bonded and free OH, which is a characteristic of liquids not seen in ices19,21. Water in PCP-1 has both characteristics of liquids (broken H-bonds) and ices (ordered structures). From the absorbance area of H4 peak normalized by oscillator strength, we can estimate the number of free-OH in PCP-1. Table 1 shows that the absorbance area of H4 peak is almost the same as that of H2 (Free-OH of coordinated water). Because of the small difference in peak position between H2 and H4, their oscillator strengths are comparable to each other16. Therefore, the number of free-OH in PCP-1 is almost the same as that of coordinated water (four per unit cell). This is consistent with the inferred H-bonds structure in Fig. 3d which shows the four free (or considerably weak H-bonded) OH-arms of guest water in site I per unit cell. The mean number of H-bonds per hydrogen atom in guest and coordinated water molecules is estimated as (24-4)/24 = 0.83, which lies between those of liquids (0.558) and ices (≈112). Interestingly, this indicates that despite the periodical array of oxygens that is characteristic in ices, confined water does not satisfy Pauling’s ice rule which states that each water molecule offers/accepts two hydrogen atoms to/from two neighboring molecules and has four H-bonds22.

We also investigated a dynamic property of confined water molecules by time-lapse IR spectroscopy. We prepared PCP-1 adsorbing the mixture of H2O and D2O. We exposed the sample to D2O gas vapor and measured the spectral changes in H3, which provides information about the diffusion dynamics of guest water molecules. The result is shown in Fig. 4a. D3 gradually increased in height with time (t), whereas H3 gradually decreased. These spectral changes are attributed to the exchange of guest H2O/HDO with D2O molecules outside the nanochannels. The H3 area in Fig. 4a normalized by H3 area in Fig. 3b provides the number of hydrogen atoms per unit cell, nH-atom(t), which is shown in Fig. 4b. It decreases exponentially (green curve), as expected from the 1D diffusion model23:

$$n_{{\mathrm{H}} \mbox{-} {\mathrm{atom}}}\left( t \right) \propto {\mathrm{exp}}\left( { - t/\tau } \right).$$
(1)
Fig. 4: Dynamic properties of confined water.
figure 4

a The series of IR absorption spectra after exposing to D2O vapor. The small peaks (2800–3100 cm−1) corresponding to the CH stretching modes of the framework show no spectral changes. b The number of hydrogen atoms per unit cell (nH-atom(t)). The green solid line shows a curve fitting result with an exponential function.

Full size image

The time constant τ = 3.8 × 103 s determined from the fit allows us to deduce the diffusivity (D) of water in the nanochannel using the following relation23:

$${\it{\uptau }} = l^2/\left( {3D} \right) + l/{\upalpha}.$$
(2)

Here, l and α are the lengths of the 1D nanochannel and surface permeability, respectively. The root-mean-square of the channel length <l2>1/2 is determined as 0.8 μm by using scanning electron microscope (SEM) (Supplementary Fig. 5). In general, the surface permeability influences the molecular exchanges with gas molecules in the case of small powder particles. It hinders the molecules from entering and leaving the pores. The surface permeability of PCP-1 is unknown, but at least a positive number. Therefore, this allows us to determine the lower limit of the proton diffusivity as 0.6 × 10−16 m2/s, which is higher than that of ice VII (2.6 × 10−19 m2/s at 300 K, 10 GPa24). Since the microscopic hopping process is induced by molecular rotations, this result shows that the structural rearrangements of confined water molecules occur faster in PCP-1 than in solid ices at the same temperature.

Discussion

The experimental results clearly show the duality of the confined water in the 1 nm nanochannel of PCP-1: an ordered structure (characteristics in ices), a significant fraction of broken H-bonds Nbroken-H (17%) (characteristics in liquids), and large diffusivity. It is commonly believed that a solid–liquid critical point does not exist in the bulk phase diagram. This is because of the famous symmetry argument that an isotropic liquid cannot continuously transform into a solid with a discrete symmetry14,15. Therefore, in this sense, one can state that the confined water in the nanospace of PCP-1 is exotic with no corresponding states in the bulk phase diagram.

We found that the structural and dynamical properties of water in PCP-1 are quite similar to those of solid–liquid supercritical water in CNTs12. The molecular simulation study clearly showed that water in CNTs with 1 nm diameter has a solid–liquid critical point in the high pressure and ambient temperature region of the phase diagram. Nbroken-H changes discontinuously (from 0% to ~50%) during the solid–liquid phase transition that occurs below the critical point. Above the critical point, however, the molecular simulation showed that Nbroken-H changes continuously in between the limits of ices and liquid phases. Therefore, the intermediate value of Nbroken-H (17%) observed in PCP-1 is characteristics of the solid–liquid supercritical water predicted in CNTs. Furthermore, the ordered square water (Fig. 2a) is the same structure as predicted in the supercritical region12, which also supports the similarity. However, it should be noted that it remains unclear whether water in PCP-1 is solid–liquid supercritical water. Further experiments such as investigation of thermodynamic responses must be performed to clarify this point.

Finally, we comment on the origin of the exotic state. It is trivial that the ordered structure of oxygens in CH1 arises from the periodic array of the attractive potentials in PCP-1. Also, the existence of broken H-bonds (Fig. 3d) probably comes from the structure of the PCP framework that aligns guest water molecules in site I and II with separation distances comparable to the upper limit of the H-bond length (Fig. 2b). This implies that attractive potentials of inner walls are strong enough to promote square water structure but not enough to freeze them on each site, resulting in both characteristics of ices and liquids.

In summary, we observed an exotic state of water in the 1-nm-size hydrophilic nanospaces of PCP-1. We found that the structural properties of water in PCP-1 are quite similar to those of solid–liquid supercritical water in CNTs. The designed hydrophilic nanospaces of PCP will open up new experimental systems at ambient pressure to clarify the existence of the predicted solid–liquid critical points, which is one of the unresolved issues in physics. Also, nanoconfined water is related to the number of biological processes, such as protein folding, enzymatic reaction, and biological surface hydrations. Our results will contribute to a greater understanding of an important role of nanoconfined water in biological activities as well as to control chemical reactions using the exotic state of water in PCPs.

Methods

H2O and D2O sorption isotherm for PCP-2 at 298K

The D2O and H2O sorption isotherms were obtained using a BELSORP-aqua3 instrument. All of the adsorbents were degassed via freeze–pump–thaw cycles using liquid N2. In prior to sorption measurement, the powder crystals of PCP-1 were placed in sample cells, which were evacuated at room temperature for more than 5 h and then purged with dry He to afford PCP-2. The sorption isotherm measurements were conducted, without exposing PCP-2 to air, using an equilibrium time of 600 s for taking each equilibrium point. The sorption measurements were conducted up to relative pressures (P/P0, P0 = 2.84 kPa) of ca. 0.95 (Fig. 1). The desorption isotherm of PCP-1 for water vapor at 298 K is shown in Fig. 1, in which the number of H2O molecule per Cu is calculated based on the amount of water adsorbed in PCP-2.

Samples for IR measurements

The sample was a pellet (~100 μm thickness) made by pressing microcrystalline powder sample of PCP-1 (the typical size is 0.67 μm). The pellet was placed in a chamber, wherein the humidity was regulated to control the adsorbed water amount in PCP-1. Chamber volume is π × 1.28 cm × 1.28 cm × 0.7 cm = 3.6 cm3. The low humidity below 4% is achieved by flowing dry air into the chamber. The high humidity (~30%) is obtained by the mixture of dry air and saturated steam. The humidity in the chamber is measured by means of a digital humidity meter (Model CTH-1100, Custom, Tokyo). The IR spectra were performed at ambient temperature by using Hyperion Fourier Transform Infrared (FTIR) microscope connected to the Bruker VERTEX 80v spectrometer at a spectral resolution of 4 cm−1. The steady states of coordinated and guest water were measured by 64 accumulation scans. The used D2O gas molecules were evaporated from 99% concentration of D2O (CAS. No. 7789-20-0, Cambridge Isotope Laboratories Inc., product code DLM-4-99-1000).

Isotopic substitutions

Isotopic substitution of guest and coordinated H2O with D2O was achieved just by exposing PCP-1 to D2O vapor at ~30% RH. The concentration of deuterium atoms of guest and coordinated water is controlled by exposing PCP-1 to the proper mixture of the deuterium oxide and distilled water (CAS. No. 7732-18-5, Wako Pure Chemical Industries Ltd, Osaka, product code 049-16787).