The success In attaining atomic layer epitaxy (ALE) of GaAs depends critically on the choice of the Ga precursor. Three systems were examined: trimethylgallium (TMGa) and diethylgallium chloride (DEGaCI) both of which give ALE, and triethylgallium (TEGa) which does not. We compared the surface reactions of these compounds on GaAs(100) and concluded that there was no evidence for reaction selectivity between Ga and As sites to cause ALE. Site blocking by the ligands on the Ga precursors alone also could not provide a self-limiting Ga deposition for ALE. We found evidence of a new mechanism by which self-limiting deposition of Ga resulted when the incoming Ga flux by the adsorption of Ga precursors was counter-balanced by an outgoing flux of Ga containing reaction product. For TMGa and DEGaCI with which ALE is successful, the products are CH3Ga and GaCl, respectively. For TEGa, the corresponding compound C2H5Ga was not formed.

In this paper we review three proposed mechanisms for GaAs ALE and review or present data in support or contradiction of these mechanisms. Surface chemistry results clearly demonstrate that TMGa irreversibly chemisorbs on the Ga-rich GaAs(100) surface. The reactive sticking coefficient (RSC) of TMGa on the adsorbate-free Ga-rich GaAs(100) surface was measured to be ∼0.5, conclusively demonstrating that the “selective adsorption” mechanism of ALE is not valid. We describe kinetic evidence for methyl radical desorption in support of the “adsorbate inhibition” mechanism. The methyl radical desorption rates determined by temperature programmed desorption (TPD) demonstrate that desorption is at least a factor of ∼10 faster from the As-rich c(2 × 8)/(2 × 4) surface than from the Garich surface. It is this disparity in CHs desorption rates between the As-rich and Ga-rich surfaces that is largely responsible for GaAs ALE behavior. A gallium alkyl radical (e.g. MMGa) is also observed during TPD and molecular beam experiments, in partial support of the “flux balance” mechanism. Stoichiometry issues of ALE are also discussed. We have discovered that arsine exposures typical of atmospheric pressure and reduced pressure ALE lead to As coverages ≥ 1 ML, which provides the likely solution to the stoichiometry question regarding the arsine cycle.

X-ray photoelectron spectroscopy (XPS) and reflection high energy electron diffraction (RHEED) have been applied to study the stable adsorbed Ga species and surface structures after GaAs (001) 2 × 4 As-rich surfaces are exposed to TMGa. These studies show that Ga atoms are the final adsorbed species, that Ga deposition is saturated at one atomic layer at temperatures between 360 and 530 °C and that the surface converts from a 2 × 4 to a 4 × 6 reconstruction after TMGa adsorption. To understand the surface reaction kinetics involved, reflectance-difference spectroscopy (RDS), an in situ rea-ltime optical technique developed by Aspnes et al., is applied to investigate TMGa adsorption on (001) GaAs surfaces. The kinetics of the surface reactions and reconstructions have been characterized over the temperature range from 400 to 500 °C using RDS. The transient RDS behaviors are interpreted by the application of a model that involves selective adsorption and reaction of TMGa at surface As sites and at Ga vacancies on Ga-rich reconstructed surfaces. Based upon these interpretations, rates of reaction and by product desorption are determined that suggest optimal strategies for ALE growth of GaAs.

Trimethylgallium (TMGa), Triethylgallium (TEGa) and Diethylgalliumchloride (DEGaCl) are used in combination with arsine (AsH3) in atomic layer epitaxy of GaAs. We simulated this process in a UHV system by dosing a c(2×8) or (1×6) reconstructed GaAs(001) surface with arsine and organometallics at different temperatures. After the surface reconstruction was verfied by LEED, the sample was transferred in situ to the HREELS (high resolution electron energy loss spectroscopy) chamber. We present new studies on the adsorption and pyrolysis of organometallics on GaAs(001) surfaces.

As a result we find that TMGa adsorbs molecularly on an As-rich cooled surface but decomposes upon adsorption on a Ga-stabilized (1×6) surface. These two adsorption states are identified by the geometrical position of the Gamethyl compound on the surface (Ga-C bond parallel or normal to the surface). The removal of the adsorbed species occurs on an As-rich surface at temperatures higher than 450°C and on the Ga-stabilized surface at 350°C.

For TEGa and DEGaCl we found similar results.

We have studied the adsorption of hydrogen atoms on GaAs (100) by multiple internal reflection infrared spectroscopy. The crystal was etched in 1:1:10 H3PO4/H2O2-/H2O solution and in 1:1 HCI/H2O solution, then annealed to 580°C in the vacuum chamber. Hydrogen adsorption was carried out at -90 and 45°C. At both temperatures, a monolayer forms giving rise to infrared bands for arsenic hydride and gallium hydride at 2105 and 1860 cm−1, respectively. The arsenic hydride vibration is polarized parallel to the surface, whereas the gallium hydride vibration is polarized normal to the surface. By monitoring the changes in the intensity of the infrared absorption bands with time during exposure to H atoms and during heating, the kinetics of hydrogen adsorption and desorption can be measured. At -90°C, the H atom sticking probability follows the Langmuir model, S/So = (1-θH). Upon heating the crystal, the arsenic hydride rapidly decomposes near 120°C, while the gallium hydride slowly decomposes between 150 and 400°C.

(AIAs)1/2(GaAs)1/2 fractional-layer superlattices (FLSs) are grown on (001) vicinal substrates by metalorganic chemical vapor deposition. Various kinds of GaAs substrates are used. When the substrate is misoriented to [110] direction by 1.92° and [110] by 0.10°, uniform superlattice periods in a large surface area are observed with a bright field transmission electron microscope (TEM). The results suggest ideal crystal growth from kink sites during MOCVD growth, and the distances between the kink sites are equal.

On a substrate misoriented to [110] by 1.90°, the superlattice periods exhibit an undulation. This shows that kink flow mode growth is not dominant in the [110] direction. On a substrate misoriented to [010[ by 2.0°, no superlattice periods were observed. From the above results, we discuss the growth mechanisms.

Polarization dependent photolumlnescence and optical absorption spectra of FLS were also observed. Electron wave interference devices with lateral periodic potential were fabricated.

A variety of optical methods are now available for studying surface processes and for monitoring layer thicknesses and compositions during semiconductor crystal growth by molecular beam epitaxy (MBE), organometallic chemical vapor deposition (OMCVD), and related techniques. New capabilities for surface analysis are being provided by developing techniques such as reflectance-difference spectroscopy (RDS), which use intrinsic symmetries to suppress ordinarily dominant bulk contributions. Bulk and microstructural properties such as compositions and layer thicknesses can be determined by techniques such as spectroellipsometry (SE), which return information integrated over the penetration depth of light. Recent advances include the application of reflectance to monitor dynamic surface processes, RDS to characterize (001) GaAs surfaces in OMCVD environments, and SE to control growth of AlxGa1-x, As materials and structures.

A modular spectroscopic ellipsometer, capable of both in-situ and ex-situ operation, has been used to measure important growth parameters of GaAs/AIGaAs structures. The ex-situ measurements provided layer thicknesses and compositions of the grown structures. In-situ ellipsometric measurements allowed the determination of growth rates, layer thicknesses, and high temperature optical constants. By performing a regression analysis of the in-situ data in real-time, the thickness and composition of an AIGaAs layer were extracted during the MBE growth of the structure.

The adsorption and surface reactions of trimethylgallium and tertiarybutylarsine on GaAs(100) surfaces have been investigated by Fourier transform infrared spectroscopy. Adsorbed methyl groups resulting from the dissociative chemisorption of trimethylgallium on GaAs(100) are shown to form As-H and CH2 species on the surface. The CH2 groups are stable on the surface at temperatures as high as 550 °C. The surface coverage is low (∼0.2% of a monolayer) and is reduced by the presence of hydrogen on the surface. This dehydrogenation of surface methyl groups could be a possible route to carbon incorporation in GaAs grown by atomic layer epitaxy. Tertiarybutylarsine is shown to decompose primarily by homolysis to form a tertiary butylgroup and AsH2. At temperatures below 400°C on trimethylgallium dosed surfaces, the decomposition products appear to cause the hydrogenation of methylene groups remaining from prior surface dosing with trimethylaallium. At high temperatures, the tertiarybutyl radical appears to undergo dehydrogenation reactions to an unsaturated species which is stable on the surface. In contrast, the dehydrogenation does not appear to occur on surfaces treated with tertiarybutylarsine. The data for trimethylgallium and tertiarybutylarsine support the general assertion that surface As-H species play a critical role in the removal of hydrocarbon species from the growth surface.

In-situ spectroscopic ellipsometry (SE) was applied to monitor GaAs (100) surface changes induced at elevated temperatures inside an ultrahigh vacuum (UHV) chamber (<1×10−9 torr base pressure, without As overpressure). The real time data showed clearly the evolution of the native-oxide desorption at ∼577°C, on a molecular-beam-epitaxy (MBE)-grown GaAs (100) surface. In addition, surface degradation was found before and after the oxide desorption. A clean and smooth surface was obtained from an arsenic-capped, MBE-grown GaAs sample, after the arsenic coating was evaporated at ∼350 °C inside the UHV. Pseudodielectric functions <ε>GaAs, from 1.6 eV to 4.5 eV, were obtained through the SE measurements, from this oxide-free surface, at temperatures ranging from room temperature (RT) to ∼610 °C. These <ε> data were used as reference data to develop an algorithm for determining surface temperatures from in-situ SE measurements, thus turning the SE instrument into a sensitive optical thermometer.

We have determined that the temperature for desorption of gallium oude from GaAs increases linearly with oxide thickness, for oxide layers between about 6Å and 26Å thick. Different thicknesses of oxide layers were created by varying the exposure time of the GaAs wafers to a low pressure oxygen plasma. In addition, we show by diffuse light scattering that highly polished GaAs substrates roughen during the oxide desorption. These results are interpreted in terms of a model in which the oxide evaporates inhomogeneously. The oxide desorption was also studied by monitoring the secondary electrons produced by the high energy electrons from the RHEED gun. After the gallium oxide desorption there is a reversible, order of magnitude, increase in the number of secondary electrons produced. We interpret this result as evidence for the formation of microscopic gallium droplets on the GaAs surface.

The potential applications of Atomic Layer Epitaxy of III–V compounds will be outlined. These include the growth of special structures and devices such as ordered alloys, ultra-thin quantum wells, non-alloyed contacts, planar doped FET's and HBT's. Also, the main challenges facing ALE will be outlined along with possible solutions. These include reactor design, control of carbon doping and the growth of ternary alloys. A general assessment of the ALE technology will be provided.

Two implementations of laser assisted atomic layer epitaxy(LALE) for selective area growth of GaAs using trimethylgallium and AsH3 as precursors are described. A wide range of growth parameters lead to self-limiting monolayer/cycle growth which is suited for precise layer thickness control. By combining LALE with conventional metalorganic chemical vapor deposition, A10.3Ga0.7As/GaAs double heterostructures including LALE GaAs have been grown, permitting electrical and optical characterization to be performed on the thin and small areas of the LALE deposits. The information is used in a growth parameter optimization process resulting in device quality GaAs. Quantum well lasers with active region grown by LALE are demonstrated for the first time. The application of LALE to optoelectronic integration is demonstrated by depositing small area quantum wells as the gain medium in an otherwise transparent waveguide.

Atomic layer epitaxy (ALE) using laser irradiation and digital etching of GaAs are described herein.

Epitaxy: We have succeeded in the laser-assisted ALE (laser-ALE) of GaAs using visible wavelength Art laser irradiation and an alkylgallium source. Visible wavelength photon irradiation induces surface decomposition but not volume decomposition of alkylmetal molecule source gases. ALE is realized by the enhancement of decomposition of alkylgallium molecules only on the As-terminated surface but not on the Ga-terminated surface. This site-selectivity of alkylgallium decomposition is induced by the optical absorption band broadening, which is due to the chemisorption of alkylgallium at the As-terminated surface.

Etching: In ditigal etching, etchant gas pulses and an energetic beam sequentially impinge onto the substrate surface. In the Ar+/Cl2 system, the etch rate is found to be independent of both Cl2flux and Ar+ beam density, and the etch rate saturates at a level below one monolayer per cycle. By using Cl radicals as etchants instead of Cl2, the self-limited etching characteristics of digital etching are obtained within both the Ar+ incidence time and Cl feed time of the etching cycle.

The effects of the growth temperature and exposure time to TMGa for ALE of gallium arsenide was studied using TMGa and AsH3 in a modified, vertical, atmospheric, MOCVD reactor with a rotating susceptor. It was found that the temperature range for ALE growth could -be extended from 450°C to 700°C by adjustment of the exposure time to TMGa. The maximum exposure time to TMGa was found to decrease as growth temperature increased with high temperature growth limited to exposures of only fractions of a second. Beyond a critical exposure time to TMGa, gallium droplets form on the surface. It is known that premature decomposition of TMGa in the heated gaseous boundary layer causes the formation of the gallium droplets and the consequent loss of ALE growth.

Processes limiting the growth of GaAs grown by an alternate gas supply are investigated by kinematical analysis. Based on these results, it is shown that the atomic layer epitaxial (ALE) window is expanded on the high temperature by the suppression of the decomposition of column III gas sources using a nitrogen carrier gas and on the low temperature side by the enhancement of their chemisorption to substrate surface atoms by a new method using a cracking tube. The latter enables us to achieve ALE of AlAs for the first time. Moreover, the carbon concentration is reduced by one order of magnitude by such a reaction control.

A simple kinetic model has been developed to explain the agreement between in situ and ex situ determination of phosphorus composition in GaAs1−xPx (x < 0.4) epilayers grown on GaAs (001) by gas-source molecular-beam epitaxy (GSMBE). The in situ determination is by monitoring the intensity oscillations of reflection high-energy-electron diffraction during group-V-limited growth, and the ex situ determination is by x-ray rocking curve measurement of GaAs1−xPx/GaAs strained-layer superlattices grown under group-III-limited growth condition.

Epitaxial (GaAs)1−x (Si2)x metastable alloys have been grown on GaAs (100) substrates using Migration-Enhanced Epitaxy in the composition range of 0<x<0.25. The lattice constant a0 of the alloys was found to decrease with increasing Si content from 0.56543nm at x=0 to 0.5601nm at x=0.25. Double-crystal x-ray diffraction rocking curve measurements and cross-sectional transmission electron microscopy studies made on a 10 period (GaAs)1−x(Si2)x/GaAs strained layer superlattice indicated sharp and abrupt interfaces. High crystalline quality GaAs has been grown on Si substrates using (GaAs)0.80(Si2)0.20/GaAs strained layer superlattices as buffer layers.

Reactor design considerations are discussed relevant to the two main problems, carbon contamination and low growth rate, facing Atomic Layer Epitaxy (ALE) of GaAs. A new reactor design addressing these problems is described. It utilizes the concept of rotating the substrate between streams of reactant gases. The growth chamber provides baffles and gas jets to shear off and sweep away the thermal boundary layer after exposure to the reactive gas streams. Construction is based on modification of a commercially available low pressure MOCVD reactor equipped with a. load lock. The reactor is capable of processing three, two-inch wafers. A background carbon concentration of about 1015cm−3 and a 77°K mobility of 30,000 cm2/V-sec were achieved. Self limited growth was observed for a growth temperature as high as 600 °C. Controlled p- and n-type doping was accomplished by changing growth conditions and adding silane.

GaAs p+-n+ tunnel diodes have been grown by atmospheric-pressure organometallic vapor phase epitaxy (OMVPE) using zinc as the dopant for the p+ regions and either Se or Si as the dopant for the n+ regions. At a growth temperature of 700° C, using a “cycled” growth for just the Zn-doped p++-GaAs layer both the conductance and the peak current of the tunnel diode has been increased by a factor of ˜65. The conductance of the tunnel diode, maximized at a growth temperature of 650 °C with the cycled growth, is comparable to the best reported values by MBE. Cycled growths for n+ Se-doped regions reduce the tunnel-diode conductance by more than two orders of magnitude. However, the cycled growth for n+-GaAs regions formed with Si doping shows no conductance degradation. A model for these observations is presented.