Single-crystal x-ray diffraction studies with synchrotron radiation on the honeycomb iridate α-Li2IrO3 reveal a pressure-induced structural phase transition with symmetry lowering from monoclinic to triclinic at a critical pressure of Pc = 3.8 GPa. According to the evolution of the lattice parameters with pressure, the transition mainly affects the ab plane and thereby the Ir hexagon network, leading to the formation of Ir-Ir dimers. These observations are independently predicted and corroborated by our ab initio density functional theory calculations where we find that the appearance of Ir-Ir dimers at finite pressure is a consequence of a subtle interplay between magnetism, correlation, spin-orbit coupling, and covalent bonding. Our results further suggest that at Pc the system undergoes a magnetic collapse. Finally we provide a general picture of competing interactions for the honeycomb lattices A2M O3 with A= Li, Na and M = Ir, Ru.PACS numbers: 61.05.cp,61.50. Ks,71.15.Mb In recent years, layered honeycomb 4d and 5d metal oxides, such as Na 2 IrO 3 , α-Li 2 IrO 3 , and α-RuCl 3 , have been intensively scrutinized as Kitaev physics candidates [1-6] due to the presence of sizable nearest-neighbor bond-dependent spin-orbital 1/2 Ising interactions. However, instead of the expected Z 2 spin liquid groundstate, as shown by Kitaev [1], these materials order magnetically either in a zig-zag structure [4, 7-9] (Na 2 IrO 3 , α-RuCl 3 ) or an incommensurate spiral structure [10] (α-Li 2 IrO 3 ). This magnetic long-range order has been suggested to originate from further non-Kitaev interactions and a present debate is whether the magnetic excitations in these materials nevertheless retain some of the non-trivial features of the Kitaev model, such as fractionalization [9,11,12]. It might be expected that one route to enhance Kitaev interactions would be by applying pressure or by doping. However, the physics of this structural family is much richer and there are many more instabilities that interfere with the Kitaev interactions, in particular under pressure. Indeed, Li 2 RuO 3 is nonmagnetic and strongly dimerized at ambient pressure [13][14][15], while SrRu 2 O 6 is an ultra-high-temperature antiferromagnet [16,17], despite having the same planar geometry, and shows no bond disproportionation.Many factors control the competition between Kitaev physics, magnetism, and dimerization [18] in A 2 M O 3 honeycomb networks, such as the number of transition metal M d-electrons, the strength of spin-orbit coupling, * valenti@th.physik.uni-frankfurt.de † christine.kuntscher@physik.uni-augsburg.de the strength of correlation effects and Hund's rule coupling, or the ionic radii of the buffer element A. In this context it is particularly instructive to compare α-Li 2 IrO 3 with Li 2 RuO 3 , which contains the same buffer element (Li). α-Li 2 IrO 3 is less correlated than Li 2 RuO 3 (5d versus 4d electrons, resp.) so that one would expect in the former a reduced tendency to magnetism in favor of dimerization. On the other hand, ...