M 4Au12Ag32(p-MBA)30 (M = Na, Cs) bimetallic monolayer-protected clusters: synthesis and structure

The synthesis and structure of mixed gold/silver M 4Au12Ag32(p-MBA)30 bimetallic monolayer-protected clusters is reported and compared to that of silver M 4Ag44(p-MBA)30 monolayer-protected clusters (M = Na, Cs).


Chemical context
The M 4 Ag 44 (p-MBA) 30 monolayer-protected cluster (MPC) has been studied in detail previously, where M + is an alkali metal counter-ion (M = Na, Cs) and p-MBA is p-mercaptobenzoic acid (Desireddy et al., 2013;Conn et al., 2015), along with other related 44 silver-atom species (Bakr et al., 2009;Pelton et al., 2012;AbdulHalim et al., 2013;Yang et al., 2013;Chakraborty et al., 2013). The formula has been shown to be Na 4 Ag 44 (p-MBA) 30 and K 4 Ag 44 (p-MBA) 30 in all-sodium and all-potassium preparations, respectively, and the molecular and crystal structures have been determined crystallographically (Desireddy et al., 2013). The crystal was determined to have a framework structure, with 52% solvent-filled void space, that is a consequence of both ligand bundling and interparticle hydrogen bonding (Yoon et al., 2014). The positions of the alkali metal counter-ions were not determined, and are presumably located in the solvent portion of the crystal (Desireddy et al., 2013).
Structurally related species have also been prepared with non-polar ligands, using non-polar synthetic conditions, forming chemically distinct members of the 44 silver-atom family of species, e.g. (PPh 4 ) 4 Ag 44 (SPhF 2 ) 30 along with SPhF and SPhCF 3 variants (Bakr et al., 2009;Yang et al., 2013). In the crystals of these species, the ligand bundling is not dominant and ligand interactions do not lead to framework struc- ISSN 2056-9890 tures. Instead, ligands pack more tightly, with 36% solventfilled void space, and the bulky PPh 4 + counter-cations lock into place in the crystals such that they can be located (Yang et al., 2013).
Silver and gold mix readily to form the naturally occurring alloy electrum and therefore the study of mixtures of silver and gold within these MPCs is of interest (Yang et al., 2013). Mixtures of M 4 Au x Ag 44-x (p-MBA) 30 MPCs can be obtained by co-reducing silver and gold polymers of p-MBA, where 0 x 12 (Conn et al., 2018). Gold-rich species have been synthesized and then thermally processed to destroy the species that contained fewer gold atoms, thereby enriching the samples in M 4 Au 12 Ag 32 (p-MBA) 30 MPCs.
Once high-purity samples of M 4 Au 12 Ag 32 (p-MBA) 30 MPCs had been prepared and crystallized, the locations of the gold atoms could be determined by crystallographic methods. Prior reports using non-polar members of the 44 metal-atom family of species determined that the 12 gold atoms are located in the icosahedral inner core of that molecule (Yang et al., 2013). It is not clear, however, whether different synthetic conditions, ligands, and solvent class would affect the synthetic mechanism, electronic structure, and ultimately the organization of metal atoms within the core. We present here the chemical synthetic method of producing M 4 Au 12 Ag 32 (p-MBA) 30 MPCs as well as their X-ray determined structure and verify that the gold atoms are indeed located in the core of this MPC. Comparisons with other family members are also made to examine the effects of heteroligands and heteroatoms on the structures of these species.

Structural commentary
There are four sets of chemically equivalent positions for metal atoms in the Au 12 Ag 32 (p-MBA) 30 4À molecular structure. All 12 positions in the icosahedral inner core are chemically equivalent, whereas the dodecahedral outer core contains a set of eight chemically equivalent positions (defining a cube) and a set of 12 chemically equivalent positions (a pair of atoms beneath each of the six mounts). The remaining 12 metal atoms are found in pairs in the six mounts and are chemically equivalent. In principle, then, there are three possible ways to locate 12 equivalent gold heteroatoms.
Density-functional calculations (Kresse & Joubert, 1999;Perdew, 1991;Perdew et al., 1992Perdew et al., , 1993 were performed to evaluate the energy differences upon substitution of gold atoms into each of these four distinct metal-atom positions. Each calculation was done for a M 4 AuAg 43 (p-MBA) 30 MPC, the structures of which were relaxed after substitution. In each case, the energy of M 4 AuAg 43 (p-MBA) 30 was found to be lower than M 4 Ag 44 (p-MBA) 30 . It was found that substitution of gold atoms into the icosahedral core has the biggest effect, lowering the energy by 0.71 eV per Au atom. The next most energetically favorable position was that of the eight atoms in the dodecahedral shell, lowering the energy by 0.30 eV per Au atom; these positions are of particular interest since they are the only atoms in the metal core that are exposed and capable of directly interacting and reacting with other species in solution. The least favorable positions for substitution were found to be the pairs of metal atoms in the mounts, lowering the energy by 0.170 eV per Au atom, and the pairs of metal atoms beneath the mounts, lowering the energy by 0.13 eV per Au atom. Based on these calculations, the 12 substituted gold atoms were expected to be found in the icosahedral core.
The positions of the 12 Au atoms were determined by single-crystal X-ray crystallographic methods. The full refinement of the Au 12 Ag 32 (p-MBA) 30 4À molecular structure revealed that the 12 gold atoms reside in the icosahedral inner core of the MPC. The structure consists of a 12 gold-atom icosahedron surrounded by a 20 silver-atom dodecahedron, forming a 32-atom excavated-dodecahedral bimetallic core. The metal core is capped by six equivalent Ag 2 (p-MBA) 5 mount motifs, which are octahedrally located about the core (Fig. 1). The Au 12 Ag 32 (p-MBA) 30 4À anion is located about an inversion center and exhibits point group symmetry 3 (Fig. 2).
The crystallographically determined locations of the 12 gold atoms in the icosahedral inner core of the bimetallic MPC are consistent with the expected locations based on our DFT Structure of Au 12 Ag 32 (p-MBA) 30 4À . Complete X-ray-determined structure shown in (a) space-filling view and (b) ball-and-stick view (out-of-plane ligands removed for clarity). The core structure is shown as (c) an Au 12 icosahedral inner shell, which is nested inside of (d) an Ag 20 dodecahedral outer shell, together making (e) a bimetallic 32-atom excavated dodecahedral core. Other colors: red -O; grey -C; yellow -S (H not shown). The overall diameter of the MPC was measured to be about 28 Å , while the diameter of the inorganic portion of the structure was calculations and based on previous reports (Yang et al., 2013). In addition, this result is in agreement with the known properties of gold and silver. Although gold and silver are isoelectronic and have almost identical atomic radii, their chemical properties and bonding can be quite different. For example, Au-S and Ag-S bonding is typically two-and three-coordinate, respectively (Dance, 1986;Dance et al., 1991), which makes the bonding of the gold heteroatoms incompatible with the structure of the protecting mounts (Desireddy et al., 2013;Conn et al., 2016). It is therefore unlikely that gold atoms would substitute into the ligand shell without changing the metal-atom count (Yang et al., 2014).
Furthermore, gold is known to be more electronegative and more noble than silver, so the gold atoms are expected to assume positions within the structure where they can possess the lowest oxidation state among the metal atoms. Bader analysis (Bader, 1990;Tang et al., 2009) of the electron distribution in M 4 Ag 44 (p-MBA) 30 has shown that atoms in the inner icosahedral core have an oxidation state of zero (Conn et al., 2015;Yang et al., 2013) whereas in M 4 Au 12 Ag 32 (p-MBA) 30 those atoms are slightly reduced (Conn et al., 2018). The other metal atoms were found to be oxidized, with their oxidation states increasing with distance from the center of the molecule. The X-ray-determined locations of the gold atoms in the inner core are therefore also consistent with the Bader analysis and the known properties of gold and silver.
The analyzed carefully to identify changes in the structure as a result of substituting 12 silver atoms for gold atoms. The metal-metal bond lengths within the 12-atom icosahedron, the 20-atom dodecahedron, and the mounts were compared for the two structures. The results of the bond-length analysis are reported in Table 1. The Au-Au and Ag-Ag bonds in the bulk metals have similar bond lengths (2.884 and 2.889 Å , respectively; JCPDS no. 04-0784 and no. 04-0783, respectively; ICDD, 2015), and therefore substituting the two metals might not be expected to change bond lengths within the structures. This is not the case, however. The bond lengths within the 12-atom icosahedron were found to shorten from 2.825 AE 0.012 Å to 2.795 AE 0.013 Å when gold was incorporated, indicating stronger than expected bonding within the inner core. Bond lengths within the 20-atom dodecahedron were found to be essentially unchanged (3.175 AE 0.040 Å versus 3.190 AE 0.040 Å ), however. The metal-metal bonds in the mounts were also found to be unaffected by the gold-atom substitution.
These changes in bond lengths may be the result of a change in the electron-density distribution due to the electrophilicity of the gold atoms in the inner icosahedral core, which tend to pull electron density from the outer dodecahedral core. For example, Bader analysis of the charge distribution shows that the number of excess electrons on the icosahedral core increases from 0.010 to 1.769 upon substitution of gold atoms. This reduction of the inner core is accompanied by a further

Figure 2
Structure of Au 12 Ag 32 (p-MBA) 30 4À using displacement ellipsoids that were drawn at the 50% probability level for three different views of the structure. Au atoms are depicted in orange, Ag atoms in grey, and S atoms in yellow. Views are (a) down a fourfold axis of the pseudooctahedral structure, (b) with one 31.7 rotation from (a) about the horizontal axis, and (c) with two 45 rotations from (a) about the horizontal and vertical axes. The organic portion of the molecule was omitted for clarity. oxidation of the silver atoms in the outer core, where the number of excess electrons decreases from À4.928 to À6.546 upon substitution of gold atoms. The gold-atom substitution into the core does not affect the charge density on the silver atoms in the mounts. The results of the Bader charge analysis are reported in Table 2.
Based on the Bader analysis, the redistribution of the electron density was found to be almost entirely confined to the 32-atom metal core (comprising the icosahedral and dodecahedral shells). While this appears to be the origin of the changes in metal-metal bond lengths inside the 32-atom metal core, it may also be the reason that the rest of the molecule remains essentially unchanged by this metal-atom modification to the structure.
It is also interesting to note that classical electrostatics predicts that any charges carried by a metal sphere would be located on the surface of that sphere. The Bader charge analysis for M 4 Ag 44 (p-MBA) 30 is in agreement with this classical picture, but that is not the case for M 4 Au 12 Ag 32 (p-MBA) 30 . In the former case, the inner core is neutral and all of the charge is located on the outer core. In the latter case, both the inner and outer core carry charge (in fact, the 32-atom metal core is polarized). This demonstrates the failure of the classical theory with regard to predicting charge distributions on such a small scale, because of finite screening lengths in real materials.

Supramolecular features
Like the silver-only M 4 Ag 44 (p-MBA) 30 MPCs, M 4 Au 12 Ag 32 (p-MBA) 30 MPCs crystallize as framework structures as a consequence of intramolecular ligand bundling and intermolecular hydrogen bonding. The ligand bundling is a consequence of interactions between the ligands, with the magnitude of the inter-ligand van der Waals interaction energy calculated to be À0.95 eV/mount. The ligands form six dimer bundles, which are evenly spaced in the same plane, and six trimer bundles, with three above and three below the plane defined by the dimers. Together, the twelve bundles define the connectivity of the crystal's framework structure such that the MPCs have pseudo-face-centered-cubic packing. The nature of the framework structure and hydrogen bonding in these materials was studied in detail in a previous report (Yoon et al., 2014).

Database survey
It is instructive to compare the structures of the related but chemically distinct Au 12 Ag 32 (p-MBA) 30 4À and Au 12 Ag 32 (SPhF 2 ) 30 4À species to examine the effect of ligand structure on crystal structure as well as the question of whether the composition of the outside of the MPC can affect the structure of the core. Likewise, the Ag 44 (p-MBA) 30 4À and Au 12 Ag 32 (p-MBA) 30 4À structures can be compared to address the question of whether the composition of the core can affect the ligand shell and crystal structure.
First, the crystal structures of the two species are entirely different, due to the different mechanisms of interactions between the MPCs. In the case of p-MBA, hydrogen bonding governs the interactions between the MPCs while ligand bundling within the ligand shell defines the directionality of those interactions (Yoon et al., 2014). As a result, the overall structure of the crystal is that of a framework material with large void spaces (Yoon et al., 2014). No such interactions exist in the crystals of hydrophobic MPCs, and therefore the crystal structure is more compact with less well-defined intermolecular interactions (Yang et al., 2013). The difference in crystal structures due to the different ligands is also expected to lead to entirely different mechanical properties of these two crystalline materials (Yoon et al., 2014). The observed differences in crystal structures are similar when comparing Ag 44 and Au 12 Ag 32 cores, however, indicating that the added gold did not affect the ligand shell and crystal structure. This also indicates that the chemical stability can be improved with the addition of gold without changing the overall structure and mechanical properties of the MPC crystal.
The differences in the nature of the ligands were not found to have affected the overall arrangement of gold atoms in the MPC cores, with the gold atoms occupying the same positions in both structures. The different ligands induce slightly different bonding within the metal core, however. Bond lengths in the icosahedral core in Ag 44 and Au 12 Ag 32 are similar for both p-MBA and SPhF 2 ligands, contracting 0.03 and 0.05 Å , respectively, with the addition of gold atoms. This indicates that changes in the icosahedral core are not influenced by the ligands. Changes in bond lengths are different in the dodecahedral core, however. In the case of p-MBA, bond lengths in the dodecahedron do not change with the addition of gold atoms, but in the case of the SPhF 2 ligand they contract slightly. This indicates that changes in the dodecahedral core are influenced by the SPhF 2 ligands, presumably due to their greater electron-withdrawing ability. The net effect is that the radius of the icosahedron contracts slightly in the case of both p-MBA and SPhF 2 (0.03 and 0.05 Å , respectively), but the radius of the dodecahedron does not change for p-MBA while it contracts 0.03 Å in the case of SPhF 2 .  using an Au:Ag input ratio of 14:30. For this input ratio, 72.4 mg of AuCl 3 (0.24 mmol) and 86.8 mg of AgNO 3 (0.51 mmol) were used for the metal sources. These materials were added to 33 ml of 7:4 water-DMSO solvent along with 200 mg of p-MBA (1.3 mmol). This mixture was sonicated and stirred to fully dissolve the p-MBA. The dissolved p-MBA reacted with the metals to form a precursor mixture of metal thiolates, which was a cloudy light-yellow precipitate that was dispersed in the solvent. The pH was then adjusted to 12 using 50% w/v aqueous CsOH. The metal thiolates dissolved as the pH was raised above 9, forming a clear, light-yellow solution. Next, 5.0 mmol of NaBH 4 reducing agent dissolved in 9 ml of water was added dropwise over a period of 30 min, and was then left to stir for 1 h. This formed a dark-yellow/brown solution. Once the reaction was completed, the product solution was centrifuged for 5 min (to remove insoluble byproducts), decanted, and then the supernatant was precipitated using DMF. The precipitate was collected by centrifugation. It is important not to dry this raw product. The raw product was precipitated from a basic solution; therefore it was the conjugate base (alkali metal salt) of the fully protonated species. To protonate, pure DMF was added to the precipitated particles, which did not initially solubilize. Glacial acetic acid was then added dropwise to the solution until the precipitate dissolved into the DMF, forming a golden brown solution. Protonation was repeated three times, using toluene to precipitate from DMF.

Synthesis, crystallization, and theoretical methodology
Fully protonated M 4 Au x Ag 44-x (p-MBA) 30 MPCs were next subjected to thermal processing. Capped glass vials containing DMF solutions of the MPCs were placed in a water bath at 333 K for 30 h. After incubation, insoluble material produced by thermal processing was separated from the solution by centrifugation. The supernatant was collected and was protonated with glacial acetic acid in DMF and precipitated with toluene. The protonation steps were repeated two times to ensure complete protonation of the carboxylates. The fully protonated product enriched in M 4 Au 12 Ag 32 (p-MBA) 30 was able to be dissolved in a neat solution of DMF.
With the above synthetic conditions, the counter-cations tend to be a mixture of alkali metals, namely Cs + and Na + . The counter-ion mixture has been identified by energy dispersive X-ray spectroscopy (EDS) to be approximately a 3:1 ratio of Cs:Na, despite the expectation that there might be a higher affinity for Na because of its size. The molecular formula for this MPC could therefore be written as NaCs 3 Au 12 Ag 32 (p-MBA) 30 . It should be noted, however, that the counter-ions are readily dissociated and easily exchanged such that the identities of the alkali metals play little role in the properties of the material. Nonetheless, Na 4 Au 12 Ag 32 (p-MBA) 30 and K 4 Au 12 Ag 32 (p-MBA) 30 can be directly prepared by using allsodium and all-potassium reaction conditions, respectively, if desired (Desireddy et al., 2013). Crystallization The M 4 Au 12 Ag 32 (p-MBA) 30 crystals were grown from a neat DMF solution of MPCs, dried under N 2 gas. Small rhombohedral crystals (10 mm) were obtained from this crystallization process. These crystals were used as seeds in a second crystallization step. The second solution was dried under N 2 and the seeds grew into larger rhombohedral crystals (>50 mm). The crystals were first separated and isolated on a microscope slide using paratone oil, and then were picked up and mounted with a MiTeGen MicroLoop.

Theoretical Methodology
The density functional theory (DFT) calculations and Bader charge analysis (Bader, 1990;Tang et al., 2009) were performed using the VASP-DFT package with a plane-wave basis with a kinetic energy cutoff of 400 eV, PAW pseudopotentials (Kresse & Joubert, 1999), and the PW91 generalized gradient approximation (GGA) for the exchangecorrelation potential (Perdew, 1991;Perdew et al., 1992Perdew et al., , 1993. For structure optimization, convergence was achieved for forces smaller than 0.001 eV Å À1 . The X-ray determined structure of Na 4 Ag 44 (p-MBA) 30 was taken as the starting configuration for structural relaxation. Hydrogen atoms were added to the structure and their positions were relaxed, yielding d(C-H) = 1.09 Å .
To estimate the inter-ligand van der Waals (vdW) interaction energy, the total energy of the relaxed Na 4 Au 12 Ag 32 (p-MBA) 30 MPC was evaluated with and without the inclusion of the vdW interactions, using density functional theory (DFT) (Grimme, 2006). The energy of the MPC, calculated with the inclusion of the vdW interactions between the atomic constituents of the ligand (S, C, O and H atoms) was found to be lower by Á tot (vdW) = 13.23 eV compared to that found without the inclusion of the vdW contributions. However, this vdW energy includes intramolecular and intermolecular interactions between the ligand molecules. The average intraligand (Ag-S-C 6 H 4 -COOH) vdW stabilization energy was calculated (for the relaxed configuration of the Agbonded ligand molecule) using DFT to be Á intra (vdW) = 0.251 eV. The total intermolecular vdW energy in the ligand shell (made of 30 p-MBA molecules) is therefore calculated as: Á inter (vdW) = Á tot (vdW) À 30 Á intra (vdW) = 5.70 eV. Since the ligand molecules are assembled into six Ag 2 (p-MBA) 5 mounts, we conclude that the inter-ligand non-bonded (dispersion, vdW) energy is 0.95 eV/mount.

Refinement
All of the Ag, Au and S atoms were located by direct methods. During the following refinements and subsequent difference-Fourier syntheses, the remaining C atoms and O atoms were located.
The Au, Ag and S atoms were ordered; however three out of the five crystallographically independent ligands in the asymmetric unit cell were disordered over two sets of sites. The three disordered ligands were modeled over the two positions, and their occupancies were refined with fixed atomic displacement parameters using a free variable to be 0.5. Final refinement released the fixed atomic displacement parameter and constrained the occupancies to be 0.5 for all disordered C and O atoms.
Au, Ag, and S atoms were refined with anisotropic displacement parameters, while all C and O atoms were refined with isotropic atomic displacement parameters. DFIX restraints were applied to the C-O bonds in the carboxylic acid groups, but C-atom positions in the phenyl rings were not restrained. All H atoms were geometrically determined on idealized positions (O-H = 0.84, C-H = 0.95 ), using AFIX 43 and AFIX 83 instructions, and were included as riding atoms in the final refinements [U iso (H) = 1.2U eq (C) or 1.5U eq (O)].
It is common for MPCs to have a high amount of residual electron density observed in the metal core. It is noted that the alkali metal cations and the solvent molecules were not identified in the X-ray data (highest residue density was 2.55 e Å À3 ). PLATON (Spek, 2009) was used to determine the total void volume in the unit cell to be about 52% with an estimate of 19000 electrons. Attempts to improve the refinement using the SQUEEZE (Spek, 2015) option in PLATON were not successful.

Special details
Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes. Refinement. The data reported herein were collected from a crystal of approximate dimensions 200 x 200 x 100 µm 3 , which was cooled to 100 K for data collection. X-ray diffraction data were collected on a Bruker Apex Duo diffractometer (CuKα = 1.54178 Å), which was equipped with an Apex II CCD detector and an Oxford Cryostream 700 low temperature device. The frames were integrated with the Bruker SAINT software package using a narrow-frame algorithm. The integration of the data using a trigonal unit cell yielded a total of 182413 reflections to a maximum θ angle of 62.42° (0.87 Å resolution), of which 12622 were independent (average redundancy 14.452, completeness = 100.0%, R int = 5.49%, R sig = 1.92%) and 11138 (88.24%) were greater than 2σ(F 2 ). The final cell constants of a = 25.7341 (3) Å, b = 25.7341 (3) Å, c = 124.079 (4) Å, volume = 71162 (3) Å 3 , are based upon the refinement of the XYZ-centroids of reflections above 20 σ(I).
The structure was solved and refined using the Bruker SHELXTL software package (Sheldrick, 2015), using the R3c space group, with Z = 6 for the formula unit C 210 H 150 Ag 32 Au 12 O 60 S 30 . The final full-matrix least-squares refinement on F 2 with 373 variables converged at R1 = 3.98% for the observed data and wR2 = 13.91% for all data. The goodness-of-fit was 1.094. The largest peak in the final difference electron density synthesis was 2.547 e -/Å 3 and the largest hole was -1.528 e/Å 3 with an RMS deviation of 0.253 e/Å 3 . On the basis of the final model, the calculated density was 1.458 g/cm 3 and F(000) was 28932 e -. The cations and solvent molecules are not included in the calculations of the density or F(000).