Synthesis of a nanoscale Cu( II ) 31 -oxo-carboxylate cluster, and eﬀect of Cu–oxo cluster size on visible-light absorption †

Cu 16 –Cu 31 Cu( II )-oxo-carboxylate clusters are reported, including those with condensed 1.5 nm Cu–O cores supported exclusively by O-donor ligands. A size–colour correlation is observed due to a red-shift of the charge transfer absorption band on increasing size; hence these clusters sit between small molecules and (black) CuO nanostructures.

By switching to bulkier secondary alcohols (R = i Pr or Cyp (C 5 H 11 )) grass-green crystals of larger clusters form after two weeks (in powdered form, or when dissolved, these compounds appear yellow/green). Isostructural forms were identified by X-ray crystallography, supported by bond-valence sum calculations, as [Cu 31 O 12 (OH) 18 (O 2 C t Bu) 18 (NO 3 ) 2 (ROH) 6 Án(ROH)] (R = i Pr, n = 6, 4; R = Cyp, n = 12, 5) ( Fig. 2 and Fig. S6-S8, ESI †). This structure has a pseudo-spherical Cu-O core (with 12 m 4 -O 2 , 12 m 3 -OH À and 6 m 2 -OH À ) with a maximum (Cu-O) core diameter of 1.5 nm and can thus be described as a 'molecular nanoparticle'. However, the Cu-O core is not a clear representation of the repeating structure in CuO. In CuO the copper centres coordinate to four oxygens in a close to square planar geometry (Cu-O bond lengths = 1.91-2.02 Å), 29 with longer distances (2.78 Å) to axially positioned oxygens. Similar coordination modes are found in 4 and 5 with comparable Cu-O bond lengths to the four nearest oxygens (range 1.88-2.28 Å) and most sites having a fifth oxygen occupying an apical site (range 2.29-2.89 Å, Supporting note 1, Fig S9, S10 and Table S1, ESI †). In contrast, the central Cu atom in 4 or 5 has a tilted octahedral geometry (Cu-O bond lengths: 2.05 to 2.22 Å). The cluster lies on a pseudo-C 3 symmetry axis passing through the centrosymmetric octahedral Cu and the N atoms of the two nitrates ( Fig. 2) with overall pseudo-D 3d symmetry. The d c of 4 and 5 is 0.68, showing a much more condensed core and thus greater resemblance to CuO@carboxylate nanoparticles than the smaller clusters 1-3. The surface of 4 and 5, which can be considered a model for CuO@carboxylate nanoparticles, 30 is capped by 18 anionic (m 2 or m 3 ) pivalate ligands, 18 hydroxides, six neutral ROH ligands and two loosely coordinated m 2 -nitrate anions (Cu-ONO 2 bond lengths 2.6-2.8 Å), which have disordered coordination geometries in the crystal structure and are each surrounded by the three coordinated and a further three non-coordinated ROH ligands (which likely engage in Hbonding interactions to the cluster surface, Fig. 2, Fig. S6 and S8, ESI †). Crucially these clusters have no remaining alkoxide groups (confirmed by bond valence sum calculations, Table S3, ESI †) which may provide enhanced stability to further hydrolysis and condensation reactions. The greater steric bulk of secondary alcohols appears to support condensation and enlargement of the core compared to less bulky primary alcohols. The synthesis of 4 and 5 is very condition-dependent, with brown precipitates containing CuO@pivalate nanoparticles (B8-15 nm by Scherrer analysis) and Cu 2 (OH) 3 (NO 3 ) (Fig S11 and S12, ESI †) alongside solutions containing the paddlewheel [Cu 2 (O 2 C t Bu) 4 (NEt 3 ) 2 ] able to form as by-products if the reaction is conducted at higher concentrations; left open to ambient atmosphere; or allowed to sit for longer times. The by-products likely form due to uncontrolled hydrolysis. It appears essential to maintain good solubility of intermediates Scheme 1 Proposed reaction scheme for the growth of Cu-oxo clusters via hydrolysis and condensation of a 'pre-hydrolysis intermediate'.  to allow 4/5 to self-assemble without the formation of insoluble by-products.
Based on our observations and hypothesised reaction mechanism (Scheme 1), moisture plays a crucial role in cluster synthesis. However, the copper source Cu(NO 3 ) 2 Á3H 2 O contains three equivalents of water per copper atom, thus always provides an excess of water for hydrolysis and condensation reactions. Therefore, selective cluster formation is likely defined by crystallisation of the growing clusters from the reaction medium (for example when 3 is dissolved in dichloromethane or isolated as a powder it is susceptible to further hydrolysis when exposed to moisture). To test these concepts, the synthesis of clusters using anhydrous precursors was explored. Cu(OEt) 2 was reacted with 1 equiv. of pivalic acid in dry toluene at 70 1C to produce previously reported loop-shaped [Cu 6 (OEt) 6 (O 2 C t Bu) 6 ], 6, 18,31 (Fig. S13, ESI †). This cluster is a 'pre-hydrolysis' structure as it contains no oxo or hydroxide groups (d c = 0), and could potentially form at the beginning of the reaction between Cu(NO 3 ) 2 Á3H 2 O, pivalic acid and NEt 3 in ethanol, i.e. before any hydrolysis (Scheme 1). 32 The reaction of 6 with 2.7 equivalents of water in EtOH induces a slow colour change from blue-green to green, and a precipitate of 3 forms after two weeks (confirmed by diffraction after recrystallisation), demonstrating that water promotes cluster growth. 32 Considering that water plays a crucial role in cluster formation, the reaction conditions to produce 4 were repeated, but, after indicating that this structure has formed under partial hydrolysis conditions, i.e. before 4 forms ( Fig. 3 and Fig. S14-S16, ESI †). This also indicates that hydrolysis and cluster growth is a relatively slow process as it can be arrested by removing moisture from the reaction solution after 30 minutes of reaction. The structure of 7 maintains some Cu-O i Pr functionality, suggesting that it may be susceptible to further reaction with water. 7 is a C-shaped cluster with an intriguing pocket which could act as a sterically encumbered reactive site. 7 contains 17 5-coordinate Cu centres which are supported by eight m 2 -briging pivalate anions and two unusually coordinated pivalates that each coordinate to seven Cu centres (Fig. S17, ESI †). Two chelating nitrates (Cu-O = 2.02-2.65 Å) and four bridging isopropoxide groups coordinate around the edges of the cluster, whilst a combination of two isopropoxides and two hydroxides occupy the internal cleft of the C-shape in a disordered manner (Fig. S18, ESI †). Two further nitrate anions are connected via H-bonding to four Cu-OH sites which sit in a square on the top of the cluster.
The relatively asymmetric shapes and range of different Cu carboxylate coordination geometries in these clusters highlights the flexibility of the Cu(II) centre and the unpredictability of the structures that it can form. This implies that many further possible cluster structures can be obtained by variation of the supporting ligands and degree of condensation.
The colours of these Cu-oxo clusters suggests a relationship with size ( Fig. 4 and Fig. S19, S20, ESI †), as the compounds change from blue through green to yellow/green with increasing size, eventually reaching dark brown/black CuO nanoparticles ( All display Beer-Lambert behaviour. The well-known paddlewheel structure (with two carboxylates per Cu) is blue in solution (Fig. 4) and has a high energy carboxylate (O) to metal (d x 2 -y 2) charge transfer (LMCT) band in the UV region 33,34 (8: l max = 262 nm; e B 10 300 M À1 cm À1 ) with a shoulder (''dimer band'') arising from the bridging carboxylate coordination mode at slightly lower energy, and weaker d-d transitions (to d x 2 -y 2) in the visible region (l max = 615 nm; e B 300 M À1 cm À1 ) which cause the blue colour. Compound 6 is blue/green in solution and shows a similar LMCT peak l max = 261 nm, and a shoulder which now extends into the visible region, possibly from added contributions of alkoxide to Cu LMCT, 8 and a weaker d-d absorption per Cu (Fig. 4). Upon growth of the cluster to (green) Cu 16 (2) and (yellow/green) Cu 31 (4), the  shoulder region becomes increasingly significant, extending into the visible region (and hence absorbing blue photons), and beginning to resemble the spectrum of B4.4 nm brown CuO@ carboxylate particles (Fig. S31, ESI †). 30 The carboxylate based LMCT peak remains in 2 and 4 with similar maxima (253 & 256 nm respectively) in the UV region, but with diminishing maximum extinction coefficient, when defined per copper, as the clusters grow (and the Cu:ligand ratio decreases, Fig. S32, ESI †). In 4 the shoulder region merges with the d-d excitations causing absorption across the visible region. The exact nature of excitations in the 350-550 nm region remain unknown and will be the focus of future study, but it is interesting to speculate that 4 contains the same surface ligands (pivalate and i PrOH) as 8 (+ two weakly bound nitrates in 4), yet absorbs broadly across the visible region (N.B. Cu(NO 3 ) 2 dissolved in i PrOH only absorbs UV light, Fig. S30, ESI †), hence the extra absorption may be due to charge transfer excitations within the cluster core (e.g. with O to Cu character) rather than surface effects. Such excitations would be akin to the band-gap transitions that are responsible for the black colour in bulk CuO. Therefore, it is intriguing to consider these well-defined B1.5 nm clusters, with condensed Cu + O cores, as a bridge between small Cu-carboxylate molecules and larger CuO@ carboxylate nanostructures.
The size-colour correlation is consistent with the concept of quantum confinement, or considered from a molecular perspective, that as an increasing number of atomic orbitals closely interact the resulting HOMO-LUMO energy gap of the system becomes smaller and can absorb lower energies of light. This excitingly demonstrates the tuneability of the charge transfer absorption onset of Cu-oxo clusters across the visible region and the possibilities for manipulating frontier orbital energies and redox capability for useful photoredox reactivity.
In summary, new Cu-oxo clusters are reported, including Cu 31 structures 4 and 5, which, to the best of our knowledge, are the largest and most condensed molecular Cu-oxo clusters discovered with only O-donor ligands. The high degree of condensation (0.68) and their size (1.5 nm Cu-O core) relates these well-defined clusters to CuO nanoparticles. Synthesis of these clusters operates via hydrolysis and condensation reactions, therefore, the amount of water influences products, and uncontrolled hydrolysis leads to CuO. By creating conditions which allow for a greater amount of hydrolysis to occur, i.e. maintaining cluster solubility with controlled water content, larger Cu-oxo clusters can be synthesised, and this will be explored in future studies. The colour of these clusters is influenced by their size, with the charge transfer absorption band becoming red-shifted on increasing size, consistent with concepts of quantum confinement and suggesting that Cu-oxo clusters and ultrasmall CuO nanoparticles have potential as highly tuneable systems for visible light driven photoreactivity.
The authors thank the Royal Society (URF\R1\191458), the Herchel Smith Fund and University of Warwick CDT in Analytical Science for funding.

Conflicts of interest
There are no conflicts to declare.