NAMBE2016 Session MBE-MoA: Science and Technology of MBE

In order to improve the crystalline quality of the GaSb buffer layer as a platform for device structures we propose the use of Ga-assisted oxide desorption to obtain an oxide-free smooth buffer layer on GaSb epi-ready substrates. The oxide desorption was monitored in-situ by reflection high energy electron diffraction (RHEED) and the surface morphology was confirmed ex-situ by atomic force microscopy (AFM). In particular we compare the surface morphology of GaSb films grown on flat vicinal substrates by the conventional thermal oxide-desorption method under antimony overpressure and the proposed Ga-assisted procedure. The results demonstrate that this procedure can replace the conventional method of thermal desorption. The technique described is ideal for in-situ cleaning of GaSb surfaces within the MBE growth chamber. The substrate temperature required is significantly lower than the temperatures used for thermal oxide desorption. We propose this technique as a means to achieve an oxide-free and atomically smooth buffer layer on epi-ready substrates. The study demonstrates the effectiveness of the Ga-assisted oxide-reduction technique. Figure 1 displays the AFM images of the buffer layer grown over the two oxide-desorption procedures followed.

The application of x-ray reflectivity characterization of epitaxial thin film analysis is limited by the large numbers of coupled variables associated with multilayered samples requiring numerical fitting routines to explore a large parameter space that often result in fit values with high residual errors and parameters that conflict with other measurements. Here, advanced signal processing enables high-resolution analysis of the x-ray reflectivity signal using Fourier analysis that uncouples the thickness and reflectivity strength for most layers and allows for assessments of composition and interface extent in a straightforward manner.

This analysis is applied to a metamorphic epitaxial sample with two silicon quantum wells grown within relaxed silicon germanium barriers. High resolution Fourier analysis directly determines that the quantum wells are 8.48 nm and 8.70 nm with the separating barrier between the wells being 22.02 nm, all with 95% confidence bound of + 0.02 nm. These values are verified with cross-sectional transmission electron microscopy.

For this sample high resolution x-ray reflectivity can provide valuable layer and interface information that is more accurate than x-ray diffraction because strong signals from the linearly graded buffer layer masks the contributions from the heterostructure. Details of this analysis in the context of dynamical simulations will be presented.

We have investigated various methods of preparing the surfaces of freestanding, Ga-polar, hydride vapor phase epitaxy (HVPE) grown GaN substrates for homoepitaxial GaN growth by plasma-assisted MBE. Substrate preparation using an aggressive ex situ wet chemical clean involving acid and base etching in addition to a solvent degrease reliably produces a surface that is sufficiently clean for homoepitaxial GaN growth to avoid the generation of new threading dislocations at the regrowth interface [1]. Cross-sectional transmission eletron microscopy (TEM) and secondary ion mass spectroscopy (SIMS) were used, respectively, to characterize the microstructure and measure the concentrations of impurities unintentionally incorporated in plasma-assisted MBE-grown homoepitaxial GaN layers. We have found that heating Ga-polar substrates to ~1100° C is as effective as a wet chemical clean for reducing impurity concentrations of oxygen, silicon, and carbon. Multiple cycles of gallium deposition and thermal desorption, such as is used effectively for N-polar HVPE GaN surfaces [2], can roughen the Ga-polar surface. We find that combining an aggressive ex situ wet chemical clean with gentle in situ Ga deposition and thermal desorption results in very low residual impurity concentrations and homoepitaxial GaN layer growth without generating new threading dislocations.

⁺ Author for correspondence: david.storm@nrl.navy.mil

[1] D. F. Storm et al., J. Cryst. Growth 380, 14 (2013)

[2] D. F. Storm et al., J. Cryst. Growth 409, 14 (2015)

The reliability of pyrometry for measurements of T_wafer < ~500°C is often compromised by radiation scattered from effusion cells, ion gauges and viewport heaters inside the growth chamber. Even though band-edge thermometry (BET) proved to be reliable for substrates such as semi-insulating (SI) GaAs, it is not suitable for small bandgap materials such as InAs or GaSb. We propose a simple method which allows the use of a BET instrument, here a 900-1700 nm InGaAs array, to monitor small bandgap substrate temperature down to about 200 °C. A quarter of InAs wafer was indium mounted on a 3” GaAs SI substrate. A Mo plate with a cut-out as shown in Fig.1 was used to cover the entire GaAs substrate with exception of a very small gap next to the edge of InAs substrate. The focusing optics was designed to collect light from an approximately 10 mm diameter spot, which could be sifted between the center of the InAs substrate for pyrometry measurements and the exposed GaAs for BET. BET measurements relied on the radiation from the substrate heater and the IR radiation from a halogen lamp, delivered to the back of the wafer by a quartz light pipe. For the pyrometric measurements of InAs we simply integrate the raw spectrometer signal across the opacity range of InAs (I_all), and subtract contributions from other hot sources (I_n). The thermal emission from InAs can be well described by the relation:

I_InAs(T_InAs) = I_all(T_InAs) - Σ_nI_n = C₁exp[-C₂/(T_InAs+273)],

where we assume that the T_InAs = T_BET, contributions I_n do not depend on T_InAs, and the detector response is linear. This equation can be subsequently used to derive the temperature of indium-free mounted substrates. We will discuss the methods used for deriving the individual contributions I_n and the validity of all the assumptions.

Nanoelectronics, complex heterostructures, and engineered 3D matrix materials are quickly advancing from research possibilities to manufacturing challenges for applications ranging from high-power devices to solar cells to fuel cells to any number of novel multifunctional sensors and controllers. One key aspect of enabling the transition from research to effective manufacturing is the ability to consistently produce a clean and smooth starting surface over a large area. This challenge exists for almost any material system, and often appears multiple times during device manufacturing. While this is not a new problem, it is a persistent one whose solution is becoming more complex as nano-scale features increasingly control device operations.

Many cleaning approaches have been investigated for different materials. Most can be classified as wet chemical cleaning [1], plasma treatment using either hydrogen or oxygen [2], high-temperature annealing, or some combination of these methods. Wet chemical treatments can be extremely uniform on the production scale, but often involve hazardous materials that increase the cost and risk of manufacturing. Annealing processes, either in vacuum or a controlled gas atmosphere, are not an option for many compound materials or for intermediate processing steps due to unwanted diffusion or evaporation caused by the high-temperature processing. Plasma treatments enable a wide range of chemical reactions on the surface, though most are not fully understood and thus lead to inconsistencies, ion damage, and other undesirable results.

Atomic hydrogen cleaning has been shown to remove surface hydrocarbons and oxides from a range of semiconductors including GaAs, InAs [3], and II-VI compounds by forming volatile H-compounds which desorb from the surface [4]. However, the details of this apparently direct mechanism are not easily predicted, and show the expected and yet complex interrelated dependence on factors such as surface structures, H/H₂ ratios, and temperature. We are investigating the hydrogen cleaning of a variety of material systems in order to better understand how to enable high quality, large area surface preparation in a manufacturing setting. Our substrate materials include Ge, Si, MCT, CZT, and SiC. Atomic hydrogen is generated from a variety of sources including a high-temperature atmospheric pressure hydrogen furnace, hydrogen atom beam thermal gas cracking cell operated as part of a vacuum processing system, and a UV light source operated both in atmosphere and as part of a vacuum processing system. We will show these comparisons and postulate some opportunities for unique processing and additional research.