Since the discovery of colloidal semiconductor nanocrystals (NCs) they have found tremendous amount of applications in bioimaging, 1 lasing, 2 diodes, 3 single-photon sources, 4 photovoltaics, 5 etc. owing to a wide range of techniques 6 available to tune their optical and electronic properties.Many of those applications rely on emission properties of NCs (quantum yield, emission wavelength broadening and diffusion, Stokes shift, blinking) which are strongly affected by different types of defects, supposedly residing at the surface. 7 Surface traps may also affect the multiexciton generation yields 8 and charge carriers extraction, relevant, e.g., for photovoltaics. Understanding better the source of the trap states can help to develop the synthesis procedures to reduce or ultimately eliminate the traps. In contrast to absorption properties, which are determined mainly by the bulk crystalline structure and the macroscopic properties of the NCs (size, shape), 9,10 surface and thus emission properties require a more detailed knowledge on atomic scale.The exact atomistic nature of surface defects remains unknown 11À13 and the interpretation of experimental data is thus often based on available theoretical models. Several semiempirical studies of ligated surfaces are available 14À18 but this methodology does not reliably capture surface reconstructions and charge redistributions. Few ab initio studies of the NC surfaces available to date addressed only the bare surfaces 19À25 or weakly bound (L-type) ligands. 26À30 However, more and more experimental data suggest that the main type of ligands present on the surface are the covalently bound (X-type) ligands, e.g., deprotonated carboxylic or phosphonic acids. 31À33 Theoretical studies of such ligands on CdSe and PbSe only start to emerge. 30,33À35 In this work we investigate from first principles the atomistic nature of the surface states in NCs. To do this, we choose CdSe NCs without structural defects, small enough to be treated within the density functional theory (DFT) but large enough to distinguish delocalized (core) and localized (trap) states, with carboxylic acid ligands bound covalently. We find that even such an idealized and small model is rich enough to create structures with or without surface trap states, depending on the amount of ligands. Contrary to expectations, apparently more passivated structures (with more ligands and less dangling bonds) exhibit more surface traps. Our most important finding is the presence of mobile surface ligands whose energy levels fluctuate respectively, a feature required by several phenomenological models of blinking. 36À38 We will discuss whether the observed diffusion on its own is capable of explaining the fluorescence intermittency, and whether it is capable of producing switchable longlived trap states.
' COMPUTATIONAL METHODSCalculations were performed within DFT using the SIESTA code. 39 Generalized gradient approximation in a PerdewÀBurkeÀ Ernzerhoff formulation, Troullier-Martins norm-conserving pseudopotenti...