Table of contents

The metal-oxide-semiconductor transistor has become the dominant device for very large-scale integrated circuits. The performance and reliability of an MOS device are heavily influenced by the quality and properties of the interface between the oxide and the Si region directly beneath. Inherent, process-related, and operationally and environmentally generated interface states or traps are exceedingly harmful or disabling when present. Although controlled successfully by semiempirical design, fabrication and operational regimens, ever-smaller device size-approaching the scale in which a single trap can be important-makes further extension of knowledge on the basic physical and chemical aspects of interface states essential for confident technological progress. In addition, Si/SiO₂ interface states continue to merit research in their own right. This paper briefly highlights selected topics from past and present interface-state research and describes some essential physical and chemical properties.

This paper contains a review of electron paramagnetic resonance studies on P_b-type centres at the Si-SiO₂ interface. It includes a synopsis of studies on the atomic structure of P_b defects and the chemical kinetics whereby P_b centres are passivated with H₂.

Si-SiO₂ interface traps are typically annealed at a low temperature (300-450 degrees C) in the presence of hydrogen. Despite the widespread acceptance and importance of this technique, the rates and mechanisms of trap annealing are not well understood. Several recent papers have presented kinetic descriptions of Si-SiO₂ interface trap annealing. The author investigates the application of these kinetic results to models of Si-SiO₂ interface chemical reactions. Directions for further inquiries are described.

The authors focus on radiation-induced interface traps, describing first how they fit into the overall radiation response of metal-oxide-semiconductor structures. Detailed measurements of the time, field and temperature dependences of the build-up of radiation-induced interface traps indicate three processes by which the build-up occurs. The largest of these is the slow two-stage process described by McLean and co-workers (1989) which is rate-limited by the hopping transport of hydrogen ions. Two other faster processes also contribute small interface trap build-ups in gate oxides. The processes seem to be controlled by hole transport to the Si/SiO₂ interface and by neutral hydrogen diffusion respectively. They also discuss several models which fall into three classes, corresponding roughly to the three processes observed experimentally. Other topics discussed briefly are dose dependence, field oxide effects, chemical and processing dependences and scaling effects.

The low-temperature (2<or=T<or=45 K) X- and K-band electron spin resonance (ESR) properties of the Si/SiO₂ interfacial (111) P_bO defect-identified with .Si identical to Si₃ and with the unpaired sp³ hybrid along (111)-as localised at three 'different' (111)Si/SiO₂ interfaces, are comparatively surveyed. It is observed that differences in the ESR details develop in moving from state-of-the-art thermally oxidised Si/SiO₂ over a Si/SiO₂ structure grown under strongly reduced O₂ pressure to the Si/native oxide case, indicating increasing interfacial strain or disorder. Secondly, ESR data are presented on the first observation of the .Si identical to Si₃ defect-now called (111)P_bNO-in (111)Si/Si₃N₄, thermally grown at 1100+or-20 degrees C in NH₃. The main ESR signatures of the P_bO and P_bNO centres are very similar, from which their common origin is concluded. Yet, the differences in ESR properties from state-of-the-art Si/SiO₂ to the Si/native SiO₂ structures are all exhibited by P_bNO too; in fact, they show up more intensely and reveal the more strained nature of the Si/Si₃N₄ interface. The P_bNO observation confirms the P_bO(P_bNO) defect as a prototype dangling-bond defect with the unpaired hybrid pointing out of the Si substrate. Generally, the P_bO(P_bNO) defect appears to be a highly sensitive probe apt to reflect variations in the physico-chemical nature of the interface through its low-temperature ESR features.

Deep-level defect centres occupy a unique position intermediate between isolated molecules and the solid state. Their electronic wavefunctions are often strongly localised about the defect site and resemble small-molecule orbitals; but they are also embedded in an extended host, and they interact with the host matrix. The authors discuss simple concepts and models from molecular electronic structure theory which are equally useful in the study of localised defect centres. First they outline the construction and use of Walsh diagrams, which predict conformations and conformational change as a function of electron number, electronic excitation and atomic orbital energy (electronegativity). Second, they discuss in detail the approximations of the widely used local-hybrid analysis of ESR and ENDOR spectra, point to the situations where those approximations are likely to fail and give an explicit comparison of theoretical calculations at several levels of sophistication for a particular molecular analogue, the silyl radical SiH₃. In both cases they concentrate on broken-bond systems of the form AB₃, which represent the most important intrinsic defects in silicon and silica. The importance and utility of molecular analogies in the study of defect centres are emphasised throughout.

The pressure dependence of the electronic gap states at the (111) Si-SiO₂ interface which are associated with the paramagnetic defect termed the P_b centre was studied with voltage transient spectroscopy. This paper presents detailed descriptions of the experimental techniques, theory of the measurement and data analysis. The P_b centre is an amphoteric defect with two transition energies in the Siband gap. The energy for the transition from double to single occupancy by holes was found to shift towards the valence band edge with increasing pressure, in agreement with theoretical predictions for the Si dangling orbital model of the P_b centre.

This paper discusses the defect structure of MOS device materials, in particular the evidence relating to dangling bond defects at the Si/SiO₂ interface. Over two decades and more, the characterisation of such defect states has given information on paramagnetic (S=¹/₂) centres at the interface. The most remarkable feature of the thermally oxidised Si surface is the very small density of interface defects ( approximately 10¹⁵ m^-2) compared to the bare semiconductor surface. These states have energies throughout the Si band gap, and their charge may be changed by changing the Fermi level in the band gap. Ionic contamination of the thermal oxide may affect the electrostatics of the MOS material. Some ionic defects (such as Na⁺) move under the influence of an applied field and cause device characteristics to drift with time, in addition there is a positive 'fixed oxide charge' located within about 3 nm of the Si/SiO₂ interface. Other centres exist throughout the bulk of the oxide, which can trap both positive and negative charge. The control of interface state density and oxide charge density is extremely important for the production of reliable devices. Given the very small defect levels, rather sensitive techniques have to be used to measure the concentration and characterise the structure of defects in Si/SiO₂ materials. Here the authors discuss the use of spin-dependent spectroscopies such as electron spin resonance, optical detection of magnetic resonance and spin-dependent recombination as aids to elucidating defect structure. They also discuss some experiments on p-channel MOSFETS in which an inversion layer of holes is created at the interface when turned on by a sufficiently negative gate voltage. Inversion layers in MOSFETS at low temperatures are used for the investigation of conduction in two dimensions. Such experiments at low temperatures on the localisation of carriers are shown here to give information on both positive and negative charge trapped near the interface.

The interface traps created by ionising radiation or hot-electron injection in MOS capacitors have been found to undergo significant changes with time over an extended period (many months). Immediately after radiation or hot-electron damage, an interface trap peak above the midgap ( approximately E_v+0.75 eV) invariably appears. This peak (designated peak 1), along with its background, would continuously change with time after damage, and the detailed time-dependent behaviour depends on the surface orientation of the Si substrate, processing history, gate bias and sample temperature. For samples made on (100) substrates, three separable regimes have been observed: (i) latent generation (peak 1 and its background increase with time), (ii) defect transformation (peak 1 gradually converts into peak 2 below the midgap, resulting in a double-peak interface trap distribution), and (iii) room-temperature annealing (the overall density of interface traps decreases with time). The focus of this paper is on the defect transformation process. For samples made on (111) samples, on the other hand, the most salient feature is the gradual shift of the energy position of peak 1 toward the valence band, and eventually a single-peak residing below the midgap is observed. In contrast to the (100) results, no discernible double-peak distribution has been found in (111) samples. Results on (110) samples are qualitatively similar to those on (111) samples, while (311) samples are similar to (100) samples. The various experimental parameters that affect the defect transformation process in both (100) and (111) samples will be discussed. While the (100) results are too complex to be explained satisfactorily based on existing theories at the present time, the (111) results will be interpreted in terms of the atomic relaxation of the dangling-bond defect at the (111)Si/SiO₂ interface.

Piezoresistance measurements have been obtained on narrow polycrystalline-silicon-gated silicon field-effect transistors. From the anomalous structure observed on the piezoresistance traces it has been deduced that large compressive intrinsic edge stresses are present in these devices. These are estimated from the experimental data to be approximately 180 N mm^-2, in reasonable agreement with a theoretical calculation based on a model proposed to explain the presence of such large stresses.

The authors describe the results of a study of discrete, low-frequency fluctuations in the current through very thin insulator ( approximately 20 AA), small-area (1-2500 mu m²) metal-oxide-semiconductor tunnel diodes. Findings suggest that the discrete nature of the fluctuations arises from the activated transitions of oxide defects, either localised or extended, between a number of accessible metastable configurations, each of which has associated with it a different net trapped charge. Inelastic interactions between the fluctuating defects and the diode tunnel current can result in random energy input to the defects, which can effectively heat them above the bulk oxide temperature. At sufficiently high device biases, these interactions, coupled with interactions between the defects themselves, are strong enough that they can result in the creation of additional oxide defects, and thus lead to the gradual wear-out or deterioration of the oxide. A study of the complex fluctuations which arise during thin oxide wear-out and their effects on diode characteristics reveals that after an initial period of Si-SiO₂ interface defect creation, the oxide wear-out proceeds by the diffusion of defects into the oxide. These moving defects appear to be closely associated with hole traps in the oxide. A number of the characteristics of very thin oxide wear-out can also be seen or inferred in measurements on thicker oxides, suggesting that the studies of very thin oxide diodes are providing a particularly detailed picture of the early stages of trap state formation, wear-out and breakdown in tunnel oxides in general.

Ionic vibrational properties of trap systems present at the silicon-silicon dioxide interface are expected to influence measured data and their interpretation. Similar to the case of deep bulk impurities in semiconductors, these traps require a model, where the eigenenergies of the total defect system, including the captured electrons and the vibrating ion, is taken into account in optical investigations. For thermally governed processes, the same particle picture can be brought into a thermodynamical description, to make more detailed information on interface states possible. Differences between optical threshold energies and thermally obtained energy quantities, Gibbs free energies and enthalpies, together with thermally activated carrier capture and entropy changes during carrier transitions are expected when ionic relaxations are present at the interface. Experimental results, published in the literature, are examined in the light of this model. A number of data exist, which indicate that important parts of a detailed understanding of the silicon dioxide-silicon interface states are lost when treated as bare electronic systems.

In micron-sized MOSFETS, the alternate capture and emission of carriers at individual Si-SiO₂ interface defects generates discrete switching in the source-drain resistance. The resistance changes are observed in the drain current as random telegraph signals, or as stepped transients after a strong perturbation of the trap occupation. The study of individual defects in MOSFETS has provided a powerful means of investigating the capture and emission kinetics of interface traps, has demonstrated the defect origins of low-frequency (1/f) noise in MOSFETS, and has provided new insight into the nature of defects at the Si:SiO₂ interface.

Si(100) substrates were prepared for the heteroepitaxial growth of GaAs using four simple cleaning procedures. The effects of these different techniques upon the quality of GaAs/Si material have been assessed in situ by reflection high-energy electron diffraction (RHEED); and ex situ employing double-crystal X-ray diffraction (XRD), Hall effect and current-voltage and capacitance-voltage measurements of Au-Schottky diodes fabricated on the GaAs epilayer. X-ray photoelectron spectroscopy (XPS) was utilised to investigate the Si surfaces after they had received the various etching treatments. Results indicate that an HF etched Si surface is not suitable for initiating the heteroepitaxial growth of GaAs. However, when the HF etch is used in conjunction with a reoxidation etch and/or the deposition and thermal desorption of a 500 AA GaAs prelayer, good-quality GaAs epilayers on Si may be produced. The current-voltage characteristics of the Schottky diodes test that this technique is a sensitive test of the GaAs/Si material 'quality'.

DC conductivity and thermoelectric power of bulk amorphous samples of Ge_xSe_1-x (0.1<x<0.4) have been studied as functions of temperature (77-350 K) and composition. The thermoelectric power was found to be positive. The thermo-electric power exhibits a less steep temperature dependence compared with that of the conductivity data. This discrepancy has been explained on the basis of the small polaron model, the chaotic band model and the two-channel model. It was observed that the discrepancy cannot be reconciled with the small polaron model and chaotic band model predictions but the data are best fitted to a parallel current path model, in extended and localised states at the band edges as proposed by Nagels et al. (1985). The present analysis also provides evidence for the existence of a small tail of localised states at the valence band edge.

The effect of hot-phonon production on the steady-state high-field transport properties of electrons in bulk and quasi-2D semiconductors is discussed within the framework of a simple theoretical model in which Cerenkov effects and phonon drift are assumed to be negligible. The model also assumes a single spherical parabolic band for the electrons whose distribution is a displaced Maxwellian, with scattering confined to polar optical phonons, restricted to interface modes in a deep square quantum well in the quasi-2D case. A finite phonon lifetime of 7 ps is shown to reduce the energy relaxation rate and to increase the momentum relaxation rate. The result is to lower the mobility and to delay runaway to higher fields, these effects increasing with the electron density. It is pointed out that nondrifting hot phonons tend to quench intervalley and real-space-transfer NDRS and to reduce the drift velocity for a given field. It is shown that, in general, the effective energy relaxation time is field dependent.

Steady-state and transient photoconductivity measurements have been made at different temperatures and intensities on the amorphous and partially crystallised films of Se₇₅Te₂₅ prepared by vacuum evaporation. Partial crystallisation is achieved by annealing the films at a particular temperature near the crystallisation temperature for different annealing times. The results indicate that the DC conductivity ( sigma _DC) and photoconductivity ( sigma _ph) increase but the photosensitivity ( sigma _ph/ sigma _d) decreases on partial crystallisation. The activation energy for DC conduction decreases while the activation energy for photoconduction remains unaffected on partial crystallisation. The transient photoconductivity behaviour is also affected quite appreciably on partial crystallisation. A peak in the rise of the photocurrent observed in amorphous films disappears in partially crystallised films. Moreover, the decay of the photocurrent becomes slower and slower as the extent of crystallisation continues to increase. The results are discussed in terms of the defect in the mobility gap of the present material.

The authors report the first photoluminescence measurements made under hydrostatic pressure up to 8 kbar at 4.2 on GaAs/Si samples grown by molecular beam epitaxy. They show that, in addition to the increase in the GaAs band gap, the applied pressure induces an increase of the splitting between the light- and heavy-hole bands. This effect is well described by considering the difference between the compressibilities of GaAs and the Si substrate, which causes an enlargement of the biaxial tensile strain of the GaAs layer as hydrostatic pressure is applied. They therefore demonstrate that, using hydrostatic pressure, one can study the physical properties of a given sample under different conditions of internal biaxial stress (caused by the lattice mismatch between the different layers), in materials produced by sophisticated heteroepitaxy techniques.

Spatially resolved experiments have been performed on the luminescence of bevelled Al_xGa_1-xAs structures as a function of the alloy composition. The binding energy of the isolated zinc-acceptor, E^Zn_a, with respect to the valence band edge, is determined for 0<or=x<or=0.4 and obeys the relation E^Zn_a(x)=30.7+495x^2.78 (E^Zn_a in meV). The radiative decay of free excitons is observed in the range 0.16<or=x<or=0.42 at T=38 K. While in the direct gap region the exciton binding energy slowly increases with increasing x the change in the behaviour of the free excitons is rather abrupt as one goes from the direct to the indirect-gap alloys, as is expected under the supposition of very weak coupling between Gamma and X exciton states. The donor ionisation energy was obtained in the range of Al content 0.25<or=x<or=0.42 by analysing the radiative decay of free holes with electrons bound to donors at T=80 K. The donor depth passes a maximum of about 60 meV near the direct-to-indirect gap transition of the alloy system and an effective-mass like value associated with the Gamma minimum of the conduction band can be expected for x<0.25.

The authors have used the layer method and a microscopic valence force field model to calculate both bulk, localised and resonant phonon modes for a planar interface. Dispersion curves along the symmetry directions of the two-dimensional Brillouin zone are calculated indicating that the modes are highly localised. They find interface modes both in GaSb/InAs and InAs/GaSb semiconductor heterojunctions but none in the GaAs-AlAs system.

A GaAs layer grown by MBE at a substrate temperature of 520 degrees C and containing three delta -doped planes with Si concentrations of 0.4, 1 and 4*10¹³ atoms/cm² has been post-growth annealed in a furnace, up to temperatures of 648 degrees C. Secondary ion mass spectroscopy (SIMS) and capacitance voltage (cv) measurements have been carried out to measure profile-broadening. The most lightly doped plane gave a near-Gaussian diffusion profile with a diffusion coefficient comparable with literature values for simple diffusion of isolated Si_Ga atoms. The more heavily doped planes exhibit a complex profile shape with two components, a proportion of the atoms being confined to the original plane, together with an almost square-shaped profile of fast-diffusing atoms. Comparison of the CV and SIMS data suggests that formation of Si islands is taking place during deposition of the delta -doped plane, giving electrically inactive atoms which can subsequently diffuse into the surrounding GaAs during heat-treatment. This model is supported by preliminary local vibrational mode measurements which have been made on a set of multiplane samples.