Cleavage of the N≡N Triple Bond and Unpredicted Formation of the Cyclic 1,3‐Diaza‐2,4‐Diborete (FB)₂N₂ from N₂ and Fluoroborylene BF

Abstract

A complete cleavage of the triple bond of N₂ by fluoroborylene (:BF) was achieved in a low‐temperature N₂ matrix by the formation of the four‐membered heterocycle FB(μ‐N)₂BF, which lacks a trans‐annular N−N bond. Additionally, the linear complex FB=N−N=BF and cyclic FB(η ²‐N₂) were formed. These novel species were characterized by their matrix infrared spectra and quantum‐chemical calculations. The puckered four‐membered‐ring B₂N₂ complex shows a delocalized aromatic two‐electron π‐system in conjugation with the exo‐cyclic fluorine π lone pairs. This work may contribute to a rational design of catalysts based on borylene for artificial dinitrogen activation.

The four‐membered heterocycle FB(μ‐N)₂BF is obtained from fluoroborylene, FB, and N₂. It shows a complete cleavage of the N₂ triple bond and a delocalized aromatic two‐electron π‐bond conjugated with the exo‐cyclic fluorine π lone pairs, indicating the importance of fluorine‐specific interactions.

Introduction

The cleavage of the N≡N triple bond (one of the strongest chemical bond) is a long‐standing task in chemistry.^[1] Since the discovery of the Haber‐Bosch process for producing ammonia from H₂/N₂ in the first decade of the 20^th century, plenty of transition‐metal (TM) complexes have been discovered to activate and functionalize thermodynamically stable and kinetically inert dinitrogen (N₂) under more ambient conditions.^[2] For p‐block elements the examples of N₂ binding are mainly contributed from boron compounds such as borylenes (:BR).^[3] The boron atom in borylene possess both, a lone pair of electrons (HOMO) and an energetically low‐lying empty p valence orbital (LUMO). Borylenes are therefore excellent candidate for mimicing transition‐metal reactivity.^[4] The reaction of N₂ with free phenylborylene, :BPh, under matrix conditions has been shown to yield the linear adduct PhB=NN in a triplet ground state, underscoring the use of borylenes as candidates for N₂ binding (Scheme 1 a).^[4e] A boron‐based fixation of N₂ has very recently also been reported using a phenylborylene stabilized by a bulky carbene ligand, [(CAAC)DurB] (CAAC=cyclic alkylamino carbene, Dur=2,3,5,6‐tetramethylphenyl), leading to the end‐on bridging complex [{(CAAC)‐DurB}₂(μ ²‐N₂)]^[4b] in which the N−N bond length [1.248(4) Å] lies in the range of N=N double bonds (Scheme 1 b).^[5]

Scheme 1.

Five different binding modes of N₂ to borylenes.

The reaction of laser‐ablated boron atoms with N₂ molecules upon co‐deposition onto a cooled (4 K) CsI window has previously been studied.^[6] These studies provided a variety of mono and diboron nitrogen compounds of the type BN₂, B₂N, and B₂N₂ as well as linear NNBN.^[6] It has recently been shown, that addition of CO to the boron‐N₂ deposits gives rise to CO complexes of the BN₂ isomers, such as the chain‐molecules NNBCO and NBNCO, cyclic (η ²‐N₂)BCO, and the diisocyanat B(NCO)₂.^[7]

Results and Discussion

Here we report on novel fluoroborylene (:BF) : N₂ compounds, the cyclic diaza‐diborete FB(μ‐N)₂BF(A), its linear FB=N−N=BF isomer (B), and cyclic fluorodiazaboririne, FB(η ²‐N₂) (C) which was previously predicted^[8] (Scheme 1, c–e). They are selectively formed upon co‐deposition of laser‐ablated boron atoms with elemental fluorine in an N₂ gas stream at cryogenic temperatures (4±1 K, for experimental details see the Supporting Information). Here, dinitrogen molecules act as both, reactants and host matrix. The cyclic compounds A and C are of aromatic nature due to the presence of a delocalized 2π electron bond which is in conjugation with the π lone pairs of the exo‐cyclic F atoms indicating a type of fluorine specific interactions.

Figure 1 shows infrared spectra obtained after laser‐ablated natural boron atoms co‐deposited with a 0.5 % F₂/N₂ mixture in a 4 K dinitrogen matrix. The reaction products induced by annealing and photolysis are indicated. In addition to the three novel fluoroborylene : N₂ products A–C binary boron fluorides BF_n (n=1–3) and the molecular boron nitrides NBN and NNBN^[6] were obtained in the present study, while the previously reported diboron compounds BBNN and BNBN^[6] were barely observed (Figure 1 and Table S1). Although BF₂ and BF₃ are also produced,^[9] the reaction conditions were optimized to achieve a maximum yield of fluoroborylene, BF (see experimental details in the Supporting Information). The strong bands for ¹¹BF and ¹⁰BF were found at 1370.6 and 1412.4 cm⁻¹, respectively (Figure 1, Table S2). These optimized conditions enabled us to tentatively assign three weak bands to different ^10/11B isotopologues of difluorodiborene, FBBF, at 1327.3 (¹¹B¹¹B), 1348.5 (¹⁰B¹¹B) and 1370.3 cm⁻¹ (¹⁰B¹⁰B) in experiments using a natural boron target in N₂ and ¹⁵N₂ matrices (Figure 1 and S3). Owing to its high reactivity, experimental spectra of free FBBF have not yet been reported, although complexes of FBBF with electron‐rich transition metals have been investigated.^[10]

Figure 1.

Excerpts from the IR spectrum obtained using a natural boron target with 0.5 % F₂ in N₂ matrix: (a) co‐deposition of B+0.5 % F₂ for 120 min, (b) after annealing to 15 K, (c) after subsequent λ=273 nm irradiation for 30 min, and (d) further annealing to 15 K. Unknown species are indicated by asterisks.

The absorptions associated with the novel fluoroborylene : N₂ product molecules A–C were unambiguously assigned based on their growth/decay characteristics in different experiments and on their characteristic ^10/11B and ^14/15N isotope pattern. The novel ring molecule A shows strong bands in the B‐F stretching region at 1531.0 and 1558.6 cm⁻¹, which are tracked by weaker ring vibrations at 1090.7 and 1101.5 cm⁻¹. These bands were already observed on deposition, they increased by 50 % on annealing to 15 K, and continue to grow on irradiation at λ=278±10 nm, where they reached three‐fold on further annealing to 15 K. For a natural boron (¹⁰B:¹¹B=19.9:80.1) sample, the vibrational modes of a diboron species with two equivalent boron atoms split into three absorptions with approximately 1:8:16 (¹⁰B¹⁰B, ¹⁰B¹¹B, ¹¹B¹¹B) relative intensities.^[11] In addition to the two B‐F stretching bands for the two most abundant isotopologues of A (Tables 1 and S3) the corresponding band associated with the ¹⁰B¹⁰B species is observed in ¹⁰B + F₂/N₂ mixture experiments at 1590.2 cm⁻¹, giving a ¹⁰B/¹¹B isotopic ratio of 1.0381. The ¹¹B¹¹B ring vibration observed at 1090.7 cm⁻¹ shift to 1123.8 cm⁻¹ (¹⁰B¹⁰B) in these experiments, indicating a ¹⁰B/¹¹B isotopic ratio of 1.0303, a typical boron atom involved isotopic ratio. As shown in Figure 2 the boron isotopic distribution supports the presence of two equivalent boron atoms in this molecule. In experiments using F₂/¹⁵N₂ mixtures (Figure 2) the ¹⁵N counterpart bands were observed at 1569.6 and 1108.5 cm⁻¹ (¹⁰B¹⁰B), 1542.0 cm⁻¹ (¹⁰B¹¹B) and 1521.4 and 1077.0 cm⁻¹ (¹¹B¹¹B), respectively (Figures S1). The relative intensities of the bands due to the ¹⁴N and ¹⁵N isotopomers are almost the same in experiments using 50 % ¹⁴N₂ and 50 % ¹⁵N₂ mixtures (Figure 2 f). Obviously, two B atoms bind only one ¹⁴N₂ or ¹⁵N₂ in this new ring molecule. We also note a good agreement between the observed and calculated frequencies at the B3LYP and CCSD(T) levels of theory listed in Tables 1 and S3.

Table 1.

Observed and calculated (CCSD(T)/def2‐TZVP) vibrational frequencies (cm⁻¹) and isotopic frequency ratios (ν(¹⁰B)/ν(¹¹B)) of the FB(μ‐N)₂BF, FBNNBF and FB(η ²‐N₂) molecules.^[a]

	11B11B/11B		10B11B		10B10B/10B		10B/11B ratio
	calcd[a]	Obs.	calcd[a]	Obs.	calcd[a]	Obs.	calcd	Obs.
cyclic FB(μ‐N)2BF (1A1)
	1612.7 (68) 1538.5 (802) 1114.6 (113)	1612.8 1531.0 1090.7	1642.2 (88) 1546.2 (733) 1125.4 (115)	1635.8 1558.6 1101.5	1662.9 (70) 1585.1 (850) 1140.5 (120)	1660.7 1590.2 1123.8	1.0311 1.0302 1.0232	1.0297 1.0382 1.0303
14N
	1597.1 (57) 1523.2 (785) 1093.5 (110)	hidden	1638.2 (87)	/	1659.2 (64)	1662.8	1.0387	/
15N		1521.4	1541.8 (727)	1542.0	1580.1 (844)	1569.6	1.0377	1.0317
		1077.0	1110.5 (113)	/	1125.8 (118)	1108.5	1.0295	1.0292
linear FBNNBF (1Σg +)
14N	2068.8 (927)	2078.2	2099.0 (934)	2108.8	2138.6 (1012)	2147.0	1.0337	1.0331
	976.1 (178)	981.0	977.4 (176)	/	978.9 (174)	983.1	1.0029	1.0021
15N	2049.5(933)	hidden	2079.2 (929)	2073.4	2120.2 (1017)	2125.0	1.0345	/
	962.7(168)	969.4	963.8 (166)	/	964.9 (165)	970.4	1.0023	1.0010
cyclic FB(η 2‐N2) (1A1)
14N	1704.5(356)	1710.5			1760.6(355)	1765.8	1.0330	1.0323
	1231.6(45)	1229.5			1234.9(46)	1231.0	1.0027	1.0012
15N	1689.6(398)	1700.1			1752.3(340)	/	1.0371
	1194.8(35)	1192.5			1198.6(33)	/	1.0032

[a] Scaled frequencies using a uniform scaling factor of 0.969.^[27] Intensities (km mol⁻¹) in parentheses. Band positions assigned to B‐F stretching modes are given in italics.

Figure 2.

Excerpts from the IR spectrum obtained using a ¹⁰B target with 0.5 % F₂ in (a)  ¹⁴N₂ matrix, (b)  ¹⁵N₂ matrix, and (c) 50 % ¹⁴N₂ + 50 % ¹⁵N₂ matrix. Natural boron target with 0.5 % F₂ in (d)  ¹⁴N₂ matrix, (e)  ¹⁵N₂ matrix, and (f) 50 % ¹⁴N₂ + 50 % ¹⁵N₂ matrix.

The linear isomer B (D _∞h symmetry, Figure 3) has a singlet ground state and exhibits only two infrared active vibrational modes in the mid‐IR region (Table S4, for more computational results on B see Part 1 of the Supporting Information). The antisymmetric B‐N stretching modes give rise to two bands at 2108.8 (¹¹B¹⁰B) and 2078.2 (¹¹B¹¹B) cm⁻¹ in Figure 1 using a natural boron target, while the antisymmetric F‐B vibration is observed only for the most abundant ¹¹B¹¹B isotopologue at 981.0 cm⁻¹ on deposition. Experiments performed with ¹⁰B and N₂ or ¹⁵N₂, and ¹¹B with N₂ or ¹⁵N₂ are shown in Figures 2 and S1–S3, and the observed product absorptions are compared to calculated values in Table 1 (for more details see the Supporting Information). These bands disappeared upon 273 nm irradiation, while the bands of A increased simultaneously, suggesting that isomerization is occurring. Note, linear B is isoelectronic to diisocyanate, OCNNCO.^[12] Like B also OCNNCO is photosensitive and decomposes rapidly under UV light to produce N₂ + 2 CO.

Figure 3.

Optimized structures of A FB(μ‐N)₂BF, B FBNNBF and C FB(NN) obtained at the B3LYP/6‐311++G(3df,3pd) and CCSD(T)/def2‐TZVP (italic) levels of theory. Bond distances are given in Å and angles in degree.

Cyclic C (Figure 3) is assigned to a band at 1710.5 cm⁻¹ in Figure 1, and another very small band at 1229.5 cm⁻¹, which corresponds to the F‐B and N‐N stretching modes, respectively (Tables 1 and S5). Further experiments were performed using ¹⁰B and N₂ or ¹⁵N₂, and ¹¹B with N₂ or ¹⁵N₂, and the absorptions of the corresponding isotopologues were observed at 1765.8 and 1331.0 cm⁻¹ (¹⁰B), and at 1700.1 and 1192.5 cm⁻¹ (¹⁵N, Figures S1,S3). These bands were observed on co‐deposition, but on annealing to 15 K they disappeared entirely, while the bands due to A increased by 30 %.

According to B3LYP/6‐311++G (3df, 3pd) calculation linear B is separated from cyclic A by a barrier of 25.0 kcal mol⁻¹ and higher in energy by 22.3 kcal mol⁻¹ (Figures S5, S11). At this level the predicted N‐N distance in A is 1.847 Å (CCSD(T)/def2‐TZVP: 1.886 Å), which is significantly longer than for example, the N−N single bond of diphenylhydrazine [d(N−N): 1.394 Å]^[5] and indicates a complete splitting of the N≡N triple bond by the two FB units. The computed B‐N distance (1.434 Å, Figure 3) is in the range of a conventional B=N double bond, like in aminoboranes.^[13] The ring inversion barrier of the puckered ring of A was found to be 15.8 kcal mol⁻¹ at the B3LYP/6‐311++G(3df, 3pd) level of theory (Figure S13).

The proposed aromatic π‐electron delocalization in the cyclic compounds A and C is supported by a molecular orbital (MO) analysis and computed nucleus‐independent chemical shift (NICS) values. Figures 4 and S14 shows typical π‐bonding orbitals of the cyclic π‐conjugated system A, which consists of the 4c‐2e central bonding orbital (HOMO−2) and two further π‐bonding orbitals (HOMO−6 and −7), to which the exo‐cyclic F atoms clearly contribute. Related puckered 4‐membered ring aromates have previously thoroughly been analyzed.[¹⁴, ¹⁵]

Figure 4.

Selected frontier molecular orbitals of FB(μ‐N)₂BF.

The NICS value is among the most popular aromaticity indices.^[16] While the NICS index was originally obtained for planar aromatic systems, it has recently been suggested to calculate an average NICS(1)_av index, NICS(1)_av=[NICS(−1) + NICS(1)]/2, as a probe of aromaticity in nonplanar molecular systems.^[17] The large negative NICS(1)_av index of −21 obtained at the center of gravity of A (Figure S15,b) indicates its significant aromatic character, which can be compared to the NICS(1) value obtained for planar C of −12 (Figure S15 e).

The selective formation of the products A–C is surprising at first glance, but due to a high dilution of the initially formed reactive intermediates, as well as the subsequent isolation of the products in a solid N₂ matrix at cryogenic temperatures, possible secondary reactions are efficiently suppressed. The predominant reaction of laser ablated boron atoms with N₂ molecules^[6] depends on whether the boron atoms are in their ²P(2s²2p¹) or first excited ⁴P(2s¹2p²) state, located 82.5 kcal mol⁻¹ higher in energy.^[18] Common trivalent boron compounds can usually be traced back to the first excited ⁴P state, and sub‐valent boron compounds such as Lewis‐base stabilized borylenes only recently became a rapidly emerging class of highly reactive intermediates.[³, ⁴] However, due to the very low Lewis basicity of N₂ it seems that ground‐state boron atoms are reluctant to react with N₂ molecules.[⁶, ^19a] The dissociation energy of the weakly bound B‐NN (²Π) adduct with respect to B(²P) + N₂ (X¹Σ_g ⁺) is reported to be only 1.2 kcal mol⁻¹.^[20] It is interesting that only the high‐energy isomer NBN (²Π) is observed in the present study (Figures 1, 2). The weakly bound B‐NN (²Π) adduct is almost isoenergetic with the cyclic isomer B(η ²‐N₂) (²A₁), while the linear trivalent boron isomers BNN (⁴Σ_g ⁻) and NBN (²Π) are substantially higher in energy by 7.7 and 22 kcal mol⁻¹, respectively.^[20] Given that the adduct B‐NN (²Π) is separated from the other three isomers by significant computed energy barriers,^[19b] the observation of only the highest‐energy isomer NBN (²Π) could indicate its higher kinetic stability compared to the other isomers under the experimental conditions. A plausible route to the formation of NBN is the reaction of N₂ molecules upon the deposition with excited ⁴P boron atoms, which are produced during laser‐ablation.

Prominent bands due to NNBN (¹Σ_g ⁺) and BF (¹Σ_g ⁺) in the spectra obtained after deposition (Figure 1) indicates the presence of free N and F atoms in the deposit. These atoms are commonly generated in the hot plasma plume or by photo‐decomposition of N₂ and F₂ molecules, respectively, as a result of the plasma broadband radiation that is produced during laser ablation. These atoms react very exothermically with boron atoms to yield diatomic NB and FB molecules, respectively, but only NB react further with N₂ molecules during deposition to yield NNBN.^[6] In addition, the lack of mixed ¹⁴N/¹⁵N isotopologues of NBN in experiments using 1:1 mixtures of ¹⁴N₂ and ¹⁵N₂ rule out the formation of NBN from diatomic BN and N atoms, which corroborates our assumption of NBN formation by insertion of excited B atoms into N₂ molecules.

Our search for FBNN (Figure S19) in the experimental spectra failed, and, in contrast to NBNN and PhB‐NN^[4e] free diatomic BF was observed. In agreement with previous results^[8] our calculations revealed that the linear adduct FBNN is endothermic by 23.4 kcal mol⁻¹ compared to N₂ + singlet FB (Figure S19). This observation is consistent with the larger singlet‐triplet gap of FB (+80 kcal mol⁻¹)^[20a] compared to that of diatomic NB (−0.5 kcal mol⁻¹)^[20b] and PhB (+31 kcal mol⁻¹).^[20c]

For the formation of the novel FB : N₂ adducts A–C, we have explored both the fluorination of initially formed BN₂ intermediates and the reaction of fluoroborylene FB with N₂ molecules (Figures S4, S5). Although the reaction of the BN₂ species with F atoms is strongly exothermic, preliminary B3LYP/6‐311++G(3df,3pd) calculations suggest significant reaction barriers, for example, 14 kcal mol⁻¹ for the reaction of NBN (²Π) with F atoms (Figure S4). These calculations also revealed that cyclic C is higher in energy by 8.1 kcal mol⁻¹, but kinetically stable to its decomposition into N₂ + singlet FB due to a barrier of 42 kcal mol⁻¹ (Figure S5).

Activation of N₂ molecules and weakening of its π bonds can mainly be attributed to interactions that donate electron density to its π* antibonding orbitals, and remove electron density from the π bonding orbitals of N₂.^[21] Unlike transition metal complexes, which provide a σ acceptor and a π donor orbital to form N₂ complexes (Figure 5 c), singlet FB can be viewed as an electrophilic σ donor and π acceptor (Figure 5 a). Its side‐on attack on N₂ (Figures 5 c and S12) enables electron donation from its sigma donor orbital into the π* MO of N₂ and removal of π‐bonding electrons into the π* MO of BF. These bonding interactions are supported by an energy decomposition analysis (EDA).^[22] This shows that the two major orbital interactions ΔE _orb(1) and ΔE _orb(2), which contribute with −119.3 and −35.0 kcal mol⁻¹ to the total orbital term ΔE _orb in the transition state of the FB + N₂ reaction, can be attributed to σ(FB) donation and π(N₂) back‐donation, respectively (Figure S20). Since the reaction between boron and fluorine atoms is strongly exothermic (182 kcal mol⁻¹),^[20a] this reaction energy can provide the activation energy for the singlet FB + N₂ during the deposition of the matrix. It can, however, not be excluded that C can also be produced in an exothermic reaction of N₂ molecules with triplet excited FB (³Π), which is 80 kcal mol⁻¹ higher in energy than the singlet ground‐state,[^20a, ^20c] and likely formed by UV radiation (λ<357 nm) emitted from the plasma plume.

Figure 5.

a) Scheme of the FB acceptor and donor orbitals. Note that FB has two mutually perpendicular π‐acceptor orbitals, only one of which is shown. b) Bonding interaction in the transition state of the FB + N₂ reaction, c)  end‐on complex of N₂ to a transition metal M, see Figure S12.

We recall that C is observed only in freshly deposited samples, and it disappeared already upon annealing to 15 K. Cyclic C is isolectronic to the known diazirinone, OC(η ²‐N₂),^[23a] and although C should be more stable because of a higher barrier and a lower dissociation energy,^[23b] the N=N bond in C is strongly activated and it reacts readily with a second singlet FB molecule to yield A through a very low energy barrier of 0.5 kcal mol⁻¹(Figure S5).

Since B₂N₂ isomers were barely observed, they can hardly be considered as starting compounds for the bisfluoroborylene: N₂ compounds A and B. However, A and B are most likely formed in an exothermic and low‐barrier reaction from BF dimer molecules and N₂ (Figure S5). The two lowest energy BF dimer isomers have been considered, the linear triplet FB=BF structure (³Σ_g ⁻) and a singlet trans‐bent isomer of C _2h symmetry in a ¹A_g ground state (Table S8). In contrast to the isoelectronic CO dimer, OCCO (³Σ_g ⁻),^[24] both of these BF dimers are more favourable than the diatomic fragments in their ground singlet state. The linear triplet FBBF, which formally arises from a double σ→π* excitation of the two singlet BF fragments,^[8] is more stable than the trans‐bent isomer by 8 kcal mol⁻¹ (Table S8). On the other hand, the trans‐bent isomer could be formed through mutual σ→π donor‐acceptor interactions of two ground‐state singlet BF molecules via a loose and low‐energy (<1.0 kcal mol⁻¹) C _2h symmetric transition state.^[8] Since the computed antisymmetric B‐F stretching frequencies of these two isomers are similar (Table S8), our assignment of the experimentally observed B‐F stretching frequencies to the trans‐bent isomer (Table S8) is therefore very tentative and only supported by its predicted low‐barrier formation from two singlet BF molecules. Nevertheless, the positive dissociation energy and the low barrier of formation of trans‐bent FBBF from two BF molecules combined with a low barrier for the subsequent reaction with N₂ molecules of 10 kcal mol⁻¹ at the B3LYP/6‐311++G(3df, 3pd) level provide a surprisingly selective, low‐barrier route to the title compound A on the singlet potential energy surface (Figure S5).

It was shown that the parent diborene, HB=BH, is efficiently stabilized by Lewis base ligands L to yield planar adducts L(H)B=B(H)L, with L=CO^[25] or bulky carbene ligands.^[26] Preliminary calculations at the B3LYP/6‐311++g(3df,3pd) level indeed predict that also the addition of the weak N₂ donor molecule to FBBF would be strongly exothermic and yield the corresponding adducts FB=B(N₂)F (C _s, ΔE=−25.6 kcal mol⁻¹) and F(N₂)B=B(N₂)F (C _2h, ΔE=−34.7 kcal mol⁻¹, Table S9 and S10). However, we found no spectroscopic evidence for the predicted formation of these dinitrogen adducts of FBBF in the cryogenic N₂‐matrix and we note that the adduct FB=B(N₂)F is still significantly higher in energy than the experimentally observed isomers A and B by 44 and 22 kcal mol⁻¹, respectively. The cyclic compound A is both energetically (Figures S4, S5) and kinetically very stable, so that it can be expected that, like the analogous diazirinone, OC(η ²‐N₂),^[23a] it could be viable also at ambient conditions.

Conclusion

In conclusion, the two novel cyclic fluorodiazaboririne, FB(η ²‐N₂) (C), and 1,3‐Diaza‐2,4‐diborete, (FB)₂N₂ (A), as well as the linear compounds FB=BF and FBNNBF (B) were produced from laser‐ablated boron atoms and fluorine embedded in an excess of N₂. The aromatic nature of the electron‐deficient rings of A and C, reinforced by fluorine specific interactions based on electronic contribution from the π lone pairs of the exo‐cyclic fluorine atoms, and confirmed by an MO analysis and computation of their NICS index, contributes to their high thermodynamic stability. Their surprisingly selective formation can be traced back to the high reactivity of fluoroborylene intermediates. This work may contribute to exciting applications in dinitrogen fixation and activation.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

B. Xu, H. Beckers, H. Ye, Y. Lu, J. Cheng, X. Wang, S. Riedel, Angew. Chem. Int. Ed. 2021, 60, 17205.