Due to their technological importance, point defects in silicon are among the best studied physical systems. The experimental examination of point defects buried in bulk is difficult and evidence for the various defects usually indirect. Simulations of defects in silicon have been performed at various levels of sophistication ranging from fast force fields to accurate density functional calculations. The generally accepted viewpoint from all these studies is that vacancies and self interstitials are the basic point defects in silicon. We challenge this point of view by presenting density functional calculations that show that there is a new fourfold coordinated point defect in silicon that is lower in energy.PACS number: 61.72.JiThe stability of crystalline silicon comes from the fact that each silicon atom can accommodate its four valence electrons in four covalent bonds with its four neighbors. The traditional point defects in silicon, the vacancy and the various interstitials, are obtained by taking out or adding atoms to the crystal and thus all destroy fourfold coordination. A relatively high defect formation energy for these defects is the consequence. In addition there is a point defect, that conserves the number of particles, the Frenkel pair, consisting of a vacancy and an interstitial. If the vacancy and the interstitial are close, the formation energy of a Frenkel pair is less than the sum of an isolated vacancy and interstitial (Table I). Nevertheless the formation energy is still considerable since again bonds are broken. Figure 1 shows a novel defect configuration that has, in contrast to all other point defects, perfect fourfold coordination and will therefore be called Four Fold Coordinated Defect (FFCD).Using state of the art plane wave density functional theory (DFT) calculations we will now present evidence that the formation energy of the FFCD of 2.4 eV is lower than the formation energy of all other known point defects both in intrinsic and doped silicon. Even though several calculations of this type were published for the traditional defects, we have decided to repeat them for several reasons. a) All these calculations take advantage of error cancellations to obtain energy differences that are more precise than the total energies themselves. This cancellation is obviously best if all the total energies that are compared are calculated with exactly the same method. b) Most DFT calculations used the basic Local Density Approximation (LDA) whereas we used a more precise General Gradient Approximations [1]. c) As was recently shown [2] cells of at least 216 atoms are required for a reasonable convergence. Most published calculations were done with smaller cells. d) We constructed very accurate pseudopotentials [3] based on atomic calculations with the same density functional as used in our target calculation and we afforded a very large plane wave basis set (25 Ry for the LDA and 35 Ry for the GGA calculations). e) The point defect formation energies in doped silicon were up to now obtained in ...

The effect of quenched impurities on a system having a commensurate-incommensurate (CI) phase transition is treated by real-space rescaling. A simplified form of the free energy is treated explicitly. New terms generated by the rescaling procedure do not modify the results. In D=3 dimensions, the Bragg peak shift (misfit) varies linearly with the chemical potential near the CI transition. The CI transition is killed for D<or=Dc=2. At the critical dimension, the misfit decreases exponentially for large negative chemical potentials. The model treated here may be considered a modified Ising model in a random field. The modification is a strong anisotropy of the interfaces separating the up and down domains.