Search

Czochralski (Cz) silicon wafers are used almost exclusively for the fabrication of VLSI devices. Such silicon contains excess oxygen which precipitates as oxide particles either when the initial ingot is grown or subsequently during the wafer device fabrication. Such oxide particles can produce reduced device performance or failure if they occur within the active device regions. However, they can be used to improve the device performance by a process known as internal oxide gettering. The wafers are given a series of preanneal treatments to produce controlled precipitation in which a surface zone of the wafer to a depth of typically in the range of 10 to 50 μm is denuded of oxide particles, while the remainder of the wafer contains large numbers of well formed particles. The devices are fabricated in the surface denuded zone and harmful contaminating metal impurities are attracted during the heat treatment stages away from the device regions to precipitate at the underlying oxide particles or their associated dislocations. In this way device yields can be significantly increased. Because of the importance of these oxide particles and the oxygen precipitation process for VLSI fabrication, considerable efforts have been made to develop methods to assess the numbers and distributions of such particles within the wafers. The number density range of most interest is 107 to 1010 cm-3, and the particle size range is typically 30 to 300 nm. The method that has mostly been used is surface etching followed by optical microscopy to obtain etch pit densities. Transmission electron microscopy is a research method used for obtaining detailed information concerning a small number of individual particles. However, because these methods are destructive, much attention has been given during the last few years to the development of infra-red microscopy methods to directly image the particles within the silicon wafers. Although the particles are smaller than the resolution of these methods, individual particles can nevertheless be imaged. This is because the particles are mostly further apart than the resolution limit, and the sensitivity can be sufficient high that adequate contrast occurs. The contrast arises from scattering or absorption of the light by the particle. Infra-red imaging methods developed include infra-red microscopy (IRM), laser scattering tomography (LST), optical precipitate profiler (OPP) and scanning infra-red microscopy (SIRM), all described more fully in Chapter 2. The SIRM has been developed and used to investigate a variety of semiconductor specimens in the Materials Department, Oxford University, during the last ten years. The SIRM has a good performance and flexibility making it especially suitable as a research instrument. Although all of these infra-red imaging methods have been successful to different degrees in assessing oxide particles in Cz silicon wafers, their performance has at least initially been assessed by comparing the number densities and distributions thus obtained with the corresponding results produced by etch pit studies. Furthermore, no serious attempt has yet been made to develop a rigorous theory of the imaging process and to compare the predictions with the experimental images. One of the main objectives of the present work is to do this or at least to make a significant start to such a project based on the SIRM. The outline of an ideal project which aims at a full understanding of the imaging process and the contrast mechanisms is as follows. The performance of the present Oxford SIRM should be improved and the number of imaging modes increased. The improved performance, i.e. better lateral and depth resolutions and higher sensitivity, would enable smaller particles and higher number densities to be imaged, and hence better quantitative data obtained. The larger number of imaging modes would enable the optimum method to be used to image different types of particle. A rigorous theory should be developed that can describe the imaging process and the contrast mechanisms. First, the illumination system should be studied, and in particular the structure of the focussed probe within the specimen and how the structure changes on focussing deeper into the specimen. Second, the interaction of the light with the specimen should be investigated and especially how light is scattered by individual oxide particles in silicon for the case of the particle size being smaller than the light wavelength. Third, the detection system should be considered. For example, for the reflection confocal SIRM, how the light back scattered by the particles is collected by the probe forming lens and imaged at a pin-hole aperture placed in the front of the detector. Well designed experiments are required to determine the imaging properties of the different modes and comparisons should be made between the experimental and theoretical data. The successful conclusion of such a project would enable SIRM images of the particles to be more fully interpreted and hence more detailed information obtained concerning the particles. Furthermore, the images expected from different types of particle could be more closely predicted, e.g. whether they are detectable or not, and hence materials projects could be better planned at the outset. In this thesis we describe the methods that are presently being used to assess oxide particles in bulk silicon (Chapter 2). We review the literature on scanning optical microscopy covering both visible and infra-red light, present some considerations regarding the design of a high performance and versatile SIRM, and describe the various microscope modes that have been or could be used to image particles in semiconductors with infra-red light (Chapter 3). We give a detailed rigorous theoretical analysis of the energy distributions in the probe for the case when the light is focussed by a high numerical aperture lens from air into silicon (Chapters 4, 5). Numerically computed distributions are obtained to illustrate how the probe changes under different conditions, e.g. different focussing depths (Chapter 6). The relationship between the penetration depth of the probe and the spherical aberration coefficient arising from the silicon specimen is determined (Chapter 7). The classical theory of light scattering is applied to individual spherical silicon dioxide particles embedded in silicon. Numerical results are presented and a contrast mechanism is proposed to describe how the scattered intensity depends on particle size (Chapter 8). A formal solution relevant to the reflection confocal SIRM is given to treat the backward propagation of light using a model which takes into account the polarisation state of the incident light, the spherical aberration introduced by the silicon wafer, the polarisation state of the scattered light and the size of the pin-hole (Chapter 9). Experimental results are obtained for most of the imaging modes described in Chapter 3, specimens being selected so that the wide range of the imaging capabilities of the SIRM is shown, and experimental contrast values are compared with theoretical values (Chapter 10). Finally, overall conclusions are drawn and suggestions are made for completing the work started here (Chapter 11).

This work describes results obtained from detailed TEM, TED, HREM and SIMS analysis of the as-implanted and annealed microstructures of high dose (O.lxl0 17/cm2 to 1.7xl018/cm²) oxygen implanted silicon (Si). Molecular oxygen (O2+) has been implanted into Si wafers at equivalent energies of 200keV/O + , 90keV/O +, 70keV/O+ and 50keV/O+ to form, after annealing in flowing N2 or Ar + ½ %O2, buried SiO2 layers below single crystal surface Si layers. This process is called Separation by the IMplantation of OXygen (SIMOX). The energies and doses investigated are potentially suitable for the fabrication of two types of "thin-film" SIMOX substrates, which have major potential benefits for high-performance CMOS devices. This work is concerned with investigating the as-implanted and annealed microstructures, understanding the basic processes and mechanisms taking place during implantation and annealing, and establishing optimum fabrication parameters. Similar microstructures and changes in microstructure as a function of the dose are observed for the different implant energies investigated. For all the different energies and doses investigated, SiO2 precipitates are present after implantation. Five different precipitate morphologies are observed. The precipitate morphology depends on the oxygen concentration and the depth below the surface. The local and long range strain also play a role in determining the precipitate morphology. For doses of 0.5xl018O/cm² to 0.7xl018O/cm² at 200keV defects at the wafer surface are non-uniformly distributed across the implanted area. The regions of defects are in plan-view rectangular in shape with edges parallel to <100> directions. The percentage of the implanted surface that is covered by these rectangular regions depends on both the dose and the time-averaged beam current density. This is the first known report of such non-uniform distribution across the implanted area of defects at the wafer surface and their occurrence in regions with precise rectangular shapes. Previously unreported "line" defects below the peak of the as-implanted oxygen distribution for these energies are investigated. They are considered to be platelets on {100} planes and edge dislocation loops on {110} planes. After annealing, two major types of defects are present, threading dislocations in the surface Si layer and Si islands within the buried SiO2 layer. Correlation of as-implanted and annealed microstructure suggests that the threading dislocations originate in the defects present at the wafer surface after implantation and grow down during annealing. The Si islands originate from Si isolated from the surface Si layer and the substrate during implantation or annealing. The optimum dose for forming a SIMOX structure at a particular energy with both a low threading dislocation density and a low Si island density is just greater than the minimum dose for forming a continuous buried SiO2 layer after annealing. In order to try and reduce the density of Si islands within the buried SiO2 layer, graded low energy implants and interim rapid thermal anneals are investigated. Their influence on the microstructure is reported. The experimental results enable, for the implantation and anneal conditions used, the likely threading dislocation and Si island density after annealing to be estimated for a particular dose and energy. Simple models have been proposed for calculating typical oxygen diffusion lengths during implantation, the thicknesses of the buried SiO2 and surface Si layers after annealing and conversely the implant dose and energy required to fabricate a SIMOX substrate with a certain thickness of buried SiO2 and a certain thickness of surface Si layer after annealing. Thin-film SIMOX substrates consisting of a thin surface Si layer above a thin buried oxide layer suitable for high-performance "fully depleted" CMOS devices have been successfully fabricated by implanting doses of ≥0.35 and ≥0.33xl018O/cm2 at energies of 90keV and 70keV, repectively. Thin-film SIMOX substrates with a thin buried oxide layer below a standard thickness surface Si layer suitable for radiation hard circuits with reduced circuit self-heating have been successfully fabricated by implanting doses of ≥0.56xl018 O/cm2 at an energy of 200keV.

This work describes results obtained from TEM, TED and HREM studies of MBE and MOCVD InASySb1-y, MOCVD InxGa1-xAs, MOCVD InPySb1-y and MOCVD GaPySb1-y layers which were grown over a wide range of conditions. These semiconductor layers are of considerable importance for a variety of applications in optoelectronic and high-speed devices. TEM/TED investigations showed that phase separation occurs in MBE InAsSb layers, resulting in two phases with platelet structures ~5 to ~200nm thick approximately parallel to the layer surface. Phase separation was dependent on growth temperature and layer composition. Anisotropic geometry of the platelets was observed when viewed in the [110] and [110] directions. The compositions of the two phases were derived by TED and EDX analyses. A model for the phase-separated layers was proposed based on the presence of a miscibility gap and using the lateral and island growth mechanisms. TEM results of InGaAs, InPSb and GaPSb layers showed a fine scale modulated contrast (8-20nm in scale) which is a characteristic of alloy clustering occurring by spinodal decomposition, and a fine scale speckle contrast (4-5nm in scale). TEM/TED studies showed that [110] diffuse intensity lines in [001] TED patterns of InGaAs are not related to the fine scale modulated contrast but to the fine scale speckle contrast. It was concluded that a fine scale modulated contrast due to alloy clustering coexists with a fine scale speckle contrast associated with static atomic displacements from the average lattice in InGaAs. For InPSb and GaPSb, a fine needle-like contrast was also observed, which corresponds to diffuse streaks in the [110] patterns. This fine needle-like contrast was attributed to segregation of atoms at missing rows of atoms in the reconstructed growing surface. TED investigations revealed CuPt-type ordering in some of the InGaAs, InAsSb and InPSb layers. Regardless of alloy systems and growth conditions, the ordering occurred on only two of the four possible {111} variants. The degree of ordering was strongly dependent on growth conditions. Two variants of the ordered regions in InGaAs nucleated separately. TED/HREM studies of the ordered structure in InGaAs revealed a direct relationship between the inclination and elongation of superlattice spots and the morphology of anti-phase boundaries present within the domains. Two competing processes of surface-induced ordering, and bulk-induced disordering within a transition region, were considered to interpret the growth condition dependence of the ordering in InGaAs. A model for the ordering observed was proposed based on the surface reconstruction mechanism. MBE InAsSb strained layer superlattices (SLSs) were examined by TEM and HREM techniques. Defect configuration and the atomic structure of tetragonal distortion of the SLSs were directly imaged. Defect behaviour was dependent on the geometry of the SLSs. Possible relaxation mechanisms for the SLSs were proposed.

Transmission Electron Microscope (TEM) studies have been carried out of emitter regions in polysilicon contacted emitter bipolar transistors. The preparation of suitable TEM thin foils is described. In addition a technique is developed for the observation and quant jtative interpretation of the break-up of the interfacial oxide layers observed in these samples. The effect of annealing the samples prior to emitter dopant implantation (pre-annealing) is investigated for phosphorus and arsenic doped samples, implanted into a polysilicon layer 0.4μm thick, with a dose of 1x1016cm2. Two wafer pre-cleans have been used prior to polysilicon deposition to produce a thin oxide (0-8Å) and a thicker oxide (14Å). In the presence of the thinner oxide, the phosphorus doped samples enhance epitaxial regrowth of the polysilicon layer compared with the arsenic doped or undoped samples. In the presence of the thicker oxide, no difference is observed between the samples. A mechanism is proposed to explain this. The mechanisms controlling the gain of a phosphorus doped device are investigated and a model proposed to explain the observed electrical characteristics. It is concluded that there are two mechanisms responsible for the observed supression of hole current. Firstly tunnelling through the interfacial oxide and secondly some blocking effect of the interface. Carrier transport in the polysilicon is not believed to contribute to this supression. A dopant sensitive etch has been applied to TEM thin foils containing fully processed emitters in state-of-the-art devices. The shape of the emitter dopant distribution is revealed in such devices for the first time, and a 2-D profile is obtained from the emitter. It is shown that reduction in the emitter depth to 8OOÅ or less does not alter the emitter dopant geometry. The technique is demonstrated to be capable of resolving spatial separations of dopant iso-concentration contours of 100Å or less.

Individual, well structurally characterised dislocations present in n-type silicon have been studied using the electron beam induced current (EBIC) mode of an SEM.