The physics of the growth mechanisms, characterization of epitaxial structures and device properties of GaAs and other compound semiconductors on Si are reviewed in this paper. The nontrivial problems associated with the heteroepitaxial growth schemes and methods that are generally applied in the growth of lattice mismatched and polar on nonpolar material systems are described in detail. The properties of devices fabricated in GaAs and other compound semiconductors grown on Si substrates are discussed in comparison with those grown on GaAs substrates. The advantages of GaAs and other compound semiconductors on Si, namely, the low cost, superior mechanical strength, and thermal conductivity, increased wafer area, and the possibility of monolithic integration of electronic and optical devices are also discussed.

Gallium arsenide and other compound semiconductors on silicon

Concerns about the changing environment and fossil fuel depletion have prompted much controversy and scrutiny. One way to address these issues is to use concentrating photovoltaics (CPV) as an alternate source for energy production. Multijunction solar cells built from III–V semiconductors are being evaluated globally in CPV systems designed to supplement electricity generation for utility companies. The high efficiency of III–V multijunction concentrator cells, with demonstrated efficiency over 40% since 2006, strongly reduces the cost of CPV systems, and makes III–V multijunction cells the technology of choice for most concentrator systems today. In designing multijunction cells, consideration must be given to the epitaxial growth of structures so that the lattice parameter between material systems is compatible for enhancing device performance. Low resistance metal contacts are crucial for attaining high performance. Optimization of the front metal grid pattern is required to maximize light absorption and minimize I2R losses in the gridlines and the semiconductor sheet. Understanding how a multijunction device works is important for the design of next-generation high efficiency solar cells, which need to operate in the 45%–50% range for a CPV system to make better economical sense. However, the survivability of solar cells in the field is of chief concern, and accelerated tests must be conducted to assess the reliability of devices during operation in CPV systems. These topics are the focus of this review.

https://www.spectrolab.com/pv/support/Cotal_III_V_multijunction_photovoltaics.pdf

III–V multijunction solar cells for concentrating photovoltaics

The possibility induction of light emission from silicon, an indirect bandgap material in which radiative transitions are unlikely, raises several interesting and technologically important possibilities, especially the fabrication of a truly integrated optoelectronic microchip. In this article, the natural considerations that constrain silicon from emitting light efficiently are examined, as are several engineered solutions to this limitation. These include intrinsic and alloy-induced luminescence; radiatively active impurities; quantum-confined structures, including zone folding and the recent developments in porous silicon; and a hybrid approach, the integration of direct bandgap materials onto silicon.

Light Emission from Silicon

Ion implantation was first applied to semiconductors over 30 years ago as a means of introducing controllable concentrations of n- and p-type dopants at precise depths below the surface. It is now an indispensable process in the manufacture of integrated circuits. This review gives a brief and selected overview of ion beam modification of semiconductors, treating both fundamental and technological issues of current interest. Damage introduction during ion irradiation and its removal during a thermal annealing step are key issues which are highlighted. Some semiconductors are easily damaged and amorphised (e.g. silicon) whereas others (e.g. gallium nitride) are quite resistant to damage production due to efficient dynamic defect annihilation during implantation. The conditions needed to remove implantation damage also vary dramatically from one semiconductor to another: amorphous layers in silicon can be recrystallised to completely remove disorder at ∼600°C, whereas extended defects in gallium nitride require temperatures of >1400°C to remove them. High dose implantation can result in the formation of supersaturated solid solutions, alloys and compounds, often with intriguing properties as a result of the non-equilibrium aspects of ion implantation. Formation of silicon dioxide layers directly during oxygen bombardment of silicon, even under cryogenic implantation conditions, is given as an example. From the standpoint of semiconductor technology, there are several current issues under intense study. Two of these are highlighted with respect to silicon technology: the problems of transient enhanced diffusion of dopants during low temperature annealing due to residual implantation-induced defects, and the need to remove extremely low concentrations of metals from active device regions. Finally, some recent novel applications of implantation in compound semiconductors are treated.

Ion implantation of semiconductors

Published data on the properties of metal-semiconductor ohmic contacts and mechanisms of current flow in these contacts (thermionic emission, field emission, thermal-field emission, and also current flow through metal shunts) are reviewed. Theoretical dependences of the resistance of an ohmic contact on temperature and the charge-carrier concentration in a semiconductor were compared with experimental data on ohmic contacts to II–VI semiconductors (ZnSe, ZnO), III–V semiconductors (GaN, AlN, InN, GaAs, GaP, InP), Group IV semiconductors (SiC, diamond), and alloys of these semiconductors. In ohmic contacts based on lightly doped semiconductors, the main mechanism of current flow is thermionic emission with the metal-semiconductor potential barrier height equal to 0.1–0.2 eV. In ohmic contacts based on heavily doped semiconductors, the current flow is effected owing to the field emission, while the metal-semiconductor potential barrier height is equal to 0.3–0.5 eV. In alloyed In contacts to GaP and GaN, a mechanism of current flow that is not characteristic of Schottky diodes (current flow through metal shunts formed by deposition of metal atoms onto dislocations or other imperfections in semiconductors) is observed.

Mechanisms of current flow in metal-semiconductor ohmic contacts

Carbon doping of III–V compounds grown by MOMBE

GaAs/AlGaAs heterojunction bipolar transitors (HBTs) utilizing highly Be‐doped base layers display a rapid degradation of dc current gain and junction ideality factors during bias application at elevated temperature. For example, the gain of a 2×10 μm2 device with a 4×1019 cm−3 Be‐doped base layer operated at 200 °C with a collector current of 2.5×104 A cm−2 falls from 16 to 1.5 within 2 h. Both the base emitter and base collector junction ideality factors also rise rapidly during device operation, and this current‐induced degradation is consistent with recombination‐enhanced diffusion of Be interstitials producing graded junctions. By sharp contrast, devices with highly C‐doped (p=7×1019 cm−3) base layers operated under the same conditions show no measurable degradation over much longer periods (12 h). This high degree of stability is most likely a result of the fact that C occupies the As sublattice, rather than the Ga sublattice as in the case of Be, and also has a higher solubility than Be. The effect of nearby implant isolated regions in promoting Be diffusion is also reported.

Stability of carbon and beryllium‐doped base GaAs/AlGaAs heterojunction bipolar transistors

The formation of high‐resistivity (>107Ω/⧠) regions in GaAs‐AlGaAs heterojunction bipolar transistor (HBT) structures by oxygen and hydrogen ion implantation has been investigated as a function of ion dose and subsequent annealing temperature (400–700 °C). Isolation leakage currents as low as 8 μA mm−1 at 6 V can be achieved between 100‐μm‐wide ohmic contacts separated by a 16 μm spacing. The isolation of these 1.8‐μm‐thick heterojunctions requires up to six different energy oxygen implants (40–400 keV) and three different energy proton implants (100–200 keV) with doses in the mid 1012 cm−2 range for O+ and 5×1014 cm−2 for H+ ions. Similar results can be achieved by substituting a MeV energy oxygen implant for the proton implants. The optimum post‐implant annealing temperature depends on the ion dose but is in the range 500–600 °C. The evolution of the sheet resistance of the implanted GaAs‐AlGaAs material with annealing is consistent with a reduction in tunneling probabilities of trapped carriers between deep level states for temperatures up to ∼600 °C, followed by significant annealing of these deep levels. Small geometry (2×9 μm2) HBTs exhibiting current gain of 44 and cutoff frequency fT as high as 45 GHz are demonstrated using implant isolation.

Implant isolation of GaAs‐AlGaAs heterojunction bipolar transistor structures

The first demonstration of GaAs/AlGaAs HBTs grown completely by metal organic molecular beam epitaxy (MOMBE) is reported. The The p-type dopant used for the base layer was carbon. Tin was used as the n-type dopant for the emitter as well as the collector. A common-emitter current gain of 140 was measured for 90 μm diameter devices with a base doping of 1 × 1019cm−3. Small area (2 × 6μm2), devices show a current gain cut-off frequency of 40GHz.

GaAs-AlGaAs HBT with carbon doped base layer grown by MOMBE

Changes in the sheet resistance of doped epitaxial layers of InP and In0.53Ga0.47As exposed to microwave Ar or H2 discharges were measured as a function of the exposure time (1–20 min), plasma pressure (1–20 mTorr) and the additional rf‐induced negative dc bias (50–400 V) on the sample. Changes in sheet resistance of ≤10% are only obtained for low dc biases (≤−75 V) or short exposures for either type of discharge. Hydrogen plasmas led to more substantial resistance changes than argon plasmas under all conditions, with the amount of damage introduction or hydrogen passivation being strongly dependent on dc bias and exposure time, but weakly dependent on pressure. The results indicate that high density, low pressure plasmas with low dc biases are capable of causing minimal disruption to InP‐based materials during dry etching for device fabrication.

T. R. Fullowan

Papers

Carbon doping of III–V compounds grown by MOMBE

Stability of carbon and beryllium‐doped base GaAs/AlGaAs heterojunction bipolar transistors

Implant isolation of GaAs‐AlGaAs heterojunction bipolar transistor structures

GaAs-AlGaAs HBT with carbon doped base layer grown by MOMBE

Damage introduction in InP and InGaAs during Ar and H2 plasma exposure