Polyoxocationic antimony oxide cluster with acidic protons.

The success and continued expansion of research on metal-oxo clusters owe largely to their structural richness and wide range of functions. However, while most of them known to date are negatively charged polyoxometalates, there is only a handful of cationic ones, much less functional ones. Here, we show an all-inorganic hydroxyiodide [H10.7Sb32.1O44][H2.1Sb2.1I8O6][Sb0.76I6]2·25H2O (HSbOI), forming a face-centered cubic structure with cationic Sb32O44 clusters and two types of anionic clusters in its interstitial spaces. Although it is submicrometer in size, electron diffraction tomography of HSbOI allowed the construction of the initial structural model, followed by powder Rietveld refinement to reach the final structure. The cationic cluster is characterized by the presence of acidic protons on its surface due to substantial Sb3+ deficiencies, which enables HSbOI to serve as an excellent solid acid catalyst. These results open up a frontier for the exploration and functionalization of cationic metal-oxo clusters containing heavy main group elements.

INTRODUCTION

Polyoxometalates (POMs) are anionic metal oxide clusters consisting of interconnected metal-oxygen polyhedral units, which are present in a solution or as a crystalline solid when combined with counter ions (, ). Since its first discovery in the 1800s, a large variety of negatively charged POMs has been found, including Lindqvist type [M6O19] (M = Mo, W, V, etc.), Anderson type [H(M′O6)M6O18], Keggin type [XM12O40] (X = S, P, Si, etc.), and Wells-Dawson type [X2M18O62], to name only a few. In addition to the compositional diversity of each POM, the intrinsic isomerism and ability to form lacunary and substituted structures further increase the structural diversity and expand the reaction possibilities (). The resulting structures permit a wide range of chemical and physical properties such as homo- and heterogeneous catalysis (, ), selective ion capturing (), and biochemical applications () arising from the redox activity of transition metals in POMs, as well as quantum computing ().

Contrary to the negatively charged POMs, positively charged metal-oxo clusters (metal-oxo polycations) are quite rare (, ), but in recent years, special attention has been paid to the development of the synthesis of cationic metal-oxo clusters (–), such as [Al13O4(OH)24(H2O)12]7+ (Al13) (), [Dy2(H2O)8V12O32]2+ (), and [Zr25O10(OH)50(O2)5(H2O)40]8+ (, ), and isolation of these clusters. While applications ranging from water purification to catalysis have been searched for Al13 and its analogs (–), the research on cationic metal-oxo clusters, in general, is in its infancy, and most studies focus on the synthetic and structural aspects, while their functions remain practically unexplored. Thus, the finding of a previously unknown cationic metal-oxo clusters not only provides a structural impact of its own but also promises unique functions in a solution or as various complexes when combined with negatively charged counter anions.

In this study, we report on [H10.7Sb32.1O44][H2.1Sb2.1I8O6][Sb0.76I6]2·25H2O (HSbOI) serendipitously found during the search for an antimony oxyiodide for photocatalytic applications. This product has cationic clusters of [H10.7Sb32.1O44]19.0+ (HSbO) that form a face-centered cubic (fcc) packed lattice with [H2.1Sb2.1I8O6]11.6− and [Sb0.76I6]3.7− clusters at the interstitial sites (Fig. 1A). Unlike the structurally related Sb35O44X5 cluster in Cu20Sb35O44X37 (X = Cl, Br) (, ), the cationic cluster in HSbOI carries protons, of which the surface protons are acidic, achieving excellent activity in heterogeneous acid-catalyzed reactions. Thus, this study opens a frontier in the study of metal-oxo clusters as solid catalysts, which has so far been limited to anionic transition metal oxide clusters.

Fig. 1.

Structure of dehydrated [H10.7Sb32.1O44][H2.1Sb2.1I8O6][Sb0.76I6]2.

(A) Dehydrated HSbOI [Fm-3m, a = 21.3305(1) Å]. Red and purple spheres, respectively, represent O and I atoms, while light blue, blue, and green spheres represent Sb1/2/3, Sb4, and Sb5, respectively. Cationic clusters HSbO form an fcc lattice with the anionic clusters [H2.1Sb2.1I8O6]11.6− and [Sb0.76I6]3.7− at the octahedral and tetrahedral interstitial sites, respectively (see also fig. S7, A and B). (B) A large cluster HSbO. (C) The HSbO cluster is composed of (left) the inner shell [Sb16−δO8] (δ ~ 4) and (right) the outer shell [Sb32−δO36] (δ ~ 2) {or [Sb28−δ(Sb324−δ′O36)] (δ ~ 0.8, δ′ ~ 1.2)}. The outer shell contains a [Sb324−δO36] (δ ~ 1) cage with eight [Sb21−δ] cations (δ ~ 0.1). The inner and outer shells are connected through Sb2─O1 [dashed bond in (B)]. For simplicity, the Sb2-site (32f) splitting into 96k is omitted in (A) to (C) (see fig. S7C for the structure with the Sb2-site splitting). (D) The [Sb324−δO36] cage with sodalite 4668 topology. (E) A medium cluster [H2.1Sb2.1I8O6]11.54−. (F) A small cluster [Sb0.76I6]3.7−.

RESULTS AND DISCUSSION

Synthesis and structural characterization of HSbOI

HSbOI was obtained by accident in an attempt to prepare an unreported Sb4O5I2, which we expected to be isostructural with Sb4O5Cl2 and Sb4O5Br2 (), but has a smaller bandgap, using a solution process suitable for solar cell applications. This oxyiodide (Fig. 2C, bright orange, and fig. S1) was obtained by dissolving antimony oxide in an aqueous hydrogen iodide solution at room temperature (RT), followed by dilution with water. The powder synchrotron x-ray diffraction (SXRD) pattern (Fig. 2C) differs markedly from the expected pattern for the monoclinic Sb4O5Cl2 structure, with numerous peaks at low angles that prevented us from even determining the unit cell.

Fig. 2.

ED and XRD patterns of [H10.7Sb32.1O44][H2.1Sb2.1I8O6][Sb0.76I6]2·25H2O (HSbOI).

(A) The ED patterns of HSbOI along [001] and [101] (see fig. S3 for details). (B) Structure model obtained by ED tomography and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of HSbOI, where red, blue, and purple balls represent O, Sb, and I atoms, respectively. Note that STEM measurement does not distinguish between antimony (Z = 51) and iodide (Z = 53). (C) Rietveld refinement of SXRD pattern [λ = 0.42096(1) Å] in the region of 14° to 24° (see fig. S8 for the full data). Red overlying crosses and green solid curve represent the observed and the calculated intensities, respectively. The bottom stands for the residual curve. The inset is a photo of bright orange powder of HSbOI.

Because of its particle size of <200 nm (fig. S2), to resolve the crystal structure, we used three-dimensional electron diffraction (3D-ED) tomography, a powerful technique that allows to determine structures from submicrometer single crystals and has recently been applied to a variety of materials from metal oxide to proteins (–), but not to metal-oxo cluster compounds. The combination of selected area ED patterns (Fig. 2A, fig. S3, and table S1), high-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images (Fig. 2B and fig. S4), and structure solution from 3D-ED data using charge flipping methods provides the initial structure model in the cubic Fm-3m space group with a large lattice parameter a = 21.3305(1) Å (details in the Supplementary Materials). The 3D-ED analysis revealed large pseudo-spherical Sb38−δO36 clusters with a diameter of ~13 Å (fig. S5A), assembled in an fcc packed lattice. Subsequent Rietveld refinement from the powder SXRD data (Fig. 2C) revealed additional oxygens in the large clusters and the presence of small- and medium-sized clusters. There are considerable amounts of Sb vacancies in each cluster type, with average compositions of Sb38−δO44 [δ = 5.95(5)], Sb6−δI8O6 [δ = 3.89(3)], and Sb1−δI6 [δ = 0.241(5)] for the large, medium, and small clusters, respectively (table S2).

The chemical composition of [Sb32.1O44]8.3+[Sb2.1I8O6]13.7−[Sb0.76I6]3.7–2 estimated above is not charge-neutral, but the presence of substantial amounts of protons is suggested by the intense background of a neutron powder diffraction profile originating from the incoherent scattering on hydrogen (fig. S11). To maintain charge neutrality, 12.8 protons should be present per formula unit. As discussed later, magic angle spinning nuclear magnetic resonance (MAS-NMR) experiments and ab initio calculations demonstrate that protons bind to the large and medium clusters, together with water of crystallization (fig. S12). Assuming that each cluster takes on two specific compositions around the average H or Sb value, the final formula can be given by [({(H5Sb12O8)0.5(H6Sb12O8)0.5}0.92{(H5Sb11O8)0.5(H6Sb11O8)0.5}0.08)·(Sb27)0.6(Sb28)0.4·({(H5Sb323O36)0.8(H6Sb323O36)0.2}0.8{(H5Sb322O36)0.8(H6Sb322O36)0.2}0.2)]·[{(H2Sb43I8O6)0.9(H3Sb43I8O6)0.1}0.1{(H2Sb42I8O6)0.9(H3Sb42I8O6)0.1}0.9]·[(I6)0.24(Sb5I6)0.76]2·25H2O. Unless otherwise stated, we will hereafter use a simplified chemical formula of [H10.7Sb32.1O44][H2.1Sb2.1I8O6][Sb0.76I6]2·25H2O (HSbOI). As displayed in Fig. 1A, the octahedral and tetrahedral voids in the positively charged clusters of HSbO in the fcc structure are occupied by medium and small anionic clusters, respectively.

Structural features of clusters in HSbOI

The HSbO cluster [Sb32.1O44]8.3+ (Sb38−δO44, where δ ~ 6) consists of an Sb32−δO36 (δ ~ 2) outer shell and an Sb6−δO8 (δ ~ 4) inner shell (Fig. 1B). The outer shell is composed of a [Sb24−δO36] cage and eight [Sb1−δ] cations (Fig. 1, B and C, and fig. S7c). The [Sb24−δO36] cage contains the Sb24 framework with truncated octahedron of 4668 geometry (Fig. 1D), which is also seen in other compounds (–), as discussed later. The outer shell Sb3+ (Sb2 and Sb3) cations are strongly asymmetric, with a triangular oxygen environment (SbO3E), where Sb 5s lone electron pair (E) is pointing outward, thereby stabilizing the cluster structure, which is theoretically validated as shown in Fig. 3A. A similar Sb38−δO44 cluster is found in Cu20Sb35O44X37 (X = Cl, Br) (fig. S6), where the cluster appears to be stabilized by sheets of CuX4 tetrahedra (, ), whereas the HSbO cluster is ionically stabilized with two types of anionic clusters. Another difference is that Cu20Sb35O44X37 is prepared using a solid-state reaction at 500°C in vacuum, so there are no protons present.

Fig. 3.

Calculated large cluster of HSbO polycation.

(A) Lone-pair orbitals of the outer shell Sb atoms for [Sb34O44]14+, which consists of defect-free outer shell [Sb8(Sb24O36)] and [Sb2O8] inner shell taking into account large Sb1 vacancies (see Materials and Methods), with the isovalue of 0.1 for orbital plotting. (B) 3D electrostatic potential (ESP) maps calculated from the 3D density functional theory (DFT)–optimized wave function of the extended cluster of [Sb36O50]8+, where capping O─Sb (× 2) and O─VSb (× 4) originated from the neighboring medium Sb6−δI8O6 clusters (where VSb represents an antimony vacancy) are added to the [Sb34O44]14+ cluster. All the oxygen sites show smaller ESPs (red), which attract positively charged protons (see also fig. S20). In contrast, all the antimony sites show larger ESP values (blue). The value of ESP decreases in the order of oxygen at the outer shell, inner shell, and capped O─Sb, meaning that the outer shell oxygen is the least weakly bound to protons and is easiest to release them. Each figure is visualized using the VESTA program ().

The medium [Sb2.1I8O6]13.7− cluster (Fig. 1E) resembles the face-capped M6Ch8X6 units of the well-known Chevrel-type compounds with transition metals (M = Mo, Re,…; Ch = S, Se,…; X = Cl, PEt3,…) (). The formation of Chevrel-type clusters can be interpreted in terms of electronic stabilization, where the highest occupied molecular orbitals are mainly of metal-metal bonding nature. In contrast, the d orbitals of Sb3+ ion in Sb6−δI8O6 are fully occupied (4d10), indicating that this medium cluster is not stable by itself but is somehow stabilized by the “nanospace” formed by the fcc network of HSbO clusters through covalent interactions. The O2− ions in Sb6−δI8O6 bind not only to Sb4 within this cluster but also to Sb3 in the HSbO cluster, indicating a substantial interaction between the two clusters, which is supported by bond valence sum (BVS) calculations (see Supplementary Text). Thus, the six Sb4 ions appear to be assembled merely to orient the Sb 5s orbitals toward the octahedral center. Such “clustering” of cations with lone-pair electrons has been observed in solids with Sb, Bi, Pb, and Te ().

The small [Sb0.76I6]3.7− cluster (Fig. 1F) is also unusual because the Sb deficiency (24%) implies a cation-vacant (I6)6− octahedron. For inorganic molecules and metal complexes, MX6 octahedra (e.g., [Fe(CN)6]3− and [SiF6]2−) do not permit a metal deficiency without being accompanied by a ligand defect (). The situation differs markedly from extended oxides such as perovskite oxides where octahedral metal defects are often seen (). The strong interaction with the HSbO cluster is once again expected to stabilize this cluster; when only the octahedral center (Sb5) is considered, the BVS value of I2 is as small as −0.44, but it rises to −0.86 when Sb2 and Sb3 sites of the large cluster are included.

Acidic protons and catalysis

1H MAS-NMR profile of the dehydrated HSbOI was collected to obtain information about the states of protons. As shown in Fig. 4A, the spectrum mainly consists of three peaks centered at 10, 6, and 4.5 parts per million (ppm) with a peak area ratio of about 25:10:26 (see fig. S13 for the spectrum of the as-prepared sample). We computationally determined the stable proton positions for an extended cluster of [Sb36O50]8+ that has a large cluster {consisting of outer [Sb28(Sb324O36)] and inner [Sb12O8]} taking into account Sb vacancies, plus two capping [O─Sb4] and four [O─VSb] units of the middle cluster [Sb46−δI8O6] (where VSb represents an antimony vacancy). The density functional theory (DFT) and the gauge-independent atomic orbital (GIAO) method were used to compute chemical shifts (see the “Computational methods” section). The extended [Sb36O50]8+ structure was fixed as the experimentally determined one, while the position of an added proton was fully optimized. As shown in Fig. 4B, there are roughly three stable H+ positions at 9.5 to 10.5 ppm for H+ bound to O2 and O3 in the outer shell, 4.7 ppm for H+ bound to O4 in the inner shell, and 6.0 ppm for H+ bound to the capped O1. These calculated chemical shifts reproduced the experimental data extremely well. The 1H MAS-NMR spectrum after soaking in heavy water (Fig. 4C) supports the 10-ppm signal being assigned to protons on the surface of large clusters. Moreover, a 2D 1H radio frequency–driven recoupling (RFDR) NMR spectrum (Fig. 4D), which recouples the homonuclear dipolar couplings depending on the nuclear distance, further validates the aforementioned computational assignment because the observed cross peak between 6 and 10 ppm means that protons corresponding to each chemical shift are close in distance.

Fig. 4.

1H MAS-NMR data of HSbOI.

(A) An RT 1H MAS-NMR spectrum of dehydrated HSbOI with calculated chemical shifts, where red, green, and blue bars denote protons attached to O2 and O3 in the outer shell, capped O4, and O1 in the inner shell, respectively. Three peaks centered at 10 ppm (outer shell), 6 ppm (cap), and 4.5 ppm (inner shell) have the area intensity ratio of about 25:10:26. Note that this ratio is a rough estimate because there is a small peak around 4.5 ppm (and 1 to 3 ppm) attributed to residual hydrated water (fig. S12). (B) Optimized structures and NMR chemical shifts of an extended [HSb36O50]9+ cluster, where the proton (light blue circle) is attached to O2 or O3 (open red circle) in the outer shell, to O1 (open red circle) in the inner shell, and to the capped O4 (filled red circle) (fig. S19). Black circles represent Sb atoms. The oxygen in the position overlapping the proton is omitted. (C) 1H NMR spectrum of the samples after immersing the as-prepared one in H2O (blue) and in D2O (red). The disappeared signals at 10 and 6 ppm after D2O immersion indicate H/D exchange, supporting our computational assignment of proton adsorption in the outer shell and capped oxygen. (D) A contour plot obtained by covariance processing a 2D 1H RFDR NMR spectrum of dehydrated HSbOI with a mixing time of 1.6 ms. A 2D 1H RFDR spectrum recouples the homonuclear dipolar couplings depending on the nuclear distance. Therefore, the cross peak between 6 and 10 ppm means that protons corresponding to each chemical shift are close in distance.

The relationship between 1H MAS-NMR chemical shift and acidity has been well known both experimentally () and theoretically (); the larger the chemical shift in solid-state NMR, the higher the acidity. For instance, the silanol groups of zeolites (H-mordenite) exhibit 1H MAS-NMR chemical shifts of ca. 2 ppm (), while those of the Brønsted acid sites of zeolites (H-mordenite) () and acidic Nafion () are observed at ca. 4 and 9.7 ppm, respectively. The chemical shift of the protons in the outer shell (10 ppm) of the HSbO cluster is as large as those of the typical acidic salts of Keggin-type POM () and the acidic polymer of Nafions (), suggesting that the surface of the HSbO cluster in HSbOI must be acidic.

In expectation of the high catalytic activity of the HSbO surface as a Brønsted acid, we carried out pinacol rearrangement reaction, which is a typical acid-catalyzed reaction (). In the pinacol rearrangement reaction, elimination of water from pinacol (a diol) gives pinacolone (a ketone) via 1,2-methyl shift as a major product and 2,3-dimethyl-1,3-butadiene (an olefin) via simple dehydration as a minor product (). As shown in Fig. 5A, HSbOI exhibits high catalytic activity as a solid catalyst (entry 1, 100% conversion and >99% selectivity to pinacolone) at 373 K and 1 hour in toluene [see figs. S14 and S15 for the 1H NMR and gas chromatography (GC) chart, respectively]. The catalytic performance of HSbOI is comparable to those of Keggin-type heteropolyacids (100% conversion and 43 to 78% selectivity to pinacolone at 423 K and 1 hour), which are well-known superacids (). The reaction under the same conditions using the well-known POM-based solid acid catalyst Cs2.5H0.5PW12O40 (Cs2.5) () showed 100% conversion of pinacol and 93% selectivity to pinacolone, demonstrating that the performance of HSbOI is comparable. The powder XRD pattern of HSbOI after the reaction is analogous to that before the reaction (fig. S16), showing that the crystal structure of the catalyst is maintained during the reaction. As shown in Fig. 5A, HSbOI can be easily recovered from the reaction mixture by simple filtration and washing with toluene and can be reused at least once (entry 2, 100% conversion and >92% selectivity to pinacolone).

Fig. 5.

Catalytic reaction with HSbOI.

The result of (A) pinacol rearrangement reaction in toluene and (B) Barbier-Grignard reaction in aqueous media. TON and TOF were calculated from the equation TON = [yield (%)]/[amount of catalyst (mol)/amount of substrate (mol)] and TOF (hour−1) = TON/reaction time (hours). Asterisk indicates that 300 mg of HSbOI corresponds to 0.01 mmol based on its unit cell (Z = 4). Note that the amounts of metal-oxo clusters (HSbO or [PW12O40]3−) are the same in entries 1 and 2. Dagger indicates that no other products were detected in entry 2 (2,3-dimethyl-1,3-butadiene was not detected). Unidentified peaks at long retention times were observed in the GC chart of the reaction solution of entry 2.

In addition, we carried out allylation of aldehydes using the Barbier-Grignard reaction, which is a typical acid-catalyzed reaction in water and an environmentally friendly C─C bond formation method (). Figure 5B lists the results of the Barbier-Grignard reaction catalyzed by HSbOI; 1-phenyl-3-butene-1-ol was identified as the sole reaction product of this reaction by GC and 1H NMR. Removing the catalyst stopped the reaction, indicating that the observed catalysis is truly heterogeneous (fig. S17). Elemental analysis of the reaction solution by inductively coupled plasma mass spectrometry (ICP-MS) showed that the amount of Sb that leached from HSbOI after 24 hours was ca. 0.4% and negligible. The yield and turnover number (TON) with 0.2 mole percent (mol %) of HSbOI (based on benzaldehyde) in H2O at 353 K and 24 hours were 63% and 315, respectively. Notably, the turnover frequency (TOF) reached a high value of 50 per hour with 0.1 mol % of HSbOI at 1 hour, which is higher than that of a typical solid acid catalyst, a scandium triflate–grated mesoporous silica (TON ~ 200 and TOF < 1 at 353 K) (). It is known that the activity in the Barbier-Grignard reaction increases by increasing the acid strength of POM-based solids as catalysts [i.e., pKa (where Ka is the acid dissociation constant) of the metal aqua ions as counter cations] (), so these results confirm the highly acidic property of HSbOI. Because HSbOI is not porous (fig. S18), the activity per proton is high, given that the reaction with benzaldehyde in this case occurred only on the small surface of HSbOI particles. We note that no by-products due to the redox property of iodine were observed in both reactions.

While metal-oxo cluster materials have been extensively studied as catalysts for a variety of acid-base and redox reactions (, , ), most of the clusters are composed of transition metals and/or light main group metals. In this regard, HSbOI composed of late main group elements (Sb) is unprecedented in demonstrating excellent properties as a solid acid. In addition to the large chemical shift, the unusually weak binding of the protons in the large cluster’s surface (attached to O2 or O3) can be presumed from the computed O─H bond length of 1.013 to 1.045 Å, which is longer than 1.010 Å for O1─H and 0.979 Å for O4─H) (fig. S19). To obtain further insights into the bonding nature of H+, we calculated electrostatic potentials (ESPs) (). Figure 3B shows that the DFT-optimized ESPs of the extended [Sb36O50]8+ cluster increase in the order of outer shell oxygen, inner shell oxygen, and capped oxygen. The reason for the presence of acidic protons attached to the outer shell is not immediately available but is speculated to be as follows. A large number of vacancies are created in the inner shell to relieve the overbonding of antimony sites (Sb1), which, in turn, generate a large number of charge balancing protons (three H+ for each Sb3+ defect). However, because there are only a limited number of oxygen sites (O1) in the inner shell that can bond more strongly with hydrogen than oxygen sites of the outer shell, the extra protons should bond to oxygen of the outer shell, resulting in high Brønsted acidity. In addition, the stereochemically active lone-pair electrons of the HSbO surface (Fig. 3A) may somehow contribute to the reaction process.

Outlook

We have successfully synthesized an all-inorganic hydroxyiodide containing a large cationic antimony-based cluster, featuring acidic protons. Considering that acid reactions have mainly been catalyzed by anionic POM clusters containing early transition metals (W, Mo, V, etc.) (, , ), using positively charged clusters with heavy main group elements and controlling the metal ion defects as in this study may open up possibilities in the design of solid-state catalysts. As there are many metal-oxo clusters containing Bi and Sb (, ), they should also exhibit acidity by protonation via introducing defects.

It is noted that sodalites [Al6Si6O24] framework (), uranium clusters [{(UO2)O2(OH)}24]24− (), and M8P16 (M = Cu, Ni) in clathrate compounds AM2P4 (A = Sr, Ba) () have a structure with similar topology (truncated octahedron, 4668) as the [Sb24−δO36] cage in the outer shell of the HSbO clusters (fig. S6). In these compounds, this framework can be modified to yield, e.g., the inner space to be used as a reaction field (). Likewise, our HSbO clusters could be structurally modified, for example, we may be able to synthesize antimony oxide cage structures with different topologies (, ) and higher-order structures (). In addition, dissolving and isolating HSbO clusters would permit the use of single clusters for homogeneous catalysis and a broader deployment as building blocks for various organic-inorganic composites. For example, if we can introduce permanent porosity into HSbO-based polycation composites, then the catalytic activity would be greatly enhanced because the reactant molecules can access not only the HSbO clusters on the crystal surface but also all the HSbO clusters inside.

Recall that various functions have been explored in anionic POM clusters such as the Keggin and Wells-Dawson types by modifying the clusters themselves, e.g., through elemental substitution () and addition of organic functional groups (), and by forming organic-inorganic complexes in combination with various cations (). Such synthetic approaches should be possible in our Sb-based cluster system, which can adjust the acidity and introduce porosity to improve the acid catalytic properties or to add other functions such as proton conductivity. It would also be interesting to examine how the catalysis of HSbOI is altered by elemental substitution, e.g., halide substitution at the iodine sites. To this end, it is important to understand the reaction mechanism, including the condensation process during the cluster formation.

MATERIALS AND METHODS

Synthesis

The powder sample of HSbOI was prepared as follows: Sb2O3 (99.9%, Wako; 2.0 g, 6.86 mmol) was dissolved in a mixture of HI (Wako; 6.8 M, 15.6 ml) and water (10 ml) and was stirred for 1 hour at 75°C. Adding water (500 ml) to this solution resulted in the precipitation of an orange powder. The suspension was stirred at 75°C for 1 hour and then at 0°C for 1 hour. The resulting product was recovered by filtration, washed with filtrate, and dehydrated through drying at 100°C under vacuum.

Characterization

The specimen for TEM study was prepared by grinding the powder in anhydrous hexane and putting a few drops of the resulting suspension onto Cu grids covered with a holey carbon layer. The ED experiment was performed using a FEI Tecnai G2 microscope operated at 200 kV. Energy-dispersive x-ray spectroscopy was performed using a FEI Tecnai Osiris microscope (200 kV), equipped with a Super-X detector. High-resolution STEM images were acquired using a FEI Titan 80-300 “cubed” microscope (300 kV). A Gatan double-tilt vacuum transfer holder was used to prevent the oxidation of the sample. 3D-ED tomography was used for the structure solution. Reciprocal space was scanned over an angular range of 85°, resulting in 696 reflections that were used for refinement with 590 reflections being independent. The PETS software () was used for the 3D reconstruction of reciprocal space.

Powder XRD data were collected using a D8 ADVANCE diffractometer (Bruker AXS) with Cu-Kα radiation. High-resolution powder SXRD experiments were performed at RT using a Debye-Scherrer camera, installed at SPring-8 BL02B2 of the Japan Synchrotron Radiation Research Institute. The incident beams from the bending magnet were monochromatized to λ = 0.42096(1) Å. Sieved powder samples (<32 μm) were loaded into a Pyrex capillary with an inner diameter of 0.2 mm. The sealed capillary was rotated during the measurement to reduce the effect of preferred orientation of the crystallites. Powder neutron diffraction (ND) data were collected at RT using the high-resolution powder diffractometer BT-1 (λ = 1.54060 Å) at the National Institute of Standards and Technology (NIST) Center for Neutron Research. The collected SXRD patterns were analyzed by the Rietveld method using RIETAN-FP program ().

1H MAS-NMR spectra were recorded on a home-built spectrometer (), operating at 400.23 MHz with a 3.2-mm T3 probe (Varian) in a magnetic field of 9.4 T. Tetramethylsilane (TMS) was used for the external chemical shift reference. The 1H spectra were acquired using a single-pulse sequence with background suppression and a RFDR pulse sequence. The typical acquisition delay was 3 s; the 90° and 180° pulse widths were 2.7 and 5.4 μs, respectively; and the MAS rate was 20 kHz. To study proton exchange behavior, the as-prepared samples were immersed in light or heavy water at 80°C overnight, followed by drying at RT.

N2 (77 K) adsorption and water vapor sorption (288 to 303 K) isotherms were measured using a volumetric gas sorption apparatus BELSORP-mini II and BELSORP-max (MicrotracBEL Corp), respectively. Before the sorption measurements, the weighted sample (ca. 0.1 g) was ground and treated in vacuum at 373 K for 3 hours. Sorption equilibrium was determined on the basis of a pressure change of ±0.3% over 5 min, and the sorption enthalpies (ΔHabs) were calculated using the Clausius-Clapeyron equation.

Thermogravimetry was measured under dry N2 flow (100 ml min−1) using a Thermo Plus 2 (Rigaku) thermogravimetric analyzer, with α-Al2O3 as a reference. The temperature was held at 30°C for 30 min, increased to 500°C with a heating rate of 10°C min−1, and held at 500°C for 30 min.

Certain commercial equipment, instruments, or materials are identified in this document. Such identification does not imply recommendation or endorsement by the NIST, nor does it imply that the products identified are necessarily the best available for the purpose.

Catalytic reaction

Pinacol rearrangement and Barbier-Grignard reactions were carried out in a glass reactor equipped with a magnetic stirrer. In a typical run of pinacol rearrangement reaction, a mixture of pinacol (0.668 mmol), naphthalene (0.267 mmol, internal standard), and HSbOI (300 mg, 0.01 mmol) in toluene (2 ml) was stirred under air at 373 K. In a typical run of Barbier-Grignard reaction, a mixture of benzaldehyde (0.50 mmol), allyltributyltin (1.0 mmol), and HSbOI (15 or 30 mg, 0.5 or 1.0 μmol) in water (1 ml) was stirred under air at 353 K. After the reaction, 1,2-dimethoxybenzene was added to the reaction mixture as an internal standard. The reaction progress was followed by GC using a GC-2014 (Shimadzu) fitted with an InertCap 5 (pinacol rearrangement reaction) or TC-WAX (Barbier-Grignard reaction) capillary column (GL Sciences) and a flame ionization detector, and 1H NMR measurements. TON and TOF was calculated from following equation; TON = [yield (%)]/[amount of catalyst (mol)/amount of substrate (mol)], TOF (hour−1) = TON/reaction time (hours). For ICP-MS measurement (Agilent Technologies, Agilent 8800), 30.8 mg of HSbOI was dispersed in 1 ml of water, and the Barbier-Grignard reaction was carried out for 24 hours. The concentration of Sb in the reaction solution was 18 μg ml−1 or 0.4% of the catalyst added.

Computational methods

DFT computations for the large HSbO cluster were performed with the Gaussian 09 package. Calculations were carried out using the extended cluster of [Sb36O50]8+ (and protonated [HSb36O50]9+), which is composed of [Sb34O44] from the HSbO cluster plus O─Sb (× 2) and O─VSb (× 4) from the Sb6−δI8O6 cluster (where VSb represents an antimony vacancy).

For the outer shell of the HSbO cluster, a small deficiency (occupancy, g ~ 0.95) in Sb3 (Wyckoff site 96j) was ignored, while the three-split Sb2 sites (96k, g ~ 0.31) were placed at the ideal site (32f, g = 1), yielding [Sb32O36]. For the inner shell, reflecting the experimental defect (g ~ 0.319), we removed four of the six antimonies for Sb1 in [Sb6O8], resulting in [Sb2O8]. Thus, outer [Sb32O36] and inner [Sb2O8] clusters are combined to give a [Sb34O44] composition. Furthermore, the capping O4─Sb4 from Sb6−δI8O6 was added on the square face of the HSbO truncated octahedron to account for the sizable interaction between the HSbO and Sb6−δI8O6 (see section Structural features of clusters in HSbOI). In addition, the “capping” oxygen atoms of the Sb6−δI8O6 cluster and the Sb atoms (Sb4) were combined into a [Sb34O44] cluster, taking into account the large defect (g ~ 0.351) in the Sb4 site ([O─Sb] × 2, [O─VSb] × 4). The two [O─Sb] and the two Sb1 atoms are arranged vertically in Fig. 4B.

As shown in Fig. 4B, O atoms of the Sb36O50 cluster can be divided into three layers according to their distance from the center of the cluster. The O atoms of HSbO cluster can be classified into five groups and the capping O atoms into two groups (fig. S19). One proton is placed near any of the seven O atoms to form a [HSb36O50]9+ cluster. The most stable orientation of the protons bound to each O atom was determined. 3D XYZ coordinates of each proton in the [HSb36O50]9+ cluster were fully optimized using the B3LYP functional, while keeping the XYZ coordinates of the framework Sb36O50 cluster fixed. The basis sets are LanL2DZ for Sb atoms and cc-pVDZ for a proton and O atoms. The structure of TMS molecule was fully optimized without any geometric constraint in the B3LYP/cc-pVDZ level of theory.

Total energies and NMR properties were calculated using the B3LYP functional and triple-ζ basis set. The basis sets are LanL08 for Sb atoms and cc-pVTZ for proton and O atoms in the [HSb36O50]9+ cluster, and cc-pVTZ for all the atoms included in the TMS molecule. The nuclear magnetic shielding tensors of the optimized [HSb36O50]9+ clusters and the reference TMS molecule were computed by the GIAO method () implemented in the Gaussian09 package. Each isotropic chemical shift of proton in the [HSb36O50]9+ cluster was calculated by setting the average of the computed NMR isotropic shielding constants of the H atoms in the TMS molecule as origin.

The 3D ESPs were made from the converged Kohn-Sham orbitals by the utility cubegen program of Gaussian 09 package. The bonding and lone-pair orbitals of [Sb34O44]14+ clusters were also calculated by the aforementioned B3LYP functional and triple-ζ basis sets. This is the core part of the [Sb36O50]8+ cluster in Fig. 4. The rendered wave function was analyzed by Gaussian Natural Bond Orbital version 3.1 () implemented in the Gaussian09 package. The obtained bonding and lone-pair orbitals were visualized by GaussView 5.0 package with the isovalue = 0.02, except for the isovalue = 0.1 used to show the lone-pair orbitals of the outer shell Sb atoms (Fig. 3 and fig. S20).