Search Documents

By C. S. Duncan, S. Scuro, D. R. Muss

Work is vepurted which was aimed at accentuating the pec14liar attributes of the GaAs laser diode, namely its sutzull size and its high efficiency. This has been done by reducing- transport losses in the bulk p and n regions of the diode and combining many of. these diodes into series stacks with common optical cazlities. The result of this effort is a nearly monochromati source at about 8450A With a luminous areu 0.05 by The GaAs laser diode is an attractive pulsed light source in the near ir because of its ability to emit very large pulse powers and its small size. In addition to this, the device is efficient; and the higher the 0.05 cm emitting peak pulse po we rs at greater than 300 LC at repetition rates below 200 pps into a cone about 0.01 steradians when operated at 77°K. The results of nzeasuvetnents of peak optical power and spectra frorr~ these stacks are reported and some conclusions are drawn about the vole the individual diodes play in the stacked operation. peak output power, the more efficient it is. The limit for the amount of power that these diodes can emit is the transient and average dissipation associated with the pumping pulse. This power dissipation arises primarily from transport across the bulk regions of the diode; this fact, along with the relatively low thermal conductivity of the GaAs, led us to the conclusion that higher power

Jan 1, 1968

By Louis G. Bailey

This paper .reviews some of the key developments which have been made in the synthesis of the III-V compound semiconductors and the associated progress in obtaining high-quality material for device development. The history of the investigation of the semi-conducting properties of these materials has indicated the need for synthesis techniques capable of producing rnaterial of high chemical purity and physical perfection. The progress in the area is described, and the present status of quality of several compounds is discussed. In addition, recent studies of localized epitaxial deposition of GaAs into semi-insulating GaAs substvates for the development of arrays of devices ave reviewed. In conclusion, some important areas for future research and developrnent are discussed. INTEREST in III-V compounds as semiconductors dates back to the period 1950 to 1952, when Goryunova, Ioffe, and, in particular, Welker determined that 111-V compounds were structural analogs of the group IV elements carbon, silicon, germanium, and tin and that they were also semiconductors.'-3 Since that time interest in the properties of these materials has ex- panded rapidly. During 1953 and 1954, the first reports of rectification, photoconductivity, and optical absorption in these materials came from several investigators. In 1955, recombination radiation from samples of GaAs, InP, and GaSb was reported by ~raunstein .' This radiation was produced by injected minority carriers from point and large-area contacts. From 1955 to 1960, efforts increased on preparing large, pure specimens of these materials for the study of basic physical phenomena and device exploration. By 1960, materials preparation and the understanding of the physics of some of the materials had advanced to where some very interesting and potentially useful phenomena came under study. The first observation of electric-field-induced changes in the optical absorption edge (Franz -Keldysh effect) of 111-V compounds was reported for GaAs by Moss in 1961.' In 1962 the study of recombination radiation advanced rapidly. Keyes and Quist6 and Pankove and Berkey-heiser7 reported high radiative efficientices obtained from recombination, and then the first injection electroluminescent lasers in GaAs were announced by Hall ef al.,' by Nathan et al.,O and by Quist ef al.,'O and in Ga(As, P) alloys by Holonyak and Bevacqua." Hil-sum has reviewed the early work leading to the first observations of laser action in GaAS." Recombination processes in GaAs, Gap, and the other 111-V com-

Jan 1, 1968

By Arnold S. Epstein

Tlze green electroluminescence occurring at liquid-nitvogen temperature in epitaxial gallium phosphide diodes is exarnined using the donor dopants silicon and sulfur. Zinc is used as the p-type diffusant. Two tt~ain peaks are found, namely til 0.56 and 0.535 p. The intensity of the gYeen peak occurring at 0.56 is found to oary as -4"' wlzere Jf is the forward current density and to first order appears to be dependent on the concet~tvation of ionized donovs. Tile vadintive lifetinze fov this process is -10-'sec. The intensity of the 0.56-p lu~ninescence is nt least several ovdevs of magnitude greate,, at liquid-nitvogen temperature than that noted jor the 0.538 radiation for current densities up to about 35 amp per sq cm. The variation oj- the intensity of. the 0.538-p peak with current density caries according to the donor dopant. With sulfur as the donor one has a J12 dependence, whereas when silicon is the donov a J1.5 dependence is noted. Tl~e above can be fitted to a In molecular process (in first order) such as a band to band recombination. The constant B fir band to band vecombination) is found from our experitt~etlt to be of the order of 2 CZI cm per sec, zcl~ereas for tlze impurity vecombination peak (0.56 p]) tile A cotzstor~t is of the order of. 7 x 10-' cu cm per sec. The structuve o/. the diodes at liquid-nitvog-en tett~peruture is oj tlze double injection type, and by carefi~lly selecting. the thickness of the structure one can regulate the onset of the 0.538-p radiation. The diode clzavaclevistics follow an I-Vx laiv where x is >1. Llregatiue resistance in these structures con be correlated with saturation of tlze 0.56- peak. ThE development of high-purity, large single crystals of epitaxially grown gallium phosphide is the ultimate aim of the studies to be reported. While a number of papers and reviews' have been presented on properties of gallium phosphide grown from gallium solution and from the melt, less attention has been paid to the properties of epitaxial gallium phosphide grown from the vapor. In the studies to be reported we have examined the effects of shallow donor impurities, primarily silicon and sulfur, at liquid-nitrogen temperature, on both the electrical and green electroluminescent properties of carefully characterized single-crystal epitaxial Gap. I) EXPERIMENTAL PROCEDURES A) Material Preparation and Properties. The samules used in this investigation were single-crystal n-type epitaxial gallium phosphide. he single crystals were synthesized from the elements by vapor transport and epitaxial deposition on single-crystal gallium arsenide substrates. Typical transport agents used were chlorides. The growth procedures are described in more detail in Ref. 2. The samples were grown in the (111) orientation and had thicknesses up to 480 . Following the growth, the GaAs substrate was removed from the epitaxial Gap by etching in nitric acid. The Hall effect and resistivity of these n-type samples were measured as a function of temperature and analyzed for impurity content and compared with spectroscopic emission analysis. The results are given in greater detail in Ref. 3 but a summary is noted in Table I. In Table I, AED denotes the donor ionization energy, Nd is the donor impurity concentration, lvA is the acceptor impurity concentration, NI is the total impurity concentration, Rip vs temperature is the measured Hall mobility-temperature relationship from liquid nitrogen to about 400"K, n(77"K) is the measured carrier concentration at liquid nitrogen, and R/p(77'~) is the measured Hall mobility at liquid nitrogen. Good agreement is noted between electrical and emission analysis for the major donor impurity, silicon, in the two materials where it is a major constituent. In the case of the two samples SC90-6 and AP138-3(111) B which were sulfur-doped the ionization energy found from the electrical data is in good agreement with that reported for sulfur.4 The acceptor impurities are not known, although from the presence of magnesium in the samples as indicated by emission analysis and its known role as an acceptor in GaP,4 magnesium could be present as an acceptor. Similarly, since silicon is known to be an amphoteric impurity4 it may also be present as an acceptor in the Gap.* The *A small band has been noted at 0.625 11 in samples SC90-6, SC76-4, and SC77-4, suggesting according to Ref. 5 that sillcon could be present as an acceptor In these samples since these authors have found this peak to represent a donor-silicon emission. presence of calcium deduced from emission analysis is not correlatable with the electrical analysis and its presence as an electrically active impurity should be questioned. B) Sample Preparation and Diode Fabrication. The original slices were of the order of 2 sq cm in area witi the (111)B face up. The material was cleaved and sectioned and the best pieces selected (i.e., those which visually appeared to have the least strain and defects). The samples were polished, lightly etched, and cleaned in trichloroethylene. Zinc was diffused in a closed tube at 800°C to form a p-n junction. Diodes were made from these samples. The depth of junction varied, but did not exceed 10 p depth. The back side of the slice was lapped back and a low-resistance n-type contact was formed (at 600°C in forming gas) to an Au-Mo tab with a hole in the middle for

Jan 1, 1968

By C. J. Frosch, J. A. May, H. G. White, C. D. Thurmond

Gallium phosphide epitaxial layers were grown from the vapor phase on undoped single-crystal galliurn arsenide substrates in silica tubes by an open-tube wet-hydrogen process. The epitaxial layers were grown over a range of water pressures at three substrate temperatures. Excess donor concentrations were determined by surface barvier capacitance measurelrzents without removing the layers from the substrates. The excess dmlor concentration, ND-NA, is fo~ind to vary approxilnately inversely with the pressure of water added to the hydrogen carrier gas. This is the relationship that would be expected for singly ionized silicon donors on gallium sites in extrinsic galliunz phosphide, with the silicon coming from the SiO generated by the reaction of hydrogen with the silica tube. An increase in the partial pressure of water in the hydrogen stream decreases the SiO pressure. The results indicate that ni, the intrinsic hole and electrmt concentration for gallium phosphide at the three substrate temperatures, is smaller than the concentration estimated from available data for the density of states effective masses and the energy gap. Mass-spectrographic measurements confirm that the dono?, introduced into gallium phosphide is silicon. The equilibrium concentrations of silicon in vapor-flown gallium Phosphide have been estimated from available thernzodynamic information that includes the solubility measurements of silicon in gallium phosphide in equilibrium with a gallium-rich liquid phase. Satisfactory agreement with the measured silicon concentrations is obtained. FROSCH1 has described an open-tube process for growing single-crystal Gap from the vapor phase by a GazO transport mechanism. The method depends upon the reaction of H20 in an H2 carrier gas with a heated source of polycrystalline Gap which provides the necessary vapor species. When the temperature of these vapor species is lowered, super saturation occurs and single-crystal Gap will deposit on a suitable substrate. Unintentionally doped single crystals of Gap grown by the wet H2 process in silica tubes are n type. Evidence is presented to show that the donor introduced is silicon, and that a qua si-equilibrium model accounts for the inverse dependence of the donor concentration on the water partial pressure and predicts the magnitude of the donor concentrations. Ainslie et al. experimentally showed a similar inverse relationship between the carrier density and oxygen pressure for GaAs. Emission-spectrographic analyses showed a decrease in the silicon concentration with increasing oxygen overpressure for GaAs. Cochran and Foster suggested the theoretical possibility of suppressing silicon contamination by using Ga20 generated by the reaction of gallium with water vapor. 1) EXPERIMENTAL The apparatus and procedures are essentially the same as those described by Frosch.' The apparatus consists of a 25-mm-ID SiO2 tube extending through a controlled high-temperature flat zone for the location of the polycrystalline Gap source and a downstream temperature gradient falling at a rate of about 14°C per cm. The latter provides the region of super saturation for the location of the single-crystal substrate. The partial pressure of water in the inflowing hydrogen stream, pA2, O was controlled by mixing me-tered proportions of dry H2 with H2 saturated with H2O vapor at 0°C. The total gas flows were about 200 cu cm per min in all experiments. The Gap sources were prepared by pulverizing boat-grown polycrystalline ingots to pass a 20-mesh sieve. The substrates were cut from an undoped single-crystal boat-grown GaAs ingot purchased from Monsanto. This ingot had a carrier concentration of about 1015 atoms per cu cm, a resistivity of about 5 ohm-cm, and a mobility of about 5000 sq cm per v sec at 25°C. Substrates with dimensions of 1 by 1 by 5 x lo-' cm were employed. The growth faces were chemically polished (111) arsenic faces. Epitaxial layers, at least 7.5 x 10-3 cm thick, were grown,. This required from 1 to 24 hr depending upon the Pii2Q values and the temperatures. In all of the runs, the source temperatures were 50°C higher than the substrate temperatures. Samples were prepared

Jan 1, 1968

By D. M. Blacknall, N. N. Winogradoff, E. W. Williams

Low-temperature measurements of photolumines-cence were used to evaluate the progvess in materials development. Variation of the impurity type, impurity concentration, and method of growth were used to clarify the chemical origin of defects in GaAs. Melt-g~ozun, epitaxial vapor-, and solution-grown samples were studied. A comparison of measurements at 77" and 20°K shows that more defects can be observed at the lower temperature. Th'vee residual defects were identified in relatively pure GaAs. These were: silicon, copper, and a gallium vacancy-donor complex. Solution growth from a gallium solution inhibits the formation of the latter two. Eleven defects were studied and the "optical" activation energy of some of these, namely, silicon, cobalt, and chromium, has not been reported in the literature before. THE discovery of the injection laser in 1962 prompted extensive measurements of the photoluminescence of GaAs. The measurements were made at 77°K on GaAs doped with shallow impurities at various doping levels and compared with electroluminescence1'2 and absorp- tion data.374 In addition, lower-temperature studies were made on relatively pure GaAS' and on GaAs doped with oxygen,5'" manganese,' and copper.' The observation of defects* in GaAs with photolu- *The word "defects" ~efers to both chemical impurities and native defects. minescence at low temperatures makes photolu mines-cence an invaluable tool in evaluating the progress in materials development. The only previous evaluation was done at 77°K when solution-grown and melt-grown GaAs were compared.O No identification of defects was made but it was shown that some of the deep-level luminescence seen in melt-grown GaAs was absent in solution-grown GaAs. The present work extends the comparison of preparation techniques to lower temperatures and to a study of epitaxial vapor-grown GaAs. It is particularly valuable for thin epitaxial layers since they cannot be analyzed by conventional analytical techniques. In addition, by varying the impurity type and impurity concentration the chemical origin and nature of defects in GaAs was demonstrated. Eleven defects were studied in all. Three of the defects were identified as residual defects which often contaminate GaAs prepared from the melt and the vapor. Only one such defect was observed in GaAs grown from a gallium solution and this was identified as silicon.

Jan 1, 1968

By Mark D. Tabak

Measurments of the transient photoconductivity in fillns of amorphous selenium with blocking- contacts haue been used in studying the transport properties. The results shozu that the transport of free carviers cannot be explained in terms of a model using bulk trapping. Measurments of the photocurrent-applied field characteristic as a unction of sample thickness do not pvoduce the scaling lau,s approps-iate lo a range-limited model. THE investigation described here was undertaken to explore the mechanisms of transport in amorphous selenium films. These films are used in xerography, where a photoconductor is charged electrostatically and then exposed to light to form a latent image. The sensitivity that one can expect when operating in this mode is directly related to the transport properties of the photoconductor material. The study of photoconductivity at high fields in insulating materials is often complicated by contact effects. Injection from the contacts can easily mask the fundamental transport processes that are being studied. A technique is described in this paper that allows a study of photoconductivity and transport processes in amorphous selenium films without the complicating effects introduced by contacts. Accordingly, the samples are prepared using insulating films between the sample and the metal electrodes, where the electrodes are capacitively coupled to the surfaces of the sample under study. A transient measurement is performed with pulsed light, and the photocurrent measured near the beginning of the pulse before the field is distorted by polarization at the blocking elec- trodes. A similar method has been described by Manyl using pulsed electric fields in studying the high field transport in insulating cadmium sulfide. EXPERIMENTAL It is possible to eliminate undesirable contact effects and maintain uniform high fields across the sample by using blocking electrodes and measuring the current-field characteristic under transient conditions. The sample configuration is shown schematically in Fig. 1. Each of the gold electrodes is separated from the amorphous selenium film by a layer of insulating plastic (Formvar) approximately 2000~ thick. The electrode on the front surface is semi-transparent to allow illumination of the sample. The whole configuration is mounted on a KEL-F substrate, which is a transparent insulating plastic having a thermal expansion coefficient that allows low-temperature measurements without cracking the selenium film. Fig. 2 shows a block diagram of the measurement apparatus. Light was obtained from a 450-w Xenon arc lamp, the intensity of which was controlled by stainless-steel screen filters and monitored by a pho-tomultiplier which was calibrated periodically against a thermopile. A camera shutter operated at 1/200 sec provided the light pulse and a synchronized signal to the oscilloscope. The photocurrent transient is ob-served with a high-speed picoammeter (response time of 1 msec) the output of which is displayed on an oscilloscope. The applied voltage is constant throughout the transient measurement. The selenium films were prepared by evaporation of four 9's pure selenium shot supplied by Canadian Copper Refiners. Spectrographic analysis showed about 1 ppm of Te and Fe, 5 ppm of C1, and 0.05 ppm

Jan 1, 1968

By Charles C. Peterson, Ronald E. Enstrom

Significant improvements have been made in the ursine systern for epitaxial vapor gvowtlz of Gds. The electron concentration has been reduced to below 1015 cm-3 with electron-mobility values as high as 7200 sq on per v-sec at 30PK and 60,000 sq cw7 per v-sec at 77°K. These high-purity layers have been successfully gvown on n&apos; layers with thicknesses as snzall as 1 p and as large as 125 p. Furthermore, high-purity layers have been incorporated into an structure without degradation of electrical properties. Different impurity gradients at the n&apos; to n interface were achieved. Using these structures, Gunn oscillators have been operated at frequencies as low as 0.4 GHz and as high as 40 GHz, which equals the highest valzle yet reported. The frequencies nleasured are in reasonable apeenlent with those calculated fron7 the n-layer widths and the transit times. CW operatiolz has been achieved from 3 to 20 GHz. In Gunn devices for >40 GHz operation (i.e., witk an n-layer tkickness below 2.5p), the n layer has been generally too pure to permit oscillations to occuv. ThE mechanism of the Gunn oscillator&apos; is now generally understood.2-5 Electron transfer from the (000) minimum to the ( 100) minima leads to a voltage-controlled negative resistance, which results in the formation of domains of high and low electric fields which travel through the crystal with a characteristic velocity of about 107 cm per sec. When a constant voltage is applied to the sample, this domain motion generaliy leads to oscillations at the transit time frequency, fT = 107/L, where L is the thickness of the active n layer.&apos; In this case, a single high-field domain is nucleated at the cathode and grows until it reaches the anode, where it is absorbed; a new domain then forms at the cathode.5 There are related modes of oscillation of the same device in which a tuned circuit causes a periodic variation of the voltage across the device. The frequency then may be either above or below fT. It has been shown that oscillations will occur when n • L 2 1012 cm"2, where n is the carrier concentration in the active region. At lower doping densities, 10" 5 n . L 5 lo&apos;&apos;, the doping is said to be subcritical for oscillation and the description of the device operation changes since the domains become comparable in thickness to the device thickness, L. Nevertheless, such devices have been found to be "active" and, in fact, are useful as amplifiers.778 For CW operation, one wishes to minimize the power dissipation, which is related to the electric field, E, and the low-field resistivity, p. Below threshold, the power dissipation density is approximately Typically, at threshold where E -- 3 x 103 v per cm, the dissipation density is 107/p watts per cu cm, p in ohm-cm. Furthermore, it is desirable to provide good conduction of the heat generated in the active region. The temperature rise, AT, across the active layer of a sandwich structure, with a heat sink on one side, is given by AT = (E&apos;/P)(L~/~~) where it is assumed that the thermal conductivity, u, and resistivity, p, of GaAs do not vary appreciably over the range of temperature and electric fields existing in the layer. This expression shows that, subject to maintaining an appropriate (n . L) product, CW operation is favored by a higher resistivity, a lower threshold field, and a thinner device. The above device requirements place some severe restrictions on the materials to be used, especially for high-frequency (>10 GHz) operation. Here, it is necessary to prepare very thin (<10 µ) layers of GaAs, in a state of high purity (-1 ohm-cm), complete with very low-resistance ohmic contacts. Problems of handling such thin and fragile samples become inereasingly severe as the frequency is raised. Furthermore, while good low-resistance ohmic contacts are easily applied to low-resistivity GaAs, contacting difficulties are encountered for these higher-resistivity materials.

Jan 1, 1968

By Raynor Linzey, Karl M. Busen

The formation of thin films of silicon oxide by anodic oxidation of silicon and the subsequent removal of these films by an etch is a process often used for the evaluation of concentration distributions Profiles) in silicon layers by the differential sheet conductance method. The accuracy of the resulting profile is very strongly influenced by the uniformity of the thickness of the reacted silicon. Normally, it would be expected that for a constant number of coulombs passed the thickness would be the same from oxidation to oxidation. Investigations show that in certain electrolytes, for a given number of coulombs passed through an n-type silicon sample, the thickness of the reacted silicon increased with increasing resistivity. Even for the same resistivity the thickness varied sometimes by a factor of 1.5. When an electrolyte was used which consisted of 10 pct water by volume in ethylene glycol with 4.0 g KNO per 1000 ml of solution, anodiza-tion at 5 ma per sq cm led to satisfying results. Short-time anodizations gave oxide layers of a higher apparent density than those experienced from thermally gown silicon oxides. THE functioning and the electrical characteristics of semiconductor devices are based upon the incorporation of "impurities" into a single-crystalline body of suitable material and on the concentration distribution (profile) of this impurity. The incorporation can be achieved by well-known processes as, for example, by diffusion, epitaxial growth, alloying, or ion implantation. Often the profile resulting from these processes is not known. A powerful tool to learn about a concentration distribution is given by the method of differential sheet conductance which employs successive four-probe measurements on a layer subjected to stepwise removal of thin sublayers. Differential sheet conductance vs position (or penetration depth) of a sublayer is plotted and a smooth curve is drawn through the data. From this curve the profile is then calculated. When the semiconductor material is silicon, the sublayers are removed suitably by anodic oxidation of the silicon and subsequent dissolving of the formed silicon oxide by an etch. The accuracy of the resulting profile is very strongly influenced by the uniformity of the sublayer thickness. Normally, it would be expected that, for a constant number of coulombs passed, the thickness would be constant from oxidation to oxidation. Investigations showed that for sublayers several hundred angstroms thick the reproducibility can be rather poor. Therefore, efforts were made to obtain a reliable technique for uniform removal. The present paper describes such a technique and certain phenomena which were encountered during the investigations. APPARATUS AND TECHNIQUE The first report on the investigation of concentration distributions, where for differential sheet conductance measurements thin silicon sublayers were removed from a diffused layer by anodic oxidation, was given by Tannenbaum. The author reports the removal of sublayers which for the most part were 400 thick. More advanced device designs now require much narrower layers. When it was tried in these laboratories to determine profiles within such layers, difficulties were encountered with respect to sublayers which by necessity had to be thinner than the ones reported by Tannenbaum. The apparatus used for the investigations is sketched in Fig. 1. Two cylindrical containers connected by a wide tube are filled with an electrolyte. The left container receives the cathode whereas the right container is closed at the bottom end by a sample support consisting of a Teflon base and a copper pedestal. The silicon sample (1 by 1 cm) is mounted to the pedestal embedded flush in the Teflon base using silver paint (Degussa) for electrical contact. Pyseal is applied to the edges of the sample to protect the copper pedestal from the electrolyte. The base is mounted to the container using a water-tight silicon rubber gasket. Fig. 2 gives a view of the sample support. The copper pedestal is connected to the positive side of a power supply operating at a constant current output (Kepco Model ABC 425M). The electrolyte which has been reported by Duffek et al. consisted of either 2 or 10 pct water by volume in ethylene glycol and 4.0 g KNO3 in 1000-ml solution. The electrolyte is best prepared

Jan 1, 1967

By F. M. d’Heurle

Films of molybdenum have been prepared by sputtering onto oxidized silicon substrates. The resistivity. lattice parameter, orientation, and grain size were studied as a function of substrate temperature and substrate bias. Under normal sputtering conditions, the resistivity of the films was found to be quite high (600 x 10 ohm-crn). However, with the use of the negative substrate bias of 100 v and a substrate temperature of 350°C, films weve produced with a resistivity of ahout twice that of bulk molybdenum. The lattice parameters measured in a direction nornzal to the surface of the films weve found to be gveatev than the bulk value. This was interpreted as being at least partly due to the presence of compressive stresses. The effects of annealing in an Ar-H atmosphere were studied in terms of diffraction line width, lattice parameter, and resistivity. BECAUSE of its relatively low bulk resistivity (5.6 x 106 ohm-cm)&apos; molybdenum is potentially interesting as a thin-film conductor in integrated circuits. An additional feature which makes it attractive for this purpose is its low coefficient of expansion (5.6 x KT6 per "c),&apos; which is fairly well matched to that of silicon (3.2 x 10 per c). It is possible to deposit molybdenum films by evaporation but generally films produced in this manner have a high resistivity. In order to achieve resistivities close to bulk value, Holmwood and Glang found it necessary to operate in a vacuum of about 107 Torr and to maintain the substrates at 600 C during film deposition. Sputtered molybdenum films have been examined by Belser et a1.7 and, recently, by Glang et al.&apos; This paper describes the results of an attempt to extend some of that work and examine the effects of annealing and getter sputtering on the physical and structural properties of the films produced. SPUTTERING APPARATUS AND PROCEDURE The apparatus used for most of the film sputtering work described here consisted of two "fingers" serving as anode and cathode, respectively, which were mounted within an 18-in.-diam glass chamber. A liquid nitrogen-trapped 6-in. diffusion-pump system was used to achieve a vacuum of about 1 x 107 Torr within the chamber prior to sputtering. The essential features of the equipment are shown in Fig. 1. Cathode and anode fingers are stainless-steel tubes isolated from the top and bottom plates by Teflon collars. In order to limit the discharge to the space between anode and cathode, each finger is surrounded by an aluminum hield, at ground potential, having an internal diameter 18 in. larger than the outside diameter of the finger. The cathode and anode fingers are 6 and 4 in. in diam, respectively. A 116-in.-thick sheet of molybdenum is brazed with a 10 pct Pd, 58 pct Ag, 32 pct Cu alloy to a copper disc which is mounted by means of screws and a large 0 ring onto the lower end of the cathode finger. The disc is cooled during sputtering by water circulation inside the finger. The use of several feet of plastic tubing for the water input and outputg reduces leakage to ground to less than 1 ma when the cathode potential is raised to 5 kv. The upper end of the anode finger is terminated by a brazed-on copper block. A variety of specimen holders can be easily mounted on the upper face of this block. Substrate heating or cooling is achieved by use of an appropriate unit attached to the lower face of the same block. Heating is achieved by means of cartridge-type heaters and cooling by copper coils fed with forming gas under pressure. The inner chamber of the specimen finger constitutes a small vacuum chamber of its own which is evacuated by an auxiliary mechanical pump in order to limit heating element oxidation and heat transfer by convection currents. An advantage of the finger arrangement is the absence of cooling and heating coils and wires within the main chamber. The stain less-steel shutter is useful to establish a discharge for cleaning the cathode at the beginning of each sputtering run. Water cooling of the shutter reduces heating and the out-gassing of impurities which might condense on the nearby substrates. Unless otherwise specified, the substrates used in these experiments were 1-in.-diam oxidized silicon wafe:s, 0.007 in. thick, having an oxide thickness of 6000A. The substrate holders were large copper discs onto the surface of which a number of molybdenum discs, 116 in. thick and 78 in. in diam, were brazed. The wafers were clamped to the molybdenum discs

Jan 1, 1967

By Kurt Kennedy

A method is described by which three elements can be condensed simirltaneously on a common substrate in such a way that the composition varies with position on the substrate. Almost all possible combinations of three solid elements can be formed in a single evaporation. The three elements are heated by magnetically deflected electron beams. The alloys and cornpounds formed by this method generally correspond to those trade by other methods; however, many exceptions exist. Simple symmetric crystallographic phases tend to form in preference to more complex phases. The Nb-Sn. Cr-Fe-Nt, and other systems of alloys made by this method will be discussed. Phases that form very slowly, such as the o phase in Cr-Fe alloys, may be formed rapidly by this method when certain substrate temperatures are used. STUDYING all the mixtures and compounds of two or more elements by the preparation of individual samples forces the investigator to make sampling sufficiently close together so that the properties of the intermediate compositions can be inferred. A completely unknown system will require a dis c our aging ly large number of preparations in order to fully know all its properties. For the purpose of expediting this type of investigation, methods have been devised to prepare samples that have a continuous variation in composition over some predefined area. The code-position of two elements onto suitable substrates has been used to produce alloys or compounds that have variable compositions depending on their positions on the substrates Composition gradients have been formed by depositing a wedge-shaped layer of one element and then depositing a second wedge-shaped layer on top of it such that the total film thickness is kept constant. Concentration equilibrium perpendicular to the surface is then established by annealing.4 This paper describes a method for the simultaneous condensation of three volatile components on a substrate such that there is a regular variation of composition with position on the substrate. The arrangement of the evaporation sources and the substrate is shown schematically in Fig. 1. Materials to be evaporated are placed in three water-cooled crucibles located at the corners of a 10-in. equilateral triangle. Electron beams from cathodes located below the crucibles are bent 180 deg by magnetic fields and focused on the material to be evaporated. The cathodes are connected in parallel to a constant-current 12-kv power supply. The distribution of emission current from each cathode is controlled by filament temperature. When materials that sublime are to be evaporated, the electron beam is deflected by an auxiliary ac magnetic field so that the beam moves over the surface of the evaporating material rather than digging a hole in one spot. The substrate is a 10-in. triangle of 2-mil stainless steel or molybdenum which is stretched against a slightly curved heat sink by a frame as shown in Fig. 2. This frame and heat sink are located so that each corner is directly above an evaporation source. The heat sink has three radiant heaters at each corner that can be separately controlled permitting the

Jan 1, 1967

By C. W. Mueller, P. H. Robinson

A technique was developed for depositing single -crystal films of silicon on single-crystal sapphire substrates via the pyrolytic decomposition of SiH4/H2 mixtures. Electron diffraction and X-ray Laue reflection examinations of these films revealed single-crystal patterns. These films were characterized by measuring conductivity type, thickness, resistivity, and Hall mobility. Chemical etching, dislocation staining, and electron-microscopy examination have indicated the presence of low-angle grain boundaries and microtwinning in the silicon films. The role of the sapphire substrate with respect to its quality, orientation, surface finish, and thermal and chemical treatment was investigated. Hall mobility as a function of sapphire orientation was measured. A mobility equal to 88 pct of bulk silicon mobility was attained. Insulated-gate field-effect transistors which have a transconductance of 4000 pmho at 6 ma were fabricated using these films. THE technology of the fabrication of microelectronic circuits with active and passive components directly on a single wafer of semiconductor material has attained a high degree of perfection during the last few years. One inherent difficulty with the single-wafer approach has been undesirable coupling between components arising from the common base material. Several techniques have been developed for minimizing this coupling and obtaining good electrical isolation between components; however, these result in extra capacitance being added to the circuit. The extraneous capacitance of the connecting leads, resistors, and capacitors degrade circuit performance and limit the high-frequency operation of these circuits. Recently there has been considerable interestl-3 in the deposition of silicon films upon insulating substrates. The use of silicon as the semiconductor allows the subsequent utilization of the large amount of technology which has been developed for silicon for the fabrication of good active components. A sapphire substrate supplies an excellent insulating material which at the same time gives a thermal-conductivity improvement of an order of magnitude over glass-type substrates. This paper will describe the results obtained for the deposition of single-crystal silicon layers upon single-crystal sapphire substrates as well as the characterization of these films. I) APPARATUS The &apos;system used for the growth of silicon films on sapphire substrates by means of the pyrolysis of SiH4/H2 mixtures is shown in Fig. 1. The all-quartz reaction chamber is water-cooled and includes a quartz optical-flat window for use in conjunction with a 5 kw, 2.5 Mc per sec constant-temperature controlled rf generator. The temperature sensor for the rf generator is a thermopile which is focused through the quartz optical flat onto the susceptor surface. With this system the rf generator will hold the temperature of the susceptor to about +0.3°C at 1200°C. The susceptor is supported on a quartz sled inclined to the gas flow to increase the uniformity of the deposition. The water jacket surrounding the reaction chamber is used to keep the quartz wall temperature low enough to limit the decomposition of the silane to the rf-heated susceptor which contains the sapphire substrates. This eliminates the reaction of silicon with the hot quartz to form SiO which deposits out on the apparatus walls, thus obscuring vision, and interferes with the optical measurements of the temperature as well as with the growth process itself. Provision is also made for "n"- and &apos;p"-type doping by using mixtures of either phosphine or diborane in hydrogen which are added to the silane during the growth process. The hydrogen used is the purest commercially available tank hydrogen further purified by use of a palladium diffuser. The susceptors which were used include 1) spectro-scopically pure graphite, 2) high-purity molybdenum, 3) high-resistivity silicon using a low-resistivity insert to facilitate coupling to the rf generator, 4) silicon carbide-coated graphite. The latter two susceptor materials produced silicon films on sapphire substrates with more reproducible resistivities. Prior to silicon deposition all connecting lines as well as the reaction chamber are thoroughly flushed with argon and then hydrogen. After the hydrogen flow is set, the growth temperature is adjusted to 11009 to 1150°C, and when thermal equilibrium is reached the silane flow is started. The flow rate of the hydrogen must be fast enough so that turbulent flow occurs in the size of reaction tube used, otherwise a streaked nonuniform silicon deposit occurs. After the deposition

Jan 1, 1967

By Eugene S. Meieran

The effects of methods of crystal growing, wafer sawing, polishing, routine handling, diffusion, and epitaxial growth on the defects in GaAs are reviewed and studied using reflection and transmission X-ray topographic techniques. In general, it was found that boat-grown crystals exhibited fewer defects than Czochralski crystals, although all crystals showed large numbers of precipitates visible when examined in the electron microscope. Mechanical surface treatments such as sawing and mechanical polishing introduce damage to a depth of about 5 µ, most of which can be removed by suitable chemical or chem-mechanical polishing. In addition, defects can be introduced through routine handling of wafers, for example with metallic tweezers. These defects can be quite severe, and have been observed 20 µ below the wafer surface. Defects can also be introduced through diffusion and epitaxial growth. These defects, which include precipitates, growth pyramids, stacking faults, dislocations, and so forth, can be detrimental to device fabrication. It is shown that wafers or films which appear defect-free optically can contain defects visible in the X-ray topographs. WHILE the use of GaAs in the semiconductor industry has increased very rapidly in the last few years, due mainly to the recent development of many important GaAs devices,1,2 the major limit to the production of commercial quantities of many GaAs devices remains a severe lack of suitable materials technology. This lack is apparent in two critical areas. First, production quantities of high-quality GaAs crystals, reproducibly doped and precipitate-free, simply are not available commercially, although some reasonable quality material is available on a limited first-come, first-serve basis. Second, in comparison to silicon technology, little is known about the effects of processing variables on the defects either present in as-grown GaAs or introduced through processing and handling of wafers. These areas are now receiving some attention from semiconductor device manufacturers, who are studying defects in GaAs in order to better understand how either to prevent their occurrence or to cope with their existence. Most investigations of the defects in GaAs have been made by optical microscopy3-5 or transmission electron microscopy techniques.&apos;-&apos; Recently, however, the imaging techniques of X-ray topography, electron mi-croprobe analysis, and scanning electron microscopy are being applied to the study of GaAs.9-14 In the case of X-ray topography, a one-to-one image is obtained that must be photographically enlarged. In compensa- tion, the defects within entire wafers may be imaged by simple scanning (Lang technique15) if the wafer is reasonably perfect, or by using the scan oscillation technique developed by Schwuttke16 if the wafer is warped or distorted. The purpose of this paper is to both review and extend the general application of X-ray topographic techniques to GaAs. Emphasis will be placed on the effects of growth and process variables on the quality and perfection of both bulk and epitaxial GaAs. Reference to optical or electron microscopy results will be made when useful. Since the effects on defects of a wide variety of processing variables such as crystal growing, sawing, polishing, diffusion, and epitaxial growth will be somewhat superficially reviewed, a fairly extensive bibliography of the most important recent results in these areas is included. However, for completeness, important defects will be illustrated here, although such defects have been previously shown by others. While this paper is concerned with defects rather than with the physics of X-ray scattering, the mechanisms of contrast formation in the topographs will of necessity be briefly mentioned. EXPERIMENTAL GaAs crystals, both boat-grown18 and Czochralski-grown,&apos;8 containing a variety of dopants of various concentrations, were purchased from outside vendors. Wafers were sliced from the crystals using a Hamco ID saw and were mechanically polished using 1 µ diamond paste. Chem-mechanical polishing was done in bromine-methanol as described by Sullivan and Kolb.18 Chemical polishing was done using a modified sulfuric-peroxide solution, 11 parts H2SO4, 1 part 30 pct H2O2, 1 part DI water.5 Zinc diffusion was carried out in a closed tube, using a 10 pct Zn-In source at 825°C for 1 hr. Oxide masking techniques were used to select the area to be diffused. Epitaxial wafers were either purchased or prepared here. All epitaxial runs prepared here were carried out using a Ga-GaAs-AsC13 source in a closed tube at a substrate temperature of 750°C. Wafers were chem-mechanically polished and gas-etched prior to deposition. The X-ray topographs were taken on a Krystallos Lang camera, operating in the transmission scanning geometry (Lang technique15) or in the reflection scanning geometry (modified Berg-Barrett technique20,21). MoKa, radiation was used for all transmission topographs using a Jarrell-Ash 100-µ spot focus. CuKal radiation was used for all reflection topographs using a General Electric CA-7 1-mm spot focus X- ray tube. Topographs were printed from an intermediate contrast inversion film, so the contrast shown in all figures here is the same as that of the original 50-µ-thick emulsion L4 Iiford nuclear plate used to record the topograph.

Jan 1, 1969

By J. M. Eldridge, P. Balk

Phosphosilicate glass films were formed, by reacting gaseous P2O5 with SiO2, over a large range of temperature (800° to 1200°C) and gas phase composition (nearly two orders of magnitude of effective P2Ospressure). The film compositions generally corresponded with the liquidus curve, delineating the maximum solubility of the tridymite Phase of SiO 2 in phosphosilicate liquid solution at the temperature of film formation. It is shown that the P2O5 concentration of the phosphosilicate liquid film tends to decrease by reaction with the underlying SiO 2 layer until the liquidus curve is reached. The validity of the thermodynamic argument used in this explanation is supported by the results of a determination of the composition of borosili-cute films, prepared by reacting gaseous B2O3 with SiO2 at different temperatures. The kinetics of phosphosilicate film formation were described by a model predicated on a steady-state diffusion of P2O5 through the film. UNDERSTANDING of the processes leading to formation of phosphosilicate and borosilicate glasses is of great importance for producing passivating layers on FET devices. Passivating films with optimum characteristics are preferably formed in a separate step, independent of the doping of the semiconductor.&apos; The results of an investigation carried out to gain improved insight into the mechanism of glass formation are presented in this paper. It would be expected that application of the known Pz05-Si02 and B 2 O 3-SiO2 phase diagrams should be useful in extending understanding of the glass-forming processes. However, the question of the propriety of treating thermally grown SiO2 in these binary oxide systems by the methods of equilibrium thermodynamics must be considered when this application is attempted. Although Sah et a1.&apos; and Allen et al. 3 investigated the kinetics of formation of phosphosilicate glass (PSG), they failed to adequately relate their diffusion models to the occurrence of experimentally observed phases in the PSG/SiO 2/Si system. Horuichi and yamaguchi4 investigated the diffusion of boron through an oxide layer and described their results in terms of a model similar to that of Sah and coworkers. More recently, Kooi 5 and Snow and Deal6 reported the compositions of PSG films formed by depositing P2 O 5 onto SiO2. These compositions apparently coincide with those at the liquidus curve delineating the maximum solubility of crystalline SiO2 in phosphosilicate liquid solutions. These authors did not discuss the thermodynamic implications of their results on the structure of thermally grown SiO2 films. The structure of thermally grown Sio2 films and that of vitreous silica are generally thought to be quite similar. Since the solubility of a substance depends on its structure, it is relevant that the solubility of vitreous silica in water7 is highly reproducible, like the solubility of thermally grown SiOz in phosphosilicate liquid. Furthermore, the vitreous silica-water system appears to be in true thermodynamic equilibrium (viz., the same solubility value can be approached from both supersaturated and under-saturated solutions). Sosman7 suggested that a type of two-dimensional lattice may form at the silica/solution interface, resulting in the observed solubility behavior that is characteristic of a crystalline solid. An alternative explanation may be that vitreous silica has a microcrystalline grain structure. Other investigators have suggested that vitreous silica has essentially the structure of B cristobalite,&apos; or is composed of microcrystals of p tridymite or cristobalite, or a mixture of both. Presumably the grain size would be sufficiently large to minimize any appreciable contribution of the grain boundaries to the solubility of the crystalline matrix. The present investigation was carried out to clarify the significance of the boundaries in the Pa,-SiO, and B2O3-SiO2 Systems in determining PSG and BSG (borosilicate) film compositions. Furthermore, the kinetic data for PSG film formation were extended, using a wider range of formation parameters than were previously reported. One model describing the kinetics of film formation will be presented that is compatible with the thermodynamics of the Pa5-Si02 system. EXPERIMENTAL PROCEDURE Glass Film Preparation. SiO2 films (1000 to 8000A thick) were obtained by oxidation of silicon substrates in dry O2 at 1100°C. PSG and BSG films were prepared by exposing these layers to gaseous oxides obtained by reacting high-purity POC13 and BBr3, respectively, with O2. A double-columned saturator was used to ensure complete saturation of the N 2 carrier

Jan 1, 1969

By Robert C. Linares

Cubic zinc sulfide has been grown from molten salt solutions substantially below the hexagonal-cubic phase transition of 1020°C. Crystals free of birefringence have been grown from molten lead chloride at temperatures from 500° to 800°C. Factors affecting the optical transmission and the birefringence are discussed. CUBIC zinc sulfide is of interest for electrooptic devices and as an optical coating material. However, it cannot readily be produced as large crystals which are free of birefringence. The birefringence arises from the fact that the cubic ZnS is usually grown as the high-temperature hexagonal phase and is transformed to the lower cubic phase by annealing.&apos; This results in stacking faults, stress, and birefringence. A requirement for the growth of cubic ZnS is that it be grown below the phase transition of 1020°C.2 Crystals of cubic ZnS have been grown below 1020°C by the hydrothermal,3 flux,4 and vapor transport5 techniques. However, these all appear to have limitations in size or quality of material produced. It was felt that a new flux method might have a good chance of overcoming some of the previous problems. Mita6 reported the growth of hexagonal ZnS from molten NaCl and wilke7 reported the growth of iron sulfides from molten lead chloride solution, indicating that molten chlorides make good solvents for sulfides. An investigation of a number of possible molten chlorides as solvents showed that molten PbCl2 has an appreciable solubility for zinc sulfide and that crystals can be grown at temperatures from 500° to 800°C. EXPERIMENTAL Crystal growth was performed in open or sealed Vycor or fused silica tubes, in an argon atmosphere or vacuum. There were no deleterious effects due to using vacuum; however, a sealed essentially isothermal system must be used to prevent vapor transport mechanisms from dominating. This will be discussed later in detail. Noble metal containers were not usable as they were severely attacked by these melts. The materials used were: reagent-grade PbC12, reagent-grade ZnS, optical-coating-grade ZnS, and Irtran 2 (ZnS). Crystal growth was accomplished by both slow cooling techniques and gradient growth techniques. In the slow cooling techniques, 15 g of ZnS and 100 g of PbCl2 were used. The components were placed in the silica tube, heated at 800°C for 2 hr, and then slowly cooled (l° to 10° C per hr) to 500°C. The furnace was constructed so that the tube could be inverted, thus allowing the liquid to drain off the crystals at the end of the run. The gradient runs were made by placing the above mixtures in a silica tube at 550° to 600°c, Fig. 1. A temperature gradient was maintained such that the floating mass of undissolved sulfide was 10° to 50°C warmer than the bottom of the tube. Growth occurred on spontaneously nucleated sites at the cooler bottom. RESULTS When using high cooling rates (5° to 10°C per hr), crystals form in thin sheets perpendicular to the {111} face. Crystals 5 by 5 by 1.2 mm, Fig. 2, have been ob-

Jan 1, 1969

By H. M. Manasevit, R. L. Nolder, L. A. Moudy

The observation that silicon-on-sapphire crystallo-graphic relationships exist which are unrelated to one another implies that there are regions of sapphire orientations containing these relationships which are separated by boundaries. The presence of these boundaries is demonstrated, initially on rounded cylindrical surfaces md more definitively on flat sapphire substrates with orientations along primary crystallograPhic zones. By means of reflection electron and X-ray diffraction techniques the crystallographic relationships have been studied. Two new relationships have been found, viz., (111) Si (1014) sapphire and (001)Si (1012) sapphire. Also found were multicrystalline relationships. The approximated boundaries between the crystallographic regions for each relationship are shown by means of a stereo-graphic projection. These studies also clarify discrepancies in the literature concerning the silicon-sapphire relationships on the (1120) sapphire substrate. SINCE the first report of the deposition of single-crystal silicon on sapphire in 1963&apos; this epitaxial system has been the object of study by many semiconductor laboratories. In particular, there has been strong interest in the crystallographic relationships between the two materials. In 19652,3 we demonstrated that at least four distinct modes of orientation relationships can occur, namely, (100)Si (0112) sapphire, (111)Si 11 (1124) sapphire, (111)Si (1120) sapphire, and (111)Si (0001) sapphire. Each relationship could be obtained when the substrate plane was at the particular orientation indicated. Other workers as well as our laboratory concurred in general with the initial results. However, two discrepancies were observed. First it was observed by Joyce et a1.4 that on a (1120) substrate a (110) silicon deposit could be obtained. This was actually consistent with the (111)Si (1124) sapphire mode. Second, two other laboratories,5,6 as well as our own,7 have found a deposition on the basal plane which is different than our initial findings. Even though a (111)Si deposit was obtained, the crystallographic relationships were not identical to those previously found. Subsequent work in our laboratories has confirmed that the form found in these later investigations of the basal plane is by far the more predominant, and the original finding was an anomaly. The relationships which have been previously estab1.ished are summarized in Table I. It is seen that at present a rather small number of possible substrate orientations has been investigated. Therefore, it seemed appropriate that a systematic investigation of silicon epitaxy on various sapphire substrate orientations would be of value. It was expected that the answers to three questions could be obtained from such an investigation. These are as follows: 1) Since we know of at least four distinct modes of orientation relationships, over what range of substrate orientation do these modes occur and what is the form of the boundary expected between the different regions in which each mode occurs? 2) Are there other modes which occur on substrate orientations not previously investigated? 3) What is the explanation for the discrepancy between our laboratory&apos;s findings and those results reported by Joyce et al.4 concerning the relationships occurring on (1120) substrates? APPROACHES AND EXPERIMENTAL METHODS Two different approaches were used. Initially, depositions were carried out on rounded cylindrical surfaces which lay perpendicular to the (0112) plane.* By this means the presence of boundaries and their approximate positions could be determined. Finally and more definitively, depositions on flat sapphire substrates with a variety of orientations, mostly along primary crystallographic zones and about the (1120) plane, were studied by means of reflection electron diffraction and X-ray diffraction (using back-reflection Laue patterns and full circle goniometer techniques) in order to determine crystallographic relationships. The sapphire orientations investigated by these means

Jan 1, 1969

By S. Razi

Injection luminescence of single crystals of rare-earth-activated CdS was investigated. Crystals of CdS:Nd, CdS:Er, and CdS:Yb, grown by the vapor transport technique, were suitably etched and hetero-junctions of Cu2S-CdS were fabricated on one side of the CdS substrate by chemical reaction. The device was then heated to 200°C in an inert atmosphere for 5 min. Static V-I measurements show that at low voltages (V < 1 v) the current is controlled by the junction, while at higher voltages the bulk Properties of the CdS substrate effectively control the current. Because of the large thicknesses no negative resistance was found. Reverse breakdown voltage varied from 5 v, for junctions that were not heat-treated, to 100 v, for diodes that received the proper heat treatment. The rectification ratio, measured in the range of 2 to 10 v, varied from 300 to 3000. Capacitance- voltage measurements indicate a weak dependence of the junction capacity on applied voltage. The capacity was found to increase appreciably with room light illumination. The variation of capacitance with frequency is indicative of a p+-i-n structure. An open-circuit voltage of 0.4 v was measured under an illumination of 100 mw per sq cm. Under forward bias, the bluish-green edge emission characteristic of CdS was observed. No sharp rare earth emission line was observed, which is in contrast to our results using ZnS:Tb, where, under high-voltage excitation, the rare earth emission was dominat over the lattice emission. The absence of rare earth emission implies that either the nature of the rare earth ion is different at the het-erojunction than in the bulk or the probability of energy transfer to the rare earth ion is small. RARE earth elements are playing an increasingly important role in the Synthesis of materials for a variety of purposes, e.g., solid-state lasers, color television screens, and so forth. Interesting systems result when wide-band-gap photoconducting lattices are activated with rare earths. A study of the lumines-

Jan 1, 1969

By T. H. Ramsey, Richard Shield

A process has been developed for fabricating thin-film resistors that are compatible with integrated circuits. Films are produced by evaporation of single refractory metals using an electron beam heating source. Resistor paths are then made using standard photoresist etching techniques. A low-temperature coefficient of resistance is obtained by control of nucleation during deposition of the film. One application of the process has been the development of a discrete resistor used for current setting in linear integrated circuits. Multiple thin-film resistors were formed on oxidized silicon substrates using molybdenum metal as a source. An evaporated protective layer of SiO2 over the resistor prevents resistance changes which might occur during subsequent high-temperature processing. Properties of a typical resistor are as follows: R = 15.7 ohm per sq, tolerance surface dimensions = 0.028 by 0.028 in., stable after 4000 hr at 0. 75 w without heat sink. THIN-FILM resistors have been investigated for many years with a large selection of usable resistors resulting from a wide variety of materials. Many of the processes necessitate a high degree of complexity but in nearly all cases the conclusions have been that the materials used in each case were acceptable provided the deposition process was reproducible and could be controlled sufficiently to give the required tolerances. It is our opinion that simplicity is still the key to thin-film resistor fabrication. In this regard, the single metal is obviously the simplest process to control, provided all the necessary properties can be obtained. The idea is not original in this laboratory by any means. Others have taken this approach.&apos; It has been demonstrated that a fairly wide range of resistivities can be obtained by careful control of deposition parameters for any given metal, and by selection of several single metals a very broad range of stable resistors can be obtained. The transition from one metal to the next is made when a film becomes too thin for adequate process control or long-term stability. A great deal of emphasis is once again being placed on the thin-film resistor. Although they have been thoroughly studied in the past, new applications have been found that require the properties obtainable only from thin-film resistors. Integrated circuits, in general, are advancing toward higher speeds and radiation environments which require replacement of the present diffused resistors. Linear circuits in particular are requiring high-power, high-precision resistors for the more advanced applications. The particular requirements of some linear circuits are such that a discrete resistor is required. The resistor described in this paper was developed as a 15.7-ohm precision discrete resistor for use in a current regulating circuit requiring a +2 pct end of life overall tolerance, a dissipation of 750 mw, and the ability to operate over a fairly broad temperature range (i.e., a low temperature coefficient of resistance, TCR). State-of-the-art integrated circuit and silicon processing technology lends itself well to the fabrication of thin-film resistors. Not only is the resistor thin film well-suited to integration on individual circuits, but the oxidized silicon surface provides one of

Jan 1, 1969

By E. H. Snow, B. E. Deal

Instability effects in semicanductor devices have long been attributed to the motion of charges on or within oxide layers on the surface. These effects are of critical importance in metal-insulator-semiconductor MIS) field-effect devices. For this reason, the capacitance-voltage or the conductance-voltage characteristic of these devices can be used as a sensitive detector to study charge transport and polarization effects in the oxide or insulator. A review is given of the results of such studies of thermally grown SiO2 as well as of other insulating films of importance in silicon technology. Three types of effects are distinguished. The first of these is the drift of mobile cations within the dielectric, examples being thermally grown SiO, which is often contaminated with sodium ions, and a variety of other glasses in which the mobile ions are a part of the glass composition. The second effect is a dipole-type polarization which occurs in phosphosilicate glass films obtained by reacting P2O5 with thermally grown SiO2. The third effect involves the transfer of charges between the dielectric and the silicon electrode. This occurs in silicon nitride and other deposited dielectrics. It is concluded that MIS studies have provided a powerful technique joy the study of charge transport and polarization effects in insulating films. The knowledge gained from these studies has led to an understanding of surface effects on conventional transistors and diodes as well as making possible stable MIS transistors. THE metal-insulator-semiconductor field-effect transistor is conceptually the oldest type of active semiconductor device.&apos; The earliest attempts at making this device were frustrated because of high surface state densities at the interface between the semiconductor and the gate insulation.&apos; However, by using a silicon substrate with thermally produced silicon dioxide as the gate insulation, this problem was solved and metal-silicon dioxide-silicon devices with good characteristics were made as early as 1960. 3 Yet it was still over 5 years before these devices became a commercial reality. This delay was largly due, not to surface states, but to stability problems associated with polarization effects within the insulating layer which caused the threshold voltage of the device to drift under temperature and bias treatments. The solution to these problems has not only made possible stable MIS devices, but it has added immensely to our understanding of failure mechanisms in conventional bipolar transistors and has added to the reliability of ali types of planar devices. In this review, we shall first describe the effects of various types of polarization phenomena on MIS device characteristics. Then, since thermally grown SiO, is by far the most important insulator used in these devices, we shall review historically the type of instabilities which have been observed in thermal oxides, the attempts at understanding and eliminating them, and the present status of the problem. We shall then turn our attention to the various deposited insulators which have been used, including lead glasses, phosphosilicate glass, vapor-deposited silicon oxide, and silicon nitride. Interestingly enough, many of these materials show polarization effects which are quite different from those generally observed in thermally grown SiO2. THE EFFECTS OF POLARIZATION PHENOMENA ON MIS CHARACTERISTICS The simplest MIS device and the one which has been most frequently used in the study of polarization effects is the MIS capacitor. Two modifications of this structure with single- and double-layer dielectrics are illustrated in Figs. l(a) and (b), respectively. The capacitance of this structure as a function of voltage applied to the metal gate electrode is plotted in Fig. 2 for the case of an n-type silicon substrate. When the silicon surface is accumulated (positive bias) the measured capacitance is just that of the insulating layer C. When the surface is inverted (negative bias), the capacitance is that of the insulator and a silicon depletion layer in series CoCs/(Co + Cs). Indicated on the horizontal axis of Fig. 2 is the voltage VT at which the silicon surface becomes strongly inverted. This voltage corresponds to the threshold or turn-on

Jan 1, 1969

By Billy L. Crowder, John M. Fairfield

The implantation of B+ , P+, and As&apos; into silicon has been studied with the purpose of making shallow p-n junctions. The influence of such parameters as 1) ion energy, 2) target orientation and temperature, 3) total dose, and 4) annealing schedule was investigated. An energy range of 70 to 300 kev was used for boron and phosphorus implants and up to 500 kev for arsenic. It is found that the experimental projected range agrees well with theory and that shallow junction depths can be made reproducibly. ION implantation has received much attention recently as a technique for doping semiconductors. Specifically, it has the potential of supplementing or replacing the diffusion process as a method for making p-n junctions. In a few specific cases it has been used successfully to make semiconductor junction devices. Potential advantages of ion implantation doping over diffusion techniques are: 1) It affords greater control of shallow junction depths (< 0.2 µ) while maintaining high peak concentrations. This is particularly important for high-speed switching devices, since lower junction capacitances and resistances can be achieved. 2) More precise registration of small planar structures can be realized if proper masking procedures are employed. This advantage is especially useful in the design of high-density integrated circuits. It has been used to advantage in FET fabrication since the edge of the source or drain can be aligned precisely at the edge of the gate electrode.&apos; 3) Ion implanatation permits lower temperatures than diffusion techniques. This factor alleviates the problem of compatibility of diffusivities often encountered when designing multiple-junction structures. Also, the lower temperatures create fewer thermal defects and dislocations, which may account for the high efficiency of some ion-implanted solar cells.2 4) Impurity profiles can be more easily tailored to resemble ideal distributions. Successful exploitation of the potential advantages of ion implantation techniques will depend on increased knowledge and understanding of the subject. The factors likely to be influential in determining impurity distribution profiles in ion-implanted single-crystal targets have been reviewed by J. F. Gibbons.3 In addition to the mass and energy of the implanted ion, the total dose, target orientation, and target temperature are important parameters. The annealing temperature required for removing lattice damage and incorporating the implanted species on an electrically active site is very important. This paper describes an investigation of some of these factors. Implants of boron, phosphorus, and arsenic into silicon have been studied. Energy ranges of 50 to 300 kev were used for boron and phosphorus and up to 500 kev for arsenic. In addition to the implantation energy, the effects of total dose, target temperature, and post implant anneal have been investigated. EXPERIMENTAL PROCEDURE The implantation targets were silicon wafers cut from Czochralski-grown crystals, lapped, and chemically polished. The orientations were (111), (110). and (100) with misorientations of up to 7 deg from the principal axis. For this study, accurate target alignment (i.e., within 0.1 deg) was not available and quoted misorientation values should be regarded as approximate . The implantation equipment consisted of an ion source, a 300-kev linear accelerator tube, an electromagnetic separator, and the associated target supporting and beam focusing assemblies. The ion source was a simple oscillating electron type source,4 which has been described elsewhere.5 The gaseous compounds BF3, PF5, and AsH3 were used as ion sources for B+, P+, As+, and AS+&apos;. Analyzed current levels of up to 20 pamp could be obtained; however, for this investigation target current levels of 1-3 µ amp were usually employed. The analyzed ion beam was collimated through a double slit (1.4 x 0.4 cm) and swept perpendicularly to the long axis of the slit such that an area of about 2 sq cm on each target was covered. Dosages of around 1015 cm-2 were normally employed, but smaller amounts were also used for comparison. A uniform flux density over the bombarded area was assured by the continuous use of profile monitors similar to those described by Wegner and Feigenbaum.6 Post-implant annealing was accomplished in an argon atmosphere in a temperature range of 600" to 950°C. It was not part of the purpose of this investigation to study the annealing kinetics; however, some isochronal and isothermal anneal experiments were conducted to determine the time and temperature necessary to render a reasonably high portion of the implanted ions electrically active (i.e., higher than 50 pct). Post-implant anneal temperatures of around 900° and 600°C were required for boron, and arsenic and phosphorus implants, respectively. Arsenic and phosphorus implants increased in conductivity rather abruptly at the proper anneal temperature of the isochronal curve, but boron increased more gradually over a wider range. Isothermal anneal curves were reasonably flat after 10 min, so an anneal time of 1/2 hr was used for the experimental results described below. The profiling techniques were: 1) neutron activation analysis, 2) differential sheet resistance,7 and 3) junction staining.8 The differential sheet resistance technique is commonly employed in this type of study. Its principal disadvantage is the uncertainty of the ef-

Jan 1, 1970

By Richard O. Carlson

Ions of p31 and B 11 were implanted in natural insulating diamond macles. The thin (-0.4µ) layers showed sheet resistances of 107 to 1011 ohm per sq and activation energies of 0.17 to 0.34 ev above room temperature. However, no Hall effect could be measured, indicating that mobilities were less than I to 10 sq cm per (v-sec). Such low mobilities may be due to excessive scattering due to radiation damage in the bombarded layer or to very high concentrations of active compensating impurities. Annealing did not materially change the room -temperature properties, but the temperature was limited to -800°C by surface graphitization, the latter process probably being accelerated by lattice damage in our diamonds arising from the ion bombardment. A weak n-type thermoelectric power was detected after p31 irradiation, but it cannot be presumed that we have made the phosphorus ions active as donors in diamond. The natural diamonds have a high and uncontrolled concentration of impurities and, when coupled with the radiation damage and graphitization problems, would appear to seriously limit the quality of semiconductor that we can presently achieve by ion implantation in diamond. ALTHOUGH diamonds are usually thought of as insulating in terms of their electrical conductivity, it was found about 15 years ago that natural semiconducting diamonds do occur rarely, and these were designated as type IIb diamonds. These p-type semiconducting diamonds were found to be dominated by an impurity level 0.37 ev from the valence band. Evidence today based on correlation of the concentration of acceptor states from Hall effect measurements with the impurity concentrations determined by neutron activation analysis point to aluminum as the dominant acceptor impurity.&apos; The compensating donor is believed to be nitrogen, which has a donor level 1.6 ev above the valence band.&apos; However, only a small fraction of the total nitrogen content in the diamond is electrically active. An infrared absorption band at 7.8 p and ultraviolet absorption near 4 ev have been associated with nitrogen, the former providing a quantitative measure of nitrogen content.3 The nitrogen content is -l020 cm-3 in insulating or type I diamonds, but is less than 10&apos;&apos; in natural semiconducting diamonds. Much of this nitrogen is distributed in platelets oriented in (100) planes and not atomically dispersed in the diamond lattice.4 Semiconducting diamonds have been deliberately formed by incorporating impurities into the graphite charge in the high-pressure apparatus used to form diamonds.5 These manufactured semiconducting diamonds are always p-type, and while some show the 0.37 ev level due to aluminum, most such samples have been ascribed to impurity banding effects.6 The best natural or manufactured semiconducting diamonds have hole mobilities near 1500 sq cm per (v-sec) at room temperature, and those with the 0.37-ev level can be analyzed to reveal an acceptor concentration of 3 to 8 X 1016 cm-3 and a donor concentration 3 to 10 times lower.&apos; NO bulk n-type diamonds have ever been reported, but the electron mobility has been measured as 1900 sq cm per (v-sec) by irradiating a diamond with ultraviolet to excite electrons out of the deep nitrogen donor level or other levels.7 New hope for the formation of n-type diamond has emerged from the ion-implantation method whereby desired impurities are introduced into a crystal lattice by bombardment with a high-energy beam of the impurity ion. It was found in the case of silicon that the usual donors such as phosphorus and arsenic and acceptors such as boron and gallium could be implanted into silicon. Wentorf and arrow&apos; produced semiconducting layers on diamonds by an ionic bombardment in a glow discharge at potentials of about 2 kv. The observed typeness from thermoelectric probing seemed to depend on the atmosphere gas (nitrogen, argon, or hydrogen) rather than on the electrode material, but the nature of the conduction process in the thin damaged surface layer is completely unknown. A Russian group under Vavilov has attempted high-energy ion implantation in natural diamonds using boron and lithium ions9 and later also phosphorus, aluminum, and carbon ions.10 Their papers claim n-type layers from lithium, phosphorus, and carbon implant and a p-type layer from boron and aluminum implant, though the methods of type determination are not described in detail. Under government contract support, the Ion Physics Corp. has studied ion implantation in several semiconductors. While the bulk of their study1&apos; was devoted to the irradiation of silicon, they did carry out a short study on boron and phosphorus implantation into natural diamonds. They did observe surface conducting layers but did not determine the typeness of the layers. For our experiments, phosphorus and boron were chosen as the dopant ions because of their respective donor and acceptor behavior in germanium and silicon. Moreover, they are the lightest mass dopants of the shallow level donors and acceptors from columns 111 and V of the Periodic Table (excluding nitrogen which is already present in diamond), and will have the largest depth penetration into the diamond lattice. EXPERIMENTAL Diamonds. As the target diamonds for our ion implantation study, we chose commercially available macles, which are flat, twinned diamond crystals. By scanning through the stock of a wholesale distributor

Jan 1, 1970