Metal–Metal Bonding in Uranium–Group 10 Complexes

Heterobimetallic complexes containing short uranium–group 10 metal bonds have been prepared from monometallic IUIV(OArP-κ2O,P)3 (2) {[ArPO]− = 2-tert-butyl-4-methyl-6-(diphenylphosphino)phenolate}. The U–M bond in IUIV(μ-OArP-1κ1O,2κ1P)3M0, M = Ni (3–Ni), Pd (3–Pd), and Pt (3–Pt), has been investigated by experimental and DFT computational methods. Comparisons of 3–Ni with two further U–Ni complexes XUIV(μ-OArP-1κ1O,2κ1P)3Ni0, X = Me3SiO (4) and F (5), was also possible via iodide substitution. All complexes were characterized by variable-temperature NMR spectroscopy, electrochemistry, and single crystal X-ray diffraction. The U–M bonds are significantly shorter than any other crystallographically characterized d–f-block bimetallic, even though the ligand flexes to allow a variable U–M separation. Excellent agreement is found between the experimental and computed structures for 3–Ni and 3–Pd. Natural population analysis and natural localized molecular orbital (NLMO) compositions indicate that U employs both 5f and 6d orbitals in covalent bonding to a significant extent. Quantum theory of atoms-in-molecules analysis reveals U–M bond critical point properties typical of metallic bonding and a larger delocalization index (bond order) for the less polar U–Ni bond than U–Pd. Electrochemical studies agree with the computational analyses and the X-ray structural data for the U–X adducts 3–Ni, 4, and 5. The data show a trend in uranium–metal bond strength that decreases from 3–Ni down to 3–Pt and suggest that exchanging the iodide for a fluoride strengthens the metal–metal bond. Despite short U–TM (transition metal) distances, four other computational approaches also suggest low U–TM bond orders, reflecting highly transition metal localized valence NLMOs. These are more so for 3–Pd than 3–Ni, consistent with slightly larger U–TM bond orders in the latter. Computational studies of the model systems (PH3)3MU(OH)3I (M = Ni, Pd) reveal longer and weaker unsupported U–TM bonds vs 3.


■ INTRODUCTION
The nature of the bonding in f-block metal−ligand bonds is still far from fully understood, and bonding between f-block metals and other metal cations even less so. By contrast, studies of the bonding between d-block and other metal cations are 50 years old and have furthered our understanding of d-orbital interactions and generated some unique small molecule activation chemistry and catalyzed reactions not seen in single metal chemistry. 1 The few complexes that feature bonds between an f-block and d-block cation 2−4 have begun to help to improve our understanding of metal−metal bonding and to challenge and help the development of computational methods; however, the challenges associated with their synthesis and characterization have precluded the systematic study of any families of heterobimetallics that would enable the prediction of trends in other systems. The 5f orbitals have a suitable spatial extension but not yet a predictability of participation in bonding that makes the d−f heterobimetallic bond a particularly interesting target to improve our understanding of the relative involvement of f-and d-orbital participation. A better understanding of the subtleties of the 5f/6d contributions to actinide bonding in general is important in the handling of nuclear materials, where differences in behavior are dominated by small covalency differences in bonding.
The groups of Bart and Thomas reported uranium−cobalt compounds with bridging heterobidentate monoanionic PN ligands. 10 At 2.874(3) Å, ICo(μ-Ph 2 PNPr i -κ 1 P,κ 1 N) 3 UI (A) exhibits the shortest uranium−transition metal bond reported prior to this study; in this most instructive work, the analogue with Pr 2 i PNMes ligands (Mes = C 6 H 2 Me 3 -2,4,6) was proposed from voltammetric experiments to have a stronger Co → U dative interaction than the relatively modest one in A, although the changes at both ends of the bidentate ligand make the components difficult to separate, and the latter complex was not structurally characterized. One of the arguments for using heterobidentate ligands has been to expand the variety of synthetic routes to M−M′ bonds; the photolytic release of CO from the isocarbonyl U−OC−Co moiety upon photolysis is arguably the most inventive synthesis yet, forming a 3.0319(7) Å U(IV)−Co(I) bond within the rigid NP scaffold N[ο-(NHCH 2 PPr 2 i )C 6 H 4 ] 3 and with the suggestion of a close contact between the U center and another Co-bound CO ligand, while vibrational data suggest stronger donation of Co electron density to uranium through the bond than through the original isocarbonyl link, although the coordination of additional phosphines has also changed the ancillary ligand set somewhat. 11 For comparison, the shortest distance yet found between an f and a d block metal is in the lutetium−platinum complex (C 5 Me 4 SiMe 2 CH 2 PPh 2 )Lu(μ-CH 2 SiMe 2 CH 2 )(OC 4 H 8 )PtMe 2 , 2.7668(5) Å, 20 which shows interesting intramolecular C−H bond cleavage chemistry at elevated temperatures.
Even though the heterobimetallic chemistry of rare-earth transition-metal compounds has now begun to receive considerable attention, examples with late transition metals are still rare. 2−4 Roesky also used phosphinoamido ligands to combine palladium with yttrium and lutetium in bi-and trimetallic compounds with 2.9898(6) Å (Y−Pd), 2.9712(8) Å (Lu−Pd), and 3.141(13) Å (Y−Pd 2 ) bond lengths for the trimetallic compound. 21 We have targeted phosphine-functionalized aryloxide analogues of U(IV)(OAr) 3 X (OAr = 2,6-di-tert-butylphenoxide) first reported in the 1980s, anticipating binding of a second metal by the incorporated phosphine groups, and since then work by others and us has shown that the U−OAr bonds are sufficiently robust to allow many X-substitution reactions without ligand scrambling that can dominate f-block coordination chemistry. 22−27 Other robust ligand sets, such as bis(permethylcyclopentadienyl) or polydentate chelates, can be insufficiently mobile to allow the M−M′ distance to change according to metal size or electronic preference. We also considered that the X ligand in the trans position to a ligated metal ion would provide the possibility to exploit the inverse trans influence (ITI) in the formation of stronger bonds to an atom (here the second metal) in the position trans to X, the phenomenon whereby mutually trans-ligands bind closer and more tightly to a uranium center than they would in a d-block system, since the available (pseudocore) U 6p orbitals can mix with the valence 5f. 28−31 We present a set of new heterobimetallic uranium−group 10 metal complexes using these simple ligands, the first study of an actinide−M bond for a complete transition-metal group and the first set of differently trans-ligand-functionalized uranium− metal bonds. We show how the ligand supports the shortest 5f−nd metal−metal bonds yet stabilized and preserves the metal−metal bond while allowing steric/electronic variation of the metal-bound X-ligand. These features have enabled a thorough study of the electronic structure of the metal−metal bond and its variation from 3d to 5d, and with ancillary ligand, for the first time.
The reaction of 3 mol equiv of 1 with uranium(IV) iodide etherate in THF gives IU(OC 6 H 2 -6-Bu t -4-Me-2-PPh 2 -κ 2 O,P) 3 [IU(OAr P -κ 2 O,P) 3 , 2] as a bright green powder in 78% yield (Scheme 1). 25 Compound 2 is moderately soluble in benzene and toluene; 1 H NMR spectroscopy shows broad overlapping resonances at ambient temperature but seven resonances at elevated temperatures, although even at 370 K these are still broad. No 31 P NMR resonance is observed at 300 or 370 K, similarly to other uranium phosphine complexes, e.g., U(dmpe) 2 X 4 (X = Cl, OPh, Me). 33,34 This could indicate a persistent coordination of the phosphine groups to the paramagnetic uranium or a more dynamic process that broadens the resonances to baseline. The room-temperature magnetic moment (Evans' method) of 2 is 2.4 μ B .
Green crystals of 2 suitable for single-crystal X-ray diffraction were grown from a benzene solution at ambient temperature.

Journal of the American Chemical Society
Article terized trialkylphosphine uranium compounds. The U−P bond distances in 2 are similar to those of U(dmpe) 2 X 4 (X = Cl, OPh, Me). 33,34 The related U(IV)(Pr 2 i PNMes-κ 2 P,N) 3 UI has U−P distances between 2.8662(12) and 2.8828(4) Å. 10 Preparation of the Bimetallic Compounds. The reaction of 2 with low oxidation state group 10 metal compounds incorporates the respective metal through phosphine ligation. The uranium(IV)−nickel(0) derivative is prepared by treatment of 2 with an equimolar amount of Ni 0 (cod) 2 (cod = 1,5-cyclooctadiene) in toluene at ambient temperature (Scheme 2). An immediate color change to dark red indicates fast displacement of cod by the triarylphosphine donor groups. The new heterobimetallic IU(μ-OAr P -1κ 1 O,2κ 1 P) 3 Ni (3−Ni) is isolated as dark red crystals in excellent yield and exhibits similar solubility to the parent compound 2.
Upon introduction of the group 10 metal centers, the magnetic moment decreases from 2.4 μ B in 2 to 1.9 μ B (3−Ni), 1.8 μ B (3−Pd), and 2.0 μ B (3−Pt), respectively. The similarity of the values for the bimetallic systems might indicate analogous interaction between the different d 10 -metals and uranium. However, a comparison of this group trend with previously reported systems is not possible, since no magnetic data were given for the only pair of complexes in which the ligand system remained unchanged between different transition-metal derivatives, Cp 3 UMCp(CO) 2 (M = Fe, Ru). 5 Our observation contrasts with that of the groups of Bart and Thomas and of Arnold and Lu, who reported an increase of the magnetic susceptibility after introduction of the cobalt center into the amidophosphine-ligated uranium compounds. 10,11 Single crystals suitable for X-ray diffraction studies of the bimetallic compounds were grown at ambient temperature by vapor diffusion of hexane into benzene solutions. The solidstate structures of the bimetallic complexes 3−Ni, 3−Pd, and 3−Pt are similar and are shown in Figure 2. Selected distances and angles are collected in Table 1, along with the data from the DFT calculations described below. The agreement between experiment and theory is very good.
Despite the C 3 -symmetry of the individual molecules, the structures of 3−Ni and 3−Pt could be refined only in the triclinic space group P1. The asymmetric unit of each contains nine molecules of the compound along with three benzene molecules each. The individual molecules show slightly different orientations and could not be refined using a space group of higher symmetry. Complex 3−Pd was solved and refined in the trigonal space group P3 2 and contains three molecules of complex, with three benzene molecules per complex, in the asymmetric unit. The U−M bond distances increase from 2.527(2)−2.540(2) Å in 3−Ni to 2.686(2)− 2.694(1) Å in 3−Pd and 2.706(1)−2.709(1) Å in 3−Pt ( Figure  2). The intermetallic bond distances are significantly shorter than those of any other crystallographically characterized d-and f-block bimetallic compound previously reported. [The shortest, Lu−Pt, noted above, is 2.7668(5) Å. 20 ] The only other actinide−group 10 derivatives reported are Cp*Th(μ-PPh 2 ) 2 Ni(CO) 2 and Cp*Th(μ-PPh 2 ) 2 PtPMe 3 , featuring intermetallic distances of 3.206(2) and 2.984(1) Å, respectively. 18,19 To the best of our knowledge, no other uranium−group 10 derivatives have been reported; comparison with the bimetallic U IV −Co I ICo(μ-Ph 2 PNPr i -1κ 1 P,2κ 1 N) 3 U[η 2 -Ph 2 PNPr i ] and ICo(μ-Ph 2 PNPr i -1κ 1 P,2κ 1 N) 3 UI reported by the groups of Bart and Thomas is most instructive. 10  The U−I bonds of 3−Ni, 3−Pd, and 3−Pt are shorter than those in the parent monometallic derivative 2 by around 0.04 Å in all cases. It is tempting to attribute this to the ITI. However, it could simply be a result of the lower coordination number in the bimetallic complexes (5-coordinate at U, including the M− M bond) compared to the phosphine-ligated parent compound (7-coordinate at U) since the effective ionic radius of the 5coordinate U will be up to 0.1 Å smaller. 35 The compounds 3−Ni, 3−Pd, and 3−Pt show paramagnetically shifted 1 H NMR resonances in the range 3.56−11.82 ppm Figure 2. Thermal ellipsoid plot for 3−Ni, 3−Pd, and 3−Pt. Solvent molecules and hydrogen atoms are omitted, and selected carbon atoms are depicted as a wireframe, for clarity. Thermal ellipsoids are drawn at 50% probability, and only one independent molecule out of nine in the asymmetric unit is shown. Selected bond distances (Å) and angles (deg) are shown in Table 1.  (4)  . Again, the C 3 -symmetry of the solid-state structures is evident in solution, but the interconversion of the Δ and Λ isomers can now be observed using variable-temperature (VT) NMR spectroscopic experiments in the range 300−370 K ( Figure 3 for 3−Ni), with coalescence relating to the interconversion shown in Figure 4, which requires the breaking of the U−M bond.
In the case of 3−Ni, two sets of distinct resonances are observed at ambient temperature for the two phenyl groups on the diphenylphosphine substituents. These can be accounted for by the two orientations of the rings either close to perpendicular or along the metal−metal axis, which are visible at elevated temperatures (see the SI). For 3−Pd, no phenyl resonances are observed at ambient temperature and for 3−Pt they are strongly broadened.
The coalescence temperatures (T C ) were highest for 3−Ni at ca. 332 K followed by 3−Pt at around 314 and 300 K for 3−Pd.
In the case of 3−Pt, strong broadening and overlapping of unrelated shifts allow only a rough estimation of T C . This agrees with the other data, making U−Ni the hardest bond and U−Pt the easiest bond to break to interconvert the isomers. Unfortunately, the presence of the paramagnetic U center precludes the calculation of the energy associated with these dynamic processes due to the additional temperature dependence of the chemical shifts.
The broadened 31 P{ 1 H} resonances appear at 300 K at 93. , but the influence from the uranium paramagnet cannot be quantified. In the absence of paramagnetism, a related shift to higher frequency on incorporation of the more electropositive metal has been used as an (unquantified) indication of group 10 metal → metal electron donation. 17,18,36,37,18,19,38 The significantly lower 31 P− 195 Pt coupling constant in 3−Pt compared to Pt(PPh 3 ) 3 can also be taken as an indication of a 4-coordinate platinum(0). 19 Computational Investigation of 3−Ni and 3−Pd. In order to probe the uranium−TM (transition metal) bonding within 3−M, we turned to density functional theory (DFT) at the PBE level. Calculated geometric data for 3−Ni and 3−Pd are collected in Table 1. 39 As noted above, the agreement between the calculated and experimental structures is excellent, the largest discrepancy being <0.04 Å (for the average Ni−P distances in 3−Ni).
Natural population analysis (NPA) data are presented in Table 2, from which it can be seen that the two systems have very similar electronic structures at the NPA level. The spin densities are very much in keeping with a U(IV) system. Partial charges rarely tally well with formal oxidation state, but those calculated here show that the actinide atoms are much more positive than the transition metals and are very close to zero for the latter, in keeping with an M(0) formalism. The population analysis shows that the 10 electrons expected for Ni(0) and Pd(0) are located mainly in the 3d and 4d orbitals, with a small 4s/5s population. The uranium populations show the expected buildup in 5f and 6d, 1.12/1.08 and 1.50/1.46 electrons, respectively, above the value expected for U(IV) (data for 3− Pd in italics). Such buildups are often taken as a measure of the extent to which the 5f and 6d orbitals are involved in covalent bonding with the surrounding ligand framework. 39− 41 We are happy to adopt this approach for the early part of the 5f series, and the present data indicate significant involvement of both fand d-orbitals.
Further insight into 3−Ni is provided by analysis of the valence natural localized molecular orbitals (NLMOs), the compositions and characters of an α spin selection of which are collected in Table 3 and shown pictorially in Figure 5. NLMOs 110 and 111 are the two U 5f electrons. NLMOs 113 and 114 are strongly iodine-localized (ca. 92%) and pπ in character, while NLMO 128 is I pσ, with a significantly larger (ca. 20%) uranium contribution than the pπ levels. The remaining orbitals (115−118 and 129) are nickel 3d-based. Together with the five β spin equivalents, these NLMOs house the 10 nickel electrons located by the NPA. They separate into σ + 2π + 2δ with respect to the U−Ni axis, with differing contributions from the actinide. The δ orbitals (115 and 116) have essentially no uranium contribution, while the π orbitals (117 and 118), also strongly nickel-localized, have slightly larger uranium contributions (similar to those of the iodine pπ-localized orbitals). Finally, NLMO 129 is nickel dσ, with ca. 10% uranium character. The uranium contributions to the iodine-based NLMOs are more 6d-based than 5f, while the reverse is true for the nickel-localized orbitals.
In order to probe further the nature of the U−Ni and U−Pd bonds, we turned to the quantum theory of atoms-in-molecules (QTAIM) approach, which we have used extensively to study the electronic structure of 5f molecules. 26,41−43 Bond critical point (BCP) data are collected in Table 4, together with five different measures of U−TM bond order. The BCP electron and energy densities and the electron density Laplacian (ρ, H, and ∇ 2 ρ) are very similar for the two target systems and very

Journal of the American Chemical Society
Article much in keeping with the extensive previous QTAIM studies of metal−metal bonds, in both bulk metals and polynuclear complexes. 44−57 These have suggested that metal−metal bonding should not be pigeon-holed as either closed-shell or shared-shell, but that "metallic" bonding has a topological behavior of its own, possessing neither ionic nor covalent features; metal−metal bonds are identified by relatively low electron density at the BCP and positive ∇ 2 ρ (normally associated with closed shell or ionic bonding) and negative H (usually typical of shared shell or covalent bonding). The BCP ellipticity ε is a measure of the cylindrical symmetry of a bond. Values close to zero are associated with either single or triple bonds, while significant deviations from zero (up to ca. 0.45) are typical of double bond character. 58 For both 3−Ni and 3−Pd, ε is very close to zero. The highly nickel localized nature of NLMOs 115−118 and 129 of 3−Ni strongly suggests that these ellipticities are not indicative of the higher, i.e., triple bond order, and this is supported by the QTAIM delocalization indices δ(U,TM), which are measures of bond order and which are below 1 for both 3−Ni and 3−Pd. Table 4 also provides four further U−TM bond order metrics. All of these agree that the U−Ni bond order is less than 1 and that that of U−Pd is smaller than for the 3d analogue. The lower bond orders found for the Pd system are in keeping with the composition of the NLMOs (Table 3). Specifically, the TMbased dσ and dπ orbitals are even more localized on the transition metal in 3−Pd than in 3−Ni, leading to reduced U− TM covalency. Although metal−metal bonding interactions typically increase down a transition-metal group, 59 the present NLMO data are consistent with the electronegativities of Ni, Pd, and U, 1.91, 2.20, and 1.38, respectively, on the Pauling scale. 60 The more electronegative 4d element has a more polar interaction with the actinide than does Ni, leading to reduced bond order.
To the best of our knowledge, there are no comparable computational analyses of the bond order in zero oxidation state group 10 complexes containing a homobimetallic metal− metal bond, so we carried out our own calculations on a previously reported low oxidation state system with an  61 Unfortunately, while geometry optimization of the 3d system proceeded smoothly, that for the Pd dimer did not, collapsing to a nonsensical solution. Thus, comparative M−M bond data are not available.
In summary, we conclude that the U−Ni and U−Pd interactions have topological features typical of metal−metal bonds. Analysis of the localized orbital structure locates MOs of σ and π symmetry between the actinide and the transition metals, but these are heavily polarized toward the latter, resulting in small orders. 3−Pd features consistently smaller bond orders than 3−Ni, in agreement with greater TM σ and π NLMO localization.
In order to probe the extent to which the U−TM interaction is a function of the geometric constraints placed on the metal atoms by the bidentate ligand framework, we have optimized the geometries of the model compounds 3−Ni(m) and 3− Pd  Table 4, which are all smaller (in an absolute sense) than in 3−Ni and 3−Pd. As with the full molecules, all of the bond orders are smaller in the model Pd system than the Ni one.
The data on these model compounds therefore indicate that the very short uranium−TM bonds observed in 3−Ni and 3− Pd are partly a function of the ligand framework. In the absence of constraining ligands, the 5f−nd bonds lengthen, though uranium−TM interactions are clearly still present. The bond orders in the unchelated compounds are, in general, a little more than half of those calculated for 3−Ni and 3−Pd.
Derivatization of the Bimetallic Compounds. In addition to the variation of the d-metal center, we investigated

Journal of the American Chemical Society
Article the effect on the metal−metal bond of exchanging the iodide for other ligands, focusing on the smaller Ni because of its stronger U−Ni bond. For this purpose it seemed reasonable to substitute the large, polarizable iodide for the more electronegative and strongly bonding fluoride. In order to differentiate between electronic and steric effects, we also included a sterically demanding and hard O-donor ligand trimethylsiloxide (OSiMe 3 ). Treatment of 3−Ni with sodium trimethylsiloxide yields the corresponding uranium(IV) siloxide compound 4 (Scheme 3) and sodium iodide. The reaction of 3−Ni with cesium fluoride results in the elimination of cesium iodide to give the F−U IV −Ni 0 complex 5 (Scheme 3). Reaction monitoring via 1 H and 31 P NMR spectroscopy shows quantitative formation of 4 and 5, respectively, within 24 h. Attempts to use silver fluoride instead of cesium fluoride resulted in decomposition of the bimetallic species and release of an oxidized nickel(II) complex Ni II (OAr P -κ 2 O,P) 2 (6), which was characterized crystallographically (see the SI). Adaptation of a published preparation allowed 6 to be prepared independently by reaction of 2 equiv of HOAr P with 1 equiv of Ni(cod) 2 in toluene (see the SI). 62 Reactions of 3−Ni aimed at the formation of a cationic compound using silver tetraphenylborate or potassium tetraphenylborate gave 6 or no conversion, respectively (see the SI). A bis(trimethylsilyl)amido derivative of 3−Ni was also targeted, but the reaction between 3−Ni and potassium bis(trimethylsilyl)amide did not show any conversion (see the SI).
Dark red crystals of 4 suitable for X-ray crystallography were grown from a benzene/hexane solution at ambient temperature. Single crystals of 5 were obtained from a benzene-layered THF solution. The two bimetallic complexes feature U−Ni bond distances of 2.556(1) Å (4) and 2.520(1) Å (5) (Figures   (4) and Fluoride Derivatives (5) Journal of the American Chemical Society    Figure 7 to highlight the C 3propeller shape and similarity of the overall structures.

Scheme 3. Preparation of Bimetallic Trimethylsiloxide
According to NMR spectroscopy, both 4 and 5 differ significantly in solution from the parent compound 3−Ni. Having studied the NLMOs of the complexes involved and not found any significant differences, we attribute the large chemical shift difference to two factors. First, the extensive electronic differences of I vs F/TMSO. Second, the slightly shorter Ni−P bonds in 4 than 5 that indicate stronger Ni−P backbonding. Both of these would combine to enhance the through-bond paramagnetic influence from the f 2 -uranium center. Resonances for the heteronuclei 19 F and 29 Si could not be observed within the spectral range from −740 to 620 and −1050 to 870 ppm, respectively. VT NMR experiments of 4 and 5 show that both have rigid structures in solution as high as 100°C. This represents a significantly higher energy barrier to the interconversion of Δ and Λ isomers compared with 3, supporting the ITI-induced stronger U−M bond being formed when the more electronegative X-ligands are uranium-bound. The replacement of the iodide changes the magnetic moment values from 1.9 μ B (3−Ni) to 2.8 μ B (4) and 2.1 μ B (5), respectively. Similarity of magnetic moment throughout series of U(IV) aryl oxide and amide complexes with different halides has been reported. 71−73 The value of the trimethylsiloxide derivative 4 is similar to that of other R 3 SiOligated uranium complexes, U(OSiBu 3 t ) 4 (2.83 μ B ) and U(OSiMe 3 ) 2 I 2 (bipy) 2 (bipy = 2,2′-bipyridine) (2.7 μ B ). 65,74 To account for these increases, and the strongly paramagnetically shifted 31 P chemical shifts for 4 and 5, we compared the composition of the two f-based NLMOs for the iodide (3−Ni) and fluoride (5) to look for different U 5f contributions that would lead to larger paramagnetic shifts for the Ni-bound atoms. For the iodide, they are 99.18% U (99.56% 5f) and 94.81% U (97.24 5f, 2.05 s, 1.75% total contribution from P) (see also Table S4 of the SI for a comparison of the predominantly Ni d-σ and π NLMOs). For the fluoride they are 99.22% U (99.68% 5f) and 92.54% U (97.58 5f, 1.63 s, 2.48% total contribution from P). While these data indicate a marginally greater through-bond mixing of unpaired 5f electron with the phosphorus, there is really very little difference between the two systems.
The electronic absorption spectra of toluene solutions of the compounds 2−5 were recorded to locate potential metal− metal charge transfer bands (see the SI). UV−vis spectra of monometallic 2 show several weak U(IV) f−f transitions 75 up to ca. 700 nm and more intense π−π* charge-transfer processes below 500 nm. 76,77 The second metalation to form 3−Ni, 3− Pd, and 3−Pt causes a bathochromic shift of the predominately ligand-based absorption at short wavelengths. In the visible region, the nickel derivative 3−Ni differs significantly from 3− Pd and 3−Pt, with an absorption at 511 nm (ε 598 M −1 cm −1 ) that is much stronger than that in 4 and 5, with weaker absorptions at 527 nm (ε 103 M −1 cm −1 ) and 533 nm (ε 90 M −1 cm −1 ), respectively. In the NIR region the monometallic complex 2 shows several absorption bands in the 850−2060 nm region (ε 18−45 M −1 cm −1 ). The NIR spectra of 3−Ni, 3−Pd, and 3−Pt are similar to each other but feature fewer absorption bands. As such, an unambiguous assignment of absorptions in this region to a metal−metal charge transfer appears to be not possible.

■ ELECTROCHEMISTRY
The electronic structures of complexes 2−5 were investigated using a range of voltammetric techniques ( Figure 8 and the SI).
In the electrochemical window provided by THF/[ n Bu 4 N]-[BPh 4 ], a single reduction process was observed for 2 during the cyclic voltammetry (CV) experiment, at E p c −2.87 V versus Fc + /Fc, assigned to the U(IV)/U(III) couple and it is irreversible. The U(IV)/U(III) redox couple is sensitive to the ligand environment and has been reported in the range from −1.83 to −2.78 V for metallocene and acetylacetonate (acac) complexes. 78−80 Incorporation of the group 10 transition metal alters the electrochemistry significantly. The CVs of 3−Ni, 3−Pd, and 3−Pt are qualitatively very similar, having three reduction processes each, suggesting a common electronic structure. The

Journal of the American Chemical Society
Article electrochemical behavior of 3−Pt will be described as a representative example. The first reduction (denoted I) at E p c −1.92 V is irreversible; two further quasi-reversible reduction processes are observed as overlapping cathodic waves at E p c −2.39 and −2.55 V, denoted II and III, respectively. Determination of the peak areas in the CV of 3−Pt reveals that the charge passed during reduction I is equal to that passed during II and III combined, indicating that process I is a twoelectron reduction, whereas II and III are single-electron reduction processes.
The electrochemical experiments with 3−Ni and 3−Pd generated identical conclusions, and in the series 3, the reduction potentials for all three processes are cathodically shifted when the transition metal is changed from Pt to Pd to Ni. The Kohn−Sham α spin LUMOs of both 3−Ni and 3−Pd are primarily U−M antibonding, so we ascribe reduction I to the filling of this orbital and conclude that the cathodic shift of the reduction process moving up the group 10 metals is due to a strengthening of the metal−metal bond. 81 This agrees with the computational results that showed higher bond order for 3−Ni compared to 3−Pd and also correlates with the shorter M−M′ distance determined crystallographically.
The CV of 5, the fluoride analogue of 3−Ni, shows only an irreversible reduction at E p c −2.39 V, but square-wave voltammetry (SWV) reveals a second process at the edge of the electrochemical window, E p c −2.81 V. This implies that 5 displays similar electrochemical behavior to the iodo complexes, albeit at more negative potentials; i.e., replacing the iodide with a fluoride strengthens the metal−metal bond trans to it.
The replacement of iodide with siloxide to make 4 cathodically shifts the reduction I further still, to E p c −2.50 V; no other reduction processes are observed by CV or SWV. This reduction is now quasi-reversible and suggests that the reduced species is stabilized to a certain degree. A reversible oxidation process is also observed at E 1/2 −0.20 V, denoted IV. The area of the anodic CV wave for IV is approximately equal to the area of the cathodic wave for reduction I, and both processes therefore involve two electrons. It is not known whether this oxidation IV is unique to 4; it may be that this oxidation is possible for all but lies outside of the electrochemical window.
Thus, we infer from the electrochemical data that the metal− metal bond strength increases in the series 3−Pt < 3−Pd < 3− Ni < 5 < 4.

■ CONCLUSIONS
The use of a relatively rigid heterobidentate phosphinoaryl oxide ligand that forms strong U−O bonds and weak, labile U− P bonds in the new complex IU IV (OAr P -κ 2 O,P) 3 has allowed the systematic incorporation of Ni(0), Pd(0), or Pt(0) via phosphine coordination, and the replacement of the iodide anion with Me 3 SiO − or F − , to form a set of five heterobimetallic U−M complexes XU IV (μ-OAr P -1κ 1 O,2κ 1 P) 3 M 0 (X = I, OSiMe 3 , F; M = Ni, Pd, Pt), all of which have shorter An−TM bonds than any previously reported example. The synthesis of a complete set of adducts from a single group for the first time and the solution and solid-state structural characterization of the complexes have enabled a thorough study of the uranium− metal bond. The U−I bond length in the starting material 2 becomes significantly shorter upon formation of the U−M complexes 3, but the coordination number changes from 7 to 5 (replacing three phosphines with one metal center), so inferences of the inverse trans influence (ITI) cannot be made here. Upon introduction of the group 10 metal centers, the magnetic moment decreases from 2.4 μ B to around 1.9 μ B in 3, respectively, an opposite change in moment to that reported upon secondary metalation of U complexes by Co I as a donor. 10 Although the changes in magnetic moment and UV−vis− NIR spectra cannot yet be interpreted in terms of bonding trends in the series, the combination of experimental electrochemistry and computation is particularly informative. A cathodic shift of the first reduction process observed upon moving from U−Pt up to U−Ni indicates a strengthening of the metal−metal bond in the order 3−Ni > 3−Pd > 3−Pt. This correlates with the shortening of the internuclear distance determined crystallographically. Natural population analysis and natural localized molecular orbital compositions indicate that U employs both its 5f and 6d orbitals in covalent bonding to a significant extent, and this agrees with experimental data that the oxidation states of the metals are best described as U(IV) and zero for the group 10 atoms. Quantum theory of atoms-in-molecules analysis yields bond critical point properties in keeping with many previous studies of transition-metal− metal bonds in both bulk metals and polynuclear clusters (relatively low electron density, positive ∇ 2 ρ, and negative H).
Replacing the uranium-bound iodide trans to the nickel center with the more electronegative fluoride and siloxide also results in NMR spectroscopic and electrochemical responses consistent with a strengthening of the U−Ni bond and with the existence of an ITI. If an ITI is influencing the M−M′ bond strength, then this is also borne out by the crystallographic data for 3 and 5, which show a shorter U−Ni bond in the F−U−Ni (5) than in I−U−Ni (3) complexes. Despite the short U−M distances, the bond orders are calculated by five different approaches to be small; less than 1 in all cases. All bond order metrics are smaller for U−Pd than U−Ni, in agreement with the electrochemical and QTAIM bond critical point data and with population analysis of the U−TM σ and π NLMOs which, while heavily localized on the TM in both cases, are even more so for the 4d system than the 3d, in keeping with the larger electronegativity difference between U and Pd vs U and Ni. Calculations on a monodentate analogue of 3 show that in the absence of the constraining ligand geometry there is clearly still a U−TM interaction, but it is enhanced by about 0.25 Å in the constraining ligand framework.
Thus, by combining the spectroscopic, computational, electrochemical, and structural studies, the U−M bond strength can be placed in increasing order: ■ EXPERIMENTAL SECTION General Details. All manipulations were carried out under a dry, oxygen-free atmosphere of nitrogen using standard Schlenk and glovebox technique. Benzene was distilled from potassium and stored over 4 Å molecular sieves. Hexane, THF, and toluene were degassed and purified by passage through activated 4 Å molecular sieves or activated alumina towers and stored over 4 Å molecular sieves. Deuterated solvents, benzene-d 6 and toluene-d 8 , were boiled over potassium, vacuum-transferred, and freeze−pump−thaw degassed prior to use. 1 H, 13 C, 19 F, 29 Si, and 31 P NMR spectra were recorded on Bruker AVA400, AVA500, or PRO500 spectrometers at 300 K. Variable-temperature NMR spectra were recorded on a Bruker AVA400 spectrometer between 300 and 370 K. Chemical shifts are reported in parts per million, δ, referenced to residual proton resonances, and calibrated against external TMS. Magnetic moment Journal of the American Chemical Society Article values were determined by Evans' method using a sealed benzene-d 6 capillary as reference. 82−85 UV−vis−NIR spectra were recorded on a JASCO V-670 spectrophotometer using a sealed quartz cuvette with 0.02−5 mM toluene solutions. Artifacts at 1650−1750 nm relate to solvent absorption. Electrochemical measurements were made on 1− 10 mM of the analyte in 12 cm 3 THF, 0.1 M [ n Bu 4 N][BPh 4 ], in a N 2filled glovebox using an Autolab ECO Chemie PGSTAT potentiostat, glassy-carbon disk (d = 3 mm) working electrode, Pt-gauze counter electrode, Ag-wire quasi-reference electrode, and ferrocenium/ ferrocene (Fc + /Fc = 0 V) standard. 86 Scan details are in the SI, and data were processed using GPES Manager 4.9. Elemental analyses were carried out at London Metropolitan University, London, UK, and Pascher Labor, Remagen, Germany. UI 4 (dioxane) 1.5 , 87 HOC 6 H 2 -6-Bu t -4-Me-2-PPh 2 , 32 and AgBPh 4 88 were prepared according to published procedures. All other reagents were from commercial sources and used as received.
Single-point calculations, at the Gaussian-optimized geometries, were run in the ADF code in order to obtain Mayer 99 and Gopinathan−Jug 100 bond orders. For these calculations, the zerothorder regular approximation (ZORA) Hamiltonian was used. Slatertype orbital ZORA basis sets of TZP quality were used for U, Ni, Pd, and I, with DZP ZORA basis sets for all other atoms. The frozen core approximation was employed, with U(5d), I(4p), Pd(3d), Ni(2p), P(2p), and 1s for all other atoms, bar H. The default SCF convergence criteria were used, together with an integration grid of 4.5.