Table of contents

Photoluminescence (PL) scanning is used to assess a number of material- or process-related concerns in semiconductors. Since it is a non-destructive, non-contact technique, it can be used at any stage of the fabrication process. It was first used to examine dislocation distributions in GaAs and its value was well established when good correlations were found between PL contrast and device electrical properties. With the advent of whole ingot annealing it became possible to determine the 'quality' of the annealing step by the residual PL contrast across the wafer; this too correlated with electrical properties. The theory of PL output efficiency at room temperature was developed and this efficiency was shown to depend most on the surface recombination velocity. PL scans thus became valuable for looking at surface passivation effects and the effects of chemical treatment. The theory was extended to explain the phenomenon of PL decay as a function of atmosphere, surface treatment and passivation.

Two optical techniques, high-spatial-resolution scanning photoluminescence spectroscopy (sPL) and scanning reflectance spectroscopy (sRS), have been developed to provide rapid, non-destructive, whole wafer assessment of III-V materials grown by molecular beam epitaxy (MBE) and metal organic chemical vapour deposition (MOCVD). These techniques have been combined in a single measurement system which is used to evaluate epitaxial layer quality, composition (or emission wavelength) and thickness. The combination of low excitation power density and high data density in the sPL intensity maps produced by this measurement system has been shown to be important in revealing those defects that control the performance of minority carrier devices such as p-i-n detectors. High-resolution sPL wavelength maps may also be generated. These can be used to infer spatial variation in ternary composition and quantum well (QW) thickness. High-density thickness maps of non-QW epitaxial layers are produced by this instrument in the sRS mode. Thickness variations of less than 20 nm are clearly distinguishable. Examples from MOCVD and MBE growth experiments demonstrate the utility of sPL wavelength and sRS thickness maps in the optimisation of epitaxial reactor designs.

The authors have developed a whole wafer, cryogenic imaging system for making rapid absorption and photoluminescence maps of semiconductor wafers up to 100 mm in diameter. The cryostat requires no cryogens and can maintain a temperature of 15 K indefinitely. Its design facilitates the rapid (15 min) exchange of samples. Furthermore, the transmitted or emitted light from the sample is imaged onto a cooled Si CCD array rather than by using raster scanning. The net result is a system ideally suited for the routine characterisation of numerous wafers. This system has been used to map the EL2 concentration and donor-acceptor pair band photoluminescence intensity for a number of semi-insulating GaAs wafers.

The correlation between the distribution of minority carrier diffusion lengths monitored using surface photovoltage measurements and the distribution of photoluminescence intensity has been investigated in silicon and III-V semiconductor materials. Photovoltage measurements have been performed using a non-contact probe and constant photon flux method at room temperature, while the scanning photoluminescence measurements have been done in the room to liquid helium temperature range. To demonstrate the combined capabilities of the two methods, assessments of a liquid phase electroepitaxial In_xGa_1-xAs single crystal and 450 degrees C annealed silicon have been performed. The discussion emphasises the effects of non-uniform carbon distribution in InGaAs and oxygen-related defects in silicon.

Low-temperature photoluminescence topography has been used to monitor local variations in composition of In_1-yGa_yP and Al_xGa_1-xAs layers which were grown lattice-matched on semi-insulating LEC GaAs substrates by MOCVD. Whereas the in content varies by less than +or-1% over the whole 2 inch water, the Al concentration changes by about +or-3%. In Al_xGa_1-xAs/GaAs quantum well (QW) structures the dependence of photoluminescence energy on QW thickness is used to estimate spatial variations in the thickness of the well material. The resulting thickness topogram is compared with numerical simulations of the gas flow profile in the MOCVD reactor.

The scanning photoluminescence technique at room temperature was applied to characterize the quality of ion-implanted indium phosphide as well as to investigate the defect recovery as a function of the annealing temperature. Doses ranging from 10¹⁰ to 10¹⁴ ions cm^-2 were implanted with either neon or magnesium ions. The intensity of the photoluminescence response was found to be related to the ion dose. Two different substrates were used: an iron-doped semi-insulating one and an undoped one. The results were similar. Nevertheless, the sensitivity was higher for the undoped substrate. After thermal anneals ranging from 200 up to 800 degrees C the average PL intensity was again recorded and was found to increase with the progressive recovery of the defects. After an anneal at 750 degrees C the PL response was comparable to the signal obtained in the case of an annealed reference. The contribution of a doping impurity in a semi-insulating substrate was not clearly evidenced, the PL signal being dominated by the remaining non-radiative centres.

A novel technique, double-beam photoluminescence (DBPL), is introduced. This is used to investigate photoquenching effects of deep levels which affect near-band-edge photoluminescence (PL) in bulk-grown semi-insulating (SI) GaAs. The authors show that there is a remarkable increase in band-to-band radiative transition efficiency after photoquenching of near-band-edge absorption (also known as Reverse Contrast or RC) at low temperatures (less than 35 K) and suggest that RC absorption does indeed map concentrations of the dominant associate this effect with a reduction in non-radiative recombination centres. They suggest that RC absorption does indeed map concentrations of the dominant recombination centre in SI GaAs and that the observed increase in luminescence after low-temperature illumination with light of near 1 mu m wavelength is due to photoquenching of the RC defects.

It is well known that processing-induced defects in semiconductors can adversely affect device performance. This is particularly important for silicon, where oxidation processes can give rise to stacking faults along with other extended defects. A report is presented here of the relationship between the 'D' line dislocation-related photoluminescence (PL) lines, stacking fault length and the power of the laser employed as the excitation source. The authors have measured by PL a series of samples with different stacking fault lengths. Irrespective of stacking fault length they always observe lines D1 and D2 but no other higher-energy D lines; the intensity of the D lines increases with increased stacking fault length. The authors also find that the relative intensity of D2 to D1 depends upon the laser power employed in the PL experiment. The effect is observable in all their stacking-faulted samples and suggests that D1 and D2 are luminescence lines that originate from a related, if not the same, radiative process.

Three independent but complementary methods (DSL photoetching combined with etch rate profiling, spatially resolved PL and LST) were employed to study the distribution of microdefects and electrically active centres in commercially available SI undoped, LEC-grown GaAs after different ingot-annealing treatments. A one-to-one correlation was obtained on comparing the microscale distribution of decoration precipitates (DPs), matrix precipitates (MPs) and microdefects (MMs) by the DSL and LST methods. Clustering of MPS is also revealed by PL intensity profiling and photoetch rate measurements. Two types of matrix microdefects are distinguishable by DSL photoetching in samples after two-stage annealing. A high degree of homogeneity across the dislocation cells (after low-temperature and multiple annealing) was clearly revealed by a drastic decrease in the uniformity parameter (U) after DSL photoetching and by a significant decrease in the PL intensity measured at the cell walls with respect to the cell interiors. Using the present results and recent data from the literature, an explanation of phase transitions during ingot annealing is proposed. This is based on the assumption that the final properties and structures after annealing are the result of two competitive processes: (i) formation of decoration and matrix microdefects; (ii) generation of electrically active point defects (clusters), presumably EL2.

Spatially resolved photoluminescence (SRPL) is a useful tool for evaluating the quality of semiconductors used within electronic and photonic devices. SRPL can be done in two ways; scanning with a focused laser beam or imaging with a Vidicon. Laser-scanned SRPL can provide both high spatial resolution and sensitivity in semi-insulating gallium arsenide, GaAs:Si, and iron-doped indium phosphide, InP:Fe. Imaging SRPL can provide rapid analysis at video rates. Video rate laser scanning with a 1 mu m laser spot combines the advantages of both approaches. The contrast of SRPL scans in GaAs:Si wafers is improved by a plasma passivation treatment. Scans clearly reveal dislocated cell structures as bright bands, arsenic precipitates as dark dots and polishing damage such as scratches. Several techniques, including XPS, SIMS, TEM and DSL etching, have been used to analyse optical contrast and the passivation mechanism. SRPL scans must be done non-destructively to allow for subsequent device fabrication and yield analysis. Photodegradation is minimised if a pumping wavelength near 720 nm is selected. This observation is apparently due to a tradeoff between surface recombination and photon recycling. The small dwell time during video scans assures a non-destructive measurement, leading to improved contrast relative to systems with longer dwell times. SRPL scans in InP:Fe reveal bright spot defects, without any need of passivation. Imaging SIMS, TEM and DSL etching have been used to analyse contrast in these scans. The bright spots are dislocations and iron phosphide precipitates that have gettered significant levels of silicon contamination.

The authors have investigated spatially resolved photoluminescence spectra around several alpha - and lambda -type oval defects in GaAs/AlGaAs multiple-quantum-well structures. In their low-temperature photoluminescence set-up they combine a high spectral resolution with a 2 mu m spot size on the sample. In the alpha -type defects the authors observe a 25-45 meV low-energy tail of the main quantum well recombination peak near the core of the oval defect as well as an approximately 1 meV line broadening originating from two regions which are symmetrically located at either side of the (110) symmetry axis of the defect. These line-broadening effects are explained by vertical growth rate variations within the oval defect.

The effect of misfit dislocations on the electro-optical behaviour of strained III-V heterostructures has been studied by high-resolution spatially resolved photoluminescence (PL) imaging. Both epitaxial layers under tensile stress (in this study GaAs on InGaAs) as well as layers under compressive stress (InGaAs on GaAs) have been examined. Dislocation lines in both (011) directions are observed in both systems. In contrast to the compressive system, where no as-grown morphological asymmetry is observed, the dislocations in the tensile system have been preferably formed in the (011) direction. This is explained by assuming (i) that the relaxation of the strained layer is initiated by the most mobile dislocation in GaAs, which is the arsenic glide As(g) dislocation, and (ii) that cross-slip of these As(g) dislocations is only possible in the compressive system. PL imaging reveals the presence of many non-radiative deep levels around the dislocation lines and also shows an opto-electronic asymmetry of the two (011) directions for both systems, in contrast to the as-grown morphology. For the compressive system this is explained by the difference in depletion between the dislocation lines along the (011) direction, of which the core is believed to consist of Ga atoms, and those along the (011) direction, formed with As atoms. For the tensile system just the opposite holds. DSL (defect-revealing) etching confirms the results obtained with PL imaging.

Spatially resolved room temperature photoluminescence is a non-destructive optical characterisation technique which allows the measurement of relevant epitaxial layer properties over the whole wafer. In this work the authors describe application of spatially resolved and spectrally resolved photoluminescence mapping to determine the spatial variation of the composition of ternary epitaxial layers growth by MOCVD and MBE. They show results from device-quality epitaxial layers which are used for the manufacturing of optoelectronic devices. In particular they concentrate on InGaAs layers deposited on InP substrates and AlGaAs layer deposited on GaAs substrates. The authors discuss some of the suspected mechanisms for these variations and the use of the technique as a diagnostic aid for MOCVD and MBE growth.

Room temperature scanning photoluminescence (SPL) measurements were performed on n-type (Sn-doped, S-doped and undoped) InP crystals provided by various suppliers. It was found that defects and non-uniformities such as doping striations, dislocations and subsurface defects are associated with specific 'signatures' in SPL images, regardless of the type of doping and the origin of the samples. SPL measurements appear to be useful for routine inspection of this material because they are fast and non-destructive.

Makyoh, the magic mirror method, has proven to be one of only a few production-worthy surface characterisation tools. This technique, very simple and non-destructive, transforms latent damage, scratches, waviness and other flaws on mirror-like surfaces into visual images. It has recently been used to characterise highly finished, mirror-polished, large-diameter Si wafers for ultralarge-scale integration (ULSI) applications and III-V compound semiconductor wafers, replacing the laborious, unstable, naked-eye wafer inspection line. The authors studies have shown that the technique is also very useful for optimising various wafering processes and monitoring various process steps required to manufacture advanced substrate structures such as silicon-on-insulator, silicon-germanium alloy-on-silicon, etc for future devices. It has been successfully applied to characterise various Si wafers used in ULSI applications with the highest finish. This paper describes the recent development of the method, reviewing some of the results obtained in the evaluation of mirror-polished Si and compound semiconductor wafers and of complex substrate structures for ULSI device fabrication processes.

In this study, 150 mm diameter p-type (100) silicon wafers (with no screen oxide) were implanted in a batch implanter with 5*10¹⁵ cm^-2 of arsenic at 80 keV. After implantation the wafers were subjected to rapid thermal annealing for 10 s at temperatures: wafer warpage measurements, optical imaging inspection (magic mirror method), X-ray transmission topography and thermal wave modulated optical reflectance. This paper summarises the measured results of wafer defects and damage due to the rapid thermal processing.

Makyoh (magic mirror) technology has been applied to the observation of slip lines and wafer warpage of Si wafers. Although Makyoh topography gives a bright field image and is not suitable for observing microflaws of micron-sized width and length, it has been shown that the method is capable of revealing slip lines as sensitively as X-ray topography. This is due to the ability of Makyoh topography to observe the slight waviness associated with microflaws in a defocused background image of macroflaws. This contradictory micro/macro observation capability was found to be due to different focal lengths of micro- and macroflaws (when the former is sharp, the latter is vague, and vice versa) and to the formation of composite images with neighbouring bright and dark areas which emphasise the image features vividly and impressively. Makyoh topography has also proven to be the most effective method to determine the optimum conditions for growing epitaxial wafers without slip lines and to check even a very slight drift from the optimum.

The technique based on the concept of the ancient Chinese holy brass mirror, the magic mirror or 'Makyoh', is utilised to inspect and characterise the surfaces of polished GaAs wafers. This non-destructive method was first introduced to examine the polished surfaces of silicon wafers and epitaxial layers. Undulations and latent damage on the mirror-like polished surfaces are easily visualised over the entire substrate in real time. Thus study extends the magic mirror method to inspect III-V compound crystals, Some polished GaAs, GaP and InP wafers are examined and discussed. Saw marks, dimples, polish scratches, texture defects and other damage, e.g. twin lines, facet boundaries and striations, hidden in the polished surface are observed over the entire wafer. Comparison between etched wafers and unetched wafers observed by the magic mirror and revealing similar deformations is discussed.

The authors studied defect states introduced in the Zn-doped AlGaAs active layer of high-brightness red-light-emitting diodes with double heterostructure (DH-LEDS) fabricated by liquid phase epitaxy (LPE) with a sliding boat and determined the correlation between LED brightness and deep levels due to defects in the active layer. Deep levels comprising a majority carrier (hole) trap and two minority carrier (electron) traps have been observed by capacitance transient spectroscopies (DLTS and ICTS). The correlation that DH-LEDS with a lower density of deep levels exhibit a higher radiative efficiency has been revealed by measurements of the brightness and by scanning ICTS (SICTS). A unique pattern of high- or low-intensity regions perpendicular to the sliding direction has been observed in both the spatial distribution of the brightness and the density of deep levels. Furthermore, by using DH-LEDS fabricated by LPE without the sliding boat, it has been revealed that the introduction of deep levels depends on the crystal growth process. These experimental results suggest that the unique distribution pattern observed in the brightness and the density of deep levels is introduced by fluctuations of the interface at the p-n junction caused by melt carry-over in the sliding boat method.

A new approach to reduce ion implantation damage at low temperature is demonstrated for the first time. Damage in Be-ion-implanted GaAs p⁺-n diodes and its reduction by minority carrier injection have been evaluated. Ion implantation increases the recombination current in the diodes, but the current can be reduced by minority carrier injection even at room temperature. The reduction is due to minority carriers themselves, not to the thermal annealing. Traps which act as recombination centres and decrease in concentration by injection are detected in the diodes, indicating that the reduction in recombination current is due to the recombination-enhanced defect reaction. The traps are different from those observed in electron-irradiated GaAs. The rate of recombination current reduction is also discussed. Minority carrier injection holds considerable promise as a low-temperature annealing method to reduce damage induced by ion implantation.

The development of the scanning infrared microscope (SIRM) at Oxford during the last few years and its application to the examination of inhomogeneities in a variety of bulk semiconductor specimens are described. The present SIRM uses a 1.3 mu m wavelength semiconductor laser and operates in bright field in either the standard, confocal or polarized light modes and in dark-field in the standard mode. It can reveal individual precipitate particles and dislocations and also point defect concentrations and local strain fields. It has a spatial resolution of 1-2 mu m, a depth of field of 7-50 mu m and can image particles down to 30 nm in size. Three-dimensional distributions of inhomogeneities can be determined. The method is non-destructive and non-contacting. Examples of its application to bulk In-doped GaAs, undoped GaAs and Si, both as-grown and post-annealed, are given. Quantitative data obtained include precipitate particle number density, particle denuded zone depth, dislocation geometry, dislocation particle decoration and dislocation Burgers vector.

Extended defects in melt-grown GaS and GaSe single crystals have been characterised by convergent beam electron diffraction techniques. The Burgers vector of dislocations on GaS and GaSe single crystals have been determined by analysis of the modifications induced both in Kikuchi lines and in the first-order Laue zone reflections observed in low-camera-length convergent beam electron diffraction patterns. The observations have shown that the dislocations in GaS are perfect (edge, screw and mixed). In contrast, in GaSe the dislocations are partial or edge. The displacement vector of stacking faults has been determined by the modifications in Kikuchi lines observed in large-angle convergent beam electron diffraction patterns recorded from faulted regions according to the Tanaka technique. The observations have not indicated the presence of stacking faults in GaS crystals, whereas stacking faults have been observed in GaSe crystals.

Metal organic molecular beam epitaxy (MOMBE) In_xGa_1-xP/GaAs (0.45<x<0.55) interfaces and epilayers have been investigated using SEM, EDX and cross-sectional TEM microscopy techniques. No oval defects were observed by SEM, while occasional surface roughness was associated with phase decomposition of the corresponding epilayer. The In_xGa_1-xP/GaAs interfaces were perfect, with no dislocations or planar defects. Only very few clusters attributed to imperfection of the GaAs substrate were detected and their density was estimated to be less than 10⁶ cm^-1. In the In_xGa_1-xP overlayer with x=0.45 ( Delta a/a=3.6*10^-3) TEM has shown a structure completely free of linear or planar defects and with only a fine-scale contrast of 100 AA periodicity perpendicular to the interface along the (001) crystallographic direction. This is attributed to statistical fluctuations in the distribution of atoms with different covalent radii. For x=0.55 ( Delta a/a=3.7*10^-3) the In_xGa_1-xP overlayer exhibits spinodal decomposition along the (001) direction, while for x=0.49 a columnar structure in the first 700 AA followed by a mottled contrast resembling that of polycrystalline films was detected.

The methods of light-induced absorption and diffraction have been used to investigate the uniformity of Si and GaAs wafers and ion-implanted Si. Metastable annealing of defects, lifetime modification by swirl and radiation defects and anomalous depth penetration of ion-implantation-created defects have been demonstrated in Si. The mapping of GaAs wafers has been performed with picosecond and nanosecond light pulses and light absorption mechanisms are discussed.

A method for revealing point defects in silicon single crystals has been investigated. D or A defects could be revealed by a preferential etching technique using Secco's etchant. Wedge-shaped flow patterns and etch pits were recognised in the D and A regions, respectively. The flow patterns were so characteristic that the D region could be distinguished very easily.

(111)- and (001)-oriented InP bulk single crystals double-doped with Cd and S were investigated by scanning electron microscopy in the cathodoluminescence mode, X-ray topography and transmission electron microscopy. Both cathodoluminescence micrographs and X-ray topographs showed that the crystals contain a low concentration of dislocations; however, in the less dislocated regions precipitate-like defects have been revealed. Transmission electron micrographs confirmed this result, showing the presence of defects exhibiting moire fringes; analysis of the fringe spacing and diffraction patterns evidenced the presence of crystalline CdS particles as responsible for the precipitate-like contrast.

The laser-scanning tomography technique has been used to explore the immediate vicinity of implanted MESFET structures in SI GaAs. From high-resolution tomography images it has been observed that apart from the well known microprecipitates which decorate the dislocations there are also process-induced particles; some are related to the n⁺ contacts, others are located in the bulk immediately below the grooves between the gate and source (or drain). It is shown that limiting specifications (breakdown and side-gating voltages) are correlated and both are controlled by the bulk microprecipitates, whereas the threshold voltage deviation could be attributed to the process-induced defects in the internal space between the gate and source or the gate and drain.

The possibilities of remote contact EBIC (REBIC) for semiconductor characterisation are discussed. Results of investigations on CdTe, PbTe and HgCdTe crystals are presented. It is shown that this technique is very suitable for in situ investigations of electron beam effects on semiconductor properties.

In scanning electron microscopy the backscattered electron yield has been shown to depend integrally on the specimen surface roughness. This means that if the surface contains many small inhomogeneities with a typical size of a few nanometres or less, then although the signal line remains smooth (without local perturbations), its level increases compared with that of an ideally flat surface. The authors study this effect by using a model of signal formation developed previously and a Monte Carlo simulation. The results obtained are compared with experimental data and are used for surface flatness mapping.

The accurate measurement of lattice strain in heteroepitaxial semiconductor layers is fundamental to the development of electronic devices, since it is often the structural properties of a material which limit its range of application in this type of structure. High-resolution X-ray diffraction techniques provide some of the assessment procedures for measuring the state of strain in these often complex structures; their use over recent years is briefly reviewed. The advantages and limitations of X-ray diffractometry and topography for measuring the onset and progress of lattice relaxation are discussed and demonstrated with examples from III-V material systems. The application of these techniques to mapping the homogeneity and defect structure in highly mismatched II-VI epitaxial systems is also described.

Thermal simulation of CMOS processing has been done. It was found that LPCVD Si₃N₄ layer deposition and subsequent thick oxide growth generate process-induced defects. These defects are demonstrated by double-crystal X-ray reflection topography.

A theoretical model of X-ray spectrum formation in the triple-crystal diffraction technique for the description of interference phenomena of X-ray scattering in crystals is presented. The proposed numerical algorithm of the solution of the corresponding inverse problems enables one to obtain a detailed two-dimensional strain distribution in the specimen crystal from the data of triple-crystal diffractometry.

The possibilities of depletion region width (W) and minority carrier diffusion length mapping using the electron-beam-induced capacitance (EBIcap) mode and measurements of d/_c/dW are discussed. A model of EBIcap signal formation is proposed. It is shown that these techniques are more suitable than electron-beam-induced current (EBIC) for local measurements and mapping of W.

A new non-contact measurement method using a millimetre wave is proposed to observe the photoconductivity decay of an epitaxial silicon layer grown on a substrate with very low resistivity. The photoconductivity decay of the epitaxial layer was measured for p/p⁺ and n/n⁺ samples. The apparent carrier lifetime (or the effective lifetime) was determined directly from the gradient of the decay. The surface recombination velocity at the epilayer surface was decreased by using an ion exposure technique. Pairs of epilayer lifetime and effective recombination velocity at the interface between the epilayer and the substrate were determined by curve-fitting the measured photoconductivity decay to a theoretical one calculated by taking the skin effect of the electromagnetic wave into consideration.

Over the past decade, secondary ion mass spectrometry (SIMS) has developed into a powerful technique for studying the surface chemistry of solids. The technique has been successfully employed in studying a wide range of materials, including those of particular interest in the electronics industry. The versatility of the SIMS technique in its three major modes (static, dynamic and scanning) is demonstrated in examples of the surface characterisation of semiconductor materials.

The authors present the principles and application examples of a recently refined, computerised, surface photovoltage (SPV) method. This new method is capable of wafer-scale, non-contact mapping of metal contaminants in the bulk and on the surface with sensitivities as high at 10¹⁰ atoms cm^-3. They demonstrate the unique ability of SPV to measure product wafers with finished integrated circuits.

The observation of oxidation-induced stacking faults (OISFs) in SIMOX structures with an optical microscope is achieved with the aid of a new method consisting of chemical etching and HF decoration. Comparison of SF lengths obtained by this method with transmission electron microscope measurements shows very good agreement. As an illustration the authors compare the SF lengths obtained in SIMOX structures with those obtained in bulk silicon under the same oxidation conditions.

The authors have developed a novel scanning deep level transient spectroscopy system based on laser-excited junction capacitance or current transient measurements under isothermal conditions (scanning isothermal transient spectroscopy or SICTS). By applying the system to the measurement of deep levels in GaAs, they have successfully obtained precise information on the spatial distribution of EL2 in n-type GaAs wafers.

The extrinsic photoconductivity response of semi-insulating GaAs at low temperature is characterised by two effects, photoquenching (PQ) and enhanced photocurrent (EPC), which present metastable properties. An experimental arrangement for photocurrent measurement under a microscope with 5 mu m spatial resolution is described. PQ and EPC are found to be spatially anticorrelated: PQ is more intense at the walls of cellular dislocation networks, while EPC is more clearly seen inside the cells. This anticorrelation rules out a relation between EL2 and the acceptors emitting the EPC holes. A relation between EPC and reverse contrast (RC) is suggested.

Knowledge of the excited area is the basic requirement for reliable quantitative interpretation of SDLTS experiments. The dependence of the spatial resolution of SDLTS experiments on experimental parameters is realistically calculated for the case of minority carrier trapping within a Schottky diode. The resolution is shown to continuously degrade with increasing excitation intensity but can be optimised using a 'pulse width scan' procedure. Such a procedure and other special investigations can easily be performed by using a specialised SDLTS unit now available.

The authors present results of non-contact, wafer-scale measurement of the deep level thermal emission realised in GaAs on a native surface barrier using surface photovoltage (SPV) transients. The principle of corresponding optical deep level transient spectroscopy (DLTS) is discussed and the approach is applied to non-contact wafer mapping.

It is well known that there are many arsenic precipitates in LEC GaAs, the dimensions of which are 500-2000 AA. The authors have recently found that these arsenic precipitates affect the device properties of chloride epitaxial-type MESFETS. They also affect the formation of small surface oval defects on MBE layers. To reduce the density of these arsenic precipitates, a multiple-wafer-annealing (MWA) technology has been developed in which wafers are annealed first at 1100 degrees C and then at 950 degrees C. By this annealing, highly uniform substrates with low arsenic precipitate densities, uniform PL and CL, uniform microscopic resistivity distributions and uniform surface morphology after AB etching can be obtained. These MWA wafers showed low threshold voltage variations for condensed ion-implantation-type MESFETS. In the present paper recent works are reviewed and the mechanism of arsenic precipitation is discussed from the viewpoint of stoichiometry.

Gas analysis was used to monitor the atmosphere of a high-pressure crystal during the growth of semi-insulating gallium arsenide by the liquid-encapsulated Czochralski process. Apart from the background argon, the only gases detected were carbon monoxide and hydrogen. These were formed by reaction between the graphite components of the puller and residual water vapour. The quantities of CO and H₂ increase with atmospheric humidity and this results in an increase in the electrical resistivity of the crystal. Diffusion through the boric oxide encapsulant appears to act as a rate-limiting step in the carbon transport. The CO concentration was increased by injection and decreased by gettering or by dilution, and by these means it was found possible to grow crystals with very uniform electrical properties and tightly controlled resistivities at selected values within the range 10⁷-2*10⁸ Omega cm.

The influence of dislocation density on the I-V characteristics of InP diodes under reverse bias has been investigated. It is observed that the leakage current increases with increasing dislocation density whereas the breakdown voltage is reduced. Further, precipitates are observed on dislocations in the p-regions formed by the in-diffusion of Cd into n-type InP layers. It is argued that the increase in leakage current is due to the presence of space charge cylinders around dislocations threading the p-n junction. On the other hand, precipitates may be responsible for premature avalanche breakdown. In addition, for the range in energy scanned, levels within the bandgap due to dislocations are not observed by photoluminescence measurements.

Positron lifetime results in semi-insulating and p-type materials after various heat treatments are presented. In as-grown and subsequently annealed (at 950 degrees C) materials a long-lived component of approximately 300 ps of substantial intensity is observed, signifying the possible presence of small vacancy agglomerates, possibly divacancies. In specimens quenched from 1100 degrees C this long-lived component increases to approximately 330 ps, then decreases to approximately 285 ps upon subsequent annealing of the samples at 950 degrees C. Evidently the quenched-in defects undergo some kind of rearrangement during the annealing. It is likely that this rearrangement of defects is in some way responsible for the homogeneous redistribution of EL2 that occurs under such heat treatment conditions.

Phase-stepping microscopy (PSM) is a useful optical metrology technique for the qualification of electronic and optoelectronic devices. Based on interference microscopy, the method uses a CCD camera and a digital image-processing system for automatic fringe analysis, leading to the calculation of the three-dimensional relief of a surface. A vertical resolution of 1 nm is achieved, with a horizontal resolution of 0.6 mu m. The authors describe the PSM system, the performance of the instrument and some of its limitations. To illustrate the uses of PSM, some applications are given in the analysis of defects in semiconductor materials. Results are given of a microprecipitate-related hillock defect on a quaternary layer of InGaAsP on InP made by MOCVD used in the fabrication of lasers. Surface analysis results of an optical guide etched on a quaternary layer are also described. Finally, some initial work is presented in the study of the cross-sectional geometry of the gate recess in FET devices.

'Preferential polishing' is a special Si wafer thinning technology, and by applying preferential polishing to device-fabricated Si wafer thinning, thin film Si devices such as thin film MOSFETs are easily fabricated. The authors have studied the mechanical properties of preferential polishing, especially the mechanism of 'overpolishing', by which the Si back-surface of the thin film device becomes concave. It was found that thin film PMOSFETs with a low overpolishing value have exactly the same electrical properties as bulk ones before thinning, and a fresh polishing pad without long nap fibres is desirable to avoid overpolishing.

The growth by molecular beam epitaxy (MBE) of GaAs and AlGaAs, although a well controlled process, has nevertheless suffered from poor reproducibility and uniformity across large substrates (greater than 2 inch) as would be required in practice for the fabrication of devices, the MESFET being a point in case. Recent advances in the design of MBE deposition chambers coupled with the availability of high-purity starting source materials have enabled the growth of unprecedentedly high-quality, very uniform GaAs and AlGaAs with electronic defect densities below 10¹⁴ cm^-3, even for the highly reactive AlAs binary, and with a compositional uniformity for the ternary alloy (AlGaAs) of better than +or-0.5%, corresponding to temperature variations of less than 1 degrees C at 700 degrees C. The design of a new Ga hot-lip cell in the VG V90H MBE system has enabled total structural defects of less than 400 cm^-2 to be routinely achieved on 2 inch substrates. The ability to deposit well characterised single-crystal Al films on GaAs and AlGaAs has also led to a new dimension in the reproducibility and control of near-ideal Schottky diodes for MESFET and HEMT applications. The advantages of such structures include uniformities of better than +or-5 meV in the barrier height across 2 inch substrates and thermal stabilities up to 500 degrees C. The practical significance of these results will be emphasised for the GaAs-AlGaAs system.

The results of a defectoscopic study of large-scale impurity accumulations in InP and GaAs single crystals are expounded. This has been carried out by means of the low-angle mid-IR-light-scattering technique ( lambda ₀=10.6 mu m). Estimates of the main parameters of the accumulations (size and bulk free carrier concentration) and the activation energies of the point centres constituting them are given. A model of the inner structure of the accumulations is presented, according to which they are spherical regions enriched with In or Ga microinclusions and In_P or Ga_As antisites in InP and GaAs respectively.

Presents a study of the electrical properties of deep states in silicon containing oxidation-induced stacking faults and dislocations. In general the electrical activity of stacking faults is associated with the Frank partial dislocation bounding it rather than the extra plane of the fault itself. In addition, the deep states associated with the partials of very clean stacking faults are rather ineffective as either generation or recombination centres. The authors have studied the effect of annealing oxidation-induced stacking faults in nitrogen and of decorating them with silver. After low-level decoration the defect states increase in concentration and the electron binding energy changes, tending to move the states towards the middle of the gap (hence increasing their effectiveness as generation centres). This level of decoration is not detectable with present-day TEM techniques. However, if the level of decoration is increased further, precipitates can be observed (using TEM) but the concentration of deep states decreases as does their electron binding energy. During this precipitative phase the intensity of the photoluminescence D lines is also seen to decrease substantially. These factors have a very considerable significance in relation to the imaging of extended defects using techniques which rely on recombination-generation effects and imply that the visibility of such defects is crucially dependent on the level of decoration.

The authors have carried out a uniformity investigation of undoped, semi-insulating GaAs grown by the vertical Bridgman method. They have employed A/B etching, EL2 mapping and near-bandgap absorption (reverse contrast, RC) mapping to reveal dislocations and non-uniformities in concentrations of two specific point defects: neutral EL2 centres and another defect whose identity is unknown but can be revealed by optical absorption imaging when the sample is cooled below 150 K and viewed with light of energy within 50 meV of the bandgap energy. The RC images were obtained with the sample at about 80 K mounted in a cryostat, using light of 830 nm wavelength obtained with a narrow-bandpass interference filter. The EL2 mapping was performed with the sample at room temperature with light of 900 nm wavelength. In both cases a sample thickness of about 500 mu m was employed. The authors demonstrate that although the dislocation density is over an order of magnitude less than that of conventional LEC GaAs (about 10³ cm^-2), most dislocations are formed into polygonised structures with cells of the order of 1-2 mm in diameter. In addition, concentrations of both EL2 and the defects that give rise to the RC image are very non-uniform in the vicinity of the cell walls. In contrast to conventional LEC GaAs the EL2 absorption near dislocations is much higher than the RC absorption.

GaAs wafers processed through various stages of active channel formation are characterised by photoinduced microwave reflectometry using 2D maps. A parasitic waveguiding phenomenon is corrected by post-measurement signal processing. Results similar to those seen in previous experiments are repeated, while the added sensitivity of an improved system reveals implant damage, annealing effects and EL2 loss associated with activation. Point defect interactions during the quench of the activation anneal dominate the observed changes in the measurement.

Distributions of phonons in ZnS_xSe_1-x, ZnTe_xSe_1-x mixed crystals were observed by Raman-scattering tomography as stripes which were caused by the lack of some elements of the Raman tensor due to rotational twins around the (111) axis. The technique was also applied to the analysis of phonons on the surfaces of Si-doped GaAs crystals, Si-implanted GaAs crystals and epitaxial layers of ZnSe on GaAs crystals.

Laser-scanning tomography (LST) was proposed by Moriya and Ogawa in 1983 as a non-destructive and non-invasive method for imaging small grown-in precipitates in the bulk of semiconductor materials. It has been shown that particles as small as 1 nm can be revealed by this method. The geometrical location of these defects has been obtained with 1 mu m precision in the (x,y) plane and with 0.1 mu m precision in the z direction. Extrapolations have also been performed in order to analyse implanted layers, epilayers and completed structures. Gallium arsenide has been the material studied most extensively by the method of LST, although InP and silicon have also been profitably investigated. Microprecipitates (MPS) contribute to an internal 3D structure which is typical of the material and also of its previous thermal history. It has been shown that annealing is able to render the electrical behaviour of SI GaAs ingots uniform and also to induce nucleation and growth of silica gettering centres in silicon wafers which are initially doped with oxygen. These MP structures are observed and quantitatively evaluated in such materials in a way which is far more convenient than traditional destructive etching. In this survey the authors recall the basis of the LST method and its metrological applications. Recent results in materials knowledge will be developed, especially in the field of GaAs, InP and silicon wafer technology.

The Raman microprobe is used for characterising chemically revealed defects in as-grown GaAs. Different microstructures were scanned with 1 mu m spatial resolution. Emphasis is paid to the breaking down of the Raman-backscattering selection rules. Precipitates and twins are studied in semi-insulating GaAs. Coupled phonon-plasmon modes are analysed in Te-doped GaAs (N_D-N_A approximately=10¹⁷ cm^-3); dislocation grain boundaries and hillocks are scanned by the Raman microprobe, revealing misorientation and free carrier effects.

Three high-spatial-resolution methods, namely DSL (dilute Sirtl with light) photoetching, electron-beam-induced current (EBIC) and photoluminescence (PL), were used to study GS defects (grown-in dislocations which have moved from their original position leaving behind a Cottrell atmosphere and a trace) in Si-doped GaAs crystals grown from Ga-rich melts by the LEC method. A peculiar behaviour of these defects during DSL photoetching was revealed by interference contrast microscopy and surface profiling of the local etch rates: the start and end points of the GS defects were etched more quickly and the trace more slowly than the defect-free matrix. EBIC measurements showed a decrease in the net ionised impurity concentration at the Cottrell atmospheres but were ineffective in determining any compositional changes at the traces of the GS dislocations. PL mapping showed all components of the defects as well as remarkable spectral shifts at their different parts. On the basis of these results and previous data a model of the point defect equilibrium within the Cottrell atmospheres is proposed. The possible mechanisms of movement of the dislocations are considered, taking into account three different processes: climb, glide and climb/glide. It is argued that antisite As_Ga defects (or related complexes) are formed during movement of the dislocations independently of the deviation from stoichiometry.

Dislocation loops in Si-doped GaAs crystals have been studied by conventional transmission electron microscopy (CTEM) and high-resolution transmission electron microscopy (HRTEM). In addition to the prismatic dislocation loops lying on (110) planes and the faulted loops lying on (111) planes, the existence of prismatic dislocation loops lying on (111) planes with ¹/₂ (110) Burgers vectors and of interstitial character was identified for the first time. Some of the dislocation loops exhibited partly faulted and partly prismatic characteristics. This observation indicates that the prismatic dislocation loops lying on (111) planes may be converted from faulted loops by the generation and movement of two ¹/₆ (112) Shockley partial dislocations across the stacking fault. Because all the dislocation loops studied by HRTEM have an extra plane, it was concluded that interstitials are responsible for the formation of dislocation loops.

Growth striations in CZ and MCZ silicon wafers were studied by the striation-etching method, spreading resistance measurements and the micro-FTIR mapping method. Microvariations in oxygen concentration were successfully measured using the micro-FTIR mapping system and the results were in good agreement with other results obtained by etching and resistance measurements. Two kinds of striations were observed in CZ and MCZ wafers with high oxygen concentration: weak striations at regular intervals and strong striations at irregular intervals. It was concluded that the former were related to the microsegregation of dopant impurities caused by remelting of the crystal during growth, while the latter were related to the fluctuation of oxygen concentration in the silicon melt.