Showing papers in "Solid-state Electronics in 1965"

Investigation of thermally oxidised silicon surfaces using metal-oxide-semiconductor structures

Abstract: The results of a comprehensive study of the overall electrical characteristics of thermally oxidised silicon surfaces are presented, and interpreted on the basis of a simple physical model of the MOS structure. Extreme care was taken throughout this study to insure the validity, significance and reproducibility of the quantities measured. It is shown that the charge in the surface states is constant over a wide range of variation of the surface potential, and that this charge is positive for both n and p-type oxidised silicon. The density of surface state charge is about 2 × 1011 cm−2, and is essentially unaffected by a twenty-fold variation in the oxide thickness and by a 250-fold variation in the concentration of boron in the p-type samples. The effects of measurement frequency, illumination and temperature on the space charge capacitance of an inverted surface were investigated. This capacitance was shown to follow one of three simple models depending on the measurement frequency and the rate of generation of minority carriers.

The Richardson constant for thermionic emission in Schottky barrier diodes

Abstract: The Richardson equation appropriate to thermionic emission in Schottky barrier diodes is derived. For a semiconductor having an energy band with ellipsoidal constant-energy surfaces in momentum space, the Richardson constant A 1 ∗ associated with a single energy minimum is A ∗ 1 =4φ qk 2 h 3 (l 2 m y m z +m 2 m z m x +n 2 m x m y ) 1 2 where l, m and n are the direction cosines of the normal to the emitting plane relative to the principal axes of the ellipsoid and mx, my and mz are the components of the effective mass tensor. In the Ge conduction band, summation of emission from all the energy minima gives maximum and minimum ratios of A∗ to the free electron value A (= 120 A/cm2/°K2) of 1·19 and 1·07 for the 〈100〉 and 〈111〉 directions respectively. In the silicon conduction band, maximum and minimum ratios of 2·15 and 2·05 occur for the 〈111〉 and 〈100〉 directions respectively. The theoretical predictions are in good agreement with experimental results from W-Si and Au-GaAs diodes.

Chemical vapour deposition promoted by r.f. discharge

Abstract: In the vapour deposition method investigated, chemical reactions take place in an r.f. discharge instead of being promoted thermally. Glassy layers are deposited at 2–4μ/hr on cold or heated substrates. Materials prepared and evaluated include: (a) silicon from silane, (b) silicon dioxide from silane and nitrous oxide, (c) silicon nitride from silane and anhydrous ammonia. The method is expected to have a much wider application.

Minority carrier injection and charge storage in epitaxial Schottky barrier diodes

Abstract: Minority carrier injection by so-called noninjecting metal-semiconductor contacts is considered under conditions of moderate to heavy forward bias. The injection ratio, γ (ratio of minority carrier current to total current), rises linearly with forward current for sufficiently large applied bias. The reason for this rise in γ is that the minority carrier current is enhanced by a drift-field component much larger than the diffusion current which dominates at low bias. In this range the injection ratio is given by γ = n 2 i j bN D 2 j ns where ni and ND are the intrinsic and doping concentrations, b the mobility ratio, jns the Schottky diode saturation current density, and j the diode forward current density. As an example a 5 Ω-cm n-type silicon-gold diode will obtain an injection ratio of 5% at a current density of 350 A/cm2. The minority carrier stored charge per unit area (Q), for Schottky diodes made on thin epitaxial layers, depends upon the characteristics of the epitaxi-substrate interface and can become very significant when this interface is highly reflecting (i.e., has a low value of surface recombination velocity). For large applied bias and negligible bulk recombination the stored charge is given by Q = qn i 2 D p j N D j ns σ where q is the electronic charge, Dp the diffusion constant, and σ the surface recombination velocity. In measurements on experimental epitaxial diodes the interface is not found to be highly reflecting but is characterised by a recombination velocity of about 2000 cm/sec. This value applied to the 5 Ω-cm silicon-gold diode yields a storage time ( Q j ) of about 1 3 nsec. Normalised curves of γ and Q j vs. forward current are presented with parameters chosen in the range expected for silicon epitaxial Schottky barrier diodes. The theory has been verified by experiments on diodes made by evaporating gold contacts on silicon.

Physical limitations on the frequency response of a semiconductor surface inversion layer

Abstract: The physical phenomena associated with the frequency response of the surface inversion layer of an MOS capacitor are examined. It is shown that the equivalent circuit presented by Lehovec and Slobodskoy for an MOS capacitor in the depletion-inversion mode of operation may be simplified when the capacitor is biased in the heavy inversion layer mode. Further it is shown that an additional resistance, to account for generation-recombination in the depletion region, must be included. This new resistance dominates the response for silicon units, which is shown to be as low as I–5o cps at room temperature. Experimentally, there are at least two observations which cannot be explained by the first order one dimensional model on which these equivalent circuits are based. The frequency responses of complementary units, fabricated on almost identical p- and n-type silicon differ by orders of magnitude, it is typically in the range 1–100 cps for n-type units, but may be as high as 10 Mc/s for equivalent p-type units. In addition, pronounced hysteresis in the bias dependence of the capacitance of some p-type units has been observed. A second-order two dimensional model is proposed to explain these anomalies. This model includes an “external” inversion layer, surrounding the gate electrode, produced during the formation of the oxide. The gate inversion layer, in this model, is coupled to the bulk through an additional RC network, representing the external inversion layer, which is in parallel with the equivalent network of the first order model. An approximate analysis of this second order model predicts a frequency response which is in agreement with experiment. The possibility of charge migration on the surface of the oxide, which serves to decouple the gate inversion layer from the external inversion layer, can account for the hysteresis in the capacitance vs. bias characteristics observed in these units.

Gallium arsenide MOS transistors

Abstract: The feasibility of using gallium arsenide for an insulated-gate field-effect (MOS) transistor was investigated. A fabrication process for n-channel depletion-type units was developed, and devices were constructed using a simple experimental geometry which had a channel length of 7 μ and a width of 126 μ. The source and drain regions were formed by a planar diffusion of tin; pyrolytically deposited silicon dioxide was used as a mask. The channel region was formed by a special diffusion technique developed for low surface-concentration n-type diffusions. Silicon dioxide was used as the gate insulation. The channel saturation current, which ranged between 50 μA and 25 mA, could be modulated by biasing either the gate or the substrate. The low-frequency (120 c/s) gate transconductance varied up to 700 μ mhos. This parameter was measured as a function of frequency up to 5 Mc and was found to increase with frequency. Substrate transconductances as high as 4000 μ mhos were measured. Analysis of the device characteristics indicated an effective channel mobility of approximately 600 cm2/V-sec and an interface state density of the order of 1012 states/cm2. The increase in gate transconductance with frequency within the measured range was attributed to interface states having response times between 0·1 sec and 1 μsec. The effect of fast states on the device characteristics was investigated by use of a model similar to that proposed by S hockley for junction field-effect transistors.

Gold as a recombination centre in silicon

Abstract: The recombination of holes and electrons through gold centres in silicon involves two recombination energy levels, a donor and an acceptor, and four capture probabilities. By comparing the low-level photoconductivity decay of gold-doped silicon samples with a theoretical expression derived from a transient solution of the recombination problem, we have determined the following values for these probabilities: Acceptor level Donor level α n = 1·65 × 10 −9 cm 3 /sec β n = 6·3 × 10 −8 cm 3 /sec α p = 1·15 × 10 −7 cm 3 /sec β p = 2·4 × 10 −8 cm 3 /sec These results have been used to calculate, under high-level conditions, minority carrier lifetimes, which have then been compared with such lifetimes determined empirically from diode storage time measurements. Good agreement has been found, indicating that diode storage times can be successfully predicted.

Low-frequency generation noise in junction field effect transistors.

Abstract: A new type of noise mechanism, due to charge fluctuations of Shockley-Read-Hall generation centers in a depletion region, is described as the dominant low-frequency noise source in junction field effect transistors. The theory of this noise mechanism and its characteristics are demonstrated on a simple one-dimensional field effect transistor model. The two-dimensional analysis in the gradual channel approximation is shown to be in excellent agreement with experimental results obtained from both conventional and gold-doped field effect transistors over a wide range of bias conditions. Finally, circuit optimisation criteria and various special effects associated with this new type of noise mechanism are discussed along with possibilities of observing this type of noise on conventional bipolar transistors.

On the relationship of semiconductor compound properties and the average heats of atomisation

Abstract: The average heat of atomisation of semiconductor compounds, as computed from the heats of atomisation of the elements, is proposed as the basis for correlating the physical properties of these compounds and in particular the corresponding energy gaps. Inconsistencies encountered by employing other criteria (electronegativity differences, heats of formation etc.) in correlating semiconductor properties are pointed out and the general applicability of average heats of atomisation as reflecting the relative bond strength is illustrated. The average heat of atomisation is rather unique in correlating the properties of ternary and higher order semiconductor compounds.

P+IN+ silicon diodes at high forward current densities

Abstract: The current voltage characteristics of silicon junction diodes at high current densities (>100 A/cm2) are discussed in terms of a two junction P+IN+ model. Recombination within the central high resistivity layer and minority carrier injection into both the outer layers are considered, as is the effect of carrier-carrier scattering. It is found that the voltage dropped across the “I” region is proportional to (current) 1 2 only if the ratio of the thickness of this layer to the diffusion length w/L is 3 4 , and if w/L >1, the voltage can increase linearly with current or more rapidly. Probe measurements on sectioned diodes agree reasonably well with the theory.

Mechanism for reverse-biased breakdown radiation in p-n junctions

Abstract: The origin of reverse-biased breakdown radiation in germanium and silicon p - n junctions has not been definitely established. This radiation was first reported by Newman for silicon p - n junctions and somewhat later by Kikuchi and Tachikawa for germanium p - n junctions, and has a visible “spotted” spatial distribution. It has usually been assumed that this radiation originates from recombination occurring between energetic electrons and holes in microplasmas. The measured spectral distributions of the radiation from silicon and germanium junctions are quite different, with the silicon emission spectrum showing no structure, whereas the germanium one does. Existing explanations, including intraband transitions, do not adequately account for the presence or lack of structure. A more recent proposal by Figielski and Torun that the radiation is from the bremsstrahlung of hot carriers in the coulomb field of charged centres is extended to include a more suitable high field distribution function due to Yamashita and Inoue. It is found that the spectral distribution of the radiation should be of the form U v ∝ [1 - erf ( hv /√(2 p ) kT e )], where T e is the electron temperature, and p is a factor depending on mobility, electric field, and the velocity of sound.

Studies on optical coupling between silicon p−n junctions☆

Abstract: The noise pulse rate of a silicon avalanche diode is increased after another diode diffused into the same silicon slice switches to its reverse or forward conducting state. Artificial microplasmas with an extremely low background pulse rate are used to investigate this coupling mechanism quantitatively. An analysis of the experimental results leads to the conclusion that the interaction between the diodes is due to infrared radiation with a quantum energy of < 1·10 eV. Phonon assisted band-to-band generation rather than generation from defect centers is dominant. From measurements of optical coupling effects it is estimated that the gamma-sensitivity of a silicon avalanche counter is determined by direct absorption within the sensitive detector volume. Recombination radiation from other parts of the crystal is negligible. For 300 keV gamma-radiation the sensitivity is estimated to be of the order of 104 r/hr.

Electron-optical-phonon scattering in the emitter and collector barriers of semiconductor-metal-semiconductor structures

Abstract: A theoretical analysis is made of the effects on the emitter-collector current transfer ratio of optical-phonon scattering of electrons in the emitter and collector semiconductors of semiconductor-metal-semiconductor structures. The collector and emitter efficiencies are shown to increase appreciably with increasing collector and emitter electric fields respectively. At temperatures such that kT ≈ the optical-phonon energy, E0, the collector efficiency varies only slightly with increasing emitter-collector barrier height difference, Δϕ, but the emitter efficiency is greatly decreased when Δϕ E0. The current transfer ratios predicted by this theory for SiAuSi and GaAsAuSi structures are 0·68 and 0·55 respectively at 298°K and 0·85 and 0·61 respectively at 105°K with a collector field of 105 V/cm. This calculation does not treat collisions in the metal or quantum-mechanical reflection of electrons at the collector barrier.

The conduction properties of GeGaAs1−xPxn−n heterojunctions

Abstract: n -type Ge has been grown epitaxially on GaAs 1− x P x substrates of 4 × 10 17 cm −3 donors at 360°C. The substrates have 〈111〉 orientation and a composition range of x = 0, 0·1 and 0·3. Junctions made with x = 0 or 0·1 exhibit diode behaviour that can be interpreted by the metal-semiconductor emission theory modified to take into account the effects of both electron tunnelling and voltage distribution between the semiconductors. The barrier heights have been found to be 0·62 and 0·53 V for the A - and B -surface of substrates with x = 0·1 and 0·50 V and 0·43 V correspondingly with x = 0. The apparent barriers at zero bias are smaller than these values by about 0·2 V due to tunnelling reduction. For substrates with x = 0·3, various measurements tend to suggest a structure of back-to-back diodes. The effect of interface states, as likely produced by the relatively large lattice mismatch in this case (>1 per cent), is believed to be responsible for this structure.

The diffusion of silicon in gallium arsenide

Abstract: The rate of diffusion of silicon from a sputtered film of silicon into gallium arsenide has been studied as a function of the ambient arsenic pressure at 900 and 1000°C. The diffusion appears to be non-Fickian. The saturated surface concentration of silicon in diffused layers is about 0·3 wt per cent and the average electron concentration of the layers is about 5 × 1018 cm−3. Silicon has been used as an emitter diffusant in a double diffused GaAs transistor.

The gate-triggered turn-on process in thyristors

Abstract: A theoretical and experimental investigation has been made into the gate initiated turn-on process in thyristors. It has been found possible to predict the behaviour of a given structure with reasonable accuracy provided the turn-on process is divided into three parts. These parts are: (1) the delay time; (2) the rise time; (3) the conduction spreading time. Of these three components the delay time is the most complicated. For the cases of greatest interest, namely when a fairly high gate drive is used to obtain as short a value as possible, it was found that the delay time is proportional to the carrier transit time of the base to which gate connexion is made. The rise time depends upon the geometric mean of the carrier transit times of both of the bases and is inversely proportional to (α 1 + α 2 − 1) . Spreading time arises since conduction is first initiated near the gate and spreads out by diffusion of the current carriers. It thus takes some time before the forward voltage settles down to its final value. It was found that spreading velocity is proportional to ( D t r ) , where D is the mean of the diffusion constants for holes and electrons and tr is the rise time for turn-on from a low voltage.

Optical phenomena in GeGaP heterojunctions

Abstract: Heterojunctions of epitaxially grown Ge on GaP have been investigated optically. Measurements of the photovoltaic effect of Ge-GaP heterojunctions in which the direction of the light is parallel to the junction area, show a sign reversal of the photovoltage. With a photon energy of hν > 1·2 eV it was found that in all our n-n Ge-GaP heterojunctions, the sign is such that the GaP becomes negative with respect to the Ge. The opposite sign is found with photons having an energy hν > 1·2 eV. The critical wavelength at which the sign reversal occurs depends neither on the dope in the GaP nor on that in the Ge, and is such that it corresponds to an optical transition over an energy difference a little more than half the band gap in GaP. From these measurements one must conclude that the Fermilevel at the heterojunction interface has a fixed position independent of the position of the Fermilevel in the bulk materials.

Thermal characteristics of GaAs laser junctions under high power pulsed conditions

Abstract: The temperature rise at the junction of a GaAs laser has been calculated for a variety of pulse currents and base temperatures. The problem is first solved analytically by an approximation assuming constant thermal parameters, and then by a machine calculation taking into account the temperature variation of the thermal conductivity and the heat capacity. These solutions are compared and discussed in terms of practical laser design parameters. The effects of bulk resistivity and contact resistance are included.

The maximum blocking capability of silicon thyristors

Abstract: The blocking capability of thyristors in forward and reverse direction is discussed under such aspects as they are of interest especially for the design of power thyristors. As in the ideal case the blocking capabilities in both directions ought to be equal, the considerations are confined at first to the simpler evaluation of the reverse blocking capability, and to this end a quantitative method is developed. The influence of certain thyristor parameters on the dependence of the blocking capability upon the resistivity of the starting material is considered. Finally, the conditions for the ideal case of identical blocking capabilities in forward and reverse directions are explained, and the steps to be taken in order to realize it are mentioned.

Expanded contacts and interconnexions to monolithic silicon integrated circuits

Abstract: The requirements for an ideal expanded contact system for oxide-passivated silicon devices are listed and discussed. Problems associated with the conventional aluminum contact-gold lead system are discussed—namely, “purple plague” and the thermodynamic instability of aluminum and silicon dioxide. Evidence is cited which shows that a bi-metal film of molybdenum-gold can be combined with ball-bonded gold wire to form a highly reliable, near ideal, expanded contact system.

Electrical properties of n-type aluminium arsenide

Abstract: Hall effect and resistivity measurements have been made between 50 and 400°K on single crystal n -type epitaxial layers of AlAs. The maximum mobility observed at room temperature is about 180cm 2 /V sec. for a carrier concentration of 8·5 × 10 17 cm −3 . An energy level 0·018eV below the conduction band is tentatively associated with silicon and from an analysis of the carrier concentration data, an effective mass of 0·5 m e for electrons is suggested, together with a ground state degeneracy of 2, or an effective mass of 0·8 m e with a degeneracy of 4.

Switching processes in alloyed pin rectifiers

Abstract: A theory for the switching of power rectifiers from forward to reverse state is derived, and confirmed by measurements. In cases interesting especially in application, one may confine oneself to strong injection in the forward state, and to strong reverse currents during the switching process. Under these circumstances, the reverse voltage is built up during the switching by the space charges of the flowing off charge carriers. The following results, which contradict a conception frequently voiced, may be pointed out: ( a ) The flowing off of the charge carriers is not limited by “reverse-biased” doping junctions. ( b ) For that reason, the duration of the switching process is not determined by the finite carrier lifetime if the sweeping out current is strong enough. It follows from the theory that during switching processes the rectifier can be destroyed by essentially smaller voltages than by those it resists under stationary conditions.

On the relationship between photocurrent decay time and trap distribution in CdS and CdSe photoconductors

Abstract: The method of thermally stimulated currents has been used to determine the electron trap distributions in single crystals of CdS and polycrystalline thin film layers of CdS and CdSe. Seven fundamental trapping levels were found and four have been tentatively correlated with the activation treatments. Photoconductive decay times have been measured over the temperature range of −192°C to +100°C, and are found to correlate directly with the thermally stimulated current in the same sample. The response time of our polycrystalline thin film layers has been found to be shorter than that of single crystals. The faster response of these polycrystalline films is explained by the observed difference in trap distribution between the single crystal and the polycrystalline film.

Structure and galvanomagnetic properties of two-phase recrystallised InSb-In layers☆

Abstract: Two-phase alloys of InSb-In have been prepared by the controlled melting and recrystallisation of polycrystalline InSb layers overcoated with thin, metallic In films. X-ray diffraction shows that recrystallisation brings about a preferential crystallographic ordering of {111} in the plane of the film. Electron micrographs, etching and metallographic studies reveal webbed InSb dendrites several mm long separated from each other by discontinuous In filaments embedded in an InSb matrix. The magnetoresistance (Δρ/ρ0)t of such a two-phase film is dominated by the microgeometry of the dendrites, their density, and distribution. It is independent of macroscopic geometrical factors; the Hall coefficient 〈Rh〉 has the expected theoretical dependence on specimen shape and electrode configurations. An essentially qualitative interpretation of the measured temperature dependence of (Δρ/ρ0)t and 〈Rh〉 suggests a superposition of two charge carrier transport mechanisms: the geometrical (Δρ/ρ0) of the near-intrinsic dendrites is shunted by the physical magnetoresistance of the acceptor-doped matrix. At 295°K, typical two-phase recrystallised films 1·5 to 2· μ thick have a (Δρ/ρ0)t ⋍ 4 at H = 2 × 104 G and a virtual electron mobility (calculated from Hall measurements) in the range 2·2 × 104 ⩽ μn ⩽ 3·2 × 104 cm2/(V-sec.).

Double-injection experiments in semi-insulating silicon diodes

Abstract: The pre-breakdown V-I characteristics of double-injection p - i - n silicon diodes doped with indium have been obtained over the temperature range 64 to 100°K. This has resulted in experimental determination of the capture cross-section for electrons and holes, by neutral and singly negatively charged indium atoms respectively, as a function of temperature. The electron capture cross-section of 2·4 × 10 −14 T −1·3 cm 2 compares extremely well with 1·1 × 10 −14 T −1 cm 2 previously determined by a different method. The magnitude of the hole capture cross-section is very close to the 1·5 × 10 −9 T −2 cm 2 previously reported, but the temperature dependency is considerably greater. High field effects are present over a considerable portion of the pre-breakdown region in the longer devices. The theoretical model has therefore been extended to include the field-dependency of mobility in the pre-breakdown region, and the observed characteristics explained. The analysis, in conjunction with the experimental results, allows determination of the field-dependencies of electron and hole capture cross-sections, and these are found to be E −2·4 and E −2·8 respectively and in good agreement with reported theory and experiments.

Approximate calculation of the spectral function for the stimulated recombination radiation in semiconductors

Abstract: The spectral function of the stimulated emission in semiconductors, proposed in integral form by Lasher and Stern (1), is explicitly evaluated through a simple approximate expression. The approximation is valid for those values of the photon energy that make the stimulated emission dominant over the absorption if it is assumed that the distances between the quasi-Fermi levels and the band edges, F'n and F'p, meet the condition: (F′ n + F′ p )/2kT ⩽ 1 . Explicit expressions for two quantities which are of interest in studying injection lasers i.e. peak energy of the stimulated emission and gain per unit length in the active region are also derived. All the calculations have been performed in the hypothesis of parabolic bands.