Metal–organic layers stabilize earth-abundant metal–terpyridine diradical complexes for catalytic C–H activation

Metal–organic layers stabilize FeII or CoII-terpyridine diradical complexes to catalyze alkylazide Csp3–H amination and benzylic C–H borylation, respectively.


Introduction
Over the past two decades, metal-organic frameworks (MOFs) have attracted great interest among scientists and engineers owing to their potential in various applications including gas storage and separation, 1-6 heterogeneous catalysis, 7-16 nonlinear optics, 17,18 chemical sensing, [19][20][21] biomedical imaging, 22,23 and drug delivery. 24,25 In particular, MOFs have provided an excellent platform for designing single-site solid catalysts for many important organic transformations. [26][27][28][29][30][31][32] By shutting down intermolecular deactivation pathways via spatial isolation of active sites, MOFs have afforded turnover numbers (TONs) several orders of magnitude higher than their homogeneous analogs. 26,29 The catalytic performance of MOFs is, however, still limited by the diffusion rates of large substrates and products within the 3D frameworks. 33 Although many strategies have been devised to overcome this diffusion limitation of MOFs, for example, by elongating functional ligands 26 or diluting them with catalytically inactive spectator ligands to construct MOFs with larger channels and pores, 34 only moderate success has been achieved to date. MOFs constructed from elongated ligands tend to suffer from interpenetration as well as framework distortion, whereas MOFs built from mixed functional and spectator ligands have diminished atom efficiency.
We recently showed that diffusional constraint of MOFs could be lied by reducing one dimension of the MOF crystals to only a few nanometers in thickness to afford a new category of 2D materials, metal-organic layers (MOLs). 35 Unlike 3D MOFs, the active sites in ultrathin 2D MOLs are readily accessible to substrates during catalytic reactions. On the other hand, MOLs still inherit the heterogeneous nature, ordered structure, and molecular tunability of MOF catalysts, [36][37][38] and have the potential to provide a rare 2D molecular material platform for designing a new class of single-site solid catalysts without diffusional constraints. We report here the synthesis of a new metal-organic layer, TPY-MOL, based on Hf 6 (m 3 -O) 4 (m 3 -OH) 4 (HCO 2 ) 6 secondary building units (SBUs) and 4 0 -(4-carboxyphenyl)-[2,2 0 :6 0 ,2 00 -terpyridine]-5,5 00 -dicarboxylate (TPY) bridging ligands and the metalation of TPY ligands in TPY-MOL with CoCl 2 and FeBr 2 to afford highly effective recyclable and reusable MOL catalysts for challenging benzylic C-H borylation and intramolecular sp 3 C-H amination reactions (Fig. 1). Spectroscopic and computational studies identied unprecedented Co II /Fe II -terpyridine diradical complexes as catalytic active sites for the borylation and amination reactions.
Owing to their distinct coordination, redox, and photophysical properties, terpyridines (tpy) and their metal complexes have been explored for potential applications in many elds, including polymer science, 39,40 optoelectronics, 41,42 medicinal chemistry, 43,44 nanotechnology, 45 and molecular catalysis. 41,46,47 Although tpy derivatives provide a potentially interesting ligand platform for designing earth-abundant metal catalysts, few examples have been reported in the literature, [47][48][49][50] in part due to their strong propensity to undergo disproportionation reactions to form catalytically inactive M(tpy) 2 complexes. 48,49 Installation of bulky groups on the 6,6 00 -positions of tpy could prevent such bimolecular deactivation processes in M-tpy catalysts but oen at the expense of their catalytic activities. 48 By incorporating a tpy derivative into the TPY-MOL, we effectively shut down the disproportionation decomposition pathway without relying on steric protection at the 6,6 00 positions and obtained highly effective MOL catalysts based on Mtpy complexes (M ¼ Co or Fe) for benzylic C-H borylation and intramolecular sp 3 C-H amination reactions. The MOL-based M-tpy catalysts displayed at least 20 times higher catalytic activity and distinct chemoselectivity in benzylic C-H borylation reactions and 50 times higher TONs in intramolecular sp 3 C-H amination reactions over their homogeneous analogs.

Results and discussion
Synthesis and postsynthetic metalation of TPY-MOL TPY-MOL was synthesized in 76% yield by heating a mixture of HfCl 4 , H 3 TPY, and formic acid in DMF and water at 120 C for 24 h. The PXRD pattern of TPY-MOL matched the simulated pattern based on the (hk0) reections only that are characteristic of 2D MOL structures and aligned well with that of isostructural BTB-MOL (BTB is 1,3,5-benzenetribenzoate, Fig. 2a). 35 Transmission Electron Microscopy (TEM) images showed ultra-thin lms of TPY-MOL whereas the high resolution TEM (HRTEM) images of TPY-MOL showed a clear lattice  with the dark spots corresponding to Hf 6 clusters ( Fig. 2b and c). The distances between adjacent spots on the HRTEM image (20.1Å) matched well with that between two adjacent Hf 6 SBUs (20.0Å) in the MOL structural model. Atomic Force Microscopy (AFM) images of TPY-MOL indicated monolayer thickness for many nano-sheets with an average measured thickness of 1.2 nm, corresponding to the van der Waals size of Hf 6 SBUs ( Fig. 2d and e).

Co-TPY-MOL catalyzed benzylic C-H borylation
We rst investigated C-H borylation of m-xylene by Co$TPY-MOL. Organoboronic compounds are a useful class of intermediates for forming carbon-carbon and carbon-heteroatom bonds through coupling reactions. C-H borylation with boron reagents such as B 2 pin 2 is one of the most direct and convenient methods for the synthesis of organoboronic compounds. Although C-H borylation with arenes has been developed in the past two decades, benzylic C-H borylation is still rare (Table S7, ESI †). 27,51-56 Upon activation with NaEt 3 BH, CoCl 2 $TPY-MOL (0.5 mol%) catalyzed m-xylene borylation with B 2 pin 2 at 100 C over 3 days to afford 42% yield of borylated products, with a 4.2 : 1 selectivity favoring the benzylic position ( Table 1, entry 1). The borylated products were obtained in 95% yield with a slightly higher selectivity for benzylic borylation (4.6 : 1) when the catalyst loading increased to 1.0 mol% ( Table 1, entry 2). The activation of CoCl 2 $TPY-MOL with NaEt 3 BH is necessary for the borylation reaction (Table 1, entry 3). Under identical conditions, a TPY-MOF control, which is isostructural to the previously reported BTB-MOF in which 2D layers stack in a staggered arrangement to result in a 3D MOF, 35 gave no conversion, likely due to slow diffusion of the substrates and products (Table 1, entry 4). The homogeneous analog gave 2% borylated products with a 5.7 : 1 selectivity favoring the arene C-H bond ( Table 1, entry 5). Such moderate arene borylation activity was recently reported for homogenous tpy-Co derivatives. 49 Active site isolation in MOLs thus not only increases the TON by more than 20 times (over the homogeneous analog) but also afforded unusual selectivity of borylation for the benzylic C-H bond.
We further investigated the substrate scope for Co(THF) 2 -$TPY-MOL catalyzed C-H borylation reactions. Benzylic borylated products were produced exclusively for p-xylene, 1-t-butyl-4-methylbenzene, and mesitylene in >90% yields ( Table 2, entries 2-4). For p-methoxytoluene, a high selectivity of 59: 6: 1 was obtained for the benzylic borylated product (   a NMR yield based on CH 3 NO 2 as an internal standard. b 1.0 mol% Co. c Without the addition of NaEt 3 BH. 5). For toluene, borylated products were obtained in 92% yield, but the selectivity for the benzylic borylation product was moderate ( Table 2, entry 6). These results indicate the inuence of steric hindrance on the selectivity of benzylic vs. aromatic borylation by Co(THF) 2 $TPY-MOL. Co$TPY-MOL was recovered and used for at least 10 times without any loss of activity in C-H borylation of p-xylene (Fig. S32, ESI †). We conducted several tests to demonstrate the heterogeneity of Co$TPY-MOL. First, we showed that the PXRD of Co$TPY-MOL recovered from C-H borylation of p-xylene remained the same as that of freshly prepared Co$TPY-MOL (Fig. S33, ESI †). Second, we used ICP-MS to show that the amounts of Co and Hf leaching into the supernatant during the C-H borylation of p-xylene were only 0.092% and 0.037% respectively. Finally, we observed that the removal of Co$TPY-MOL from the reaction mixture aer several hours stopped the C-H borylation of p-xylene (Scheme S2, ESI †).

Identication of the Co(THF) 2 $TPY-MOL catalyst
We studied the catalytically active species by hydrogen quanti-cation, infrared (IR), UV-Vis-NIR, XPS, and electron paramagnetic resonance (EPR) spectroscopy, XANES, EXAFS, and density functional theory (DFT) calculations. One equiv. of H 2 was generated upon treatment of CoCl 2 $TPY-MOL with NaEt 3 -BH, suggesting the formation of Co(THF) x $TPY-MOL via reductive elimination of H 2 from the putative CoH 2 $TPY-MOL intermediate. This 2-electron reduction process was also conrmed by titration of Co(THF) x $TPY-MOL with ferrocenium hexauorophosphate which resulted in the generation of two equiv. of ferrocene w.r.t to CoTPY-MOL (Fig. S6, ESI †). IR spectra showed no characteristic band of N^N, ruling out the coordination of dinitrogen to Co. XANES analysis indicated +2 oxidation state for the Co center (Fig. 3a). This oxidation state assignment was further supported by XPS spectroscopy which gave a Co 2p 3/2 binding energy of 781.2 eV with the expected shake-up peak for the Co II centers (Fig. 4).
The EXAFS spectra at the Co K-edge were well tted with a structural model in which Co coordinates with three N atoms of TPY and two THF molecules (Fig. 3e) , but similar Co-N bond distances to a reported low-spin Co II (tpy)(BH 4 ) complex with the (tpyc) À ligand (Co-N c ¼ 1.810Å, Co-N t ¼ 1.925Å). 58 The Co-N bond distance analysis thus supports the formulation of the Co II -(tpycc) 2À electronic structure for Co(THF) 2 $TPY-MOL.
We used UV-Vis-NIR spectroscopy to discern the diradical nature of TPY ligands in CoTPY-MOLs (Fig. 5). Co(THF) 2 $TPY-   MOL exhibited two intense, broad bands centered at 552 and 759 nm and a weak but broad band at 1105 nm, indicative of p to p* and p* to p* transitions for the reduced tpy ligand. [59][60][61][62][63] In contrast, these bands are absent in CoCl 2 $TPY-MOL with the neutral TPY ligand (Fig. 5). The proposed (tpycc) 2À species was previously observed in reduced M(tpy) 2 complexes, such as Cr III (tpy) 2 , V IV (tpy) 2 , and Ti IV (tpy) 2 , by Wieghardt and coworkers. 62,63 However, we are not aware of any example of Mtpy complexes featuring the (tpycc) 2À species. Our XANES, EXAFS, and XPS results clearly indicate the Co II oxidation state for Co(THF) 2 $TPY-MOL whose electronic structure is best described as Co II (THF) 2 $(TPYcc) 2À -MOL. The (tpycc) 2À diradical dianion can have either a singlet (S ¼ 0) or a triplet (S ¼ 1) ground state, which can potentially be experimentally differentiated by EPR spectroscopy. EPR spectroscopy of Co(THF) 2 $TPY-MOL gave an isotropic signal with g iso ¼ 2.003 at r.t. in toluene suspension. The same MOL sample frozen at 20 K exhibits a stronger isotropic signal with g iso ¼ 2.003, con-rming that the same species was detected at r.t. and 20 K (Fig. 6). More interestingly, the g value falls in the range of 2.003-2.005, 59,64,65 where radicals in extended organic p systems were oen observed. The EPR signal intensity was temperaturedependent, which can be tted with the Bleaney and Bowers equation 66 typically used for organic diradicals (Fig. 6). The tting of temperature-dependent EPR signals indicates that the (TPYcc) 2À diradical has a singlet ground state with singlet-totriplet energy gap of 0.04 kcal mol À1 . The observed EPR signal is thus attributed to the thermally populated TPY triplet excited state. 67 Moreover, a weak signal g iso z 2.04 was observed at 20 K, consistent with low-spin Co II centers. Therefore, our EPR data provide strong support to our proposed electronic structure Co II (THF) 2 $(TPYcc) 2À -MOL. We have ruled out the possibility of SBU-based free radicals because TPY-MOL treated with NaEt 3 -BH exhibited no signal at r.t. or 20 K (Fig. S16, ESI †).
Density functional theory (DFT) calculations and natural population analyses with the B3LYP/6-311G(d) basis set on Co(THF) 2 $tpy gave a doublet ground state (GS) with high positive charge distribution (1.24) on the Co center and negative charge distribution (À1.34) on tpy (Table S9, ESI †). A comparison charge distribution on CoCl 2 $tpy revealed that the Co center in Co(THF) 2 $tpy maintains +2 oxidation state. A Mulliken spin population analysis and spin density plot revealed that 0.996 unpaired electron resides on the Co center, affording a ground state with a low-spin Co II , d 7 doublet (S Co ¼ 1/2) and a tpy diradical dianion singlet (S tpy ¼ 0) (Fig. S47, ESI †). The singlet tpy diradical dianion is not expected to give any EPR signal. Interestingly, the energy of quartet state of Co(THF) 2 $tpy is calculated to be only 0.40 kcal mol À1 higher than that of the doublet GS. This small energy gap is consistent to that deduced from temperature-dependent EPR signals of Co(THF) 2 $tpy. The charge distribution of the quartet state is similar to that of the doublet GS with positive charge (1.29) on the Co center and negative charge (À1.40) on tpy (Table S9, ESI †). The calculated bond distances are similar between the quartet state and the doublet GS (Table S11, ESI †). A Mulliken spin density population and spin density plot of the quartet state revealed the residence of the 1.091 unpaired spin on Co center and 1.887 unpaired spins on tpy, affording a low-spin Co II , d 7 doublet (S Co ¼ 1/2) and a tpy triplet diradical dianion (S tpy ¼ 1) (Fig. 7). The energetically accessible low-lying triplet excited state of (tpycc) 2À was previously proposed for the hypothetical [Zn II (tpy 2À )(NH 3 ) 2 ] 0 . 62 DFT calculations thus support the origin of the experimental tpy diradical dianion EPR signal as thermally populated quartet state of Co II (THF) 2 $tpycc. Moreover, we believe that conjugation of Hf 6 SBU to TPY can further stabilize TPY diradical dianion and lower the energy difference between doublet and quartet states of Co II (THF) 2 $TPYcc-MOL.
We also investigated the activation of CoCl 2 $tpy molecular complex with NaEt 3 BH. Upon treating CoCl 2 $tpy in THF with 10 equiv. of NaEt 3 BH, the mixture turned dark green immediately with concomitant formation of Co nanoparticles as black precipitate ( Fig. S7 and S9, ESI †). The solution was ltered through Celite and evaporated to afford Co(tpy) 2 (HR-MS calculated for C 30

Mechanistic studies of Co(THF) 2 $TPY-MOL catalyzed C-H borylation
To gain insight into the mechanism of the C-H borylation reaction, we carried out several experiments. First, we performed kinetic isotope effect (KIE) studies in order to afford information on the rate-determining step of the C-H borylation  reactions. The initial rates of C-H borylations with p-xylene and p-xylene-d 8 were determined by running parallel reactions in separate vessels, and the comparison of the initial rates gave a KIE value of 2.7 (Scheme S3, ESI †). Such a primary KIE indicates the involvement of the C-H bond breaking in the ratedetermining step.
Second, we detected the presence of HBpin by gas chromatography-mass spectrometry (GC-MS) at the end of the C-H borylation reactions. Third, we determined the resting state of the catalyst by EXAFS studies. By treating Co(THF) 2 $TPY-MOL with 20 equiv. of B 2 pin 2 , we obtained the Co(Bpin) 2 $TPY-MOL product in which Co coordinates to three N atoms of TPY and two Bpin groups according to EXAFS tting (Fig. S13, ESI †). To determine the resting state of the catalyst, the C-H borylation reaction was stopped at 70% conversion and the organic volatiles were evaporated. EXAFS studies indicated that the remaining residue had the same structure as Co(Bpin) 2 $TPY-MOL (Fig. S14, ESI †). Finally, EPR spectra of Co(Bpin) 2 $TPY-MOL did not show any signals corresponding to a TPY-based radical EPR signal (Fig. S16, ESI †), suggesting a typical Co II $TPY complex with negative charge localized on the Bpin ligands.
On the basis of these experimental and calculation results, we propose a catalytic cycle for the C-H active borylation of methylarenes as shown in Scheme 1. The CoCl 2 $TPY-MOL(I) is activated by NaEt 3 BH in THF to give the CoH 2 $TPY-MOL(II) intermediate, which quickly undergoes reductive elimination of H 2 to produce the Co II (THF) 2 $(TPYcc) 2À -MOL(III) catalyst. Oxidative addition of B 2 (pin) 2 to III results in Co(Bpin) 2 $TPY-MOL(IV), which is the catalyst resting state for the C-H borylation reactions. s-Bond metathesis between IV and methylarene proceeds as a rate-determining step to form Co(H)(Bpin)$TPY(V) and the benzylic borylated product. The reaction of V with B 2 pin 2 regenerates the intermediate IV and forms HBpin as a byproduct via s-Bond metathesis. The transformation of V to IV could alternatively involve a two-step process of reductive elimination of HBpin from V followed by oxidative addition of B 2 Pin 2 to the intermediate to form IV. We are not able to differentiate between the concerted one-step s-bond metathesis and the two-step reductive elimination/oxidative addition process.
DFT calculations and natural population analyses with the B3LYP/6-311G(d) basis set on Fe(THF) 2 $tpy gave a triplet GS with high positive charge distribution (1.29) on the Fe center and negative charge distribution (À1.39) on tpy (Table S10, ESI †). Spin density plot of the GS revealed that 2.013 unpaired electrons reside on the Fe center, affording an intermediate-spin Fe II , d 6 center (S Fe ¼ 1), and a tpy singlet diradical dianion antiferromagnetically coupled to each other (S tpy ¼ 0) (Fig. S51, ESI †). The GS of Fe(THF) 2 $tpy again is not expected to give any organic radical EPR signal, which contradicts our experimental results. We believe that the experimental tpy EPR signal comes from thermal population of the quintet state of Fe(THF) 2 $tpy which is only 5.26 kcal mol À1 higher in energy than that of triplet GS, consistent to our EPR analysis. The charge distribution of the quintet state is similar to Scheme 1 Proposed mechanism for the Co(THF) 2 $TPY-MOL catalyzed C-H borylation of arenes with B 2 pin 2 .
that of triplet GS with positive charge (1.34) on the Fe center and negative charge (À1.44) on tpy (Table S10, ESI †). A Mulliken spin population analysis and spin density plot revealed that 2.094 unpaired spins reside on the Fe center and 1.887 unpaired spins on tpy, affording an intermediate-spin Fe II , d 6 compound (S Fe ¼ 1), and a tpy triplet diradical dianion (S TPY ¼ 1) (Fig. 7), which is consistent with our experimental EPR results. The coordination of Hf 6 SBUs to TPY is expected to further stabilize TPY diradical dianion and lower the energy difference between triplet and quintet states of Fe II (THF) 2 $(TPYcc) 2À -MOL.
Upon activation with NaEt 3 BH, 2 mol% of FeBr 2 $TPY-MOL catalyzed intramolecular C sp 3 -H amination of 1-azido-4phenylbutane (1a) in the presence of two equivalents of ditert-butyl dicarbonate (Boc 2 O) at 90 C to form Boc-protected aphenyl pyrrolidine (2a) in 89% yield. This level of activity is 9 times as high as that of the MOF control (Table 3, entry 4). Under identical conditions, the homogeneous tpy-Fe catalyst only afforded the product in 3% yield, probably due to the deactivation of tpy-Fe catalyst via bimolecular pathways (Table  3, entry 5). Indeed, treatment of FeBr 2 $tpy with 10 equiv. of NaEt 3 BH produced a mixture Fe(tpy) 2 and Fe nanoparticles; such a disproportionation reaction was previously observed for a series of (PDI)FeBr 2 complexes. 69,70 A higher TON of 76 was achieved when the Fe loading was decreased to 1 mol% (Table 3, entry 2). With a much simpler ligand, Fe$TPY-MOL outperformed Betley's Fe-dipyrrinato homogenous catalyst by 13 times 71 and our recently reported NacNac-MOF catalysts by 4 times 28 in TONs. It is worth noting that FeBr 2 $TPY-MOL, without activation with NaEt 3 BH, showed low activity (Table 3, entry 3), suggesting that the formation of Fe-nitrene compound might be a key elementary step of the intramolecular C sp 3 -H amination reaction. [71][72][73][74][75][76][77] We further explored the substrate scope of intramolecular C sp 3-H amination reactions (Fig. 8). At 2 mol% catalyst loading and in the presence of 2 equiv. of Boc 2 O, the 2,2-dimethylpyrrolidine (2b) was formed in 57% yield. Due to reactivity of the vinyl substituent in 2c, 5 eq. of Boc 2 O was required to give modest yield at 2 mol% Fe. Since the MOL catalysts are free from diffusion constraints, substrates with a bulky substituent such as 3,5-diphenylphenyl was also tolerated and gave 75% yield at 5 mol% Fe and 2 eq. of Boc 2 O.
PXRD pattern of Fe$TPY-MOL catalysts recovered from C sp 3 -H amination reactions suggested that the integrity of the MOL was maintained under reaction conditions. ICP-MS of the supernatant showed <0.1% of Fe and <0.1% of Hf had leached into the supernatant. Furthermore, The Fe$TPY-MOL catalyst could be recovered and reused four times (Scheme S4, ESI †).

Conclusions
We have synthesized a terpyridine-based TPY-MOL and metalated TPY-MOL with CoCl 2 and FeBr 2 to generate M$TPY-MOL   catalysts for benzylic C-H borylation and C sp 3 -H amination reactions. Interestingly, M$TPY-MOL catalysts showed signicantly higher activity and different chemo-selectivity than homogeneous and MOF controls. Spectroscopic studies and DFT calculations indicated the formation of unprecedented MOL-stabilized M II -(TPYcc) 2À species featuring divalent metals and TPY diradical dianions. We believe that the formation of novel M II -(TPYcc) 2À (M ¼ Co or Fe) species endows them with unique and enhanced catalytic activities in C sp 3 -H borylation and intramolecular amination reactions. Our work demonstrates the ability to engineer MOLs as single-site solid catalysts without diffusional constraints and to elucidate intricate electronic structures of MOL-stabilized metal complexes.

Conflicts of interest
There are no conicts to declare.