Palladium-catalyzed allene synthesis enabled by β-hydrogen elimination from sp2-carbon

The rational design based on a deep understanding of the present reaction mechanism is an important, viable approach to discover new organic transformations. β-Hydrogen elimination from palladium complexes is a fundamental reaction in palladium catalysis. Normally, the eliminated β-hydrogen has to be attached to a sp3-carbon. We envision that the hydrogen elimination from sp2-carbon is possible by using thoroughly designed reaction systems, which may offer a new strategy for the preparation of allenes. Here, we describe a palladium-catalyzed cross-coupling of 2,2-diarylvinyl bromides and diazo compounds, where a β-vinylic hydrogen elimination from allylic palladium intermediate is proposed to be the key step. Both aryl diazo carbonyl compounds and N-tosylhydrazones are competent carbene precursors in this reaction. The reaction mechanism is explored by control experiments, KIE studies and DFT calculations.

P alladium catalysis has proved to be a powerful synthetic tool, which is demonstrated by numerous useful transformations and highlighted by the 2010 Nobel Prize in chemistry [1][2][3] . Although the mechanism involved in those reactions have been extensively explored, efforts to acquire a deep understanding of the current mechanistic hypothesis and apply them to design new transformations have never ceased. As an elementary reaction in palladium catalysis, β-hydrogen elimination has been well studied (Fig. 1a) [4][5][6][7] . Theoretically, the hydrogen elimination can be divided into two categories according to the hybrid state of the attached carbon atom. Hydrogen elimination from sp 3 -carbon is the most common pattern, and both alkyl and alkenyl palladium complexes [8][9][10][11] can undergo this elimination pathway, affording olefins and allenes respectively. In contrast, the second hydrogen mode, where the eliminated hydrogen is attached to a sp 2 -carbon (also means β-elimination of vinylic hydrogen from 1 η-δ-allylic palladium) and allene would be generated, has not been reported yet (Fig. 1b).
On the other hand, allenes are of great importance due to their wide existence in natural products 12 , pharmaceuticals 13 , and molecular materials 14 . The active nature imparted by its unique orthogonal cumulative π-system also makes them highly versatile and useful building blocks in organic synthesis [15][16][17][18][19] . Although numerous methods for the preparation of allenes have been developed [20][21][22][23][24] , they still lag far behind the growing demand in the application. At present, the majority of the existing methodologies rely on the utilization of elaborate alkynes. Therefore, it is highly desirable to develop new approaches via new mechanistic pathways, which may deliver the allenes efficiently from easily accessible starting materials and complement the current methodologies [25][26][27][28][29][30][31][32][33] . Therefore, β-hydrogen elimination of allylic palladium from sp 2carbon represents an attractive new strategy for allene synthesis.
Here, we report the successful application of the β-hydrogen elimination from sp 2 -carbon for the allene synthesis. In our research plan, the desired δ-allylic palladium intermediate is planned to be produced from the classic allylic alcohol derivatives, which can undergo an oxidative addition/isomerization sequence in the presence of Pd 0 (Fig. 1c, pathway I). A second pathway was also devised where the cross-coupling of alkenyl halides and diazo compounds offer the desired δ-allylic palladium intermediate via migratory of palladium carbene IV (Fig. 1c, pathway II) [34][35][36][37] . In these two pathways, there is an equilibrium of 1 η-δand 3 η-π-allylic palladium intermediates. The π-allylic palladium normally showed higher stability compared with the corresponding δ-allylic one. However, The lack of a syn coplanar arrangement of C-H and C-Pd bonds, a key factor for most βhydrogen elimination, would make the hydrogen elimination from π-allylic palladium rather difficult [38][39][40] .
Allene synthsis via hydrogen elimination from C(sp 2 ) (this work)  Fig. 1 Allene synthesis based on β-hydrogen elimination from sp 2 hybrid carbon. a β-Hydrogen elimination in palladium catalysis. b Two modes of β-hydrogen elimination. c Allene synthesis via β-hydrogen elimination from C(sp 2 ) (this work).

Results
Initial study. With these considerations in mind, we set out to explore the feasibility of the planned strategy. A small amount of allene 2 was observed when allyl acetate 1 was treated by Pd (OAc) 2 /PPh 3 at 100 o C with poor conversion (Fig. 2a). However, attempts to further improve this reaction were unsuccessful, and a complicated mixture was observed when a full conversion was achieved by changing ligands or solvents. Next, we tested the cross-coupling of 2,2-diarylvinyl bromide 3a and diazoacetate 4a in the presence of Pd(OAc) 2 /PPh 3 (Fig. 2b). These two model substrates were selected because the planned elimination is expected to be promoted by the generation of stable multi-aryl substituted allenes, and the competitive elimination from a sp 3carbon will be avoided. Delightfully, the desired allene 5a was generated in high yield, and its structure was unambiguously confirmed by X-ray analysis.
Reaction conditions development. Encouraged by the above results, more reaction conditions were screened for a higher reaction yield (Table 1). Other mono-phosphine ligands, with either electron-withdrawing fluorine (L2) or electron-donating MeO group (L3), gave reduced reaction yields (entries 2 and 3). Bis-phosphine ligands were also competent to promote this reaction, and the ligand bearing a linkage of six carbon atoms further improved the reaction yield to 87% (entries 4-7). Instead of CsOAc, several other bases were also examined, but offered inferior results (entries 8-10). The reaction also went well in other ether solvents, but was rather sluggish with toluene or DCE as solvent (entries 13 and 14). While a comparable result was obtained in an elevated reaction temperature of 90 o C (entry 15), an obvious loss in reaction yield was observed at a lower temperature (entries 16 and 17).  Substrate scope of 2,2-diarylvinyl bromides with diazo carbonyl compounds. With the optimal reaction conditions in hand, we began to explore the generality of this cross-coupling reaction (Fig. 3). First, a variety of 2,2-diarylvinyl bromides 3 were used in the coupling with phenyl diazoacetate 4a. All of them afforded high yields, with a deleterious effect on the reaction outcome by introducing electron-withdrawing groups to the phenyl ring, or moving the substituents from para-to meta-or ortho-position (5d-i). Vinyl bromide with a flat terminal fluorene substitution, instead of two separate aryl groups, also proceeded well (5l). Next, the variation of the aryl diazoacetates 4 was also investigated. The methyl ester could be successfully replaced by an ethyl or benzyl ester, as well as an ethyl ketone (5m-o). Introduction of different substituents onto the para-or meta-position of the phenyl ring was well tolerated, albeit in slightly reduced reaction yields (5p-w). Compared with the vinyl bromide substrates, the diazoacetate part was more sensitive to the steric properties, as the ortho-methyl substituted phenyl ring completely blocked the coupling reaction (5x). Delightfully, other aromatic rings, like 2thienyl or naphthyl group, could provide the desired products in good yields (5y and 5z).
Substrate scope of 2,2-diarylvinyl bromides with N-tosylhydrazones. Encouraged by the above success, we sought to use diaryldiazomethanes to produce tetra-aryl-substituted allenes, which showed some unique properties in material science 41 , catalysis 42,43 , and molecular recognition 44,45 . Although a preliminary experiment with diphenyldiazomethane furnished the tetra-phenyl-substituted allene 7a in moderate reaction yield under the standard reaction conditions, further efforts were hampered by the relatively lower stability of this kind of diazo compounds. Therefore, we switched to the corresponding Ntosylhydrazones 6, a family of stable carbene precursors [46][47][48] . Gratifyingly, the slight adjustment of the base and ligand to cesium pivalate and dppe could lead to the desired cross-coupling products in good to excellent yields (Fig. 4). While electronic variation on the phenyl ring of the 2,2-diarylvinyl bromides showed marginal effect on the reaction outcome (7a-h), the introduction of electron-withdrawing groups to the diaryl ketones derived N-tosylhydrazones gave slightly reduced yields (7j-l).
Ortho-substituted phenyl rings on either vinyl bromides or Ntosylhydrazone part resulted in an obvious loss in reaction yield, consistent with results from diazoacetate species (7g and 7n).
Conversion of the obtained product. The conversion of allene 5a was tested (Fig. 5)  allenic esters can be selectively borylated by B 2 (pin) 2 to afford vinyl boronate pinacol ester 8a in 78% yield. According to the previous report 49 , the allenic esters can also undergo a sequential nucleophilic attack/cyclization process to give polysubstituted αnaphthol 8b in moderate yield. Treatment of 5a with TfOH afforded allenic carboxylic acid 8c in 78% yield, which may be used to attach this unique allene architecture to other molecules.
Control experiments. To probe the mechanism of this catalytic reaction, palladium complex 9 was prepared and subjected to several control experiments (Fig. 6). Palladium complex 9 was prepared by reaction of 2,2-diphenylvinyl bromide 3a and Pd (PPh 3 ) 4 , and the structure was verified by X-ray crystallographic analysis. See SI for details. When the mixture of complex 9 and diazoacetate 4a in THF was heated at 80 o C for 2 h, the reaction solely afforded olefin 10 (Fig. 6a). The reaction mixture was also analyzed by SAESI-HRMS, which is a direct and reliable method for the characterization of reaction intermediates in situ through a gentle ionization process [50][51][52][53] . The obtained MS spectrum showed a signal of palladium complex [C 59 H 49 O 2 P 2 Pd] + . The peaks in MS spectrum labeled as experimental m/z-relative percentage abundance matched the theoretical shown in brackets, unambiguously indicating the existence of allylic palladium species 11. The relative abundance of the isotopic ion at m/z 959.2242 was higher than the theoretical value due to the influence of the background signal nearby.
A small amount of allene 5a could be observed upon elevation of reaction temperature, with olefin 10 still as the major product (Fig. 6b). However, the preference of reaction products was completely inverted when cesium acetate was added, and only allene 5a was produced even at 80 o C (Fig. 6c). These experiments hint that the reaction generated an allylpalladium intermediate, which could undergo either protodepalladation 54 to afford olefin 10, or hydrogen elimination to give allene 5a. Such a hydrogen elimination step can be facilitated by the basic carboxylate salts.

KIE and deuterium-labeling experiments.
To gain more mechanistic insights, two deuterium labeling experiments were carried out (Fig. 7). The kinetic isotope effect (KIE) was measured in two parallel reactions using 3a and deuterium-labeled d 1 -3a. A KIE value of 1.02 implicated that the final hydrogen elimination was not involved in the rate-limiting step 55   Proposed reaction mechanism. Based on the above investigations and literature precedents, a plausible mechanism is outlined in Fig. 8. Initially, oxidative addition of vinyl bromide 3a to the Pd 0 catalyst offers the Pd II species A. Then the Pd II species A reacts with the diazoacetate 4a to form Pd II carbene species B. A subsequent migratory insertion of carbene into the Pd−C bond affords π-allylpalladium species C 56-59 , which is followed by a hydrogen-elimination to provide the desired product 5a.
DFT calculations. To gain a deeper understanding of the palladium-catalyzed cross-coupling of vinyl bromides with diazo compounds, DFT calculations were carried out for the envisioned reaction intermediates and related transition states (Fig. 9, see computational details in the Supplementary Information). As can be seen from Fig. 9a, the oxidative addition of vinylbromide 3a to the Pd(0)L 2 complex (L = PPh 3 ), which is exergonic by 9.1 kcal/mol with an energy barrier of 12.7 kcal/ mol, initiates the reaction and offers bromide coordinated species Int2. Isomerization of Int2 by exchanging the positions of vinyl ligand and PPh 3 ligand forms a more stable isomer Int3. We also explored the reaction pathway initiated by the addition of diazo compound with Pd(0)L 2 . However, as the dediazonation process in this pathway requires a high free energy barrier of 33.2 kcal/mol (see Supplementary Fig. 4 in Supplementary Information), such a mechanism will not be further considered. The subsequent addition of diazoacetate 4a to Int3 with simultaneous dissociation of one PPh 3 group is endergonic by 22.3 kcal/mol and leads to Int4 in which diazoacetate weakly occupies the vacant site of Pd(II). Further reaction of diazoacetate with Pd(II) center releases a molecule of nitrogen and leads to a Pd(II) carbene intermediate Int5. This step is calculated to be 18.8 kcal/mol exothermic with an energy barrier of 9.8 kcal/mol. A subsequent migratory insertion of the generated carbene into the Pd−C bond of Int5 affording the π-allylpalladium species Int6 is further exothermic by 47.7 kcal/ mol without any energy barrier. We also investigated an alternative reaction pathway for π-allylpalladium species formation, in which ligand exchange of bromide with the base CsOAc happens before dediazonation and migratory insertion (see Supplementary Fig. 2 in Supplementary Information). However, with an overall energy barrier of about 36.4 kcal/mol, this reaction pathway seems to be unfavorable in practice and thus will not be discussed further. Nevertheless, as shown in Fig. 9b, the ligand exchange indeed takes place after the formation of Int6 to produce base coordinated π-allylpalladium species Int7 with an endergonic reaction energy of 3.2 kcal/mol. The so-generated Int7 then undergoes β-hydrogen elimination to provide the desired allene product. At this stage, two alternative pathways, of which one involves the coordinated base (path A in Fig. 9b) while the other involves the palladium center (path B in Fig. 9b), were investigated respectively. As depicted in path A, isomerization of Int7 to δ-palladium tetra-coordinated intermediate Int8 is endergonic by 11.2 kcal/mol, with the base AcO − acting as a bidentate chelate ligand and β-hydrogen of allyl ligand getting close to one of the oxygen atoms of AcO − . The resulting Int8 then undergoes base promoted β-hydrogen elimination through a bicyclo[4.1.0] transition state (TS4) to afford allene coordinated complex (Int9) with a small energy barrier of 7.8 kcal/mol accompanying with an exothermicity of 6.1 kcal/mol. Finally, the Pd(0)L 2 coordinated with allene is restored by ligand exchange between the coordinated AcOH of Int9 and PPh 3 with an exergonic reaction energy of 12.6 kcal/ mol indicating that the whole reaction is thermodynamically favorable. In contrast, due to the sterically more crowded structure possessed in the isomer Int10 and the larger tension of the four-membered ring transition state (TS5), the reaction mechanism via path B is energetically unfavorable. As such, the ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20740-w results from our DFT calculations suggest that palladiumcatalyzed cross-coupling of 2,2-diarylvinylbromides with diazo compounds to produce allenes involves base promoted βhydrogen elimination mechanism. The rate-determining step is found to be dediazonation with the overall energy barrier of 32.1 kcal/mol (TS2 vs Int3 in Fig. 9a), which is somewhat high considering the reaction temperature of 80 o C, possibly due to the insufficient accuracy of DFT methods in some cases. The experimentally observed small KIE supports the DFT results that β-hydrogen elimination is not the rate-determining step (for computational details, see Supplementary Data 1).
In summary, We have developed a highly efficient palladiumcatalyzed cross-coupling of 2,2-diarylvinyl bromides with diazo compounds for the modular synthesis of tetrasubstituted allenes. The reaction can be promoted by either mono-or bis-phosphine ligands, and ligand dpph with a flexible six-carbon linkage proved to be the optimal choice. Under optimized conditions, both aryl diazoacetates and N-tosylhydrazones are competent coupling partners. To gain insight into the reaction mechanism, control experiments, KIE studies, and DFT calculations were carried out. The key step in the catalytic cycle is believed to undergo a βvinylic hydrogen elimination from allyl palladium intermediate, where acetate anion acts as an inner base to form a bicyclo[4.1.0] transition state. The computational study also indicates that the rate-determining step is dediazonation with the overall energy barrier of 32.1 kcal/mol. Notably, the β-hydrogen elimination mode revealed by the present work deepened our understanding of this elementary step in palladium catalysis and paved a new way for the allene synthesis.

Data availability
Detailed experimental procedures and characterization of compounds can be found in the Supplementary Information. The X-ray crystallographic coordinates for structures reported in this study have been deposited at the Cambridge Crystallographic Data Centre (CCDC) under deposition numbers CCDC 1918015 (5a), 1918277 (9).