The Investigation of III-V Semiconductors using SuperSTEM

In recent years, the performance of electron microscopes has been greatly improved through the implementation of practical aberration correction technology. This has allowed the creation of instruments that can form Å-scale electron probes at acceleration voltages of only 100kV. As an example, SuperSTEM 1 was the first UK based aberration-corrected 100kV field emission gun scanning transmission electron microscope (FEG-STEM) that was capable of achieving a spatial resolution of 1Å. Instruments, such as SuperSTEM 1, permit a wide range of nanostructures to be studied at scales that were not previously possible in commercial microscopes.

The introduction of aberration-corrected instruments has been an important development for the characterisation of state of the art semiconductor materials. For instance, some III-V semiconductor structures already incorporate layers that are only a single atom in width. However, due to the limitations of the techniques that have been previously used to characterise such materials, it remains unclear exactly how successful growth methods (such as MBE i.e. molecular beam epitaxy) actually are at producing sharp interfaces. Hence, the ability to study semiconductor materials at the atomic scale has become ever more crucial for technological and economic reasons.

In this project, SuperSTEM 1 was used to study several MBE grown III-V semiconductor nanostructures. These materials have applications in present, and possibly future, semiconductor devices. However, in order to improve the performance of such devices, a more in-depth appraisal of the associated growth techniques is necessary. Hence, the aim of this project was to provide atomic scale information on the composition and interfacial sharpness of the various layers that were present in the MBE grown III-V semiconductor nanostructures. This project also required a greater understanding of some aspects of probe scattering and the HAADF (high angle annular dark field) imaging technique due to the exceptional Å-scale spatial resolution of SuperSTEM 1.

Background information on the type of III-V materials that were studied in this project is given in Chapter 1. This information involves a description of superlattices and doped heterostructures. In addition, the molecular beam epitaxy growth technique is explained along with previous estimates of the sharpness of AlAs / GaAs and InAs / GaAs interfaces. At the end of Chapter 1, a small section is devoted to the background of the SuperSTEM project.

Chapter 2 is mainly concerned with the experimental apparatus and techniques that were employed to characterise the various materials. For instance, the chapter begins with a short introduction to the foundations of electron microscopy such as electron lenses and their associated aberrations. The CTEM (conventional transmission electron microscope) and STEM imaging techniques are also explained along with the principles of the HAADF STEM imaging process and EELS (electron energy loss spectroscopy). The two electron microscopes that were utilised in this project are also outlined. These instruments comprise a Tecnai F20 and SuperSTEM 1. In the case of SuperSTEM 1, the method of aberration correction is also outlined. It should be noted that the majority of the results in this project were obtained using SuperSTEM 1 and the HAADF imaging technique. Finally, a summary of the specimen preparation techniques are given at the end of Chapter 2.

In this project, all of the specimens were orientated along the <110> direction to give the familiar dumbbell configuration characteristic of zinc-blende crystals. In addition, the first material that was studied through the use of SuperSTEM 1 was a MBE grown III-V semiconductor heterostructure that formed part of a high frequency modulation doped field effect transistor (MODFET). The results from this investigation are presented in Chapter 3. In order to enhance the analysis of the SuperSTEM 1 data, HAADF images were converted into maps of the dumbbell column ratio. These maps gave an indication of the dumbbell composition of the various layers that were present in the heterostructure. HAADF images also revealed that several growth defects were present throughout the heterostructure.

Frozen phonon multislice computer simulations were also performed in order to gain a deeper understanding of the scattering behaviour of Å-scale electron probes in III-V semiconductor materials. Real space electron intensity distributions in crystals of GaAs [110], AlAs [110] and InAs [110] were calculated as a function of probe size, probe position and crystal thickness. It was found that the strong channelling depth was largest in atomic columns constructed from low Z number atoms such as Al. The extensive set of results from the simulations is discussed in Chapter 4.

In addition, the HAADF image contrast and the HAADF dumbbell column ratio of AlAs [110] and GaAs [110] were calculated as a function of thickness for the SuperSTEM 1 probe. These HAADF imaging attributes were compared against experimental values obtained using SuperSTEM 1. It was established that the simulated and experimental values of the dumbbell column ratio were in good agreement. However, there was a poor correspondence between simulation and experiment when the image contrast was considered. The comparison between simulated and experimental values is shown in Chapter 5.

In the case of MBE grown AlAs / GaAs structures, interfacial sharpness was investigated as a function of specimen thickness. This is presented in Chapter 6. It was demonstrated that AlAs-on-GaAs interfaces were associated with large [110] surface step lengths and GaAs-on-AlAs interfaces were associated with elemental diffusion. These results were obtained from a single AlAs / GaAs interface and a wide layer 9ML AlAs / 9ML GaAs superlattice. Furthermore, the quality of a narrow layer 1ML AlAs / 2ML GaAs superlattice is reported in Chapter 6.

A study of InAs / GaAs superlattices is presented in Chapter 7. Despite the fact that In tends to spread over several monolayers, the results highlight that such multilayers are able to be grown using MBE. However, it is not clear whether such structures can be grown to a similar quality as MBE grown AlAs / GaAs structures. In addition, a study of Si δ-doped layers is detailed in Chapter 7. The results suggest that the detection of such a small concentration of Si in GaAs is at the limit of what can be achieved using SuperSTEM 1.

The final chapter (Chapter 8) contains a discussion of the experimental findings from this project. This chapter also considers some improvements that could be made to the experiments along with possible future work.