Introduction

Friedel-Crafts (FC) reaction is an important method to incorporate carbon skeletons into aromatic system1,2. Great successes have been achieved for the hydroarylation of neutral arenes (such as toluene, anisole, and their homologues)3,4,5,6,7,8,9,10,11,12,13,14. Because the FC reactions typically require Lewis acid catalysts, for arenes containing nitrogen atom, the substrate scope of FC reactions are quite limited due to the coordination between amine and Lewis acid catalyst, except indole and pyrrole15,16,17,18,19,20,21,22,23. Being profited from the extremely weak basic properties24,25, acid-catalyzed additions of indole and pyrrole to alkenes have obtained great achievements15,16,17,18,19,20,21,22,23. However, the hydroarylation of alkaline arenes to alkenes still remains many challenges. Some researches have shown the possibility of hydroarylation between the parent anilines C6H5NH2 and alkenes26,27,28. However, the reaction of arenes with stronger basicity (such as N,N-dimethylaniline and N,N-diethylaniline29) still is a big problem, due to their ability to coordinate with Lewis acid catalyst which can lead to deactivation of the aromatic ring30. Furthermore, alkaline arenes can trap the proton in the C-H activation process and the reaction will be terminated as result26.

Recently, Bertrand et al. reported an anti-Bredt cyclic diaminocarbene which showed increased π-accepting character without diminishing its σ-donor property31,32. We found that Gold(I) compound derived from this new carbene can be used as effective catalyst for the FC reaction between alkenes and N,N-dialkylanilines33. Now, these new FC reactions are receiving more and more research interests34,35,36,37. As we known, most of the electrophilic substitution reactions followed the Markovnikov rule38. For the FC reaction of alkenes, the reactions following the Markovnikov rule should form branched product3,4. Only several examples were reported on the formation of linear product by anti-Markovnikov rule39,40. For the FC reactions between alkenes and N, N-dialkylanilines catalyzed by carbene Gold(I), both Markovnikov and anti-Markovnikov hydroarylations were observed and all these reactions gave high para-selectivity products (Fig. 1)33. The selectivity to the branched or linear product was highly depended on the structure of alkenes. The purpose of this paper is to find out the reaction mechanism of this rare FC reaction and understand the origin of the reaction selectivity.

Figure 1
figure 1

The Friedel-Crafts reactions of N,N-Dimethylaniline and corresponding products (5a and 6a are the main products in the experiment33).

Full size image

Results and Discussion

The structure of the catalyst

To examine the reliability of the theoretical method, the structure of the catalyst 1 was fully optimized with both B3LYP and M06-2X methods. The calculated results were compared with the x-ray experiment data (Table 1). The calculated structure parameters match well with the experiment data. Most of the calculation errors for the bond length and angle are smaller than 1.0% and the maximum error is 1.7%. It suggests that the calculation method used here is accurate enough to reproduce the crystal structure even though the catalyst contain gold atom. Indeed, methods based on DFT using pseudopotential basis set have been widely used for the calculation of compound containing gold41,42,43,44,45,46,47,48,49,50,51. Our previous experimental results on the hydration of alkynes catalyzed by gold(I) isocyanide were successfully explained by theoretical calculation based on DFT41.

Table 1 Some important structure parameters of the carbene gold (I) catalyst 1 obtained by calculation and experimenta.
Full size table

According to the experimental results, the conversion of the hydroarylation between N,N-diethylaniline and α-methylstyrene is zero when only 1 is used as catalyst. However, the conversion is 97% when chlorine scavenger reagent KBArF was added in the presence of 133. It suggests that the active center in the reaction should be anti-Bredt carbene-Au+ cation 1+ (Fig. 2). Similarly, other kinds of carbene-Au+ cations have been proposed to be the active catalytic species in a series of reactions52,53,54,55.

Figure 2
figure 2

The optimized structures of 1+, 1+…2, 1+…3 and 1+…4 (Bond length in Å). All hydrogen atoms have been omitted for clarity.

Full size image

The reaction between 2 and 3

5a is the main product in the reaction between N,N-Dimethylaniline 2 and alkene 333. There has two possible pathways to produce 5a (Fig. 3). Pathway 1: the gold cation activates the alkene 3. Then, the electrophilic alkene attacks the aromatic ring. Pathway 2: the gold cation activates the N,N-Dimethylaniline 2.

Figure 3
figure 3

Different substrate activation models.

Full size image

As we known, FC reaction is a kind of electrophilic aromatic substitution reactions, in which the hydrogen atom of aromatic ring is replaced by an electrophile1,56,57,58,59. The complex 1+…3 can serve as the electrophilic reagent (Fig. 2). It has been known that the gold cation has strong ability to bind alkene60,61. The binding between alkene 3 and catalyst 1+ is highly exothermic (33.6 kcal/mol). The C=C bond length of alkene 3 was increased from 1.347 to 1.391 Å, which indicates that the C=C bond was activated after binding with 1+.

For pathway 1, both Cc and Cd of alkene 3 attacking the Ca of N,N-dimethylaniline 2 were taken into consideration, which can produce branched and linear product (5a and 5b) respectively. In the case of Cc attacking, the activation energy for the Ca–Cc bond formation (TS1A-1) is 27.4 kcal/mol and the Ca–Cc bond length in TS1A-1 is 1.998 Å (Energies obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation were discussed if not mentioned) (Fig. 4). At the same time, the C=C double bond of alkene 3 becomes almost single bond (bond length increases from 1.391 to 1.517 Å) in TS1A-1. An unstable intermediate (Int1A-1) is produced via TS1A-1 and the Ca–Cc bond length of Int1A-1 is further reduced to 1.647 Å. The formation of Ca–Cc bond makes the Ca-H bond active. Meanwhile, the Cc–Cd bond becomes single bond in Int1A-1 (length: 1.562 Å). Though TS2A-1 can lead to the final product 5a, a direct proton transfer from Ca to Cd is not energy favorable due to the high overall activation energy (40.8 kcal/mol).

Figure 4
figure 4

(a) The pathway 1 to produce 5a by the Friedel-Crafts reaction between 2 and 3 without the assistance of aromatic amine (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06–2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

It is worth to notice that there is plenty of aromatic amines in the microenvironment33. The aromatic amine is a good proton acceptor. It can abstract the proton of Ca-H in Int1A-1 with a barrier of 22.1 kcal/mol (TS2A-1′) (Fig. 5). Then, the intermediate Int2A-1′ can re-abstract the proton of ammonium via TS3A-1′ and the corresponding energy barrier is 14.7 kcal/mol. The re-abstracting proton process is vital for the catalytic cycle. If the catalyst cannot re-abstract the proton from the ammonium, the reaction will be terminated. As an example, when 25 mol% proton-trapping reagent, 2,6-di-t-butyl-4-methylpyridine, was added in the FC alkylation of aniline catalyzed by acid, no any alkylation reaction can be observed26. TS3A-1′ can lead to the final product 5a and the overall energy change from ReaA-1 to ProA-1′ is −24.6 kcal/mol.

Figure 5
figure 5

(a) The pathway 1 to produce 5a by the Friedel-Crafts reaction between 2 and 3 with the assistance of aromatic amine (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

As indicated above, N,N-dialkylaniline is not only reactant but also a promoter. It should be the reason why excess N,N-dialkylaniline is necessary for the reaction33. The rate determining step is the Ca–Cc bond formation process via the transition state TS1A-1 (barrier: 27.4 kcal/mol). This barrier height is reasonable considering that the reaction requires a temperature of 135 °C33.

Producing 5a by the pathway that the gold cation activates the N,N-Dimethylaniline 2 and the Ca carbon of 2 attacks the Cc of alkene 3 is not possible because of the extremely high activation energy (Figs 3 and 6). The binding energy between 2 and 1+ is −35.3 kcal/mol. This binding can increase the positive charge on the Ca–H proton from 0.112 to 0.214. It makes the Ca–H proton quite active. The external N,N-dialkylaniline 2 can abstract the Ca–H proton with quite low activation barrier (12.8 kcal/mol, TS1A-2). However, the following Ca–Cc bond formation is extremely energy-unfavorable and the corresponding activation barrier is 60.1 kcal/mol. It indicates that producing 5a by gold cation activated N,N-Dimethylaniline 2 is not possible. Hence, this pathway is not taken into consideration for the producing of 5b, 5c, 6a and 6b.

Figure 6
figure 6

(a) The pathway 2 to produce 5a by the Friedel-Crafts reaction between 2 and 3 (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

If the alkene terminal carbon Cd can attack the Ca of N,N-dialkylaniline 2 (Fig. 7), anti-Markovnikov product 5b should be obtained. The activation energy for the Ca-Cd bond formation is only 22.5 kcal/mol (TS1B). Though the barrier height of TS1B is lower than that of TS1A-1, the proton abstracting process via TS2B is not easy for this anti-Markovnikov reaction. The overall activation energy for the process to produce 5b is 29.5 kcal/mol. Furthermore, the energy change from ReaB to ProB is only −1.5 kcal/mol. Hence, comparing with the producing of 5a, the generation of the linear product 5b is difficult because the higher activation energy and quite small free energy change. Indeed, no 5b was observed for the FC reaction between 2 and 3 in our previous experiment33.

Figure 7
figure 7

(a) The pathway to produce 5b by the Friedel-Crafts reaction between 2 and 3 (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

Most of the known FC reactions gave both para- and ortho-products1,62,63. Interestingly, for the reaction between N,N-Dimethylaniline 2 and alkene 3, quite high para-selectivity was obtained and 5a was isolated with 93% yield33. Due to the steric hindrance of the –N(CH3)2 group, the Cc of alkene 3 attacking the Cb carbon of 2 is difficult (Fig. 8). The energy barrier for the Cb-Cc bond formation is 31.8 kcal/mol (TS1C). Furthermore, the access of extra aromatic amine is not easy due to the steric hindrance, which makes the abstracting proton from Cb not easy comparing with the process to produce 5a. The free energy change from Reac to Proc is −7.9 kcal/mol. Considering that the barrier of the rate determining step to produce 5c is 4.4 kcal/mol higher than that to produce 5a, observing a minute quantity of 5c as byproduct is reasonable33.

Figure 8
figure 8

(a) The pathway to produce 5c by the Friedel-Crafts reaction between 2 and 3 (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

The reaction between 2 and 4

The reaction of olefine ketone 4 is quite different from that of aromatic alkene 3. For aromatic alkene 3, the Ca of N,N-Dimethylaniline 2 and Cc of 3 forms C–C bond and the producing of 5a follows the Markovnikov rule. However, the reaction of olefine ketone 4 follows the anti-Markovnikov rule. The C–C bond was formed between the Ca of 2 and the terminal carbon Cd of 4 to produce 6a (Fig. 1)33. To fully understand this difference, both Cc and Cd of 4 attacking the Ca of 2 were taken into consideration, which can produce branched and linear product (6b and 6a) respectively. The binding between 4 and 1+ can form the electrophilic group and this process releases 25.5 kcal/mol energy. The C=C bond length of olefine ketone 4 was increased from 1.341 to 1.375 Å during the formation of 1+…4 (Fig. 2).

In the case of Cd attacking, the activation energy for the Ca-Cd bond formation (TS1D) is only 15.2 kcal/mol and the Ca-Cd bond length is 2.117 Å in TS1D (Fig. 9). During the Ca-Cd bond formation, the C=C double bond of 4 was increased from 1.375 to 1.439 Å. For the reaction of aromatic alkene 3, this bond was increased by 0.126 Å (from ReaA-1 to TS1A-1). As we know, the structure distortion can induce energy change and enhance the reaction barrier64. Correspondingly, the energy barrier of TS1D is 12.2 kcal/mol lower than that of the same process to produce 5a (TS1A). The smaller structure change is responsible for the low barrier of TS1D. Similarly, proton abstracting by aromatic amine and re-abstracting the proton of ammonium happen for the producing of 6a. The proton re-abstracting process from the ammonium is the rate determining step and the corresponding activation energy is 21.1 kcal/mol (TS3D).

Figure 9
figure 9

(a) The pathway to produce 6a by the Friedel-Crafts reaction between 2 and 4 (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

If the Cc of olefine ketone 4 can attack the Ca of N,N-dialkylaniline 2 (Fig. 10), Markovnikov product 6b should be observed in the experiment. In the case of Cc attacking, the activation energy for the Ca–Cc bond formation is quite high (TS1E: 30.8 kcal/mol). At the same time, the C=C double bond of 3 becomes almost single bond in TS1E (increasing from 1.375 to 1.461 Å). The proton abstracting by the aromatic amine and re-abstracting from the ammonium is easy to achieve in this pathway. However, the producing 6b is not competitive to 6a. The activation energy of the rate determining step to produce 6b is 9.7 kcal/mol higher than that of 6a. That is why the anti-Markovnikov linear product 6a was observed in the experiment33.

Figure 10
figure 10

(a) The pathway to produce 6b by the Friedel-Crafts reaction between 2 and 4 (Energy in kcal/mol). Energies out of parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ). Energies in parenthesis were obtained at M06-2X (6–311 + G*/LANL2DZ) by PCM calculation. (b) The structures of the optimized transition states (Bond length in Å). All hydrogen atoms (except the transferred H) have been omitted for clarity.

Full size image

Conclusion

The Friedel-Crafts alkylation of N,N-dimethylaniline with alkenes catalyzed by cyclic diaminocarbene-Gold(I) complex were theoretically investigated. The calculation method adopted here is accurate enough to reproduce the crystal structure of the catalyst. The gold cation can activate the C=C double bond to produce the electrophilic [R-C=C…Au-L]+. Then, the [R-C=C…Au-L]+ attacks the aromatic ring, following the electrophilic aromatic substitution mechanism. Being different from previous result that alkaline arenes will trap the proton and the reaction will be terminated as result26, herein, it was found that the alkaline N,N-dimethylaniline can assist the reaction. Based on the obtained reaction mechanism, we can well understand why the reaction was high para-selevtivity, and why branched and linear products were obtained for different substrates: (1) Producing the para-product is more energy favorable comparing with the ortho-product. (2) The reaction of α-methylstyrene follows the Markovnikov rule. The activation energy to generate the branched product of α-methylstyrene is lower than that producing the linear product. Besides, the reaction leading to branched product is highly exothermic. (3) The reaction of butanone follows the anti-Markovnikov rule. The activation energy to generate the branched product of butanone is higher than that producing the linear product.

These theoretical results are quite useful for designing more effective catalysts for this rare FC reaction using alkaline arenes as substrates. Based on the understanding of the reaction mechanism, the development of none-noble metal catalyst for this FC reaction is on going in our group.

Computational Methods

All the structures were fully optimized with B3LYP based on Density Functional Theory (DFT) (See SI for structure details). This method is a good choice for the calculation of organometallic systems65,66,67,68,69,70,71,72,73,74,75. The following combination of basis sets were used for geometric configuration optimization and frequency calculation: 6–31G basis set for all atoms except Au and LANL2DZ basis set for Au (abbreviated as 6–31G/LANL2DZ)41,75. LANL2DZ basis set includes the relativistic effect of the heavy element65,76. The calculation method adopted here can well reproduce the crystal structure of the carbene-gold used for the FC reaction of basic arenes with alkenes (Table 1). Because B3LYP method often suffers from incorrect energies, especially for systems containing non covalent bonds. A higher level method M06-2X was used to generate more accurate energies. The energy calculations were performed using 6–311 + G* basis set for all atoms except Au and LANL2DZ basis set for Au (abbreviated as 6–311 + G*/LANL2DZ). The influence of solvent was performed in condensed phase with the Polarizable Continuum Model (PCM) using 6–311 + G*/LANL2DZ basis set. This method creates the solute cavity via a set of overlapping spheres. Aniline was used as solvent to simulate the environment of N,N-dimethylaniline.

The computed stationary points have been characterized as minima or transition states by diagonalizing the Hessian matrix and analyzing the vibrational normal modes. In this way, the stationary points can be classified as minima if no imaginary frequencies are shown or as transition states if only one imaginary frequency is obtained. The particular nature of the transition states has been determined by analyzing the motion described by the eigenvector associated with the imaginary frequency. All calculations were performed with the Gaussian 03 suite of programs77.