We are also very pleased to report that a new project has recently begun sponsored by Matsushita Electric Works Co, who have an office on the Research Park. The contract which supports a study of light emission from microcrystalline silicon, is initially for two years. The contract allows Mr Takuya Komoda, the Manager of MEW, UK, to register as a post graduate student and to spend about half of his time working in the University. It also supports a full time, research studentship over three years.
During the past year, the group has made significant progress in the utilisation of the new HVEE 2 MV high energy implanter. There have, however, been numerous unexpected breakdowns and stoppages, which have delayed our programme of development. But, despite these setbacks, many SERC customers have been serviced and various development projects have been completed. Further developments and improvements remain a high priority for the coming year. In March 1993, the second Facility Workshop was held at which about 60 attendees were present. This year the programme was different from last year with five invited review papers and twelve short presentations as introductions to some of the twenty posters. The five reviews, which were all associated with the uses of ion beams in the fabrication of devices and integrated circuits, and in materials science, were presented by Professor M J Kelly (Surrey), Professor J Robertson (Edinburgh), Dr J Marsh (Glasgow), Professor A R Peaker (UMIST) and Dr P L F Hemment (Surrey). The format of invited speakers, short presentations and posters is likely to be maintained for next year's Workshop planned for March 1994. Thanks go to Chris Jeynes for his hard work in organising the event.
The group received a visit from an SERC panel in April to discuss the latest roll-on application to support the Central Facility. The panel was very pleased with our achievements over the past two years and have recommended continued support, although, due to the financial situation at the SERC, the details of the award will not be announced until after the Autumn, 1993 round of submissions. The roll-on date has, therefore, been delayed by six months, until the 1st February 1994.
In October 1992, Michael Kelly was appointed Professor based in the Department of Physics, but with the remit to work closely with our Department and with the Department of Materials Science and Engineering. It is expected that his very positive interaction with the SSD & IBT Research Group will
be extremely rewarding during the coming years. We congratulate Mike on his recent election to the Fellowship of the Royal Society.
Also during the past year: Bernard Weiss and Roger Webb were promoted to Reader and Senior Lecturer, respectively; and Peter Hemment was appointed Chairman of E-MRS Network 3: Ion Beam Processing of Semiconductors.
The group has continued to receive visitors from collaborating institutions around the world, including China, Russia, Germany, Spain, Yugoslavia, India. Most of these collaborations are supported by the British Council, the Royal Society and the European Community.
The following pages describe briefly much of the current research work of the group. Further information can be obtained from Professor Sealy.
The accelerators and associated characterization and processing facilities are available to users authorised by the SERC to have access to the Surrey Facility. A total of 2500 hours of beam time per annum is available to such users. The time used is mostly associated with work on semiconductors, (75%), the remainder being used to support studies of polymers, metals and ion beam analysis of non-semiconducting materials. In addition, we offer our facilities to industry, either as a service or via collaborative projects.
Our new HVEE 2 MV van de Graaff implanter, installed in the Spring of 1991, has serviced a variety of SERC customers as well as European collaborators. However, although we have made good progress, there have many difficult and unexpected problems to overcome before the machine will be available for turn key operation.
The 500 kV implanter has remained our workhorse and achieves an up time approaching 90%, whilst the 400 kV high current implanter is utilised about 50% of the available time being used primarily for ion beam synthesis of silicides. A 2 MV van de Graaff accelerator is available for ion beam analysis, with most work requiring Rutherford backscattering (RBS) and channelling analysis. However, also available is particle induced X-ray excitation analysis (PIXE) using a focused beam having a diameter down to about a 10 um. Light mass elements can be detected using nuclear reaction analysis (NRA) or elastic recoil detection analysis (ERDA).
An extensive range of processing and characterization equipment is also available, including rapid thermal annealing, film deposition, device fabrication, and optical and electrical assessment.
Potential users should discuss their requirements with the Senior Liaison Fellow, Dr C Jeynes.
The aim of the SIMOX work is to develop a generic, low cost integrated optics technology primarily aimed at sensor systems, although other applications are clearly possible. Thus far we have investigated planar and rib waveguide structures, and optical modulators. The research has demonstrated that guiding structures can be realised that have losses that are experimentally indistinguishable from those of pure silicon. Furthermore, optical phase modulators have been designed, with figures of merit which significantly exceed those of other technologies for some applications. The modulators are based upon refractive index changes due to injection of free carriers. It is vital to maximise the interaction between the propagating optical mode and the injected carriers in order to produce an efficient device. One example is a modulator and that has been designed to maximise this interaction in a relatively large waveguide (several um in cross-section), resulting in a device that can efficiently act as a phase modulator, but is also compatible with other optical devices with similar dimensions (see Figure 1.1).
The UV photosensitivity of proton implanted SiO2 doped with germanium which may be produced at selective locations within a sample has been determined. The presence of hydrogen creates absorption bands around 212 and 240 nm which are bleached by exposure to 249 nm excimer laser pulses. The removal of these absorption bands causes a change in the refractive index of the material at longer wavelengths and this change in refractive index is described by the Kramers Kronig transformation. The relationship between the implantation process, the post implantation annealing and the absorption spectra of the material have been determined. In addition, the waveguide loss of rib waveguides at 1.5 um has been found to be ~2 dB/cm. Experiments to characterise waveguides produced by this process are underway.
SiGe is of interest because its bandgap wavelength covers the useful optical communications wavelengths for optical fibre systems so that, by locally changing the composition by ion beam synthesis for example, devices for signal processing and detecting/generating optical signals could be fabricated on a single substrate. The optical properties of planar SiGe/Si heterostructures have been measured as a function of wavelength for Ge concentrations ranging from 1.3% to 33% with the lowest losses of ~1 dB/cm being obtained for 1.3% Ge at 1.523 um.
3 x 3 optical couplers have been investigated for applications in fibre optic gyroscopes. The gyroscope configuration that contains the 3 x 3 coupler is potentially a low cost configuration that relies on the characteristics of the coupler for suitable performance. In particular stability of amplitude and phase are crucial. A series of couplers have been experimentally evaluated, and compared with the literature, from which it has been established that excess loss of optical couplers is not a simple parameter, and as such is often misunderstood. An alternative matrix representation has therefore been developed for one family of couplers.
SCAN THIS PHOTO IN.
Figure 1.1. Proposed Modulator Geometry.
The samples used in this study had an approximately uniform atomic profile, achieved using multiple implants of silicon at energies of 100 keV, 330 keV and 900 keV. The dose was chosen so that the atomic concentration was about 1 x 10 [-18] cm [-3] over a depth of about one micron. This was done in order that carrier concentrations would not vary significantly with depth following annealing. Also, it was important to abtain carrier concentrations of the order of 10 [-18] cm [-3] so that the effect of the zero bias depletion region at the air-semiconductor interface could be ignored.
After implantation, samples were encapsulated with silicon nitride and rapid thermally annealed between 800 deg.C and 950 deg.C for times of 5 to 30 seconds. Electrical profiles were measured using the automated Hall apparatus. The electron concentration profiles have three shallow peaks corresponding to the three ion energies used. However, there is very poor agreement between the peak positions and the ion ranges calculated using LSS theory. In contrast, an extremely good fit between theory and experiment occurs when TRIM is used.Profiles of the total concentration of ionised impurities were obtained for each sample using published tables of electron concentration and mobility as a function of compensation ratio. These profiles are identical in shape to the atomic profiles calculated using TRIM, but the concentration is of the order of 50% higher than the TRIM profiles, assuming the doses implanted are accurate. In order to check this, atomic profiles will be measured using an alternative technique such as SIMS.
This work is being extended to include the use of the electrochemical C-V technique to produce reliable carrier concentration depth profiles. These, together with the theoretical atomic profile and tabulated values of carrier concentration and mobility as a function of compensation ratio may be used to estimate the mobility profile. Results will be compared with data from Hall effect and resistivity measurements which we feel are more reliable, and, of course produce measured values of mobility.
Using ion implantation to introduce known quantities of the lattice constituents Ga and As into InGaAs/GaAs quantum wells has helped to show that the diffusion is governed by thermally created vacancies. This was achieved by studying the time dependence of the diffusion following ion implantation, and the variations in this dependence with different implanted ions. In order to compare the interdiffusion on the group III and group V sublattices, quantum wells grown in the InGaAsP material system were supplied by both British Telecom and BNR Europe Ltd. Different samples were designed, each of which had only a concentration gradient on either the group III or group V sublattice, so that the diffusion in the two sublattices could be measured independently. Using these materials we have shown that the diffusion on both sublattices are controlled by vacancies on each sublattice, and that the activation energy for interdiffusion is the same for both sublattices. However, for the group V sublattice the prefactor for diffusion is two orders of magnitude greater than that measured for group III diffusion in the InGaAs/GaAs system. In addition we have found that for anneal temperatures below 700deg.C there is a low activation energy process controlling the diffusion on the group V sublattice which we believe to be due to grown-in vacancies. This low activation energy process is important as it is responsible for significant diffusion occurring on the group V sublattice during the growth of devices such as lasers. This diffusion during growth can result in changes in the operating wavelength of semiconductor lasers from their designed values.
Rutherford backscattering and channelling of 1.5 MeV helium ions has also been used to study the diffusion in GaAsSb/GaAs quantum wells. Measurements of the GaAsSb/GaAs material system suggest that the diffusion is not a simple Fickian process, but that the diffusion is strongly dependent on composition.
SCAN IN THIS PHOTO
Figure 1.2. Direct bandgap measurement in [[beta]]FeSi2.
The latter has involved the preparation of substrates at Surrey, the fabrication of MOS structures at Southampton and detailed evaluation by colleagues at Liverpool.
Surrey has a unique capability to prepare thin film SIMOX structures using low energy O + implantation (40 keV to 90 keV) with pre- and background wafer heating up to a temperature of 680deg.C. In order to evaluate these structures a comprehensive study of process induced defects in this material and in SOI structures formed by sacrificial thermal oxidation of silicon overlayer in thick film SIMOX wafers has been undertaken. To do so a new defect etchant based upon K2Cr2O7 and Cu(NO2)3 has been developed which enables extended defects to be delineated in silicon films as thin as 500 Å. By using both defect etching andplan view transmission electron microscopy it has been possible to identify small stacking fault tetrahedra at the Si/SiO2 interface and to follow the evolution of oxidation induced stacking faults during sacrificial oxidation. A model has been proposed to describe the evolution of these defects which is based upon the capture of silicon interstitials emitted during thermal oxidation by the dislocations associated with the vacancy type tetrahedra. As these extended defects (in sacrificially thinned SIMOX) will have an adverse effect upon processing yields it is argued that the direct formation of thin film structures by low energy O + implantation is to be preferred and, potentially, is a lower cost process due to the benefits of scaling the O + implantation equipment.
Isotope marker experiments involving the implantation and SIMS depth profiling of 18 O in SIMOX structures has enabled the mass transport and isotope exchange between the buried oxide layer and SiO2 cap during high temperature annealing to be quantified. For example it is found that in thin film (1000 Å) structures up to 40% of the oxygen can exchange between cap and buried oxide layer. This mass transport, which is driven by a process of diffusion controlled precipitate growth (Oswald ripening), can be so pronounced in thin film structures that all of the implanted oxygen is gettered by the capping layer thus putting tight constraints on the processing window.
High quality, low defect density Si/SiGe/Si structures have been prepared by high dose Ge + ion implantation followed by amorphisation of the complete structure by Si + self ion implantation and low temperature regrowth. As regrowth is initiated deep within the substrate, end of range defects (EOR) are located remote from the active volume. Also the alloy/Si interfaces are graded and cross-sectional TEM analysis has revealed no extended defects in the vicinity of these interfaces. This low thermal budget process is known by the acronym "EPIFAB" (epitaxial fabrication across phase boundaries).
Thermal oxidation of SiGe alloy is problematic due to different diffusivities and heats of formation of the oxides. Detailed studies of dry and wet oxidation of Si0.5Ge0.5 material has led to physical models which describe the mass transport of Ge during the movement of the alloy/oxide interface. Oxide layers with compositions from SiO2 to a mixture of (SiO2 + GeO2) can be formed depending upon the choice of processing conditions.
The electrical characteristics of heterojunction bipolar transistors and test structures fabricated at Southampton (Dr P Ashburn) in strained SiGe MBE and CVD substrates have been determined. The laterial non-uniformity of the transistor parameters has been correlated with process related geometric non-uniformities. Junction ideality factors have been determined both for e/b and b/c junctions from measured Gummel plots.
A new grant has been awarded by the SERC involving studies of dopant activation and control of strain in SiGe HBTs with the aim of enhancing the manufacturability of these devices. This is a collaborative project with GPS, Cheney Manor.
The program has been further modified to simulate the effects of a realistic surface topography. Input to the program, describing this topography, is available from the process simulator COMPOSITE and profile data can be returned to COMPOSITE for more accurate modelling. Full effects of surface sputtering and redeposition can be taken into account during the simulations.
Work currently is concerned with molecular impacts on graphite, diamond and silicon surfaces. Due to the weak inter-planar bonding of graphite the surface reacts in a completely different fashion to bombardment of silicon or diamond.