Bistability of theDX center in GaAs and AlxGa1-xAs,and experimental tests for negativeU of theDX level

We summarize a large body of experimental and theoretical work, especially in Si-doped GaAs and Al _x Ga_1-x As, regarding the bistability of theDX center. There is good evidence that theDX center is just the simple donor, and that each donor can exist in either of two distinct lattice configurations, each with its own spectrum of bound electronic states. Generally, the substitutional configuration binds electrons in shallow hydrogenic states, but many observations also indicate a deep (highly localized) state ofA ₁ symmetry. These states are to be distinguished from bound states of a lattice-distorted configuration, the lowest-lying of which is the deepDX level. The occupation of theDX level in thermal equilibrium with the states of the conduction band can be reasonably well modeled by assuming thatDX is either a one-electron or a two-electron state, and we discuss the reasons for this ambiguity. However, we then show that such thermal equilibrium results are consistent with thermal capture and emission kineticsonly if we assume thatDX is a two-electron state. Our results thus support the model of Chadi and Chang in which the distorted configuration is stabilized by capture of two electrons. In other words, the defect exhibits negative effective correlation energy (negativeU).     

Analysis of the vacancy-interstitial model ofDX centers

I describe the vacancy-interstitial model ofDX centers, in which the substitutional donor atoms relax toward interstitial sites when they capture electrons. This appears to be the only known model ofDX centers which can account for the structure observed by Mooneyet al. in DLTS spectra of dilute Al _x Ga_1-x As alloys. By comparing its predictions with the DLTS and Hall data from Al _x Ga_1-x As alloys of low Al content one can establish values for many of the relevant parameters, such as the enthalpies and entropies (degeneracies) associated with sites surrounded by differing numbers of Al atoms. The data also show that the distribution of donor configurations among the four available neighboring interstitial sites does not attain thermal equilibrium among states of differing energies before being released, although transitions (through tunneling) among states of equal energy may be possible. The model is seen to be consistent with those treated theoretically by Chadi and Chang forDX and by Dabrowski and Scheffler for EL2 centers.     

Growth and characterization of a GaAs/AlAs superlattice with variable layer thicknesses

Growth and properties of a short-period (GaAs)_n/(AlAs)_m superlattice grown by molecular beam epitaxy on (100) GaAs are reported. The structure of the superlattice with variable (n∼m = to 15) layer thicknesses was studied by double crystal x-ray diffraction. Special care was taken to exclude the effect of several parameters; e.g., the inaccuracy of the lattice mismatch between the AlAs and GaAs layers, which may cause errors in the determination of the layer thicknesses. The electronic properties of the superlattice were studied by photoreflectance and photoluminescence. The γ-X band crossing was found to occur at 13.5 monolayers at 12K.     

Preparation of CdSSe-ZnS superlattice,SrS, and CdSSe-SrS superlattice by hot-wall epitaxy,and applications to electroluminescent devices

CdSSe (manganese-doped, E_g = 1.9–2.5 eV, lattice constant a = 6.05–5.8A)-ZnS (Eg = 3.56 eV, a = 5.41A) superlattices, SrS (cerium-doped, E = 4.4 eV, a = 6.02A) layers, and CdSSe-SrS (cerium-doped) superlattice layers nave been prepared by hot-wall epitaxy, and the properties and the electroluminescent device characteristics of the active layers are reported. For the superlattices with ZnS, the maximum luminance was 800 cd/m² at an applied sinusoidal voltage (V_o-p = 200 V) with frequency 1kHz, and the wavelength of the spectral peak was 610 nm due to the large strain caused by the lattice mismatch (8–15%) between the CdSSe and ZnS layers. The maximum luminance and Comisson Internationale de Enluminure (CIE) chromaticity of CdS(Mn)-ZnS superlattices and CdSe(Mn)-ZnS superlattice devices were 557cd/m² and (x,y) = (0.58,0.41) and 982 cd/m² and (0.61, 0.38), respectively. For superlattices with SrS, the maximum luminance of the device with the SrS (cerium-doped) active layer was nearly 700 cd/m² at a voltage of 340V. Blue electroluminescent emission was observed in the photon wavelength region less than 450 nm, due to carriers dropping into the quantum wells of the device with the CdSSe-SrS superlattice active layer.     

Electron paramagnetic resonance and optically-detected magnetic resonance of donors in AlxGa1-xAs

The Electron Paramagnetic Resonance (EPR) and Optically-Detected Magnetic Resonance (ODMR) work on Si-donors in Al _{x Ga1- x} As is reviewed in the context of the shallow-deep bistability (DX) problem. Three donor states are important. Little work has been published on donors tied to theT-minimum. However, there are many results forX-donors. In AlAs/GaAs heterostructures, well-resolved spectra reveal a donor state comprised of independentX _x andX _y valleys with theXz valley unpopulated due to the hetero-epitaxial strain. As Al mole fraction decreases, intervalley coupling is evident from the line positions and linewidths. The published attempts to observe and identify the deep (relaxed) state are inconclusive. Some suggestions for future work are presented.     

Tilted superlattice composition profile determined by photoluminescence and thermal disordering

The composition profile of an (AlAs)_1/2(GaAs)_1/2 tilted superlattice is characterized for the first time. The tilted superlattice sample is thermally disordered, and the energy of the direct band gap photoluminescence peak is measured as a function of increasing layer interdiffusion. The shift in the photoluminescence peak energy after completely disordering the tilted superlattice is 39 meV. A theoretical model is used to simulate the change in band gap as a function of layer interdiffusion for several composition profiles. The profile that gives the best fit to the experimental data is chosen. The tilted superlattice composition profile is found to be sinusoidal, varying from Al_0.40Ga_0.60As to Al_0.60Ga_0.40As.     

Photocurrent anisotropy in compositional modulated superlattice of long-range ordered Ga0.5In0.5P

Photocurrent anisotropy of long-range ordered Ga_0.5In_0.5P alloys has been sys-tematically investigated. The ordered Ga_0.5In_0.5P is a compositional modulated superlattice of Ga_0.5+η/2In_0.5-η/2P and Ga_0.5-η/2In_0.5+η/2P monolayer planes, where η is a long-range order parameter. The photocurrent edge of the [110] polarization is lower than that of the . The observed anisotropy in the photocurrent spectra is due to a crystal-field splitting at the valence-band maximum in ordered Ga_0.5In_0.5P. The anisotropy shows a continous variation as a function of η. In order to make clear the effects of the valence-band splitting on the polarized photocurrent spectra, we performed theoretical calculations in which a distribu-tion of η in the epitaxial film and an order-parameter dependence of oscillator strength were considered. From these calculations, it is found that oscillator strength is a key parameter in the anisotropic character. The calculated results moderately agree with the measured data. Furthermore, the epitaxial thickness dependence of the anisotropic character in photocurrent was investigated.     

Electrical properties of nickel in silicon

Electrical activity and energy levels as well as diffusion properties of nickel in silicon have not yet been reliably established. In this paper, we investigated the diffusion and the electrical properties of nickel in silicon to confirm that nickel is electrically active and introduces one acceptor and one donor level by combined measurements of Hall coefficient and DLTS, and measurements of the distribution of electrically active nickel in various silicon diodes by DLTS. The former experiments show that bothn- andp- type silicon are compensated by nickel and that nickel introduces an acceptor level ofE _c-0.47 ± 0.04 eV and a donor level ofE _v +0.18 ± 0.02 eV. The concentrations of these two levels are almost identical over the diffusion temperatures from about 800 to 1100° C, indicating that these donor and acceptor levels are due to different charge states of the same nickel center. In the distribution measurements of electrically active nickel in silicon diodes, we inspected how nickel can be observed by DLTS. It was found that the nickel diffusion intop- n junction is rather complicated, the distribution profiles of nickel in the vicinity of thep- n junction being markedly influenced by an additional heating at elevated temperatures after the nickel diffusion. This gives evidence that the difference in silicon devices used in various studies could give rise to different results.     

Confirmation of large lattice relaxation of theDX center by extended photo-ionization cross-section measurements

Although the large lattice relaxation model (LLR) for electron capture at the donor relatedDX center in Al_xGa_1-xAs has seen wide acceptance over the last 12 years, there have been some recent proposals which have attempted to explain the experimental data with models that only require small lattice relaxation (SLR). One key piece of evidence supporting LLR is the large observed difference (in the case of Si-doped Al_xGa_1-xAs) between the optical (∼1.4 eV) and thermal (∼0.2 eV) ionization energies. The SLR model proposed that the lowest energy optical ionization was a very weak process, and that the optical transition which had been observed previously is a transition to a higher band. These arguments were supported by photoconductivity data showing a finite photo-ionization rate at energies as low as 200 meV. To resolve this question we have measured the photo-ionization cross section over 7 to 8 orders of magnitude using a tunable infrared laser as a source. A consistent optical ionization energy of about 1.4 eV was observed for 4 samples of differing alloy compositions and doping levels.In no case was there any detectable photo- ionization below 0.8 eV. A detailed discussion of these experiments examines the difficulty in obtaining such a large dynamic range optical spectrum. Of particular relevance are the issues of ionization detection, and the brightness of purity of the optical source. A thorough review of these issues and their impact on previous studies of theDX center is presented.     

The determination of impact ionization coefficients in ln0.2Ga0.8As/GaAs strained-layer superlattice mesa photodiodes

Sandia National Laboratories, Albuquerque, NM 87185 Photocurrent multiplication measurements have been performed on two different In_0.2Gao_0.8As/GaAs strained-layer superlattice (SLS)p ⁺n photodiode structures which are designed to permit simultaneous injection of electrons and holes. Initial devices were found to suffer from low quantum efficiencies produced by small electron diffusion lengths as well as mixed injection caused by lower than expected optical absorption coefficients in the SLSn ⁺ contact layers for hole injection conditions. Using a second device structure having a thicker n⁺ contact region, the electron multiplication factors are found to be larger than that of holes with a ratio of the electron to hole ionization coefficient of 1.4 for fields between 2.9 and 3.4 x 10⁵V/cm.     

Germanium as a deep level in alxga1− xas grown by organometallic vapor phase epitaxy

Optoelectronic devices require materials which exhibit extremely low trap concentrations. The Al_xGa_1−xAs system has been used extensively for optoelectronic applications despite trap concentrations in the Al_xGa_1−xAs which limit the efficiency of the resulting devices. Deep level transient spectroscopy (DLTS) performed on Al_0.2Ga_0.8As layers grown by organometallic vapor phase epitaxy (OMVPE) has revealed three traps with concentrations >10¹³ cm⁻³ -E _c-E_t = 0.3, 0.5 and 0.7 eV. The dominant source of the 0.3 eV trap has proven to be a Ge impurity in arsine. SIMS analysis of Al_0.2Ga_0.8As samples show Ge as the only candidate for the impurity responsible for the 0.3 eV trap. DLTS and SIMS analysis performed on Al_0.2Ga_0.8As samples intentionally doped with Ge displayed a proportional increase in the 0.3 eV trap concentration with the Ge concentration and establishes that Ge is indeed the source of the 0.3 eV trap in Al_xGa_1−xAs. Comparison of C-V, SIMS and DLTS measurements performed on Al_xGa_1-xAs:Ge indicate that approximately 30% of elemental Ge incorporated created the 0.3 eV trap, DX_Ge.     

Deep electron levels in undoped polycrystalline CdTe annealed in liquid Cd

The method of deep-level transient spectroscopy (DLTS) is used to determine the set of deep electron levels in undoped n-CdTe polycrystals grown by chemical synthesis from vapor phase in highly nonequilibrium conditions and then subjected to annealing in liquid Cd. After annealing, the electron concentration is found to increase from ∼10⁸ to 10¹⁵ cm⁻³. Electron traps with deep levels E ₁ = 0.84 ± 0.03 eV and E ₂ = 0.71 ± 0.02 eV with total concentration ∼10¹⁴ cm⁻³ are dominant in the DLTS spectrum. The role of grain boundaries and intrinsic point defects in formation of deep-level centers and their effect on the value of conductivity after annealing are discussed. A relation between the level E ₁ and complexes of intrinsic point defects and relation of the level E ₂ to the Cd _i point defect are considered.     

A sliding wafer-OMVPE scheme for fabricating subnanometer superlattices

A sliding wafer-OMVPE (Organometallic Vapor Phase Epitaxy) reactor suitable for growing superlattices is developed. The reactor is a two-channel horizontal reactor with a susceptor placed across the two channels. A slider is used to transport the substrate by sliding it along the susceptor surface from one channel to the other. This scheme makes it possible to set the temperature of the susceptor in each channel independently by utilizing the skin effect of radio wave in graphite. The performance of this scheme is demonstrated by growing a superlattice of ten periods of GaAs(3.5 Å)/AlAs(7 Å).     

Donor-acceptor recombination in short-period silicon-doped GaAs/AlAs superlattices

A low-frequency band is observed along with an exciton band in photoluminescence spectra of short-period GaAs/AlAs superlattices doped with Si in the barriers or in the barriers and wells. This band is ascribed to donor-acceptor recombination on the basis of the dependence of its frequency on the excitation intensity under cw excitation and on the time delay under pulsed excitation. Mainly type-II superlattices are investigated. The estimate E _A+E _D≈120 meV can be obtained from the peak energy of the donor-acceptor band with a very weak excitation intensity. The estimates E _A≈23 meV and E _D≈90 meV are obtained from the temperature dependence of the band intensity. It is suggested that the deep donor level is associated with a DX center in the AlAs layers. Fiz. Tekh. Poluprovodn. 32, 839–842 (July 1998)     

Atomic structure ofDX centers: Theory

A structural model forDX centers in Al _x Ga_1-x As alloys which explains their unusual properties in terms of two distinct bonding configurations for donor impurities is proposed. The results of ourab initio self-consistent pseudopotential calculations showDX to be a “negative-U” defect center. It results from a large lattice relaxation which is different for a Group IV than a Group VI donor. The proposed structural model provides a satisfactory explanation of the properties ofDX centers.     

Electron-phonon interaction in short-period (GaAs) m (AlAs) n (001) superlattices

The deformation potentials of electron scattering at short-wavelength phonons for intervalley transitions in the conduction band of short-period (GaAs) _m (AlAs) _n (001) (m, n = 1, 2, 3) superlattices are determined by the electron density functional method. The dependences of the electron and phonon states and deformation potentials on the layer thickness in the superlattices are analyzed. The results of ab initio calculations are in good agreement with the data of empirical calculation of the deformation potentials integrated over phonons, but differ from data on the corresponding potentials for partial scattering channels because of approximations of the phenomenological model of interatomic binding.     

Magnetoluminescence properties of Hg1−xCdxTe epitaxial layers and superlattice structures grown by metalorganic molecular beam epitaxy

We present a study of the electro-optical properties ofHg _1- xCd_xTe epitaxial layers and Hg_1-x Cd_xTe/CdTe (0.28 < x < 0.30) superlattice structures by x-ray diffraction, lateral transport and photo- and magneto-luminescence measurements. Systematic studies of the excitation intensity and magnetic field dependence of the photoluminescence revealed direct evidence of an excitonic contribution to the observed luminescence in Hg_{1- x}Cd_xTe epitaxial layers. Similar investigations of the superlattice structures indicated that excitonic corrections were required to adequately fit the luminescence data. Optical gains of 80 cm⁻¹ were obtained for an excitation intensity of 100 kW/cm² indicating suitable electro-optical properties for making efficient mid-infrared laser diodes.     

Studies of interdiffusion in GemSin strained layer superlattices

We present the results on the characterization and interdiffusion behavior of Ge _m Si _n strained layer superlattices (SLS’s) composed of alternating monolayers of pure Ge and pure Si. Such Ge _m Si _n SLS’s were grown on top of thick relaxed Ge _y Si_1-y buffer layers so as to symmetrize the strain distribution and to maintain the pseudomorphic growth of the superlattices. Samples with different superlattice periodicities (i.e. d = d_Ge + d_Si and different layer thickness ratios (i.e. d_Ge:d_Si were prepared for comparison. Raman scattering spectroscopy and x-ray diffraction were used to characterize these samples. Initial results on thermal stability of these Ge _m Si _n SLS’s are also reported     

How the type of bombarding ion affects the formation of radiation defects in silicon

The method of nonstationary capacitance was used to study how the chemical nature of implanted ions affects the creation of electrically active defects in silicon. Oxygen O⁺ and nitrogen N⁺ ions were implanted into Si at a target temperature of 300 K, in doses of 2×10¹¹cm⁻² with energies of 75 keV, and argon Ar⁺ ions were implanted in a dose of 7×10¹⁰cm⁻² with energies of 150 keV, in such a way that all the samples of n-and p-Si received approximately the same number and spatial distribution of primary radiation defects. It was observed that the spectrum of stable radiation defects depends on the nature of the bombarding ion. Thus, the DLTS spectrum of n-Si irradiated by O⁺ ions has three peaks, whereas the spectrum of n-Si implanted by N⁺ ions exhibited only one of these peaks. The DLTS spectra of samples of n-and p-Si implanted by O⁺ and N⁺ ions revealed peaks of reverse (anomalous) polarity, whose energy positions matched the most clearly defined DLTS peaks of silicon samples with the opposite type of conductivity. Fiz. Tekh. Poluprovodn. 32, 523–526 (May 1998)     

Electrically Active Defects in GaN Layers Grown With and Without Fe-doped Buffers by Metal-organic Chemical Vapor Deposition

Electrically active defects in n-GaN films grown with and without an Fe-doped buffer layer have been investigated using conventional and optical deep-level transient spectroscopy (DLTS). Conventional DLTS revealed three well- defined electron traps with activation energies E _a of 0.21, 0.53, and 0.8 eV. The concentration of the 0.21 and 0.8 eV defects was found to be slightly higher in the sample without the Fe-doped buffer, whereas the concentration of the 0.53 eV trap was higher in the sample with the Fe-doped buffer. A minority carrier trap with E _a ≈ 0.65 eV was detected in both samples using optical DLTS; its concentration was ∼40% higher in the sample without the Fe-doped buffer. Mobility spectrum analysis and multiple magnetic-field measurements revealed that the electron mobility in the topmost layer of both samples was similar, but that the sample without the Fe-doped buffer layer was affected by parallel conduction through underlying layers with lower electron mobility.