German Edition: DOI: 10.1002/ange.201806426Mixed-Valent CompoundsInternational Edition: DOI: 10.1002/anie.201806426

Electron Cartography in ClustersRaffll Hern#ndez S#nchez,* Anouck M. Champsaur, Bonnie Choi, Suyin Grass Wang, Wei Bu,Xavier Roy, Yu-Sheng Chen,* Michael L. Steigerwald,* Colin Nuckolls,* and Daniel W. Paley*

Abstract: Deconvoluting the atom-specific electron densitywithin polynuclear systems remains a challenge. A multiple-wavelength anomalous diffraction study on four clusters thatshare the same [Co6Se8] core was performed. Two cluster typeswere designed, one having a symmetric ligand sphere and theother having an asymmetric ligand sphere. It was found that inthe neutral, asymmetric, CO-bound cluster, the Co@CO site ismore highly oxidized than the other five Co atoms; when anelectron is removed, the hole is distributed among the Se atoms.In the neutral, symmetric cluster, the Co atoms divide byelectron population into two sets of three, each set beingmeridional; upon removal of an electron, the hole is distributedamong all the Co atoms. This ligand-dependent tuning of theelectron/hole distribution relates directly to the performance ofclusters in biological and synthetic systems.

Herein, we detail how the electron density distributionwithin hexanuclear cobalt chalcogenide clusters is determinedby systematically changing the nature of the substituents andthe electron count. The foundations of reactivity in organicsystems are predicated to a large extent on the electronicproperties imparted to a substrate by its substituents. Thus,electron-withdrawing or electron-donating substituents willdrastically alter the electron distribution within the moleculeand, as a consequence, shut down or promote certainreactivity modes. By contrast, this level of predictability inreactivity is absent in molecules containing several metals inproximity, mainly because the electron distribution withinpolynuclear assemblies is not well understood. To investigatethe oxidation levels of individual atoms within the cluster, weemployed multiple-wavelength anomalous diffraction(MAD).[1] We chose [Co6Se8] cluster compounds becausedetermining the individual atom valence for this clusternuclearity poses an unmet challenge in polynuclear syntheticsystems. These molecules have a structure best described by

a face-capped octahedron similar to those described byChevrel et al.[2] Members of this family of materials areknown to exhibit multivalency and superconductivity.[3]

Because MAD relies on single-crystal data, this methodologyallows us to study the relative valences of the individual atomscomprising the [Co6Se8] cluster core in two closely relatedmetal chalcogenide compounds, namely [Co6Se8(CO)(PEt3)5](1) and [Co6Se8(PnBu3)(PEt3)5] (2). We found that when thecluster core is asymmetrically substituted, as in the case of 1,the electron distribution follows what classical ligand fieldtheory would predict for the case in which the Co-CO site ismore electron deficient than the other five Co-PEt3 sites. Incontrast, when a symmetrical[4] all-phosphine ligand setting isimposed on the cluster core (2), the electron distributionremains asymmetrically grouped in a meridional fashion intotwo groups of three atoms, resembling the seam in a baseball.Furthermore, it is surprising that when 1 is oxidized the hole isdelocalized in Se-based orbitals, whereas when 2 is oxidizedthe hole is taken up by the Co atoms. This reordering of theorbital energetics can be attributed directly to the CO andPnBu3 substituent or ligand effects on the [Co6Se8] clustercore. Most importantly, this study provides an atom-individ-ual map of the electron distribution imposed by these ligandeffects.

Mixed-valent polynuclear systems are ubiquitous innature and also in synthetic materials.[5] They have uniqueproperties resulting from the electronic and magnetic inter-action of several metal atoms in close proximity.[6] Withinthose metal atoms, electron delocalization allows the clusterto distribute the electron density among the various redox-active centers.[7] As a consequence, determining the atom-individual valence of the sites involved in delocalization posesa formidable challenge. Despite the fundamental importanceof the charge distribution in clusters, there have been onlya few studies in which site-specific relative oxidation levelscan be determined.[8] Moreover, until the present study, therehas been no demonstration of how the electron densitydistribution is reconfigured upon oxidation at the atom-individual level within the same cluster framework.

The goal of the present study is to understand thefollowing questions: 1) How is the atom site-specific electrondensity distribution modulated by systematically substitutinga single ligand in polynuclear systems? 2) How is this electrondensity distribution reconfigured after the cluster undergoesoxidation? To answer these questions, we employed thecluster type [Co6Se8(L)(PEt3)5]

n+ as prototypical example inwhich the ligand field around the cluster core [Co6Se8] caneither be asymmetric (L = CO) or symmetric (L = PnBu3); inaddition, these clusters can be investigated at two differentformal electron counts (n = 0 or 1 + , see Figure 1). Morespecifically, the CO-bound asymmetric compound, [Co6Se8-

[*] Dr. R. Hern#ndez S#nchez, A. M. Champsaur, B. Choi, Prof. X. Roy,Dr. M. L. Steigerwald, Prof. C. Nuckolls, Dr. D. W. PaleyDepartment of Chemistry, Columbia UniversityNew York, NY 10027 (USA)E-mail: rh2780@columbia.edu

mls2064@columbia.educn37@columbia.edudwp2111@columbia.edu

Dr. S. G. Wang, Dr. W. Bu, Dr. Y.-S. ChenChemMatCARS, The University of ChicagoArgonne, IL 60439 (USA)E-mail: yschen@cars.uchicago.edu

Supporting information, including experimental details and crystal-lographic data, and the ORCID identification number(s) for theauthor(s) of this article can be found under:https://doi.org/10.1002/anie.201806426.

AngewandteChemieCommunications

13815Angew. Chem. Int. Ed. 2018, 57, 13815 –13820 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

(CO)(PEt3)5] (1, n = 0), has five Co atoms capped with PEt3

ligands while one is capped with CO (Figure 2a); meanwhile,the ligand sphere of the symmetric species, [Co6Se8(PnBu3)-(PEt3)5] (2, n = 0), has all six Co atoms capped withphosphines, one phosphine (tri-n-butylphosphine, PnBu3) ofwhich is different from the others to maintain the overall lowsymmetry of the molecule (Figure 2b). In addition, to explorethe electron density reconfiguration after oxidation, the one-electron oxidized compounds [Co6Se8(CO)(PEt3)5][PF6] (1-[PF6] , n = 1 +) and [Co6Se8(PnBu3)(PEt3)5][PF6] (2[PF6] , n =

1 +) were also investigated (Figure 2c,d).The synthesis of 1 and 1[PF6] has been previously

reported.[9] Following the same photochemical protocol, 2and 2[PF6] were synthesized by substitution of the CO in1 and 1[PF6] with PnBu3. High-quality single crystals of the

neutral species 1 and 2 were obtained by vapor diffusion ofdiethyl ether and pentane into toluene, respectively (Fig-ure 2e,f). Similarly, crystals of 1[PF6] and 2[PF6] were grownby diffusing Et2O into a concentrated solution of the clustersin dichloromethane (Supporting Information, Figures S5 andS6).[10] The cluster core structure [Co6Se8] in all four specieshas a geometry best described by an ideal face-cappedoctahedron, in which the Se atoms occupy the vertices ofa cube and the Co atoms reside in the center of each of thecubeQs faces.

Compounds 1 and 2 have an electron count that results ina formal mixed-valent configuration [4CoIII2CoII] (assumingall selenium atoms as Se2@), and thus 1[PF6] and 2[PF6] have[5CoIIICoII]. Consequently, all four species contain itinerantelectrons delocalized within the [Co6Se8] core. This electronsharing can be quantified by extracting the comproportiona-tion constant[11] (Kc) of each species by means of cyclicvoltammetry.[12] In this regard, the Kc values obtained for theneutral species (1, 1.3 X 1024 and 2, 8.3 X 1027), which are higherthan those of the charged clusters in 1[PF6] (5.3 X 1010) and2[PF6] (6.1 X 1011), indicate an overall large degree of electrondelocalization (Supporting Information, Table S3). To high-light this point, we can compare these Kc values to well-established strongly delocalized compounds having Kc valuesof circa 107, for example, the Creutz–Taube ion.[13] While theelectrochemical data presents an overall picture indicatinga large extent of electron sharing within the polynuclearassembly, it does not provide much insight into the electrondensity environments of the individual atoms. Therefore, todissect the site-specific relative electron densities for eachatom in the [Co6Se8] cluster core, we have examined thewavelength dependence of the atomic scattering factors onhigh-quality single crystals of 1, 1[PF6] , 2, and 2[PF6] .

Valence differentiation studies, for example, determina-tion of Cu oxidation state in the superconductorYBa2Cu3O6+x,

[8b] make use of the large variation of theatomic scattering factor (f) of an atom around an absorptionedge, which is given by f = f 0 + f ’ + if ’’, where f 0 is the

Figure 1. Synthetic strategy to modify the cluster’s external ligand fieldenvironment. Ligand substitution of L in the cluster type [Co6Se8(L)-(PEt3)5]

n+ for CO or PnBu3 produces an asymmetric or symmetricenvironment around the cluster core [Co6Se8] , respectively.

Figure 2. Polynuclear compounds analyzed by MAD: a) 1, b) 2, c) 1[PF6] , and d) 2[PF6]. Crystal structure of e) 1 and f) 2. Inset: Schematic of thecore with the Co and Se labeling scheme used throughout this report. Co (aquamarine), Se (brown), C (gray), O (red), and P (orange). Thermalellipsoids are set at 50% probability. Hydrogen atoms have been omitted for clarity.[10]

AngewandteChemieCommunications

13816 www.angewandte.org T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2018, 57, 13815 –13820

scattering factor of the unperturbed atom, and f ’ and f ’’correspond to the contributions from the in-phase and out-of-phase components of the anomalous scattering, respec-tively.[14] MAD has been utilized extensively in polycrystallinesystems,[8] but less so in single crystal studies.[1a, 15] In reports ofMAD on single crystals, the mixed-valent species under studydisplays a clear minimum in the f ’ versus energy curve foreach atom site, which indicates the edge position, thusallowing a clear assignment of the relative oxidation statesof the redox atoms involved in electron sharing.[15a–c,e] More-over, as electron delocalization increases this assignmentbecomes less straightforward.

The MAD experiments in this report were carried outafter first determining the Co and Se K-edge positions on thesame single crystal to be further examined by X-ray diffrac-tion (top part of Figure 3e–h). With this information, wedesigned a data collection strategy that consists of collectingpartial diffraction data (720 frames) every 1 eV from belowthe pre-edge feature covering the entire K-edge and untilright after the edge; additionally, below the pre-edge andabove the edge, data was collected every 3–5 eV (seeSupporting Information). Moreover, prior to collecting thepartial diffraction data sets, the full structure was determinedat high energy (30 keV). All the structure parameters from

Figure 3. Truncated molecular structures of a) 1, b) 2, c) 1[PF6] , and d) 2[PF6] . Insets (a–d): Relative oxidation levels within the [Co6Se8] core,where a Co red sphere indicates a higher oxidation level relative to the blue ones, and similarly a Se yellow sphere indicates a higher oxidationlevel relative to the gray ones. e–h) Data plots from 1 (circles), 2 (squares), 1[PF6] (diamonds), and 2[PF6] (triangles). In each panel the top curvecorresponds to the single-crystal X-ray fluorescence scan collected around the Co K-edge in steps of 1 eV at 100 K on the same crystal employedin all the diffraction data. The bottom part of each panel corresponds to the anomalous scattering factor f ’ of each cobalt center in the [Co6Se8]cluster core. The anomalous scattering data was refined using the full structure data set for each compound collected at 30 keV as the reference.

AngewandteChemieCommunications

13817Angew. Chem. Int. Ed. 2018, 57, 13815 –13820 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org

the 30 keV data were used and kept fixed when refining thepartial diffraction data except for f ’ and f ’’ for every Co andSe atom at the Co and Se K-edge, respectively. In this manner,and since every Co and Se atom in 1, 1[PF6] , 2, and 2[PF6] arecrystallographically independent, we were able to deconvo-lute the individual f ’ and f ’’ curves for all atoms within the[Co6Se8] cluster cores.[16]

In Figure 3a–d we display truncated structures of all fourspecies with the corresponding atom labels subsequently usedin the anomalous scattering plots in Figure 3e–h. Theanomalous scattering plots in Figure 3e–h provide the fullpicture of the f ’ evolution around the absorption edge.Regardless of the substitution pattern and oxidation level, allfour species display a pre-edge feature in the Co K-edge at7704 eV (insets in Figure 3e–h). The anomalous scattering f ’curve for 1 is shown in Figure 3e (bottom). Starting at lowenergy, all Co sites have superimposable f ’ values. After theminimum, which occurs around 7711 eV, all Co sites remainequivalent within experimental error, except Co1. This siteQsf ’ curve broadens and departs from the rest in the high-energyside indicating an overall higher oxidation level than the otherfive cobalt sites, exemplified in Figure 3a, inset. While the Cooxidation states can be described formally as [4CoIII2CoII],the experimental data does not support this, showing that Co1has a slightly higher oxidation level than Co2 to Co6, whichshare the same valence. This result fits with chemicalintuition, by which one expects the CO-bound Co site to bemore electron deficient than the other, phosphine-bound,sites. The resulting electron density at the Co centerscorrelates directly with the donor–acceptor properties ofCO and PEt3. For instance, CO withdraws electron densitysince it acts as a weak s-donor and strong p-acceptor, whereasalkylphosphines are known as strong s-donors and poor p-acceptors.[17] Remarkably, the electron density distributionchanges considerably in the symmetric all phosphine-boundspecies 2, in which two overall oxidation levels are observedin a meridional arrangement (Figure 3b, inset) formed byCo1, Co3, and Co5 residing at a higher valence relative toCo2, Co4, and Co6 (Figure 3 f). In terms of the Se localenvironments, 77Se NMR data shows two and one (broad)chemical environments in 1 and 2, respectively (SupportingInformation, Figure S7).

We examine the f ’ anomalous scattering data extractedfrom 1[PF6] and note that it is identical within experimentalerror to that of 1 (Figure 3g). This is remarkable since 1[PF6]has one electron less than 1. Hence, the hole in 1[PF6] doesnot sit on the Co atoms. We hypothesize the Se atoms in1[PF6] are oxidized and no longer display a unique valence asin the parent compound 1 (Figure 3c, inset). Unfortunately,the refined data at the Se K-edge for 1[PF6] (SupportingInformation, Figure S14) is lower in quality than that of 1. Incontrast, the MAD analysis on 2[PF6] shows coalescence ofthe Co environments into a single one. Remarkably, theburden of oxidation at Co falls now upon all six Co atoms asseen in Figure 3h and exemplified in the inset of Figure 3d.

Figure 4 displays the electron density evolution from 1 to1[PF6] and from 2 to 2[PF6] . The scattering factors, f ’, of1 and 1[PF6] are compared across a small energy window(7710–7718 eV), in which the departure of Co1 (CO-bound

site) from the rest is apparent (Figure 4a,b). The behavior isdifferent in 2 and 2[PF6] , in which the meridional arrange-ment of two sets of three Co atoms each, described above for2, is lost and now all Co atoms in 2[PF6] reside at the sameoxidation level (Figure 4c,d). Thus, triggered by oxidation,the binomial distribution of oxidation levels in 2 merges intoa single level in 2[PF6] that resides right in between theformer oxidation levels seen in 2 (Supporting Information,Figure S18).

At a first glance the charge distribution obtained fromMAD analysis is contradictory with the overall reduction inp-backbonding observed towards CO and the phosphines.[17]

We took a closer look at the bond metrics within eachoxidation sequence. In the sequence of 1 to 1[PF6] the Co@CO and Co@Pavg bond distances indicate an overall decreasein p-backbonding by changing from 1.728(5) to 1.740(5) c,and from 2.142(2) to 2.180(3) c, respectively. Similarly, theCO stretching frequency (nCO) is in line with the decreased p-backbonding upon oxidation by going from 1958 to1983 cm@1.[9] At the same time the Co@Seavg distances in1 and 1[PF6] remain unchanged (Table S2). While this alonewould indicate an overall increase in oxidation of all Cocenters, it appears that the electron density directed towardsp-backbonding with the peripheral ligands is now used toenhance intracluster bonding judged by an overall contractionof the Co@Coavg distances in 1 (2.946(5) c) and 1[PF6](2.906(11) c). We thus conclude that these two effects leveleach other and gives support to the electron densitydistribution obtained from MAD. In agreement, the sequencefrom 2 to 2[PF6] displays an overall expansion of the Co@Pavg

(2, 2.141(1) c; 2[PF6] , 2.176(1) c) and contraction of the Co@Coavg (2, 2.953(4) c; 2[PF6] , 2.921(10) c) distances.

The electron density redistribution seen in 1, 2, 1[PF6] ,and 2[PF6] is remarkable and points to an electronic structurechange upon a minor perturbation of the ligation pattern.

Figure 4. Zoomed-in anomalous scattering f’ factor for a) 1, (b) 1[PF6] ,c) 2, and d) 2[PF6] . The neutral and oxidized cluster core data withinthe same family (CO- or PnBu3-ligated) are aligned in the same energyrange in order to distinguish differences among these.

AngewandteChemieCommunications

13818 www.angewandte.org T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2018, 57, 13815 –13820

While metal-based redox is usually invoked, the datapresented herein demonstrates that when removing anelectron from the CO-bound species 1, the oxidation burdenrests primarily at the chalcogen, Se atoms, similar to what isobserved in biological cofactors.[18] In contrast, removal of anelectron in the ligand field symmetric all-phosphine species 2leaves behind a hole delocalized primarily within the metalatoms (Co). Thus, by deconvoluting the individual relativeoxidation levels in polynuclear systems through MAD, wewere able to map the electron density configuration within anunsymmetrical (as in 1, 1[PF6]) and symmetrical (as in 2,2[PF6]) ligand field environment. Furthermore, we were ableto determine how the electron density map is reconfiguredupon oxidation. This study demonstrates the important effecta single ligand has on the electron density distribution withina polynuclear system akin to the electron withdrawing/donating effect of substituents in organic systems. Theseresults also indicate that while a first-order, electron-countingdescription of the electronic structure of such clusters may beuseful and valuable, the details of chemical bonding withinthe clusters is clearly much more complex.

This ligand-dependent tuning of the electron/hole reser-voirs has direct relevance to how biological cofactors functionin which similar mechanisms are likely in play when biologicalpolynuclear systems bind and activate substrates (e.g. N2,H2O, CO2, or CH4), especially those comprising multiplemetal atoms acting in concert.[19] Moreover, this tuning of theelectron density distribution based on the peripheral ligandfield should have important implications towards buildingmagnetic anisotropy in magnetic materials[20] or definingpreferred conduction pathways in superconductors.[8b] Theresults on these complex clusters present an ideal system toexplore for those at the forefront of quantum chemicalcalculations. Future studies exploring the electron densitydistribution across different overall oxidation levels inbiological cofactors and materials will yield insight into themechanism of action of these systems.

Acknowledgements

R.H.S. acknowledges the support from the Columbia NanoInitiative Postdoctoral Fellowship. B.C. acknowledges sup-port by the NSF Graduate Research Fellowship under GrantDGE 11-44155. C.N. thanks Sheldon and Dorothea Bucklerfor their generous support. D.W.P. thanks A. L. Spek forhelpful correspondence on the use of Platon ShxAbs, andV#clav Petr&cek for assistance with Jana2006. ChemMat-CARS Sector 15 is supported by the National ScienceFoundation under grant number NSF/CHE-1346572. Supportfor this research was provided by the Center for PrecisionAssembly of Superstratic and Superatomic Solids, an NSFMRSEC (award number DMR-1420634), and the Air ForceOffice of Scientific Research (award number FA9550-18-1-0020). This research used resources of the Advanced PhotonSource, a U.S. Department of Energy (DOE) Office ofScience User Facility operated for the DOE Office of Scienceby Argonne National Laboratory under Contract No. DE-AC02-06CH11357. The Columbia University Shared Materi-

als Characterization Laboratory (SMCL) was used exten-sively for this research. We are grateful to ColumbiaUniversity for support of this facility.

Conflict of interest

The authors declare no conflict of interest.

Keywords: clusters · electron density distribution ·mixed-valent compounds ·multiple-wavelength anomalous diffraction

How to cite: Angew. Chem. Int. Ed. 2018, 57, 13815–13820Angew. Chem. 2018, 130, 14011–14016

[1] a) W. A. Hendrickson, J. L. Smith, R. P. Phizackerley, E. A.Merritt, Proteins Struct. Funct. Bioinf. 1988, 4, 77 – 88; b) W. A.Hendrickson, Science 1991, 254, 51 – 58.

[2] R. Chevrel, M. Sergent, J. Prigent, J. Solid State Chem. 1971, 3,515 – 519.

[3] O. PeÇa, Phys. C 2015, 514, 95 – 112.[4] We use tri-n-butylphosphine (PnBu3) as the closest relative of

triethylphosphine (PEt3) that gives high quality crystals whileaverting an inversion center in the cluster. Thus, we justify thewording “symmetrical all-phosphine ligand setting”, since theligand field imposed by PnBu3 is very similar to PEt3 anddifferent from the asymmetric environment caused by thelargely distinct CO ligand.

[5] J. C. Fontecilla-Camps, P. Amara, C. Cavazza, Y. Nicolet, A.Volbeda, Nature 2009, 460, 814 – 822.

[6] P. Day, N. S. Hush, R. J. H. Clark, Philos. Trans. R. Soc. A 2008,366, 5 – 14.

[7] H. Beinert, R. H. Holm, E. Munck, Science 1997, 277, 653 – 659.[8] a) R. S. Howland, T. H. Geballe, S. S. Laderman, A. Fischercol-

brie, M. Scott, J. M. Tarascon, P. Barboux, Phys. Rev. B 1989, 39,9017 – 9027; b) G. H. Kwei, R. B. Vondreele, A. Williams, J. A.Goldstone, A. C. Lawson, W. K. Warburton, J. Mol. Struct. 1990,223, 383 – 406; c) J. P. Attfield, Nature 1990, 343, 46 – 49; d) A. P.Wilkinson, A. K. Cheetham, D. E. Cox, Acta Crystallogr. Sect. B1991, 47, 155 – 161; e) J. P. Attfield, J. Phys. Chem. Solids 1991,52, 1243 – 1249; f) A. K. Cheetham, A. P. Wilkinson, Angew.Chem. Int. Ed. Engl. 1992, 31, 1557 – 1570; Angew. Chem. 1992,104, 1594 – 1608; g) A. P. Wilkinson, A. K. Cheetham, S. C. Tang,W. J. Reppart, J. Chem. Soc. Chem. Commun. 1992, 1485 – 1487;h) G. Artioli, A. Pavese, M. Bellotto, S. P. Collins, G. Lucchetti,Am. Mineral. 1996, 81, 603 – 610; i) M. Cianci, J. R. Helliwell, M.Helliwell, V. Kaucic, N. Z. Logar, G. Mali, N. N. Tusar, Crystal-logr. Rev. 2005, 11, 245 – 335; j) R. J. Goff, J. P. Wright, J. P.Attfield, P. G. Radaelli, J. Phys. Condens. Matter 2005, 17, 7633 –7642; k) H. Palancher, S. Bos, J. F. Berar, I. Margiolaki, J. L.Hodeau, Eur. Phys. J. Spec. Top. 2012, 208, 275 – 289.

[9] A. M. Champsaur, A. Velian, D. W. Paley, B. Choi, X. Roy, M. L.Steigerwald, C. Nuckolls, Nano Lett. 2016, 16, 5273 – 5277.

[10] CCDC 1565550 (1), 1565552 (1[PF6]), 1565551 (2), and1565553 (2[PF6]) contain the supplementary crystallographicdata for this paper. These data can be obtained free of chargefrom The Cambridge Crystallographic Data Centre.

[11] [Co6]n + [Co6]

n+2Ð2[Co6]n+1, where the equilibrium constant is

referred to as the comproportionation constant (Kc).[12] a) M. D. Ward, Chem. Soc. Rev. 1995, 24, 121 – 134; b) D. M.

D’alessandro, F. R. Keene, Dalton Trans. 2004, 3950 – 3954.[13] C. Creutz, M. H. Chou, Inorg. Chem. 1987, 26, 2995 – 3000.[14] P. Coppens, D. Cox, E. Vlieg, I. K. Robinson, Synchrotron

radiation crystallography, Academic Press, San Diego, 1992.

AngewandteChemieCommunications

13819Angew. Chem. Int. Ed. 2018, 57, 13815 –13820 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org

[15] a) Y. Gao, A. Frostjensen, M. R. Pressprich, P. Coppens, A.Marquez, M. Dupuis, J. Am. Chem. Soc. 1992, 114, 9214 – 9215;b) Y. Gao, M. R. Pressprich, P. Coppens, Acta Crystallogr. Sect.A 1993, 49, 216 – 219; c) G. Wu, Y. Zhang, L. Ribaud, P. Coppens,C. Wilson, B. B. Iversen, F. K. Larsen, Inorg. Chem. 1998, 37,6078 – 6083; d) A. R. Cowley, R. H. Jones, S. J. Teat, A. M.Chippindale, Microporous Mesoporous Mater. 2002, 51, 51 – 64;e) O. Einsle, S. L. A. Andrade, H. Dobbek, J. Meyer, D. C. Rees,J. Am. Chem. Soc. 2007, 129, 2210; f) M. Helliwell, J. R.Helliwell, V. Kaucic, N. Z. Logar, S. J. Teat, J. E. Warren, E. J.Dodson, Acta Crystallogr. Sect. B 2010, 66, 345 – 357; g) L. M.Zhang, J. T. Kaiser, G. Meloni, K. Y. Yang, T. Spatzal, S. L. A.Andrade, O. Einsle, J. B. Howard, D. C. Rees, Angew. Chem. Int.Ed. 2013, 52, 10529 – 10532; Angew. Chem. 2013, 125, 10723 –10726; h) T. Spatzal, J. Schlesier, E. M. Burger, D. Sippel, L. M.Zhang, S. L. A. Andrade, D. C. Rees, O. Einsle, Nat. Commun.2016, 7, 10902.

[16] We note that near the absorption edge, the anomalous scatteringfactors are more accurately described as second-rank tensorsthat interact with the polarization of the X-ray beam (L. K.Templeton, D. H. Templeton, Acta Crystallogr. Sect. A 1988, 44,1045 – 1051). In this work, the polarization dependence is largelyaveraged out in the final refinement because we collected 18088 f

scans with the f axis parallel to the plane of polarization. Asa future improvement, a full refinement of three-dimensionalanomalous scattering factors would contain valuable chemicalinformation, especially in the pre-edge region where symmetry-allowed electronic transitions are a significant part of the

absorption coefficient. We thank two anonymous reviewers forhelpful comments on this subject.

[17] A. G. Orpen, N. G. Connelly, Organometallics 1990, 9, 1206 –1210.

[18] a) R. Weiss, D. Mandon, T. Wolter, A. X. Trautwein, M. Muther,E. Bill, A. Gold, K. Jayaraj, J. Terner, J. Biol. Inorg. Chem. 1996,1, 377 – 383; b) M. T. Green, J. Am. Chem. Soc. 1999, 121, 7939 –7940; c) M. T. Green, J. Am. Chem. Soc. 2001, 123, 9218 – 9219.

[19] a) R. Y. Igarashi, P. C. Dos Santos, W. G. Niehaus, I. G. Dance,D. R. Dean, L. C. Seefeldt, J. Biol. Chem. 2004, 279, 34770 –34775; b) J. H. Jeoung, H. Dobbek, Science 2007, 318, 1461 –1464; c) A. Pomowski, W. G. Zumft, P. M. H. Kroneck, O.Einsle, Nature 2011, 477, 234 – 237; d) S. M. Smith, S. Rawat, J.Telser, B. M. Hoffman, T. L. Stemmler, A. C. Rosenzweig,Biochemistry 2011, 50, 10231 – 10240; e) T. Spatzal, K. A.Perez, O. Einsle, J. B. Howard, D. C. Rees, Science 2014, 345,1620 – 1623; f) J. Yano, V. Yachandra, Chem. Rev. 2014, 114,4175 – 4205.

[20] a) Z. X. Wang, J. van Tol, T. Taguchi, M. R. Daniels, G. Christou,N. S. Dalal, J. Am. Chem. Soc. 2011, 133, 17586 – 17589; b) R.Hern#ndez S#nchez, T. A. Betley, J. Am. Chem. Soc. 2015, 137,13949 – 13956.

Manuscript received: June 4, 2018Accepted manuscript online: September 4, 2018Version of record online: September 21, 2018

AngewandteChemieCommunications

13820 www.angewandte.org T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2018, 57, 13815 –13820