Coherent twin boundaries (CTBs) are widely described, both theoretically and experimentally, as perfect interfaces that play a significant role in a variety of materials. Although the ability of CTBs in strengthening, maintaining the ductility and minimizing the electron scattering is well documented [1][2][3] , most of our understanding of the origin of these properties relies on perfect-interface assumptions. Here we report experiments and simulations demonstrating that as-grown CTBs in nanotwinned copper are inherently defective with kink-like steps and curvature, and that these imperfections consist of incoherent segments and partial dislocations. We further show that these defects play a crucial role in the deformation mechanisms and mechanical behaviour of nanotwinned copper. Our findings offer a view of the structure of CTBs that is largely different from that in the literature 2,4,5 , and underscore the significance of imperfections in nanotwinstrengthened materials.CTBs formed during growth, deformation or annealing exist broadly in many crystalline solids with low or medium stackingfault energies 1,5,6 . The strengthening behaviour and other attractive properties of CTBs have been studied in nanotwinned metals (with an average twin spacing <100 nm; refs 7-9). One prevalent view is that CTB-strengthened materials have certain advantages over nanocrystalline or ultrafine-grained materials; that is, materials strengthened through traditional grain boundaries (GBs) that are considered incoherent and defective 10 . GBs not only scatter electrons, but can migrate and slide under shear stresses 11 , leading to a maximum in strength in nanocrystalline materials 12,13 . In contrast, such migration/sliding mechanisms may not be operative in CTBs despite some reports of detwinning evidence 7,14,15 and the observation of a similar maximum strength in a nanotwinned copper 3 (nt-Cu). Existing models widely assume perfect CTBs and rationalize flow softening due to CTB migrations and detwinning as caused by nucleation and motion of partial dislocations parallel to CTBs (ref. 4). These mechanisms are informative as long as CTB lengths are limited to the tens of nanometres typically used in molecular dynamics simulations 4,[16][17][18] . It still remains difficult through molecular dynamics simulations to validate the migrations/detwinning of the much longer CTBs seen in experiments (500 nm; ref. 3). There could be alternative mechanisms that are intricately related to the potential structures of CTBs and the characteristics of GBs, both of which are not accounted for in the literature.Recent studies of nanotwinned copper pillars without GBs revealed strong deformation anisotropy and a brittle-to-ductile transition behaviour (where CTBs are considered intrinsically brittle) 2 , suggesting that CTBs alone are not sufficient for increased plasticity despite their strong strengthening effect, and that a reasonable mix of GBs is helpful to mediate the plasticity and achieve high ductility. Experiments and simulations have f...