The influence of defect charge state on the magnetism of Cu doped ZnO as well as the Cu defects clustering have been investigated by the first-principles calculations. We demonstrate that p-type ZnO:Cu could have ferromagnetic (FM) property, but n-type ZnO:Cu would not have local magnetic moment. Furthermore, the neutral substitutive Cu defects are found to be favorable in clustering, which maintains the FM ordering.

We distill from first-principles spin-polarized total-energy calculations some practical rules for predicting the magnetic state (ferromagnetic/antiferromagnetic/paramagnetic) of substitutional transition-metal impurity with different charge state in various host crystal groups Ⅳ, Ⅲ-Ⅴ, Ⅱ-Ⅵ, Ⅰ-Ⅲ-Ⅵ2, and Ⅱ-Ⅳ-Ⅴ2 semiconductors. The basic mechanism is the stabilization of a ferromagnetic bond between two transition metals if the interacting orbitals are partially-occupied. These rules are then subjected to quantitative tests, which substantiate the mechanism of ferromagnetism in these systems.We discuss cases where current electronic structure calculations agree with these rules, and identify a few cases where conflicts exist. The effect of doping on transition-metal magnetic properties is also covered by these rules by considering the oxidation state changes due to doping. In addition, we systematically apply these rules to ideal substitutional impurities, contrasting our predictions with experiment. Discrepancies may be used to assess the role of various nonidealities such as presence of additional dopants, precipitates, clusters, or interstitial sites.

The efficiency of CuInSe2 based solar cell devices could improve significantly if CuGaSe2, a wider band gap chalcopyrite semiconductor, could be added to the CuInSe2 absorber layer. This is, however, limited by the difficulty of doping n-type CuGaSe2 and, hence, in its alloys with CuInSe2. Indeed, wider-gap members of semiconductor series are often more difficult to dope than lower-gap members of the same series. We find that in chalcopyrites, there are three critical values of the Fermi energy EF that control n-type doping: (i) EnF,pin is the value of EF where the energy to form Cu vacancies is zero. At this point, the spontaneously formed vacancies (=acceptors) kill all electrons. (ii) EnF,comp is the value of EF where the energy to form a Cu vacancy equals the energy to form an n-type dopant, e.g., CdCu. (iii) EnF,site is the value of EF where the formation of Cd-on-In is equal to the formation of Cd-on-Cu. For good n-type doping, EnF,pin, EnF,comp, and EnF,site need to be as high as possible in the gap. We find that these quantities are higher in the gap in CuInSe2 than in CuGaSe2, so the latter is difficult to dope n-type. In this work, we calculate all three critical Fermi energies and study theoretically the best growth condition for n-type CuInSe2 and CuGaSe2 with possible cation and anion doping. We find that the intrinsic defects such as VCu and InCu or GaCu play significant roles in doping in both chalcopyrites. For group-II cation (Cd, Zn, or Mg) doping, the best n-type growth condition is In/ Ga-rich, and maximal Se-poor, which is also the optimal condition for stabilizing the intrinsic InCu/GaCu donors. Bulk CuInSe2 can be doped at equilibrium n-type, but bulk CuGaSe2 cannot be due to the low formation energy of intrinsic Cu-vacancy. For halogen anion doping, the best n-type materials growth is still under In/ Ga-rich, and maximal Se-poor conditions. These conditions are not best for halogen substitutional defects, but are optimal for intrinsic InCu/GaCu donors. Again, CuGaSe2 cannot be doped n-type by halogen doping, while CuInSe2 can.

The need for spin-injectors having the same zinc-blende-type crystal structure as conventional semiconductor substrates has created significant interests in theoretical predictions of possible metastable “half-metallic” zinc-blende ferromagnets, which are normally more stable in other structure-types, e.g., NiAs. Such predictions were based in the past on differences △bulk in the total energies of the respective bulk crystal (forms szinc blende and NiAs). We show here that the appropriate criterion is comparing difference △epi(as) in epitaxial total energies. This reveals that even if △bulk is small, still for MnAs, CrSb, CrAs, CrTe, △epi＞0 for all substrate lattice constant as, so the zinc-blende phase is not stabilized. For CrS we find △epi(as)＜0, but the system is antiferromagnetic, thus not half-metallic. Finally, zinc-blende CrSe is predicted to be epitaxially stable for as＞6.2 Å and is half metallic.

The wider-gap members of a semiconductor series such as diamond→Si→Ge or AlN→GaN→InN often cannot be doped n-type at equilibrium. We study theoretically if this is the case in the chalcopyrite family CuGaSe2→CuInSe2, finding that: (i) Bulk CuInSe2 (CIS, Eg=1.04 eV) can be doped at equilibrium n-type either by Cd or Cl, but bulk CuGaSe2 (CGS, Eg=1.68 eV) cannot; (ii) result (i) is primarily because the Cu-vacancy pins the Fermi level in CGS farther below the conduction band minimum than it does in CIS, as explained by the “doping limit rule; (iii) Cd doping is better than Cl doping, in that CdCu yields in CIS a higher net donor concentration than ClSe; and (iv) in general, the system shows massive compensation of acceptors (CdⅢ,VCu) and donors (ClSe,CdCu, InCud).