MSc Chemistry
Molecular Sciences
Master Thesis
Heterolytic Bond Cleavage using
Zirconocene Amine Complexes
A DFT investigation
by
Titus de Haas
11030895
June 2020
42 EC
November 2019 – July 2020
Supervisor/Examiner:
Examiner:
Prof. Dr. Peter H.M. Budzelaar
Prof. Dr. Evert Jan Meijer
Prof. Dr. Bas de Bruin
Department of Chemical Sciences
Federico II, university of Naples
2
Preface
This thesis contains accounts on the research that I conducted in order to fulfil my graduation requirements at the University of Amsterdam. This work has been performed in the period between November 2019 and July 2020 and was supervised by prof. dr. Peter H.M. Budzelaar from Federico II, university of Naples (Italy). During the first four months of this thesis, I was living in Naples, where I had a workplace at the chemistry department of the Federico II university. In this period, I learned a lot about organometallic chemistry and I obtained skills required to do research in the field of computational chemistry. Additionally, I had many opportunities to explore the city of Naples and the region of Campania (south of Italy), which was very exciting for me. Unfortunately, the COVID-19 outbreak and the subsequent restrictions imposed by the Italian government, forced me to move back to Amsterdam in March 2020. Luckily, prof. dr. Peter H.M. Budzelaar was so kind to continue his supervision on the project via Skype conversations. This way I was still able to finish my thesis within the given time.
The thesis covers two distinct topics. The first two months of the internship I performed calculations on a β-diiminate rhodium aniline dimer. Experimentally, ortho-selective C-N coupling was observed to occur between the two aniline molecules. Although we did make quick progress in the beginning, it proved very difficult to find a complete reaction mechanism. Because the mechanism appeared to be too complicated to unravel and calculations were very time-consuming, it was decided to put this research on a hold and start with a new project. Prof. dr. Peter H.M Budzelaar had just published an article in collaboration with dr. Luca Rocchigiani from the University of East Anglia on heterolytic cleavage of hydrogen molecules in a cationic, zirconaziridine complex. This work appeared very interesting to me, so we decided that I would continue the research, exploring different kinds of aspects influencing the reactivity of this system. In this light, I computed reaction paths for a range of similar systems to determine steric and electronic influences of different (ancillary) ligands. In the end, we also expended the reactivity beyond hydrogen, by investigating activation of the Si-H bond. Given that the two topics that I studied during the internship are relatively unrelated, I decided to cover the research in two chapters, that can be viewed separately. The first chapter will contain the work on hydrogen activation using the metallocene amines and the second chapter the work on the β-diiminate rhodium system.
I would like to thank my supervisor prof. dr. Peter. H.M. Budzelaar for giving me the opportunity to work on this project. He was very welcoming when I first moved to Naples, and I very much appreciated his continuing support even after I had to move back to Amsterdam. Also, I would like to thank dr. Luca Rocchigiani for his feedback on my work on the metallocene amine reactivity, prof. dr. Bas de Bruin for being first examiner in Amsterdam and helping me to find this project, and lastly prof. dr. Evert Jan Meijer for being second examiner.
3
Table of Contents
Preface ... 2
Chapter I: Heterolytic Bond Cleavage Using Zirconocene Amine Complexes ... 4
Abstract ... 4
Introduction ... 5
Computational Methods ... 6
Results & Discussion ... 7
Variations in Lewis base and Lewis acid ... 7
Proton affinity versus zirconocene affinity ... 11
Activation of the Si-H bond in SiH4. ... 14
Conclusions ... 16
Chapter II: Aniline C-N Coupling in β-Diiminate Rhodium Anilide Dimer ... 17
Abstract ... 17
Introduction ... 18
Computational methods ... 19
Results & Discussion ... 19
Reconstructing the overall-reaction ... 19
Investigation of potential reaction mechanisms ... 21
Direct aminyl reactivity towards the phenyl C-H bond ... 22
Nitrene formation and subsequent reactivity towards the phenyl C-H bond ... 25
Conclusions ... 28
4
Chapter I: Heterolytic Bond Cleavage Using Zirconocene Amine
Complexes
Abstract
Recently reported activation of dihydrogen by [Cp2Zr(η2-CH2NR2)]+ in chlorobenzene, was investigated
using density functional theory (DFT).1 Within the mapped out reaction path, two consecutive
hydrogen activation steps were identified, the first displaying a σ-bond metathesis-type transition state, and the second a linear heterolytic activation-type transition state. Formation of a dicationic, Zr3H4 cluster, was found to be the driving force of the reaction. By introducing small variations in the
zirconocene and in the amine fragments, we were able to determine which variables influence the barrier-heights and the thermodynamics for the observed reaction. Increase in steric bulk on either the zirconocene or the amine fragment, results in lowering of the heterolytic activation barrier. Overall thermodynamics become less favourable when bulky substituents are attached to the zirconocene fragment, because they hamper formation of the trimeric product. Stronger bases as amines tend to also result in less favourable thermodynamics, due to liberation of two units of free base during the reaction. Substitution of Ti for Zr lowers the heterolytic activation barrier and disfavours trimer formation, due to the smaller size of Ti. These results are interesting, since we argue that formation of the final Zr3H4 dication is not necessarily desirable for catalysis. Stabilization of the Cp2ZrHH
intermediate would be more advantageous, in this respect. This computational investigation outlines the dominant factors that determine chemical behaviour in the reported FLP-like system. Variations from this thesis could be interesting for further experimental analysis. However, it is emphasized that the observed, large thermodynamic changes, will likely result in deviating reaction paths for the real system.
5
Introduction
Activation of dihydrogen has been a well-established reaction in chemistry for decades, with earliest reports already stemming from the early sixties.2,3 Initial reports showed the activation
through oxidative addition mechanisms, where the involved metal complex formally donates two electrons to form a dihydride species. Later, also systems were discovered where no formal oxidation of the metal centre would occur. Here, the distinction can be made between systems that activate dihydrogen via a σ-bond metathesis route and via heterolytic cleavage.4 The σ-bond metathesis occurs
through a concerted 2+2 reaction, resulting in cleavage of the metal-X and H-H bond to form the metal hydride. On the other hand, In heterolytic cleavage, the H-H bond is cooperatively broken between a strong electron donating and a strong electron withdrawing fragment. Both types of reactions can result in the same products, yet they differ in their mechanisms. Formally, σ-bond metathesis would display a characteristic 4-centred transition state, while heterolytic cleavage would rather exhibit a linear, highly polarized transition state structure. In reality, the two are often less easily distinguished and characteristics of both may be observed. That being said, heterolytic activation of H2 has gained
an enormous amount of attention since the discovery of so called frustrated Lewis pairs (FLPs) around 2006.5,6 Reports showed that, cooperative reactivity of strong Lewis acidic boron compounds, together
with strong Lewis basic phosphorus compounds resulted in reversible activation of dihydrogen, potentially opening the doorways for metal-free catalysis. Crucial for this type of chemistry is the introduction of sterically demanding substituents on both the Lewis base and the Lewis acid. This way, potential reactivity is not quenched by Lewis adduct formation and can instead be utilized in the activation of other substrates. Since then, a vast amount of research has been published, exploring this type of reactivity.7
In a recent publication by Budzelaar and Rocchigiani, et al., activation of molecular hydrogen by a cationic dicyclopentadienyl zirconium(IV) amine species was reported.1 Provoked by the
characteristics of [Cp2Zr(η2-CH2NR2)]+ (INT1) in olefin-polymerisation catalysis,8–11 the authors
investigated the complex’s reactivity towards molecular hydrogen. It was found that the complex was capable of splitting dihydrogen over the much strained Zr-C bond to form a zirconocene monohydrido amine complex (INT2). The resulting species, subsequently, reacted in a 3:1 ratio with another hydrogen molecule to form a dicationic, hexagonal Zr3H4 cluster (1), one equivalent of
N,N-dimethylaniline salt, and two equivalents of free N,N-N,N-dimethylaniline (Figure 1). In this second part of this reaction, the hydrogen molecule was thus effectively split into a hydride and a proton, similar to what is known in FLP-type chemistry.
Figure 1: Formation of the (Cp2Zr)3H4 cluster from initial zirconaziridine cation and molecular hydrogen.
First INT1 reacts 1:1 with H2 to form INT2. Subsequently, a second H2 molecule is effectively split over
6
In their publication, the authors explored the mechanism of the reaction using density functional theory. Although computing a complete, beginning to end, reaction mechanism appeared to be computationally unfeasible, the authors did propose a plausible mechanism for the most relevant first steps in the reaction (see Figure 2). It was found that after the σ-bond metathesis reaction to split the first H2 molecule, heterolytic activation of another molecule of H2 is possible
between the Lewis acidic zirconocene and Lewis basic amine fragments, to form a zirconocene dihydride species. Even though the system is not as sterically encumbered as traditional FLP systems, the heterolytic activation of H2 only occurs after initial dissociation of the zirconium and amine
fragments, demonstrating that large ("frustrating") steric bulk is not always a prerequisite for this type of reactivity. Additionally, the involvement of a transition metal introduces the traditional benefits associated with d-orbital type catalysis, making further investigation of this system worthwhile.
In this report, the reactivity of this system is computationally explored by introducing modifications on the original zirconocene and amine. A set of ligands was chosen to distinguish between electronic and steric effects (see Figure 3). The critical part of the reaction mechanism, involving both hydrogen activations steps, was computed for each set. This way, it was examined which characteristics have the most significant influence on the reaction barriers and on the overall thermodynamics.
Figure 2: Proposed reaction mechanism, involving first H-H σ-bond metathesis over the strained Zr-C bond, followed by heterolytic H-H activation to form the zirconocene dihydride species.
Computational Methods
Structure optimisations and transitions state searches were performed using the TPSSh functional and the def2-SVP basis set in TURBOMOLE, where an external optimizer was applied through the BOpt package.12–17 At this level, a numerical vibrational analysis was performed to check
for imaginary frequencies and to obtain enthalpy and entropy corrections. Then, the resulting structures were used in Gaussian as input for single-point calculations with the M06-2X functional and cc-pVTZ basis.18–22 In these single-point calculations, solvent corrections were approximated using the
polarizable continuum model (PCM).23–25 The Gibbs free energy profile was constructed by correcting
the electronic energies found for the stationary points, according to ΔG = ΔEelec + ΔHcorr - 0.67TΔScorr,
where the entropic scaling factor is introduced to correct for the damping effects of the solvent in the liquid phase (these damping effects are not fully captured by the solvent model, as this only takes into account electronic polarization effects).26,27 Proton affinities for the amine fragments are calculated
as: PA = ΔG°amine + ΔG°H+ - ΔG°H-amine+ . Similarly, Zirconocene affinities are calculated as: ZrA = ΔG°amine
7
Results & Discussion
Variations in Lewis base and Lewis acid
In this research, variations in both the Lewis acidic zirconocene fragment and the Lewis basic amine fragment were considered. Variations in substituents on the nitrogen were already explored in the experimental work of the original publication. There, it was found that the amount of steric hindrance that could be imposed was limited. Adding too many sterically encumbering substituents on the nitrogen prevented formation of the N-Zr bond in the zirconaziridine cation. Therefore, only variations on one single R group in the amine unit are considered here. For the amine moiety, from least to most sterically encumbered, trimethylamine (TMA), dimethyl-1-ethylamine (DMEA), N,N-dimethylaniline (DMA) and N,N-dimethyl-1-adamantanamine (DMAA) were studied. Additionally, reactions paths were computed for N,N-dimethyl-trifluoromethyl-amine (DMTFMA), N,N-dimethyl-(4-fluoro-phenyl)-amine (DMFPhA) and N,N-dimethyl-(2,4,6-triN,N-dimethyl-(4-fluoro-phenyl)-amine (DMTFPhA) to determine electronic influences. For the Lewis acidic fragment, computations were performed with variations in the actual metal centre, as well as variations in the Cp ligands. For the metal centre, paths were computed for zirconium, hafnium and titanium, and as ligands, η5-cyclopentadienyl (Cp), η5
-pentamethylcyclopentadienyl (Cp*), the ansa- SiMe2Cp2 and EthCp2, and the rac- isomer of the ansa-
SiMe2(Ind)2 (Ind=η5-indenyl) were investigated. All considered variations are depicted in Figure 3.
Figure 3: Computationally investigated set of amine ligands and zirconocene fragments. Name of the fragment is indicated below, and the letter is depicted in red at the right top of each structure. Since
Cp2Zr+ was used as the main zirconocene when the amines were varied, there is no need to specifically
address this structure.
For the remainder of this report we use a simple naming system. Structures of intermediates and transition states will be pointed out by INT or TS, together with a number which corresponds to the one depicted in Figure 2. To further specify variations in the used amine and metallocene, a letter in the range of A-M is added according to the structures depicted in Figure 3. For example, [Cp2
Zr(η2-CH2NMe2)]+, will be named: INT1A. Also for the final, Zr3H4 centred product, 1, we will use A-M to
specify each particular system. Even though this product is identical for all systems where we consider the amine variations, formation energies are different, since these are partly determined by the basicity of the dissociated amine. For the computations on these systems, it was assumed that a the mechanism that was reported by Budzelaar, Rocchigiani and coworkers was also applicable to the investigated systems in this report. For the systems with variations on the metallocene fragment, TMA was chosen to model the amine fragment. This resulted in significant savings in computational time, due to the relatively small system size and a significant reduction of realisable conformers. The
8
absence of a C3 rotational symmetry axis in the other amine ligands introduces a variety of new
orientations of the amine with respect to the zirconocene. For most of the investigated systems, multiple conformers for both local minima and saddle-points were found. For all the systems that were investigated, all required structures were found to form the complete reaction path that was discussed earlier (the only exception here being 1M). Transition states featured similar characteristics as to what was reported in the original publication. For example, substitution of Cp* for Cp did result in a larger N-Zr distance at the transition state for heterolytic H-H activation (4.44 Å vs 3.89 Å, see Figure 4), but still the same mechanism applies.
Figure 4: Optimised transition states for heterolytic H-H activation in Cp system (TS2A, left) and Cp* system (TS2M, right).
To compare the different systems, barrier heights for the crucial H-H splitting steps, as well as the Gibbs free energy associated with the hydride formation and overall reaction, are displayed in Tables 1 and 2. In most systems, multiple conformers for both transition states and intermediates were found. In these cases, the energy of the most stable conformer is displayed. For tables which include all data, the reader is referred to the excel sheets accompanying this report.
Table 1: Barrier heights for H-H activation and thermodynamics of dihydride (INT6) and Zr3H4 (1)
formation with the Lewis base variations
Amine type TS1‡ (kcal mol-1) TS2‡ (kcal mol-1) ΔG
INT6 (kcal mol-1) ΔG1 (kcal mol-1)
TMA 19.6 28.2 19.8 5.6 DMEA 18.2 27.4 17.9 4.0 DMA 19.0 24.6 22.6 0.2 DMAA 23.5 21.4 11.1 -11.1 DMTFMA 20.0 23.2 31.2 -7.8 DMTFPhA 19.1 25.2 26.3 9.0 DMFPhA 18.1 24.2 22.2 -2.5
The substituents on the amine fragment generally seem to have a limited effect on the barrier for the first H-H splitting step. A small lowering ~1 kcal mol-1 is observed when TS1B is compared to
9
TS1A. Besides this, no clear trends are established. The only exception here is the adamantyl system (TS1D), which is ~4 kcal mol-1 higher in ΔG than TS1A. In the optimized INT1D, we can see that the
sterically demanding adamantyl group forces a slight twist, which likely explains the increase in energy with respect to TMA (see Figure 5, left). Larger differences occur in the transition states for the heterolytic activation step. Progressing from TS2A to TS2B to TS2C to TS2D, the increasing substituent size results in a lowering of the associated activation barrier with ~ 1, ~4 and ~8 kcal mol-1, for each
step, respectively. Besides the transition state energies, it is also interesting to look at the energies associated with the formation of the dihydride (INT6) and the trimer (1). INT6B and 1B are more stable than INT6A and 1A (differences are ~2 and ~2 kcal mol-1). In contrast, INT6C is less stable (by ~3 kcal
mol-1), yet formation of 1C is again lower in energy than of 1A by ~5 kcal mol-1. This is likely due to the
electron withdrawing character of the phenyl group compared to the alkyl. The adamantyl system (INT6D), namely, does result in a more exergonic reaction towards INT6D and 1D, differing by a very significant ~8 and ~17 kcal mol-1, with respect to the TMA analogue.
The effects of the addition of electron withdrawing fluorine substituents are less unambiguous. Replacement of one methyl group in TS2A by a trifluoromethyl group, TS2E, significantly lowers the heterolytic activation barrier with ~ 5 kcal mol-1. The same trend is observed
when the 4-fluorophenyl group in TS2F is substituted for the phenyl group in TS2C. In contrast, a slight increase in barrier height is observed when the phenyl group in TS2C is replaced by a 2,4,6-trifluorophenyl group (TS2G). Formation of INT6E, INT6F and INT6G is less favorable compared to their non-fluorinated counterparts. Interestingly, usage of DMTFMA as the Lewis base results in a significant lowering of the formation energy of the 1E (by ~13 kcal mol-1 with respect to 1A). Contrarily, the usage
of DMTFPhA actually leads to an increase in the energy of formation of 1G (~9 kcal mol-1 more than
1C). Examination of the structure of INT2G reveals the cause of this apparent discrepancy (see Figure
5). The TPSSh/def2-SVP DFT optimization predicts a relatively strong interaction between the zirconium centre and a fluorine atom in the ortho-position of the phenyl group. The optimized F-Zr bond length of 2.38 Å is in good agreement with F(arene)-Zr bond lengths found in crystal structures of very similar zirconocene complexes.28 This would explain the additional stabilization of INT2G,
resulting in less favourable thermodynamics for the further reaction. Also a mono-fluorinated phenyl amine was analysed, where no such Zr-F interaction was possible. It was observed that for this system, INT6F and 1F did follow the same trends as observed for the DMTFMA system. When the overall-reaction (Figure 1) is considered, it makes sense that weaker bases result in more favourable thermodynamics. For the formation of 1, two molecules of free amines are released together with just one molecule of protonated amine.
Figure 5: Gas-phase optimized structures of TS1D (left) and INT2E (right). Structures were optimized at TPSSh/def2-SVP level of theory.
10
From these results, it becomes evident that an increase in steric bulk on the amine moiety lowers the activation barrier for the heterolytic cleavage of H2. Addition of electron withdrawing
groups on the amine ligand have a less clear effect, although TS2E is lower than TS2A, the formed dihydrido species (INT6E) is actually less stable. It must be noted here though, that also a structure was found where the protonated amine moiety was still coordinated to the Cp2ZrHH fragment (INT5E).
This resulted in a very significant stabilization (the structure, INT5E, was found to have a ΔG value of +23.1 kcal mol-1 with respect to INT2E). Naturally, the reduced affinity towards the zirconocene also
comes with a reduced proton accepting capability, so simply altering base strength is not expected to yield a clear improvement regarding formation of this dihydrido species.
Table 2: Barrier heights for H-H activation and thermodynamics of dihydride and Zr3H4 formation with
the Lewis acid variations
Lewis acid type TS1‡ (kcal mol-1) TS2‡ (kcal mol-1) ΔG
INT6 (kcal mol-1) ΔG1 (kcal mol-1)
Cp2Zr+ 19.6 28.2 19.8 5.6 Cp2Hf+ 21.3 29.2 18.3 7.2 Cp2Ti+ 20.1 23.6 14.5 40.5 Cp*2Zr+ 21.8 16.9 8.3 - SiMe2Cp2Zr+ 21.2 29.2 21.3 0.0 EthCp2Zr+ 22.0 29.5 21.5 2.1 rac-SiMe2(Ind)2Zr+ 28.0 28.5 21.3 44.5
Besides variations in the amine, also variations in the metallocene were investigated. As was to be expected beforehand, replacement of Zr with Hf only had a little influence on the calculated free energy profile. For Hf, energies for both H-H activation barriers are slightly raised in energy compared to the barriers for Zr, by about 2 and 1 kcal mol-1 each. Formation of INT6H appears to be slightly
favourable over INT6A, with a difference of ~2 kcal mol-1. Formation of 1A is slightly more favourable
than 1H by ~1 kcal mol-1. Contrarily to hafnium, titanium actually displayed very different
characteristics than zirconium. Where hafnium and zirconium are very similar in terms of size and charge density (due to the lanthanide contraction), titanium is much smaller than the two.29
Consequently, the barrier for heterolytic activation of H2 was significantly lower (~5kcal mol-1), and
also the resulting dihydride species was much more stable (also ~5kcal mol-1). Interestingly, the
formation of the trimeric product, 1G, was calculated to be thermodynamically uphill by 41 kcal mol -1.
Substitution of the ligands also had a pronounced effect. Specifically the replacement of Cp by Cp* leads to a rather deviating path. For this case, TS1M is almost 2 kcal mol-1 higher in energy than
TS1A. Furthermore, the activation energy for the heterolytic splitting was determined to be significantly lower by about 11 kcal mol-1. Also INT6M was much more stable than INT6A with ~12 kcal
mol-1. Most contrasting for the Cp* system is the fact that no local minimum could be found for the
11
Cp*2ZrH+ fragments. Intuitively, these trends can be understood by the increase in steric bulk. The
increase in steric repulsion between the Cp* and the amine could destabilize the INT2M adduct species, thereby decreasing the barrier height for the heterolytic activation which relies on the dissociation of the two. Along the same lines, it is easily rationalised that the Cp* ligand blocks the formation of 1M, as there would not be enough space to fit the methyl groups. The ansa-systems displayed behaviour in line with the above mentioned trends. The SiMe2 and ethylene bridged systems
were found to slightly raise the barriers for both activation steps compared to the non-bridged system. In this case, the INT6J was found to be roughly 2 kcal mol-1 higher, and 1J was found to be ~6 kcal mol -1 lower in energy than its non-bridged counterpart. Apparently, the more “open” structure does not
contribute to the H-H activation steps, although seemingly, it does facilitate a more favourable formation of the final products, 1J and 1K. This makes sense, as the Cp ligands are now pre-organised, to accommodate for a larger coordination site at the Zr cation. The reaction path was computed for the ethylene bridged system, because it was hypothesised that the mouth of the zirconocene would be more “closed” here. However, this appeared not to be the case upon closer inspection of the found transition state structures (structures are provided in SI). The rac-SiMe2(Ind)2Zr+ system experienced
an increase in barrier height for the first σ-bond metathesis step (TS1L) and ~1 kcal mol-1 decrease in
the heterolytic activation barrier (TS2L). Again, here it is observed that the addition of steric bulk severely hinders formation 1L. Although a local minimum was found (Figure 6), the endergonicity of the reaction (44.5 kcal mol-1) indicates that no such species should in fact be present in solution.
Clearly, a significant deformation of the original rac-SiMe2Idn2Zr+ structure is required to form the
trimer.
Figure 6: Gas-phase optimized structure of 1H. The indenyl ligands are pushed upwards to accommodate the formation of the trimer, resulting in an increase in energy (optimized at TPSSh/def2-SVP level of theory).
Proton affinity versus zirconocene affinity
One interesting observation is that, an increase in sterically more congesting substituents, results in a lowering of the heterolytic H-H splitting activation barrier. Presumably, the increase in steric repulsion between the Lewis acid and Lewis base results in easier dissociation, leading to a lower transition state (One raises the relative energy of the starting reagent, by introducing this steric hindrance). To further quantify the amount of steric hindrance associated with the different amine fragments, the affinity for these fragments towards their Lewis acidic CpZrH+ counterparts is
compared to their affinity towards H+. It is argued that the steric hindrance has a profound effect on
the Cp2Zr+H affinity (ZrA), while the effect on the proton affinity (PA) should be minimal. This way, the
12
and PA values are depicted in Table 3. The affinities are calculated as described in the computational methods section. All values are given relative to the values obtained for the TMA system. A higher PA thus indicates a more basic system.
Table 3: Calculated affinities for different amine fragments with the Lewis acidic Zr fragments and with a proton.
Amine fragment Cp2ZrH+ affinity relative to TMA
(kcal mol-1)
Proton affinity relative to TMA (kcal mol-1) TMA 0.0 0.0 DMEA -1.0 0.9 DMA -6.4 -9.1 DMAA -5.9 2.9 DMTFMA -14.0 -25.3 DMFPhA -7.6 -9.9 DMTFPhA -5.5 -12.0
Moving from TMA to DMEA to DMAA (not DMA), an increase in PA and a decrease in ZrA is observed, confirming that the increase in steric repulsion is indeed responsible for the more favourable Lewis adduct dissociation. It is observed that the adamantyl system has a significantly lower ZrA than TMA, whilst the PA increases (becomes more basic). This explains why, for this system, significant lowering of the heterolytic activation barrier, TS2D, was observed, while also contributing to more favourable thermodynamics for formation of the dihydride, INT6D. The calculations on the DMTFMA system show a strongly reduced PA, compared to its non-fluorinated analogue, TMA, as well as a strongly reduced ZrA. This confirms that the weakening of the Zr-N bond can indeed be attributed to the electron withdrawing effects of the CF3 group. The same effect is observed when DMFPhA and
DMA are compared, however, the effects are much less pronounced. The more favourable adduct formation for the DMTFPhA system compared to the DMA system, appears not be of electronic nature, as the PA is actually lower here. However, again here we argue that this is due to the Zr-F interaction which was already discussed previously. Certainly this stabilizing interaction is responsible for the otherwise surprising trend in reactivity.
13
Figure 7: Zirconocene affinities plotted on the steric-free proton affinity scale.
To rationalise the observed trends, it is important to take into account the overall reaction (Figure 1). Taking a look at the heterolytic hydrogen activation, three molecules of INT2 cation react with dihydrogen to form 1, two free amine molecules and the protonated amine salt. Therefore, it is expected that the driving force will be partly determined by the ability of the zirconium fragment to form 1, and partly by the ability of the amine species to accept a proton. Effects of both of these factors are recovered from the reported data. The calculations show that sterically encumbered Lewis acids such as the Cp*2ZrH+ and rac-SiMe2Ind2ZrH+ drastically change the reaction thermodynamics, as
they prevent formation of 1. On the other hand, the calculations on the amine variations show that the weaker Lewis bases result in more favourable reaction-thermodynamics. This is due to the fact that during the reaction, three amine bases are liberated for every H+ unit. Now, although at first sight
this might seem as a very effective strategy towards gaining more driving force, and also lowering the H-H activation barriers, there is a catch. The lower basicity namely severely hampers the formation of INT6 as can be observed in the DMTFMA system. It is expected that such a species will participate in the reaction mechanism for the trimer formation. It is, therefore, questionable whether this would have a positive effect after all. Moreover, the formation of 1 is actually not convenient if we consider possible applications. This form of the hydride is not easily pictured to participate in complex chemical processes, such as (asymmetric) hydrogenation or other chemically interesting reduction reactions. For this reason, rather than looking at formation of 1, it is more interesting to look at the factors that affect the relative stabilization of the INT6 intermediate. As was already discussed previously, significant stabilization is observed in the Cp* system. Here a 11 kcal mol-1 stabilization is established
compared to the Cp system (transition states are displayed in Figure 4). Similarly, introducing large steric bulk on the nitrogen also resulted in lower barriers, as was observed with the adamantyl system (INT6D was 9 kcal mol-1 lower than INT6A). Also replacing Zr for Ti yielded lower barriers and more
stable dihydride species, while simultaneously preventing formation of the trimer species 1. As a final experiment, we tried to combine some interesting features of the set of investigated ligands and metals, by computing the path for a hypothetical Cp2Ti-NMe2Adamantyl system (computed reaction
mechanism is depicted in Figure 8). This proves that combining the different features can drastically change the reaction thermodynamics.
TMA DMEA DMTFPhA DMTFMA DMA DMFPhA DMAA -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 -30 -25 -20 -15 -10 -5 0 5 10 ZrA w .r.t . T MA (kcal/mol) PA w.r.t. TMA (kcal/mol)
14
Figure 8: Computed reaction path for the Cp2Ti-NMe2Adamantyl system (Optimizations at
TPSSh/def2-SVP level of theory, and single points at M06-X2/cc-pVTZ).
Although it is tempting to draw far-reaching conclusions from such calculations, it must be kept in mind that it is unlikely that the same chemical behaviour is maintained, when large changes in thermodynamics occur due to the introduced variations. Synthesis of some of the aziridinium cations that are discussed, might be very difficult or even impossible. In particular, behaviour of the Ti system is unpredictable. Here, alternative reaction paths are very likely, for example the dimerization of INT2 to form a TiII or TiIII species and dihydrogen. Therefore, we want to emphasize that the reported
calculations are intended to gain insights into the chemical aspects which determine reactivity for the particular system that was reported by Budzelaar, Rocchigiani and co-workers. That being said, it would be interesting to experimentally investigate some of the examples that were computationally examined in this work. Besides this, we also argue that the trends that are established here, could be extrapolated to systems bearing different ancillary ligands, which could facilitate even more interesting oxidation states and/or steric properties.
Activation of the Si-H bond in SiH
4.
To further probe the potential of the zirconocene amine systems, we investigated reactivity towards the Si-H bond. As this bond is more polarised than the H-H bond, we expected that this could result in more facile activation. Although, again here it is very difficult to predict what kind of reactivity will be observed in experiments, we did perform some calculations on reactivity, in an analogous way to what was to the H-H activation mechanism. We were only interested in the heterolytic activation step, so we did not consider reactivity of the aziridinium cation. The computed mechanism is depicted in Figure 9.
15
Figure 9: hypothetical pathways for Si-H bond activation using the reported zirconocene amine complex.
Interestingly, both addition of the SiH4 to the zirconocene, as well as the decisive Si-H bond
activation steps appear to be barrierless, indicating more facile activation for SiH4 than for H2.
However, it must be said that the system demonstrates shortcomings of the used methodology. Local minima and saddle-points were approximated by applying simple corrections to minima in the electronic energy surface (see computational section). However, barrierless electronic activation does not necessarily imply that the same process is also barrierless in ΔG (for example, entropic factors could hamper addition of two fragments). Unfortunately, performing analysis on the Gibbs free energy surface is way more complicated, and unfeasible in practice.30 Another interesting aspect of the
observed calculation is the coordination mode of SiH4 to the zirconocene. Instead of binding via the
Si-H bond, a structure is preferred, where the fragment is bound via two different hydrogen “bridges” (see Figure 10). In the end, the products appear to be thermodynamically uphill by a very large margin. Presumably, the formed dihydride could again complexate into a trimeric structure to lower the thermodynamics. Also the H3Si-NMe3 product could be a determining factor here. Experimental work
16
Figure 10: Optimised structure of the Cp2Zr+H2SiH2NMe2Ph intermediate. Interestingly, the silane is only
indirectly bound to the Zr cation, via two bridging hydrogen atoms (optimized at TPSSh/def2-SVP level of theory).
Conclusions
The performed calculations show a few well balanced aspects of the system under investigation. Increase in steric repulsion between the amine and zirconium fragments leads to a lowering of the activation barriers for heterolytic H-H splitting, as well as thermodynamically more stable Cp2ZrHH species. Stronger amine bases lead to a less favourable overall reaction due to the
formation of two molecules of free base in the final product, however, influence on the dihydride formation is actually reversed. The main driving force of the investigated reaction is the formation of the hexagonal Zr3H4 centred product species. However, it is very questionable whether this type of
hydride would be useful in catalysis. It is, therefore, interesting to look at the relative energy of the Cp2ZrHH intermediate. We see that usage of Cp* instead of Cp, and large substituents on the nitrogen,
such as adamantane, should favour formation of this intermediate, while also lowering the transition state for heterolytic HH activation. Calculations also predict similar trends to occur, if Ti is substituted for Zr. It is very questionable whether these large thermodynamic changes would not lead to completely deviating reaction paths. Nonetheless, our calculations do show which factors determine the critical reaction barriers and thermodynamics of the system under investigation. In the final section, we argue that the investigated complex could also be applied for activation of different substrates. We show that activation of the Si-H bond should be relatively fast, however, in this case it remains difficult to draw conclusions purely on a computational basis.
17
Chapter II: Aniline C-N Coupling in β-Diiminate Rhodium Anilide Dimer
Abstract
A recently reported, ortho-selective aniline coupling reaction was investigated using density functional theory. By exploring a range of reactions, a plausible overall-reaction was established for the conversion of (β-diiminate-Rh)2(μ-NHPh)2 to the C-N coupled product, liberating H2. From here
multiple reaction paths were investigated. Since calculations showed that complete dissociation of the dimer is unlikely, focus was laid on intramolecular mechanisms involving a single dimer. Breaking of the Rh-N bond was considered, followed by addition of the nitrogen to the phenyl carbon atom. Our calculations show that this kind of reactivity, where the aromaticity in the phenyl ring is broken, is energetically unfeasible. Alternatively, one could picture a [2+2] mechanism, where simultaneously to the C-N bond formation, the Rh centre abstracts the hydrogen atom. Although, in this way aromaticity could be preserved, the found transition states for this type of reactivity were again too high in energy. Finally, intramolecular hydrogen atom abstraction between the two anilide units appeared feasible. This way it seemed possible to form a reactive nitrene species. However, again no possible reaction paths were found from there. Possibly, the real reaction mechanism involves an intermolecular step, where a Rh centre is exchanged between two molecules in an associative manner. We show that such an additional (β-diiminate)Rh unit can stabilize the loss of aromaticity in the direct C-N coupling mechanism. Unfortunately, an in-depth investigation of all possible mechanisms involving this type of (β-diiminate)Rh “hopping” would not be possible within the timespan of this project.
18
Introduction
C-H functionalization is one of the key steps in organic synthesis. Modern-day concepts of sustainable chemistry express the need to develop catalysts, to minimize chemical and energy waste. It is, therefore, interesting to investigate chemical reactions where, by means of a transition metal, direct functionalisation of the C-H bond is observed. In recent decades multiple transition metal catalysed aniline couplings have been reported.31–34 In this chapter, a recently reported ortho-selective
reaction by the group of Budzelaar et. al. is investigated using density functional theory.35 A good
understanding of the nomenclature of the different types of active nitrogen species is essential in this field. Therefore, a small overview of the nomenclature of the structures, is provided in Figure 1.
Figure 1: Lewis structures and nomenclature of nitrogen species reported in this thesis, starting from the left with the amine form in aniline. In the reported reactions, the nitrogen lone-pairs were coordinating to (BDI)Rh centres.
In the original publication, addition of two equivalents of lithium anilide to one equivalent of a β-diiminate (BDI) rhodium(II) bromide dimer, yielded a two-rhodium-two-nitrogen centred complex (1A, see Figure 2 left arrow). The RhII stabilized amido species, was hypothesized to have some radical
aminyl character upon opening the four-centred ring. Interestingly, after given time, conversion of 1A to the C-N coupled product (2) was observed, effectively losing two hydrogen atoms (see Figure 2, second step). Unfortunately, side reactions could not be disregarded due to the low yield of the reaction (15%). This leaves the fate of the hydrogen moieties open for speculation. Both the possibilities of formation of dihydrogen and hydrogen abstraction by a (yet unknown) reactive species in solution, must be considered. In the publication two hypothetical reaction paths were suggested. The authors argued that the C-N bond could form directly from 1A by opening of the Rh2N2 core.
Re-aromatization could then be the driving force to lose molecular hydrogen and eventually form 2. Alternatively, they suggested that one coordinated anilide could undergo hydrogen atom abstraction (HAA) from the other, resulting in a nitrene and amine complex. The formed nitrene species could then insert into the C-H bond to form the coupled product.
Figure 2. Observed reaction of aniline Rh-BDI dimer (1A) converting to ortho-coupled aniline product (2).35
In this thesis, the reported reaction is investigated using density functional theory calculations. First, a likely overall-reaction is proposed, based on thermodynamic results. Then, possible reaction mechanisms are explored. The goal of the research is to find a complete reaction mechanism from reactants to products.
19
Computational methods
Crystal structures from the publication by Budzelaar, et. al.35 were taken as starting points for
geometry optimizations of 1A and 2. These structures were then modified to find other stationary points. Geometry optimizations and transition state searches were performed using the BP86 functional and the def2-SVP basis-set in TURBOMOLE, using an external optimizer in the BOpt package.13–17,36 From here, the hessian was computed to check for imaginary frequencies and to obtain
enthalpy and entropy corrections. Then, the resulting structures were used in Gaussian as input for single-point calculations with the M06 functional and cc-pVTZ basis.18–22 In these single-point
calculations, solvent corrections were approximated using the polarizable continuum model (PCM).23– 25 The Gibbs free energy profile was constructed by correcting the electronic energies found for the
stationary points, according to ΔG = ΔEelec + ΔHcorr - 0.67TΔScorr, where the entropic scaling factor is
introduced to correct for the damping in the liquid phase.37,38 Comparison between singlet and triplet
states within the framework of Kohn-Sham DFT has shown to be difficult.39 However, the report shows
that no significant influence by these triplet state optimized structures can be expected for the overall-reactivity, making more precise investigations of spin states unnecessary.
Results & Discussion
Reconstructing the overall-reaction
Before potential reaction paths can be investigated, conclusions must be drawn on the overall-reaction and specifically the fate of the “disappearing” atoms of hydrogen. To this end, multiple possibilities were explored. In the original publication, spontaneous conversion of a relatively pure sample (90%) of 1A to 2 is reported. For this reason, we limit our consideration here, to only 1A as being the reactive species. The relatively low yield of the reaction (15%), however, does allow for side reactions to occur, which could be important. Free energies of a range of plausible reactions were calculated and are depicted in Figure 3. Dissociation of 1A into two monomeric (BDI)Rh aniline species (either into two equivalent radicals, or into one nitrene and one amine species) was considered as a reasonable activation step. The two monomer species could then act as hydrogen acceptors, facilitating the reaction of a second unit of 1A to 2 (see Figure 3, reactions A and B). Calculation predicted reactions A and B to be endergonic by 13.0 and 61.2 kcal mol-1, respectively (see Figure 3).
Alternatively, one can picture a situation where a single RhI BDI unit dissociates from 1, to leave a RhIII
di-aniline fragment (Figure 3, reaction C). The (BDI)RhI species could, hypothetically, accept two
hydrogen units. This reaction was calculated to be endergonic by 65.7 kcal mol-1. Finally, it was
pictured that 1A could effectively lose one aniline molecule and accept dihydrogen on the leftover fragment. We considered multiple possibilities, and most energetically favourable was the structure depicted in Figure 3, reaction D, with an endergonicity of 11.4 kcal mol-1. Interestingly, calculations
predicted that none of these hypothetical overall-reactions would occur spontaneously. Therefore, also the possibility was investigated of an of 1A as a whole, accepting a hydrogen atom on either one of the Rh or N moieties (Figure 3, reaction E). However, also this was predicted to be thermodynamically uphill by 26.4 and 34.5 kcal mol-1, respectively. Ultimately, formation of molecular
hydrogen was considered as the final product (Figure 3, reaction F). Although, this reaction was predicted to be endergonic by 9.7 kcal mol-1, loss of gaseous dihydrogen could, in theory, push the
20
Figure 3: Considered overall-reactions. Reactions are displayed and ΔG values are given on the right. The letter corresponds to the reference in the text above.
The hypothesis of a reaction involving H2 evolution is strengthened when possible free
energies of dissociation of dimer 1A are calculated (see Figure 4). Dissociation to form the two radicals was calculated to be uphill by 39.4 kcal mol-1 (Figure 4, A). Alternatively, asymmetric dissociation into
an amine and a nitrene, was predicted to be endergonic by 48.3 kcal mol-1 (Figure 4, B). Finally, both
dissociation of a free (BDI)RhI unit and dissociation of an aniline unit from 1A were predicted to be
endergonic by 69.5 and 32.9 kcal mol-1, respectively (Figure 4, reactions C and D). This strongly
indicates that none of these species would be detectible in solution. Additionally, the free energy of dissociation of a (BDI)RhI species from the final product (2) was calculated. This reaction appeared to
be uphill by 42.4 kcal mol-1, further strengthening our belief that no free (BDI)Rh species could be
active in solution. For the remainder of the report, intramolecular conversion of 1A to 2 is considered, releasing gaseous dihydrogen (Figure 3, F). Degasification from solution could then drive the thermodynamically uphill reaction to completion. Based on the calculation up to this point, it cannot be excluded that (BDI)Rh fragments can exchange between the different complexes in an associative manner. However, complete consideration of all possible exchange paths and subsequent potential activation routes, was not possible within the timespan that was available for this thesis.
21
Figure 4: Dissociation possibilities of 1A. Reactions are displayed and ΔG values are given on the right. The letter corresponds to the reference in the text above.
Investigation of potential reaction mechanisms
Now that a plausible overall-reaction has been established, a more thorough investigation of the mechanism can be conducted. By evaluation of the structures of 1A and 2, already three key steps can be identified. At a certain point during the reaction, bond formation between the rhodium coordinated amido moiety and the ortho-carbon on the phenyl ring occurs. Another (or possibly the same) step must liberate a dihydrogen molecule from the complex. Finally, somewhere during the reaction the two Rh centres dissociate, and one (BDI)Rh fragments coordinates to the phenyl ring that either is to react with the opposing aniline, or has done so already. Although it is understood that these steps must occur, nothing can yet be said about the order of the steps. Still a huge amount of possible reaction paths can be pictured. Therefore, it was attempted to approach this problem in a rational and systematic way. Some preliminary guidelines on reactivity can be found in literature. Free aminyl radicals have been reported as active species in aliphatic hydrogen atom abstraction (HAA) reactions. Being coordinated to formal RhII centres, the nitrogen atoms are expected to have a degree
of aminyl character upon dissociation of one of these N-Rh bonds. Possibly, sufficient radical character still is present in the nitrogen atom to allow direct hydrogen abstraction from the ortho position on the aniline fragment. Cundari and co-workers computationally investigated the reactivity of pincer 3d-metal aminyl complexes towards C-H activation in methane.40 They reported transitions states for
both concerted [2+2] type mechanisms and hydrogen atom abstraction by the aminyl with reasonable ΔG‡ values. Rather than direct aminyl reactivity towards the phenyl C-H bond, initial hydrogen
abstraction from the other aminyl moiety can be pictured. Subsequently, the resulting nitrene species could insert into the C-H bond. Both concerted and stepwise (HAA followed by radical rebound)
22
mechanisms have been found for such nitrene insertions.41 In addition, computational work by Bao,
et al. shows nitrene intermediates can also insert into aromatic C-H bonds via formation of an aziridinium intermediate, where, temporarily, the aromaticity of the system is lost.42 Perhaps,
coordination of one rhodium centre to the aromatic ring could stabilize such an intermediate. Since the observed reaction was carried out under normal reaction conditions, this computational research will look into reaction paths with barriers in the range of 20 to 25 kcal mol-1 above the starting product.
Figure 5: Structures of investigated potential reaction intermediates. Hydrogen atoms that should
eventually form H2 are depicted in red. For the remainder of this report, (BDI)Rh units will be depicted
as [Rh].
Evidently, a large amount of reaction paths can be pictured. To find a plausible mechanism, we first computed reasonable intermediates that could occur in the reaction before we searched for transition states. In Figure 5, key intermediates are depicted that could potentially play a role during the investigated mechanism. In the following section direct aminyl type reactivity towards the phenyl C-H bond is discussed. In the subsequent section, mechanisms are discussed where first a nitrene species is formed, only then to react with the phenyl C-H bond.
Direct aminyl reactivity towards the phenyl C-H bond
As was discussed earlier, some degree of aminyl character can be expected from the amido moieties in 1A. For this reason, it is not unlikely that C-N coupling can occur directly from this initial complex. Interestingly, 1A exhibits a C2 symmetric structure, where both phenyl moieties of the aniline
fragments are facing the same direction (see Figure 6). Calculations showed that conversion to the trans-isomer (1B, phenyl rings facing opposite directions) would be 16.0 kcal mol-1 higher in ΔG value.
23
Possibly, this could be an initial activation step by pre-organizing the amido site towards the phenyl ring.
Figure 6. Optimized cis (1A, left) and trans (1B, right) isomers of the starting structure (BP86/def2-SVP level of theory). Non-relevant hydrogen atoms are omitted for clarity.
From this point, C-N bond formation can be envisioned to occur from either 1A or 1B via a concerted [2+2] mechanism. Effectively, the Rh-N bond, and the C-H bond will be broken to form a Rh-H and a C-N bond (see Figure 7). Two possible stereoisomers can then be identified, differing in their configuration around the active nitrogen atom (see Figure 7, 3A and 3B). Optimized geometries of these hypothetical structures were calculated to be 28.1 and 22.7 kcal mol-1 higher in ΔG‡ value
than 1A. Based on this, these structures could possibly act as intermediates during the reaction mechanism. However, relatively large structural changes are required, and finding transition states from 1A or 1B to 3A or 3B proved to be very difficult. One saddle point was located, with a single imaginary vibrational frequency corresponding to formation of a CN bond in 1A, with the H atom having (minimal) interaction with the Rh centre. However, this structure was endergonic by 59.3 kcal mol-1 with respect to 1A. One particular problem with computing this type of bond opening is that
standard DFT lacks the capability of describing the transition from the starting closed shell singlet to an open shell singlet state. This means that our computations do not consider possible unpairing of electrons in this transition state. Although this could possibly have a significant effect, we do not believe that this effect would be large enough to lower the energy of the transition state to a kinetically feasible level.
Figure 7: Lewis structures of stereoisomers of 2+2 reaction product of 1A and 1B.
This type of concerted, one step reactions, would be the most straight-forward way to reach C-N bond formation starting from structure 1A. However, finding the corresponding transition states appeared difficult. Literature reports of [2+2] reactions in similar systems, often show formation of a N-H bond and an M-C bond, instead of the investigated N-C and M-H bond formation. To further examine whether such reactivity could potentially activate the system in consideration, we also optimized the resulting structure from a [2+2] reaction, forming a C-Rh bond and a N-H bond (see
24
Figure 5, 4). Calculations showed that this structure had a ΔG value 33.2 kcal mol-1 higher than 1A,
eliminating this type of route as well.
In the original publication by Budzelaar et al., it was suggested that a mechanism could possibly involve direct addition of the aminyl moiety to the aniline phenyl ring. A follow up step, could then eliminate H2, to regain the aromaticity. Geometry optimizations of this “adduct compound”
revealed that such a mechanism is questionable. Again, multiple stereoisomers are possible, with both the active nitrogen and carbon atoms being chiral centres after reacting. Finding all four local minima proved difficult. One stereoisomer (Structure 5 in Figure 5) was localized relatively easily, but computations revealed a ΔG value of 50.8 kcal mol-1 above 1A, making this kind of mechanism very
unlikely. As an experiment, the same structures were optimized where the (BDI)Rh was coordinated to the aniline ring already (see Figure 8, right), as would be the case for the final product. Interestingly, it is observed that these structures are only slightly higher in energy than 1A, with lowest ΔG values only being endergonic by 6.9 kcal mol-1. Likely, the flexibility in the binding mode of the phenyl ring to
the (BDI)Rh fragment, can compensate for the loss of aromaticity that occurs due to the C-N bond formation.
Figure 8: Optimized structure (left) and structure(right) of the stabilized non aromatic adduct species (BP86/def2-SVP level of theory, non-relevant hydrogen atoms are omitted for clarity).
These results raise the question whether it is actually possible to first have coordination of the (BDI)Rh to the phenyl ring, and only then have C-N coupling (clearly the reaction depicted in Figure 8 is not a one-step mechanism). The possibility of breaking a single Rh-N bond was explored, to facilitate coordination of the corresponding phenyl ring to the rhodium centre. Again, finding local minima proved difficult, and the only structure that was found had a ΔG value 35.8 kcal mol-1 higher than 1A
(see figure 9). We therefore conclude that it is unlikely that one (BDI)Rh fragment in 1A, can coordinate to the phenyl ring of one of the aniline moieties, in an intramolecular way. It was already discussed that complete dissociation of one (BDI)Rh unit, from 1A or 2, is energetically unfeasible. Perhaps, an associative mechanism can be pictured between two dimers to provide an additional (BDI)Rh centre, which could coordinate to the phenyl ring. In that case, the loss of aromaticity, after C-N bond formation, could potentially be stabilized by the changing in binding mode of the ring to the Rh (for example, η2 to η4 ). However, such a mechanism will be very hard to explore in-depth, given the
number of possible approaching orientations of the two dimers, the manifold in adduct conformers and isomers, and the diversity in possible further reaction mechanisms. Nonetheless, some hypothetical reactivity was explored, assuming that this exchange could be possible. Simple geometry optimizations of a (BDI)Rh fragment transfer between two molecules of 2 appeared to be thermodynamically feasible (endergonic by 13.3 kcal mol-1) and also experiments have shown that
25
fragment in 2 to the phenyl ring in 1A. This reaction was calculated to be uphill by 29.6 kcal mol-1.
Considering that this is a local minimum, and that transition states will thus even be higher in energy, it is not very likely that this structure participates in the reaction path. However, different conformers are easily missed for this kind of mechanism and the order of reaction steps is difficult to predict. It cannot be excluded that a similar path does take place, but we were unable to find it.
Figure 9: Optimized structure resulting from opening of the Rh2N2 centre to provide space for phenyl
rhodium coordination (BP86/def2-SVP level of theory, non-relevant hydrogen atoms are omitted for clarity).
To summarize, a variety of reaction paths were explored where the aminyl adds directly to the phenyl carbon. Although, thermodynamics for [2+2] reactions involving the Rh-N and C-H bond seemed promising, transition state searches proved that such reactions would be extremely slow (read: impossible). One issue here is that standard DFT does not consider spontaneous unpairing of electrons in a closed shell singlet state, even though this is likely to occur during the Rh-N bond elongation. Additionally to these σ-bond metathesis mechanisms, complete breaking of the Rh-N bond was considered, followed by direct addition of nitrogen to the phenyl ring. However, ΔG values of the hypothetical intermediates revealed that such a mechanism is not possible, when no additional stabilization is provided to compensate for the loss in aromaticity in the phenyl ring. Coordination of (BDI)Rh to the phenyl ring could stabilize the loss of aromaticity, however, we were not able to find a plausible route towards this coordination in an intramolecular way. Intermolecular exchange of (BDI)Rh centres would result in a very large number of possible realisations. Explicit consideration of all of these paths would, unfortunately, not be possible in the timespan of this project.
Nitrene formation and subsequent reactivity towards the phenyl C-H bond
As was discussed before, nitrene intermediates have been reported to insert into C-H bond in a variety of ways, making it an interesting possibility for this system as well. Two intramolecular reactions were investigated to form a nitrene species from 1A. A local minimum was found, where the abstracted hydrogen was μ2- bridged in between the two rhodium centres (Figure 5, 6A). However,
this triplet complex had a ΔG‡ value 35.6 kcal mol-1 higher than 1A, eliminating it from actively
participating in the reaction mechanism. Alternatively, we considered abstraction of the H atom by the opposing amido moiety. This resulted in structure 6B (see figure 10), which was endergonic by only 14.8 kcal mol-1 with respect to 1A. Additionally, the possibility of an intermolecular hydrogen
26
M06/cc-pVTZ level single points, showed that intermolecular hydrogen atom transfer, of one unit of
1A to another, is thermodynamically uphill by 34.5 kcal mol-1 (Figure 10, this was the most favourable
of multiple isomers, including one where the hydrogen is accepted over the two Rh centres). When taking the earlier reported thermodynamics from Figure 3 and 4 into account, it appears unlikely that another (monomeric) species would act as temporary hydrogen acceptor during the reaction.
Figure 10: Above - Possible reactions to form a nitrene intermediate. Abstracted hydrogen is pictured in red. Below - Optimized structure of 6B (BP86/def2-SVP level of theory). Non-relevant hydrogen atoms are omitted for clarity.
Evidently from the geometry optimized structures in Figure 9, the formed amine moiety (which thus appears to be a requisite for nitrene formation), is only bound to one single Rh centre. From here, a [2+2] reaction could be envisaged where a Rh-H bond would form, accompanying the formation of the C-N bond (see Figure 11, 7). C-N coupling was investigated where the Rhodium centre acted as hydrogen acceptor (Figure 11, above). Geometry optimizations of the effective product of this reaction resulted in complex 7 (see Figure 10, right). Interestingly, single-point calculations at the M06/cc-pVTZ level of theory showed that the overall reaction of 1A to 7 was endergonic by only 22.8 kcal mol-1, making a more in-depth investigation worthwhile. We want to mention here that we are
27
carbon and hydrogen atoms end up being bound to the inserting nitrogen (Figure 11, below). However, we argue that such a mechanism is unlikely to be active for the system under consideration in this thesis. As was shown, nitrene formation appears to only be possible via an intramolecular abstraction. Consequently, the formed amine moiety, opposite to the nitrene, has only a single lone-pair available for coordination to either one of the Rh centres. Since direct insertion of the nitrene into the C-H bond would again result in formation of another amine moiety, this would result in both nitrogen only being able to bind to a single rhodium centre The reaction forming the free (BDI)Rh here was calculated to be endergonic by 28.0 kcal mol-1 (see Figure 10, below). Therefore, this type of
insertion is unlikely to occur during the reaction mechanism.
Figure 11: Left - [2+2] type C-N formation, from 6B (above) and concerted nitrene insertion, resulting in a free (BDI)Rh centre (below). Right - Optimized structure of 7 (BP86/def2-SVP level of theory, non-relevant hydrogen atoms are omitted for clarity).
For the conversion of 6B to 7 two possible mechanisms were considered. A one-step, concerted mechanism, was investigated first, where the C-N bond was formed simultaneously with the Rh-H bond in a σ-bond metathesis reaction. For this mechanism, a saddle-point was localized with a single imaginary frequency corresponding to formation of the C-N bond. Unfortunately, the ΔG‡ of
this transition state was determined to be 73.4 kcal mol-1 higher than that of 1A, displaying the rigidity
of the system under investigation. Ultimately, the possibility of a stepwise mechanism was considered, where first the nitrene would add to the phenyl ring in a direct manner (see Figure 12), only then to have a hydrogen abstraction as a consecutive step to regain aromaticity. Multiple stereoisomers of 8 are possible after the first addition step, with the carbon atom being a chiral centre. The most favourable stereoisomer was calculated to have a ΔG value of 48.5 kcal mol-1 above 1A, ultimately
eliminating this reaction path as well. The strong interaction between the Rh centre and the phenyl ring (2.11Å), again shows that there is a tendency for the Rh centre to compensate for loss of aromaticity.
28
To the best of our understanding, there are no obvious reaction paths that have not been discussed until now. The suggestions that were put forward in the original publication were investigated, and have been shown to be less straight-forward than originally pictured. We investigated pathways to pre-activate 1A, by forming a nitrene species. Our calculations suggest that nitrene formation is only possible within one single dimer structure. This indirectly eliminates pathways where the nitrene inserts in a concerted manner into the C-H bond, limiting the number of possibilities. In the end, the chemical behaviour of the system appeared to be too complex to solve easily. Possibly, a bit of residue chemicals could have been left in solution, which could either catalyse the observed reaction, or even act as stoichiometric reagent. Alternatively (and perhaps more likely), a mechanism is active where one (BDI)Rh centre can “hop” from 1A or 2 to one molecule of 1A. This additional Rh centre could then potentially catalyse the observed reaction. Finally, also computational limitations could have been a source of error. It was mentioned that DFT is not capable of describing spin unpairing upon bond opening, and also the applied BP86 functional could be inaccurate for this system. In the end, follow-up experimental research could shed a light on the complex behaviour of this system. Perhaps, the reaction mixture can be analysed more extensively to find side-products. In addition, kinetics could be interesting, as the rate order in 1A could confirm an intermolecular mechanism (only when this step is rate-limiting).
Conclusions
Guided by literature, multiple potential reaction routes towards C-N bond formation in a dimeric [β-diiminate] rhodium anilide complex (1A) were investigated. By calculating thermodynamic properties of the initial dimer and potential product complexes, it was concluded that completion of the investigated reaction relies on formation of gaseous dihydrogen. Pathways involving direct C-N bond formation from 1A and pathways containing initial activation steps were considered. Both cis-trans isomerisation (regarding orientation of the aniline phenyl rings) and intramolecular hydrogen atom abstraction to form a nitrene species were found to be feasible activation steps. Starting from these geometry optimized structures, direct C-N bond formation was considered, which would form at the cost of loss of aromaticity in the phenyl ring. It was observed that the resulting intermediates from such an initial step are too high in energy compared to 1A to play an active role in the reaction mechanism. Interestingly, coordination of a free [β-diiminate] rhodium species to the phenyl ring appeared to have a highly stabilizing effect on these specific non-aromatic intermediates. However, breaking of the N-Rh bond to facilitate Rh-coordination to the phenyl ring intramolecularly, appeared to be energetically unfeasible. Calculations show that [β-diiminate] rhodium fragments are not expected to dissociate from either the initial substrate or product complex. We suspect that (BDI)Rh fragments can possibly be exchanged between starting and product complexes in an associative manner. This could potentially have a catalytic effect. However, the number of possible pathways grows very fast in such intermolecular exchange mechanisms, making thorough analysis of them impossible in the timespan given for this project. To circumvent the loss of aromaticity as a whole, a range of intermediates were explored, which could result from four-centred, σ-bond metathesis reactions, where the Rh centre abstracts the phenyl hydrogen atom simultaneously with the C-N bond formation. It was pictured that such reactions could occur either from 1A directly, after cis-trans isomerisation or even after intramolecular nitrene formation. Optimized potential intermediates, resulting from this hypothetical reaction step showed that this could be feasible thermodynamically. Unfortunately, all found transition states appeared to have ΔG‡ values, much higher than the starting
complex. In conclusion, no complete reaction path was found from initial complex to the final product, forming H2 during the reaction.
