Zinc Oxide as a Model Transparent Conducting Oxide: A Theoretical ...

Article pubs.acs.org/cm

Zinc Oxide as a Model Transparent Conducting Oxide: A Theoretical and Experimental Study of the Impact of Hydroxylation, Vacancies, Interstitials, and Extrinsic Doping on the Electronic Properties of the Polar ZnO (0002) Surface Hong Li,† Laura K. Schirra,‡ Jaewon Shim,§ Hyeunseok Cheun,§ Bernard Kippelen,§ Oliver L. A. Monti,‡ and Jean-Luc Bredas*,†,∥ †

School of Chemistry & Biochemistry and Center for Organic Photonics and Electronics, Georgia Institute of Technology, Atlanta, Georgia 30332-0400, United States ‡ Department of Chemistry & Biochemistry, University of Arizona, Tucson, Arizona 85721, United States § School of Electrical & Computer Engineering and Center for Organic Photonics and Electronics, Georgia Institute of Technology, Atlanta, Georgia 30332-0250, United States S Supporting Information *

ABSTRACT: The technology-relevant zinc-terminated zinc oxide (0002) polar surface has been studied at the density-functional theory level using both Perdew−Burke− Ernzerhof (PBE) and hybrid Heyd−Scuseria−Ernzerhof (HSE06) functionals. We have considered a number of surface conditions to better understand the impact of surface hydroxylation and intrinsic and extrinsic surface defects, including zinc vacancies, oxygen vacancies, zinc interstitials, and aluminum dopants on the surface electronic properties. Our calculations point to large variations in surface work function and energy band gap as a function of the surface model; these variations can be attributed to changes in surface charge carrier density and to additional surface states induced by the defects. The calculated shifts in O(1s) core-level binding energy of the surface oxygens in different bonding configurations are in good agreement with experimental X-ray photoelectron spectroscopy data and point to the presence of two distinct OH-species on the ZnO surface. Our results also show that the electron-compensation centers induced by zinc vacancies can be stabilized by intrinsic and/or extrinsic n-type doping near the surface; such n-type doping can lead to better performance of organic opto-electronic devices in which zinc oxide is used as an electronselective interlayer. KEYWORDS: zinc oxide, conducting oxide, surface defects, work-function determination, DFT slab calculations

I. INTRODUCTION Given its wide bandgap of 3.3 eV1 and the high natural abundance of zinc, zinc oxide (ZnO) represents a promising transparent conductive oxide with the potential for widespread use in both inorganic and organic opto-electronic devices and sensors. ZnO nanowires have been widely investigated as active elements of sensors and lasers2 and more recently as nanostructured electrodes in hybrid photovoltaic cells.3,4 ZnO has also been exploited as an interlayer (hole-blocking/ electron-selective layer) between an ITO electrode and an organic electron-transport/active layer in organic light-emitting diodes (OLEDs)5−10 and organic solar cells11−28 with inverted device architectures. In all of these applications, charge injection at the interface between ZnO and an active layer is one of the major features controlling device performance; as such, device performance can be expected to depend strongly on the ZnO surface electronic properties. While the most stable ZnO surface is the nonpolar (1010̅ ) surface,29 the preferential growth direction of ZnO films with a © 2012 American Chemical Society

wurtzite-structure has been found to be along the c-axis with exposed polar surfaces identified as (0002) or (0001).30−33 Other orientations can also be obtained depending on substrate and growth conditions.34 For the polar surfaces, the origin of the stability of the zinc- and oxygen-terminated surfaces has long posed a significant challenge, since the dipoles related to Zn- and O-terminated surfaces would appear to make crystals exhibiting these surfaces electrostatically unstable. The polar ZnO surfaces and their reactivity with hydrogen, oxygen, water, or carbon oxides, and the possibility of various surface reconstructions have thus been extensively researched and stabilization mechanisms have been debated in the literature.29,35−49 Previous theoretical studies based on densityfunctional theory calculations36,37 have thus mostly focused on modeling the stabilization mechanism of the zinc-terminated Received: May 23, 2012 Revised: June 28, 2012 Published: June 29, 2012 3044

dx.doi.org/10.1021/cm301596x | Chem. Mater. 2012, 24, 3044−3055

Chemistry of Materials

Article

(PAW) method71 using the Perdew−Burke−Ernzerhof (PBE) exchange-correlation functional72 for the geometry optimizations and self-consistent total-energy calculations. To describe the on-site Coulomb interaction among the localized zinc 3delectrons, we adopted the GGA+U approximation73 with an effective Hubbard U-parameter (Ueff = 8.5 eV); this methodology places the zinc 3d-orbitals of the ZnO crystal at 7.7 eV below the top of the valence band,74 in agreement with ultraviolet photoelectron spectroscopy (UPS) data.75,76 However, the calculated bandgap of the ZnO crystal based on such a GGA+U approximation remains 50% lower than the experimental value (see the “Results and Discussion” Section). Recently, the hybrid HSE06 functional, which adds a fraction of Hartree−Fock exchange to the GGA exchange within a fixed radius, has been found to present a significant improvement in optimizing the crystal structure and bandgap of many materials.77−81 However, the two adjustable parameters, that is, the fraction of Hartree−Fock exchange and the screening length, are system dependent. In several recent studies on defect states in bulk ZnO, it was found that the fraction of Hartree−Fock exchange has to be increased to α = 0.375 from its original α = 0.25 value to reproduce a bandgap value close to the experimental result for the ZnO crystal.66,82,83 However, in this case, we found that the zinc 3d orbitals remain localized rather high in energy (∼ −6 eV, see the projected density of states shown in the Supporting Information, Figure S1) vs. the experimental binding energy (7.5 ± 0.2 eV). A consequence of this too high energy for the Zn3d levels is hybridization of the zinc 3d orbitals with the oxygen 2p orbitals, especially around −4 to −5 eV binding energy. As a result, in the present work, to evaluate the total densityof-states (DOS) and its projection onto different atoms (PDOS) on the basis of the optimized geometries, we choose a less conventional approach by using the HSE06 functional (with α = 0.25) with a reoptimized Coulomb potential U (Ueff =7.3 eV); the goal here is to maintain the zinc 3d-orbitals at the experimental binding energy and well separated from the O2p and Zn4s levels, see the Supporting Information, Figure S2. (We note that taking the same U-parameter (8.5 eV) for both the PBE and HSE functionals only induces minor shifts in both band gap and work function values for all our surface models (by less than 0.1 eV), although as expected significant binding energy shifts (larger than 1 eV) are obtained for the zinc 3dorbitals; see Supporting Information, Tables S1−S4 and a detailed discussion in the Supporting Information). In our methodology, unlike the tuning of the Hartree−Fock contribution α, the optimization of the Coulomb potential U is justified based on the physically meaningful Zn3d binding energy, as determined by UPS. Calculations were performed with the Vienna Ab Initio Simulation Package (VASP);84,85 the plane wave cutoff was set at 400 eV, and the total energy convergence at 10−6 eV for the self-consistent iterations. The Gaussian smearing method with σ = 0.05 eV was considered for Brillouin-zone integrations on a 2 × 2 × 1 k-mesh. The geometry optimizations were performed using a damped molecular dynamics scheme until the forces on the atoms were