We propose a new direct mechanism for the pressure driven α→ω martensitic transformation in pure titanium. A systematic algorithm enumerates all possible mechanisms whose energy barriers are evaluated. A new, homogeneous mechanism emerges with a barrier at least four times lower than other mechanisms. This mechanism remains favorable in a simple nucleation model.PACS numbers: 81.30. Kf, 64.70.Kb, 5.70.Fh Martensitic transformations are abundant in nature and are commonly used in engineering technologies [1]. Materials from steel to shape-memory alloys are governed by their underlying martensitic transformations [2]. The pressure driven α(hcp) → ω(hexagonal) transformation in pure titanium, discussed here and reviewed extensively by Sikka et al. [3], has significant technological implications in the aerospace industry because the ω phase formation lowers toughness and ductility. This transformation was first observed by Jamieson [4], and has since been extensively studied using static high-pressure [5] and shockwave methods [6]. Because of experimental difficulties in directly observing martensitic transformation mechanisms, they are usually inferred from the orientation relationships between the initial and final phases. Such an approach may result in multiple transformation mechanisms for any given set of orientation relations, and requires one to guess the appropriate transformation mechanism. Thus, despite several attempts, the mechanism for this transformation is still unclear.We calculate the energy barrier for homogeneous transformation for different titanium α→ω transformation mechanisms and compare the values using a simplified nucleation model. We systematically generate and sort possible α→ω mechanisms by their energy barriers. A new direct mechanism emerges whose barrier is lowest both homogeneously and when considered in a simple nucleation model. Figure 1 shows our new low energy barrier mechanism for the α→ω transformation in Ti, called TAO-1, for "Titanium Alpha to Omega." This direct 6-atom transformation requires no intermediary phase, and has small shuffles and strains. The 6 atoms divide into a group of 4 atoms that shuffle by 0.63Å and 2 atoms that shuffle by 0.42Å. Combining these shuffles with strains of ε x = 0.91, ε y = 1.12, and ε z = 0.98 produces a final ω cell from our α cell. The original α matrix is then oriented relative to the ω matrix such that (0001) α (0111) ω and [1120] α [0111] ω . These orientation relations are seen in some experiments, but not others [5,6,7].Our mechanism identification method matches possible supercells of α and ω to determine the lattice strain, and atom positions to determine the necessary internal relaxations; similar to [8]. While there are infinitely many possible supercells, we consider only 6 and 12 atom supercells with principal strains less than 1.333 and greater than 0.75 = 1.333 −1 . For each supercell, there are multiple ways to shuffle the atom positions from α to ω; we consider mechanisms where, relative to the center of mass, no atom mov...