Halide Comparisons and trends

(Electro)chemical N2-splitting and (electro)chemical mechanistic investigations are now performed on three starting platforms that only vary their halide ligand, which allows us to comment on results or trends that we found during this study.

Trends in N2-splitting abilities

The determined yields of 3^Br and 3^I are listed in Table 7, with yields for 3^Cl added for comparison. In general, 3^Br and 3^I can be obtained in moderate to high yields via chemical N2 -splitting.

Table 7. Spectroscopic yields of 3^Cl, 3^Br, and 3^I via (electro)chemical N2-splitting.^a

a Average yield over at least two runs, individual yields listed in the Experimental Section IV.5. - indicates the experiment was not performed. ^b yields determined via integration of all ³¹P{¹H}NMR peaks. ^c Yields determined via addition of an internal standard either via ³¹P{¹H} NMR or ¹H NMR spectroscopy.

The reduction potential of Cr(Cp*)2 (−1.47 V)⁵⁹ should regarding the reduction potential of 1^I not be sufficient to reduce this compound; (partial) precipitation of [Cr(Cp*)2]I likely adds the required driving force. The best yields for 3^Cl and 3^Br where obtained using Na/Hg, yet 3^I is only formed in low yields. The large overpotential when using Na/Hg to reduce 1^I (circa 0.7 V) likely leads to unproductive over-reductions. Homogeneous reductant Co(Cp*)2 is performing slightly worse for synthesis of 3^Cl and 3^Br (vs. Na/Hg), which can be the effect of very efficient halide removal by precipitation of NaCl or NaBr, respectively.

It was imagined that exchanging the halide ligands from chloride to bromide and iodide would influence the efficiency of N2-splitting, based on their different leaving group- and steric properties. However, it is concluded that for both electro- and chemical N2-activation, the nitride yields are basically independent of the halide of the starting platform. The differences within chemical N2-splitting (85 % 3^Cl and 80 % 3^Br using Na/Hg vs. 60 % 3^I for Co(Cp*)2), are comparably small and might be the result of the different nature of the reductant (heterogeneous vs. homogeneous). A heterogeneous reductant without a large overpotential might be more beneficial due to higher local concentrations and therefore allows less decomposition on the Re(II)-level. In future research, it would be interesting to examine a heterogeneous reductant in the potential range required for 1^I to examine this hypothesis.

For electrochemical N2-splitting, the yield for all three platforms is around 60 % and FE efficiency is around 50 %. In case of 1^I, a slightly better reaction outcome is observed when electrolysing at the peak potential of the second reduction. This CPE potential clearly triggers the 2e/comp over the 1e/dim pathway. The higher nitride yield might indicate that this first

pathway is more efficient, although in general a more negative CPE potential will speed up reactions,¹⁸⁸ and therefore maybe afford higher nitride yields.

Without observing a preferred halide concerning nitride yields, the iodide platform clearly overrules bromide and chloride. Moving to the heavier homologues in the halogen group, a 0.15 V less electronegative reducing agent or applied potential is required for N2-splitting.

Moving to these more benign conditions upon halide exchange is an important finding for future research.

Trends in redox potentials

During analysis of the three halide-substituted platforms, we encountered three sets of redox potentials. The redox potentials of fully reversible Re(VI/V) oxidation of 3^X (X = Cl, Br, and I), the Re(II/I) reduction of [ReCl(N2)(PNP)] and [ReI(N2)(PNP)], and the Re(III/II) reduction of 1^X. In the latter case, we experimentally only have access to a peak potential for 1^Br, but for 1^Cl and 1^I, the redox potential was obtained via digital simulation. Comparing these, we clearly see a less electron rich metal centre going down the halide group.

Table 8. Redox potentials of 1^X, [ReX(N2)(PNP)], and 3^X.

a – simulation not performed for the bromide platform

As elaborated in Section I.1.4, chloride, bromide, and iodide have basically the same Lever Parameter (EL Cl/I: −0.24 V, EL Br: −0.22 V), which implies halogen-substitution should barely shift the redox potential.⁶⁴ This is in contrast to the here shown results. However, the Re(VI/V) couple was not assessed by Lever, and the Re(III/II) couple is only examined on a limited database. Within this already limited database, it is basically impossible to examine for trends amongst the halide ligands (16, 3, and 0 complexes are examined containing Cl, Br, and I, respectively).⁶⁵ Therefore, we do not rely on the Lever Parameters to interpret these Re-trends.

Figure 34. Comparison to other compound series that show different redox potentials upon halide exchange by the groups of Pickett, Nishibayashi, and Deutsch.^52,163,189

Beyond the Lever Parameters, there seems to be literature precedent that the donor properties decrease going from Cl to I. For instance, the group of Pickett examined the redox characteristics of [MoX2(dppe)2] (dppe = 1,2-bis(diphenylphosphino)ethane), and find an increasingly milder reduction potential going down the group (E1/2 = −1.68 V (X = Cl), −1.51 V (X = Br), −1.34 V (X = I) vs. the saturated calomel electrode (SCE)), see Figure 34. These values should be handled with some care: although all reductions show a reverse peak, at lower scan rates they couple to halide loss and it is not specified at which scan rate these potentials are read off.¹⁸⁹. Another example is given by Nishibayashi, who established the five-coordinate Mo nitrides [MoNX(PNpyrP)] (PNpyrP = 2,6-bis(di-tert-butyl-phosphinomethyl)pyridine) with X = Cl, Br, I.

They share a very mild, reversible Mo(V/IV) oxidation revealing an anodic shift when going from Cl to I (E1/2 = −1.25 V (X = Cl), −1.18 V (Br), −1.14 V (I)). These shifts are strikingly similar to our 3^X series. It is also reported that these nitrides can be reduced, yet these waves appear irreversible and are substantially decreased in current compared to the one electron oxidations, and are therefore not over-interpreted.⁵² Deutsch and co-workers isolated Re(III) [ReX2(dmpe)2]⁺ (dmpe = 1,2-bis(dimethylphosphino)ethane) with X = Cl and Br, and both show two subsequent and reversible reductions with an anodic shift when going from Cl to Br (E1/2 = −0.92 V for Re^III/II and −2.06 V for Re^II/I (X = Cl), E1/2 = −0.80 V for Re^III/II and −1.92 V for Re^II/I (X = Br)). These reductions were actually used to derive the Lever Parameters, partially covering for the different reduction potentials, yet not to its full extend.

An explanation for this decreased electron donating property when going down the halide group is summarised by Lautens and Fagnou, focusing on the interaction between the halide ligands and the metal.¹⁹⁰ This interaction consists of a ligand to metal σ-donation, which increases down the halide group based on decreasing electronegativity. This is accompanied by a π-donation from the ligand into empty metal d-orbitals, which decreases going down the halide group (Figure 35). When this latter interaction dominates (when a d-orbital of appropriate symmetry

is vacant), iodide is the least donating ligand within the halide series as here examined. In both Re(III) 1^X and Re(V) 3^X, the d-electron count could allow an appropriate d-orbital to accept π-donation. To substantiate this rather qualitative picture, Odom developed the Ligand Donor Parameters (LDP) for a series of monoanionic ligands, as quantified via the rotational barrier of the amide groups in [CrN(NiPr2)2X] as determined via Spin Saturation Transfer NMR spectroscopy. The stronger the ligand X donates, the lower the rotational barrier for the amide is, leading to LDP’s in the unit of energy per mole. For the here examined halides, they find decreasing donating properties in the order Cl>Br>I (LDP in kcal mol^-1 = 15.05 ± 0.29 (Cl), 15.45 ± 0.30 (Br), 15.80 ± 0.30 (I)).¹⁹¹ This could provide an explanation for the lower electron donating properties going to the heavier halide homologies. It should be noted that this system represents a high valent complex, which in our case would be only valid for the trend for 3^X. Nevertheless, this explanation line matches with our results as obtained herein.

Figure 35. Metal-halide σ- and π-interactions.

Another trend in the redox-potentials is shown in the N2-splitting mechanistic model, where two reductions are proposed, [ReX2(PNP)]^0/- and [ReX(N2)(PNP)]^0/-. For 1^Cl and 1^I, the reduction potentials could be obtained via digital simulation of the CV traces, and were additionally calculated by DFT (Table 9). The chemical difference between both reductions is a) different formal oxidation state: Re(III) for the first and Re(II) for the latter reduction, and b) the exchange of a donating halide ligand with a π-accepting N2-ligand. The balance between both phenomena cause for a potential inversion in case of bromide and chloride, and a peak deconvolution in case of iodide. This underlines the statement about the different donation properties of the halides: since chloride and bromide are more electron donating, the influence on exchanging one for a N2-ligand has a larger influence than the different formal oxidation state. The opposite must be the case for iodide, the decreased donor properties are less of an influence than the formal changed oxidation state.

Table 9. Simulated and calculated redox potentials of [ReX2(PNP)]^0/- and [ReX(N2)(PNP)]^0/-.

Visibility of dinuclear N2-bound 2^X in CV

The product of the proposed N2-binding, halogen loss and dimerization reactions is dinuclear 2^X and we probed for its presence in CV. There is precedent in literature that N2-bridging complexes with a (δ⁴)π¹⁰ configurations can be oxidised twice to the δ⁴π⁹ ([2^X]⁺) and δ⁴π⁸ ([2^X]²⁺) state. The corresponding potentials are (rather) mild, as was found for Cummins π¹⁰ system [{Mo(N(tBu)(3,5-C6H3(CH3)2))3}2(μ-N2)] (E1/2 = −1.46 V and −0.32 V (THF))²⁴, the δ⁴π¹⁰ system [{ReCl2(PNP^iPr)}2(μ-N2)] as found by Dr. F. Wätjen (né Schendzielorz) (E1/2 = −1.03 and −0.37 V (THF))⁷¹, and the tungsten π¹⁰ dimer [{W(N(CH2CH2N(4-tBuC6H4))3}2(μ-N2)] by Schrock (E1/2 = −1.63 and −0.37 V (THF)).¹⁹²

Figure 36. Exemplary CV traces of a 1.0 mM solution of 1^Cl in THF with 0.2 M (nBu4N)PF6 under an N2 atmosphere.

Measured three ranges, scanning through: a) only the oxidative area, b) the first reductive area, then the oxidative area, c) the whole reductive area, then the oxidative area.

Upon measuring CV of 1^X, we scanned a) in oxidative direction only, b) through the first reduction wave(s) which are proposed to productively form 2^X and subsequently the oxidative area, and c) through the whole reductive area and subsequently the oxidative area (see Figure 36 for the CVs of 1^Cl and A26 for the CVs of 1^Br and 1^I). This comparison allows us to identify if

k Calculations performed by Dr. M. Finger using: M06/def2-TZVP (SMD: THF) // D3(BJ)-PBE0/def2-TZVP, def2-SVP(C,H)

[ReX2(PNP)]^0/- [ReX(N2)(PNP)]^0/- E0 (sim) / V E0 (calc.) / V^k E0 (sim) / V E0 (calc.) / V

Cl −2.00 −1.98 −1.84 to −1.88 −1.94

I −1.67 −1.58 −1.72 −1.69

oxidative features of 2^X are present after scanning condition b. Unfortunately, no oxidative features are present that resemble two successive (reversible) oxidations. All new features up to a potential of 0 V are present in the CV following conditions of b and c, which indicates that the proposed oxidations of 2^X are not visible in the CV. Likely, the formation of 2^X is too slow to be observed in the CV in substantial quantities.

Another observation that triggered the search for 2^X in the CV data, is the appearance of an irreversible, small reduction under N2 at Ep = −2.4 V for 1^Cl, −2.3 V for 1^Br, and −2.2 V for 1^I (ν = 0.1 Vs^-1). This feature grasps attention in the ν-, N2- and concentration-dependent data that could hint its belonging to 2^X. In both before mentioned (δ⁴)π¹⁰systems by Cummins and Wätjen, an irreversible reduction is observed at Ep = −2.4 V (ν = 0.1 Vs^-1)^71,193. The calculated reduction potentials of 2^X are in the same range (Ep,calcd. =−2.61 V (2^Cl), −2.58 V (2^Br) and

−2.39 V (2^I))^k. When measuring 1^X ν-dependent under N2, this feature diminishes with increasing scan rate, strongly indicating that the corresponding species is a follow up product of a relatively slow chemical reaction following the first reduction(s). Secondly, upon increasing the N2-pressure, this reduction also decreases, indicating that its formation is in competition with a N2-binding reaction. Coordination of additional N2-ligands to e.g. [ReX(N2)(PNP)] or [ReX(N2)(PNP)]⁻ could block the free coordination site and prevent formation of 2^X. An explanation in this direction was successfully introduced in case of 1^Cl in Chapter II.1 to account for the N2-pressure data. The concentration dependent data of 1^I and 1^Br showed an increase of this far cathodic wave upon increasing concentrations, indicating the species forms via a bimolecular reaction. So far, this is in line with this reduction wave belonging to 2^X. In addition, due to the deconvolution of [ReI2(PNP)]^0/− and [ReI(N2)(PNP)]^0/− in case of 1^I, it is observed that the [ReI(N2)(PNP)]^0/− reduction wave disappears parallel with the increase of this far cathodic wave. At high concentrations, the 1e/dim is more favoured over the 2e/comp pathway, and it might be that more of the dinuclear species is formed. Yet, the relatively high concentrations of 2^X that would lead to these substantial peak currents of this far cathodic wave are in contrast to the here above described experiment, where no onset for the oxidative features of 2^X was observed. This remains therefore an unsolved question and in future research, either low temperature CV should be examined or chemical access to [2^X]⁺ and [2^X]²⁺ should be pursued.