Overcoming the Crystallization Bottleneck: A Family of Gigantic Inorganic {Pdx}L (x=84, 72) Palladium Macrocycles Discovered using Solution Techniques

Abstract The {Pd84}Ac wheel, initially discovered serendipitously, is the only reported giant palladium macrocycle—a unique structure that spontaneously assembles from small building blocks. Analogues of this structure are elusive. A new modular route to {Pd84}Ac is described, allowing incorporation of other ligands, and a new screening approach to cluster discovery. Structural assignments were made of new species from solution experiments, overcoming the need for crystallographic analysis. As a result, two new palladium macrocycles were discovered: a structural analogue of the existing {Pd84}Ac wheel with glycolate ligands, {Pd84}Gly, and the next in a magic number series for this cluster family—a new {Pd72}Prop wheel decorated with propionate ligands. These findings confirm predictions of a magic number rule for the family of {Pdx} macrocycles. Furthermore, structures with variable fractions of functional ligands were obtained. Together these discoveries establish palladium clusters as a new class of tunable nanostructures. In facilitating the discovery of species that would not have been discovered by orthodox crystallization approaches, this work also demonstrates the value of solution‐based screening and characterization in cluster chemistry, as a means to decouple cluster formation, discovery, and isolation.

Abstract: The {Pd 84 } Ac wheel, initially discovered serendipitously,i st he only reported giant palladium macrocycleau nique structure that spontaneously assembles from small building blocks.Analogues of this structure are elusive.Anew modular route to {Pd 84 } Ac is described, allowing incorporation of other ligands,a nd an ew screening approach to cluster discovery.S tructural assignments were made of new species from solution experiments,overcoming the need for crystallographic analysis.A saresult, two new palladium macrocycles were discovered:as tructural analogue of the existing {Pd 84 } Ac wheel with glycolate ligands, {Pd 84 } Gly ,and the next in amagic number series for this cluster family-a new {Pd 72 } Prop wheel decorated with propionate ligands.T hese findings confirm predictions of am agic number rule for the family of {Pd x } macrocycles.Furthermore,structures with variable fractions of functional ligands were obtained. Together these discoveries establish palladium clusters as an ew class of tunable nanostructures.Infacilitating the discovery of species that would not have been discoveredb yo rthodoxc rystallization approaches, this work also demonstrates the value of solution-based screening and characterization in cluster chemistry,a s am eans to decouple cluster formation, discovery,a nd isolation.
One of the most striking phenomena in inorganic chemistry is the formation of nanoscale inorganic macrocycleswhereby large structures can self-assemble from simple building blocks-and their unusual properties (such as single molecule magnet (SMM) behavior, [1] with potential applications in quantum computing). [2] One large family of nanoscale metal oxide macrocycles have been known for some time:Mo-blue wheels, [3] which can form ahuge variety of different sized supramolecular architectures. [4] Christou et al. have reported another family based on manganese:first {Mn 84 }, and more recently {Mn 70 }, both of which behave as SMMs. [5,6] Our group recently reported anew nanoscale metal oxide macrocycle, {Pd 84 } Ac (Na 56 H 14 [Pd 84 O 42 (CH 3 CO 2 ) 28 -(PO 4 ) 42 ]), the largest known polyoxopalladate (POPd) to date, [7][8][9][10][11] and potentially the parent of at hird family of macrocycles.H owever,e fforts to produce analogues have been in vain, illustrating that synthesis of these structures is not straightforward and that the underlying principles are not well-understood. Only with understanding can we make the transition from discovery to sophisticated control over macrocycle design and synthesis,a nd finally,e xploitation of their properties.
Herein, we describe our rational efforts to build af amily of POPd wheels.T he discovery of {Pd 84 } Ac was serendipitous, using single crystal X-ray crystallography to identify crystals Scheme 1. Twosynthetic routes to {Pd 84 } Ac using different palladium sources:a )the original route, using palladium acetate;b )anew route, using palladium nitrate. formed after stirring Pd(OAc) 2 in phosphate buffer (Scheme 1a). [11,12] Most large inorganic nanostructures are discovered in as imilar way,r equiring reactions not just to proceed, but to also produce pure diffraction-quality crystals. Where no such crystals are produced, no structure is determined and no compound is discovered;t hat is,a"crystallization bottleneck". Herein, we demonstrate how new metal oxide macrocycles can be discovered without the need for diffraction-quality crystals.I nstead, we exploit solution sizing techniques [12] to screen reactions for the presence of new nanoscale species,e ven where no crystals are observed. Data suggesting new species ("leads"), discovered either by screening or crystallization, are then structurally characterized from crude solution. We could then focus further efforts on isolation of the compounds by crystallization. As ar esult of these new approaches,wereport the discovery of two new members of the POPd wheel family.W ealso demonstrate the formation of at unable range of wheels with mixed acetate/ glycolate ligand functionality.
As outlined above,t he original synthesis of {Pd 84 } Ac uses Pd(OAc) 2 as as tarting material (Scheme 1a). To synthesize wheels with other carboxylate ligands,t he first challenge we faced was to design anew modular synthetic route,separating the carboxylate source from the palladium source,a llowing ligands to be varied independently.T his was accomplished using palladium nitrate as apalladium(II) source,and sodium acetate as as eparate carboxylate source,w hich was readily substituted for other sodium carboxylate salts (Scheme 1b). This new route yielded {Pd 84 } Ac crystals in similar yields to the previous approach (see the Supporting Information for the method and characterization details). Having validated this alternative synthetic route,aseries of experiments were attempted substituting different carboxylate ligands (as sodium salts). Of these,n or eaction produced crystals of sufficient quality for structural determination by X-ray diffraction (XRD) and only one reaction-substituting propionate for acetate-produced any crystalline material at all.
Despite alack of crystalline products suitable for XRD,it was clear from observing color changes and performing gel electrophoresis [12] that palladate cluster products were being formed in the reaction solutions.Inthe course of our previous studies on the formation of {Pd 84 } Ac ,w ed eveloped as ize exclusion chromatography (SEC) method in which POPd clusters of different sizes are resolved;i ns od oing we were able to show that macrocycles are present in reactions even when no crystals are formed. [12] Knowing there to be products in the reaction mixtures,w es creened crude solutions of reactions incorporating aw ide variety of carboxylate ligands for the synthesis of large species.That is,instead of setting up many reactions and waiting for serendipitous crystallization, our new approach screened for the formation of large species as leads to be further characterized. To do this,wecompared the relative amounts of large species observed by comparing the relative integrals of as creening range (from 12-17.3 min), corresponding to where {Pd 84 } Ac or other large species were expected to elute (in SEC,l arger species elute earlier).
Thel igands to be screened were chosen to incorporate carboxylate moieties (including some di-and trivalent examples), along with ar ange of other functional groups (amine,a lcohol, aromatic,a nd fluorine). Other than substituting acetate for the new ligand, the reaction procedure was the same as the new modular {Pd 84 } Ac reaction (scaled down, to preserve the expensive palladium nitrate starting material). Since {Pd 84 } Ac was previously found to form in solution over six days (persisting for weeks), nine days were allowed for macrocycle formation before screening.T he results can be seen in Figure 1b,e xpressed as relative amounts of material observed to elute in the range corresponding to large clusters. Under these conditions,t he reaction using glycolate (ligand "O") stands out as yielding considerable amounts of large species in solution. Thec hromatogram for this reaction mixture consisted of alarge peak with avery similar retention time to that of {Pd 84 } Ac (Supporting Information, Figure S1). This is consistent with the formation of a {Pd 84 } macrocycle surrounded by glycolate ligands in place of acetate ligands (denoted as {Pd 84 } Gly ). Furthermore,wesee no distinct peaks eluting beyond 16.7 min, suggesting no significant amounts of smaller clusters are present (for example,{Pd 15 }or{Pd 10 }). We note that no crystals were formed from these screening reactions and that any "leads" would not have been identified using am ore orthodox approach.
Having obtained strong solution evidence of new large palladate clusters with propionate and glycolate ligands,b ut no diffraction-quality crystals,o ur attention turned to characterization of these structures.W eh ave been developing electrospray ionization ion mobility mass spectrometry (ESI-IMS-MS) as atool for cluster discovery,characterization, and structural assignment. [13][14][15] This technique is valuable as it gives information on size and shape-as collision crosssection (CCS He )-as well as composition, m/z.E SI-IMS-MS was carried out on an aqueous solution of pure {Pd 84 } Ac crystals (Figure 2a), as ample of the glycolate reaction mixture from the ligand screening (Figure 2b,r eaction "O"

Angewandte Chemie
Communications from ligand screening), and as olution of the crystalline material obtained from the substitution of propionate as aligand (Figure 2c), all desalted to avoid ionization suppression (Supporting Information, Section 2). As can be observed in Figure 2, as ingle series of broad peak envelopes is observed in each spectrum, showing that only one major structure is present in each case.T he spectrum yielded by ions (x, y,and z vary,hence broad peaks) and reveals aCCS He of 1070 .T he product of the glycolate substitution reaction mixture yields aspectrum (Figure 2b)that is strikingly similar to that of {Pd 84 } Ac in apparent mass,c harge distribution, and CCS He .T aken together,t his data allows us to assign it as {Pd 84 } Gly ,anew {Pd 84 } Ac structural analogue.I tm ight be expected that ligand substitution with propionate would also yield an analogous {Pd 84 } macrocycle.However,ESI-IMS-MS shows this not to be the case (Figure 2c); the main product observed is smaller in mass and size (CCS He of 929 , compared to 1070 for {Pd 84 } Ac )corresponding to asmaller, but nonetheless well-defined, {Pd 72 } Prop cluster.H aving converted these "leads" into "hits", and assigned structures using solution methods,w et hen aimed to produce diffractionquality crystals from these two reactions to validate our structural assignments.
Efforts to obtain diffraction-quality crystals of the new {Pd 72 } Prop met with success more readily.Upon XRD analysis, the crystallographic data reveals an ew macrocycle with 72 palladium centers,w ith ag eneral formula Na 60 [Pd 72 O 36 -(C 2 H 5 CO 2 ) 24 (PO 4 ) 36 ]· % 200 H 2 O, denoted {Pd 72 } Prop .T he wheel has six-fold symmetry,d iffering from the seven-fold symmetry of the original {Pd 84 } Ac wheel (Figure 3). We note that this observation is consistent with our previous prediction, and that on the basis of the discovery of {Pd 84 } Ac and the existence of {Mn 84 }, the next clusters in each series would be {Pd 72 }a nd {Mn 70 }, respectively. [6,11] Structurally,w ec an observe the same {Pd 6 }s ubunits (Supporting Information, Figure S20) as those in {Pd 84 } Ac ;t he smaller size and six-fold symmetry of this new wheel arise from incorporating twelve of these {Pd 6 Figure S19). This is analogous to the {Pd 10 } Ac cluster reported previously. [12] 31 PNMR analysis of as olution of {Pd 72 } Prop crystals further confirms its structural similarity to {Pd 84 } Ac ,s howing that the new macrocycle also comprises two inequivalent phosphorous environments. [11] Interestingly, 31 PNMR also revealed minor peaks,up-field of the main peaks,which grow over time (see the Supporting Information for the difference over 24 h); these are thought to correspond to ad egradation product of {Pd 72 } Prop ,d emonstrating limited stability in pure water. This degradation was also observable in SEC (over a20hperiod; Supporting Information, Section 7) and by ESI-IMS-MS (we attribute the faint second series of peaks at longer drift times to the same degradation product, Figure 2c).  Efforts to isolate diffraction quality crystals of {Pd 84 } Gly were met with less initial success,a ttributed to the cluster being considerably more hydrophilic than {Pd 84 } Ac ;w eo nly persevered because of the strong SEC and ESI-IMS-MS evidence that an ew {Pd 84 } Gly species was present in solution. After many attempts,r homboid shaped crystals appeared from ac oncentrated mother liquor and XRD confirmed this new structure to be {Pd 84 } Gly ,amacrocycle of general formula Na 56 H 14  Theo nly difference in the reactions producing {Pd 72 } Prop and {Pd 84 } Ac is the identity of the ligand added;both produce differently sized macrocycles.T his suggests that we have not only established am ethod to produce analogues of the {Pd 84 } Ac archetype,b ut also established ligand identity as am eans of controlling macrocycle size.W here two ligands lead to the same macrocyclic size,h owever,t hey may be mixed. This was shown in proof of concept experiments using am ixture of acetate and glycolate ligands at varying ratios, (1:9, 2:8, 3:7, and so forth;S upporting Information, Section 9). As the mole fraction of glycolate ligand was increased, the overall mass of the product observed increased gradually (Supporting Information, Figures S28 and S29). This is as trong indication that we obtain mixed functionality wheels,r ather than two homo-ligand wheels.G lycolate ligands bear alcohol functional groups.T he ability to tune the amount of these functional ligands on ar ing holds great synthetic promise,b oth as am eans to tune macrocycle properties (for example,s olubility), and to incorporate moieties amenable to further functionalization, either before or after macrocycle synthesis.
In conclusion, we have reported anew modular synthetic route to produce giant POPd macrocycles,a llowing us to independently vary the carboxylate ligand. This has facilitated expansion of the family of POPd wheels by the synthesis of two new structures, {Pd 84 } Gly and {Pd 72 } Prop .F urthermore, we have demonstrated the availability of ar ange of tunable mixed-ligand structures,a nd that the macrocycle size can be controlled by choice of ligand. Crucially,t he exploration of the clusters in solution using SEC and ESI-IMS-MS facilitated screening for new species in reaction mixtures without producing diffraction-quality crystals-the "crystallization bottleneck" in the discovery workflow-thereby decoupling synthesis and isolation from discovery.The sparse distribution of "hits" available over the parameter space screened-that is,t he fact that only af ew ligands produced POPd macrocycles-demonstrates the advantage of screening over traditional discovery,w here optimizing crystallization for each individual reaction can take months.T he conditions included in this screen have been limited, resembling those producing the original {Pd 84 } Ac structure.W eexpect that by investigating awider range of conditions using the same approach, we will add many more members to the palladium wheel family. Further to this,t he SEC screening approach can be more readily automated than conventional techniques.W eh ave found some validation for the "magic number" rule we previously predicted [11] for this emerging family of {Pd x } (Figure 4; also seen in recent reports on {Mn x }m acrocycles). We hope that by expanding the range of macrocycles available for study we will further reveal links between symmetry,b uilding blocks,a nd structure in the assembly of gigantic inorganic systems,thereby aiding the transition from discovery to design.