Hypothesis
HKUST-1 consists of two types of large cages with the shape of a truncated cube (type-I and -II cages) and a type of small cage with the shape of a truncated octahedron (type-III cage, Supplementary Section 2). In a topological view, the link of type-III cages by the paddlewheel Cu centres in the fashion of a primitive cube constructs HKUST-1. Based on this topology, HKUST-1 has a unique structural feature: OMSs at the paddlewheel nodes face toward the open space of the type-I cages, which are large enough to accommodate multiply H-bonded water molecules.
With this structural insight in mind, we hypothesised the vibrational chain-connectivity of the H-bond, conceiving that H-bond between the coordinating and pore-filling H2O molecules could sensitively influence the Cu−Cu vibrational mode. When coordinating H2O is absent initially, open-state paddlewheel Cu2+ ions can be strongly associated together, exhibiting conceivably short bond length with high vibrational frequency (see Fig. 1c). Once H2O molecules coordinate at the Cu2+ centres, however, the paddlewheel Cu2+ ions can be loosely bound to each other, exhibiting conceivably longer bond length with lower vibrational frequency than that in the open state owing to the partial electron donation from coordinating H2O to an antibonding molecular orbital of Cu2+ ion (Fig. 1d)47–49. (We note that, while the Raman spectra of open-state and H2O-coordinating HKUST-1s have been reported previously39–43, no report has been made on crystallographic Cu−Cu lengths varying with the H2O-coordinations.) We hypothesised further that when the coordinating H2O molecules begin to form H-bonds with pore-filling H2O molecules, the paddlewheel Cu2+ ions can be firmly associated together reversely, exhibiting shortened bond length with higher vibrational frequency due to the weakened coordination strength by the chain connectivity of H-bonded water molecules (see Fig. 1e). Conceiving a strong association of H-bonded H2O to the coordinating H2O, we further hypothesised that the H-bonded H2O molecules could be site-selectively sat in the pores even at room temperature, differently from bulk H2O molecules that must be dynamic in their translational and rotational motions. On the other hand, the above hypothesis is difficult to demonstrate if a bulk experimental system such as a molecular Cu2(CH3CO2)4 complex dissolved in water is employed because, in an aqueous system, H2O-coordination-free complex and only H2O-coordinating complex without H-bonded water cannot be isolated. Thus, a water-endurable MOF, such as HKUST-1, with paddlewheel structure and permanent porosity is suitable for demonstrating the above hypothesis.
Sample preparation. We prepared activated HKUST-1 crystals that contain no water molecule (denoted as Act-HKUST-1). We also prepared H2O-coordinating HKUST-1 crystals that are expected to contain coordinating H2O only (denoted as H2O[C]-HKUST-1, where [C] stands for ‘coordinating’) and H2O-filled HKUST-1 crystals that contain both coordinating and pore-filling H2O (denoted as H2O[F]-HKUST-1, where [F] stands for ‘filling pores’, see Methods). Further, we prepared MeOH- and EtOH-filling HKUST-1 crystals (denoted as MeOH[F]-HKUST-1, and EtOH[F]-HKUST-1, respectively) via soaking Act-HKUST-1 crystals into pure MeOH and EtOH, respectively. The phase purities of the samples were checked by examining 1H nuclear magnetic resonance (NMR) spectroscopy and powder X-ray diffraction (PXRD, Supplementary Section 3).
Raman and crystallography studies at 220 K. We conducted Raman studies with Act-HKUST-1, H2O[C]-HKUST-1 and H2O[F]-HKUST-1 initially at 220 K. While an Act-HKUST-1 exhibited a Raman band for Cu−Cu stretching vibration at 231 cm−1, H2O[C]-HKUST-1 and H2O[F]-HKUST-1 exhibited the bands at 166 and 176 cm−1, respectively (Supplementary Section 4). To investigate the correlation between the Cu−Cu vibrational band energies and their bond lengths upon the HKUST-1 states, we conducted synchrotron SCXRD experiments with the crystals at 220 K. The SCXRD results show that the Act-HKUST-1 crystal contains no water molecule certainly (Supplementary Section 5). We expected the H2O[C]-HKUST-1 to contain coordinating H2O only because the Cu−Cu Raman band appeared at the lowest frequency (166 cm−1). In contrast to our expectation, however, the H2O[C]-HKUST-1 contained pore-filling H2O as well with 40 mol% at specific positions in type-I cages (denoted as type-1 pore-filling H2O with the abbreviation of H2O[F]1), of course, in addition to coordinating H2O (H2O[C], see Fig. 2a-c and Supplementary Table 2-3). The appearance of the Raman band at the lowest frequency despite the inclusion of the H2O[F]1 is ascribed to the dynamic coordination bond character of H-bond, by which H2O[F]1 can associate at the Cu2+ centre instantaneously after replacing the previous coordination bond of H2O[C] on the molecular time scale44–46. SCXRD result of H2O[F]-HKUST-1, where pore-filling H2O molecules are supposed to bind at coordinating H2O via H-bonds strongly, shows a type of coordinating H2O and two types of pore-filling H2O (type-1 and type-2; denoted with the abbreviation of H2O[F]2, see Fig. 2d-f). The result describes accurate positions of the H2O molecules: coordinating H2O[C] at the Cu−O[C] distance of 2.199 Å; pore-filling H2O[F]1 and H2O[F]2 at the O[C]−O[F] distance of 2.825 Å and 3.065 Å, respectively (Supplementary Table 3)1–3. Also, these crystallographic data support our hypothesis, providing the dependence of Cu−Cu lengths (2.485, 2.624 and 2.613 Å) upon the Cu2+ states.
Raman studies at various temperatures. To observe the influence of temperature on vibrational energy, we also conducted the Raman experiments at 298 K, slowly removing H2O from an H2O[F]-HKUST-1 under vacuum conditions. As a result, the pattern of spectral changes at 298 K was the same as that observed at 220 K. Whereas the Raman band of H2O[F]-HKUST-1 appeared at 176 cm−1, the band redshifted approaching 163 cm−1 as the filled H2Os were removed from the pores (Supplementary Section 6). However, the band was blueshifted to 231 cm−1 in reverse after all coordinated H2Os were thoroughly removed from the sample. To ascertain the absence of temperature effect on the vibration, we further performed the Raman analysis at 170 and 250 K with an H2O[F]-HKUST-1 (Supplementary Section 7). As predicted, the Cu−Cu Raman bands at both 170 and 250 K were the same as those measured at 220 and 298 K. However, as the temperature decreases, the O−H vibration of pore-filling H2O (3110 – 3146 cm−1) redshifted simultaneously with band sharpening due to the ice H-bond. On the other hand, we examined an isotope effect on the vibrational band after introducing D2O and H218O into HKUST-1 crystals, respectively. However, no isotope effect was observed (Supplementary Section 8). Thus, these Raman results explicitly support the absence of temperature and isotope effects on the chain connectivity.
Raman studies with EtOH and MeOH. We found that the chain connectivity is generic for EtOH and MeOH. EtOH[F]-HKUST-1 and MeOH[F]-HKUST-1 showed the Cu−Cu vibrational bands at 188 and 180 cm−1, respectively (Supplementary Section 9). However, these bands redshifted to 178 and 172 cm−1 when the pore-filling EtOH and MeOH were removed, respectively. After thorough removal of even coordinating EtOH and MeOH, both bands were rather blueshifted to 231 cm−1. Meanwhile, EtOH and MeOH confined in nanopores have shown a behaviour different from their bulk states in terms of C−O and C−C vibrations (Supplementary Section 9-10)50. Given that the binding of a molecule governs its internal vibrational energies, the above vibrational changes can be ascribed to the enhanced H-bond strength of confined EtOH and MeOH molecules.
In situ Raman at room temperature. We wondered if the water-ingress will give a result similar to the above water-egress in Raman spectral changes. We performed in situ Raman experiments to address this question, exposing an Act-HKUST-1 crystal to moist air at room temperature. As a result, the in situ Raman spectra have shown a pattern similar to those observed in the above ex situ (water-egress) experiments (Figure 3a and Supplementary Section 11). The spectra show that the vibrational band of the Act-HKUST-1 redshifts from 231 cm−1 to 163 cm−1 as the exposure time increases. Whereas the band intensity at 231 cm−1 gradually decreases, the band intensity at 163 cm−1 gradually increases. A notable feature is that these bands are ratiometric at discrete positions rather than continuously shifting. This result indicates that although paddlewheel Cu2+ centres are interconnected through BTC3− linkers, the Cu−Cu vibrations are not interconnected. When the exposure time increased more, by contrast, the band was blueshifted, approaching 176 cm−1 in a continuous fashion rather than a discrete one. Also, the band was broadened as the quantity of filling H2O increased. We speculated that the continuous shift and broadening of the Raman band could be attributed to the increase in the number of H-bonds around a coordinating H2O. Based on this speculation, we questioned how many H-bonds would be present around a coordinating H2O because the pattern of the band shift would be discrete if only one or two H-bonds are present.
To this end, we monitored the weight changes of Act-HKUST-1 crystals with a microbalance, exposing the crystals to moist air. The test showed a continuous increase in weight by 52.7 wt% (Fig. 3b). This weight change indicates that the stoichiometric molar ratio of [H2O]/[Cu2+] is 5.9. We also noted an inflexion point in the curve that appeared at ~8.3 wt% because this value is very close to the value theoretically calculated for the [H2O[C]]/[Cu2+] ratio (1.0). Taking into account the above results, we concluded that the molar ratio of [H2O[F]]/[H2O[C]] (the average number of H-bonds around H2O[C]) should be ~4.9. We also found that this ratio is close to a value (4.666) theoretically calculated based on in situ SCXRD results (vide infra). Therefore, the seemingly continuous Raman band shift can be ascribed to a result that ~4.9 H-bonds are reflected.
Additionally, we monitored the colour changes of the crystals using an optical microscope. While the colour of the Act-HKUST-1 crystal was a deep navy blue, the colour was turned to pale blue as the water exposure time increases (Fig. 3c and Supplementary Section 11). Using UV-vis absorption spectroscopy, we confirmed that the colour change comes from the changes in the d-d transition band of Cu2+ centres (see Supplementary Section 12)39–41.
In situ SCXRD at 298 K. We conducted in situ SCXRD experiments, exposing an Act-HKUST-1 crystal to moist air at 298 K to observe the correlation between the vibrational chain connectivity and Cu−Cu length. SCXRD results show that while the Cu−Cu length of the Act-HKUST-1 was 2.484 Å, the length was gradually increased as the exposure time increased and subsequently maximised to 2.632 Å when the contents of H2O[C] and H2O[F]1 reached 100% and 34%, respectively (Fig. 4a-c; Supplementary Section 13 and Supplementary Table 14-15). However, the length was inversely decreased as the exposure time increased further and finally minimised to 2.617 Å when the exposure time was 240 min. When the exposure time was 180 min, type-2 and type-3 pore-filling H2O molecules (abbreviated as H2O[F]3) began to fill in the type-II cages (Fig. 4d-f). Thus, these in situ crystallographic results agree well with the in situ Raman results (vide supra). Concretely, the vibrational band was redshifted from 231 to 163 cm−1, simultaneously with the Cu−Cu length elongation from 2.484 Å to 2.632 Å. However, the band was reversely blueshifted from 163 to 176 cm−1, simultaneously with the Cu–Cu shortening from 2.632 Å to 2.617 Å. Therefore, the H-bond reflects on the Cu−Cu length and its vibrational mode, exhibiting chain connectivity. Meanwhile, the inclusion of 34 mol% H2O[F]1 can be ascribed to the dynamic coordination bond character of the H-bond, as described above. Based on the single-crystal structure of a fully H2O-filled HKUST-1 (i.e., H2O-HKUST-1(9th )), we calculated the theoretical numbers of coordinating and pore-filling H2O molecules in a unit cell. The calculation resulted in the maximum values of 1.00 and 4.666 for the ratios of [H2O[C]]/[Cu2+] and [H2O[F]]/[Cu2+], respectively, showing a good agreement with the above water uptake result (Supplementary Table 16).
Distortion of paddlewheel node. Based on the Cu−Cu length variation after H2O-coordination, we expected that the OBTC−Cu−OBTC angle could also be altered, giving rise to a geometrical distortion around the Cu2+ centres. Fig. 5 shows a pattern that the OBTC−Cu−OBTC angle dramatically decreases as the Cu−Cu length increases. We speculate that the increase of Cu−Cu length and subsequent decrease of the bond angle is attributed to the partial electron transfer from the coordinating H2O to the antibonding molecular orbitals of Cu2+ such as b1(x2-y2) or a1(z2)47–49. A decrease in Cu−O[C] length demonstrates this speculation because the decrease can be attributed to the strong association of coordinating H2O to Cu2+ that can arise from its electron donation (Fig. 5). Nonetheless, we could not observe any pattern in the changes after the substantial formation of H-bond from H2O-HKUST-1(6th ). Although a more thorough study of the H-bond effect on the OBTC−Cu−OBTC angle and Cu−O[C] length is necessary, we speculate that these are ascribed to the only slight alteration of Cu−Cu length after H-bond formation (2.632 and 2.617 Å, respectively).