Search Documents

By L. D. Jaffe, F. W. Cotter, E. Cordon

The hardenability of titanium-base alloys was studied by metallographic examination and hardness survey of Jominy specimens end-quenched from the B range. Analyses of the data led to the equation log J = -0.57 + 0.25 @ct Fe + pct Mn + pct Mo) + 0.19 @ct Cr) +0.16 @ct V) + 0.03 @ct Zr). Here J is the distance, in sixteenths of an inch, from the quenched end of a Jominy hardenability specimen in the position of peak hardness, for material quenched from the B range. This equation fitted the experimental data with a standard deviation of approximately 0.29. The effects of the elements Al, Sn, W, Cu, Ni, B, C, N, 0, and H, and of pain size, were not statistically significant or not practically significant. A check against hardenability measurements in the literature showed agreement within the stated standard deviation. The equation should be useful in estimating hardenability of new or modified titanium alloys. HARDENABILITY in a titanium-base alloy is the ability of the alloy to retain the B structure on quenching. An alloy with high hardenability will retain the /3 structure even when cooled relatively slowly from a temperature at which B or P plus a is stable. A low hardenability material will retain P only if quenched extremely rapidly from the range of p or 0-plus-a stability, or will not retain it at all, at room temperature. High hardenability is desirable in titanium alloys to be heat-treated to high-strength levels. Its value is by no means limited to large section sizes. With high hardenability, a material can be solution-treated and cooled at a variety of rates, either to give high strength directly or, more generally, to give a soft ductile condition from which high strength can be obtained by subsequent aging. With low hardenability, high strength can be obtained, if at all, only by very rapid quenching, and there will generally be little increase in hardness on subsequent aging; an alloy of this type is limited in its applicability. On the other hand, alloys of very low hardenability have some advantages in weldability; essentially, they are always in the annealed condition, after welding as well as before. For commercial alloys, hardenability data are usually available, in the form either of property data after cooling from the solution temperature at various rates, with or without subsequent aging, or of results of a standard hardenability test, such as that originally developed for steels by Jominy and Boegehold.' When modifications of an available alloy are considered, or preparation of new alloy compositions, it would be Very convenient to be able to estimate the hardenability of the new material without having to make and test it. A method of estimating hardenability of titanium alloys from their composition was suggested by one of the authors some time ago, on a preliminary basis, utilizing scattered data found in the literature.' It seemed worthwhile to carry out a systematic experimental study of the effect of composition upon hardenability. EXPERIMENTAL PROCEDURE Approximately fifty heats of various compositions, weighing 8 to 10 Ib apiece, were melted in a small inert-gas tungsten-arc furnace with water-cooled copper walls. The starting material was 110 Brine11 titanium sponge, with high-purity metals added for alloying. Each heat was bottom-poured under vacuum through a molybdenum burnout strip into a cold graphite mold, to form an ingot approximately 4-1/2 by 3-1/2 by 3 in.* From each ingot were cut *The material was melted and cast by Pitman-Dunn Laboratory, Frankford Arsenal, to whom the authors must express their thanks. two pieces 4-1/2 by 1-1/2 by 1-1/2 in. These were forged, at temperatures adjusted to the composition, into 1-1/4-in. rounds, from which standard 1-in.-diam hardenability specimens3 were machined. A number of small samples were also prepared from forged materials of each heat, annealed, quenched from various temperatures, and examined metallographically. The P-transus temperature was determined by observation of the degree of resolution of primary a in these pieces. samples for chemical analyses were also taken from the forgings. One hardenability specimen of each heat was solution-treated for 1 hr approximately 50°F above the measured transus temperature, and the other for 1 hr approximately 250°F above the transus. (An additional hour was allowed for the specimens to reach furnace temperature.) These are not necessarily the temperatures that would be selected for

Jan 1, 1964

By L. D. Jaffe

From data found in the literature, a method has been derived for calculating hardenability of titanium alloys from their composition. A single graph gives the contributions of each alloying element. These are simply added to a base hardenability for unalloyed titanium. Results agree satisfactorily with measured harden-abilities. RECENTLY the importance of hardenability of titanium alloys in permitting attainment of high strength and hardness has become apparent. If a titanium-base alloy has adequate hardenability for the section size and quenching practice used, it can generally be heat treated to high strength and hardness either by simple quenching or cooling from temperatures near the ß transus, by quenching and tempering (aging), or perhaps by direct isothermal transformation of ß1,2 If the hardenability is inadequate, high strength cannot be attained. Preliminary charts translating hardenability requirements for shapes quenched in various media into terms of Jominy end quenched hardenability specimens or ideal round sizes are now available for titanium alloys." The method proposed in 1942 by M. A. Grossmann for calculating hardenability of steels from composition has proved to be of great practical value.' It appeared worthwhile to see whether, using information available in the literature, a corresponding method for calculating hardenability of titanium-base alloys could be derived. Data No systematic experimental study of the hardenability of titanium-base alloys has yet been made. There is available, however, a considerable quantity of scattered data bearing on hardenability. Some investigators determined Jominy curves on one or several alloys. Others reported as-quenched hardness for specimens of a few alloys cooled at various rates or in various media. Still others gave the as-quenched hardness for systematically varied composition, using one or several cooling mediums. The measurements utilized in this work were limited to hardness. Strength measurements were not used because a low strength may reflect high hardness combined with brittleness. Since, in general, the hardness of titanium alloys goes through a peak as the cooling rate is varied from extremely fast to extremely slow, the cooling conditions at which peak hardness was found were taken as the measure of hardenability.1,2 When as-quenched hardness was found to drop as the content of an alloying element (believed to increase hardenability) was raised, this was taken as an indication that the cooling rate was faster than that giving peak hardness; otherwise, the hardness should go up because of solid solution hardening. Also, if the hardness increased on tempering at low temperatures, this was again taken to indicate that the cooling rate had been faster than that giving peak hardness. Many of the data did not indicate precisely the cooling conditions for peak hardness but merely set a maximum or a minimum bound, or a bracket. Since the sizes of specimens quenched and the location of the hardness readings were not always given, it was occasionally necessary to estimate reasonable limits for them. Only the nominal compositions were available in some cases. As quenching from below the ß transus is equivalent to starting with a mixture of a and ß phases, whose individual composition would differ from the overall composition of the alloy, only measurements after quenches from above the ß transus were used. Hardness measurements on samples heated during metal-lographic mounting were also discarded. Because of their bulk, the data used will not be reproduced. They have been or will be published elsewhere.5-22 Approach All data were first translated into terms of ideal round size. For conversion from Jominy or quenched round or sheet, the author's graphs were used." For a few furnace cools or helium quenches where cooling rates were given, conversion was made on the basis of equal cooling coefficient.23 Plots of ideal round size for peak hardness against percentage of alloying element were next made for

Jan 1, 1956

By J. L. Meijering

Oxidation hardening of cylindrical and spherical specimens first decreases with depth below the surface, but then increases again as the center is approached. This is in agreement with the view that this change in hardness is mainly determined by the change in velocity of the oxidation boundary, which affects the dispersion of the oxide. WHEN a thick slab of silver containing 0.3 pct Mg by weight is internally oxidized, the hardness measured on a cross section is found to decrease considerably with distance from the surface.&apos; This phenomenon has been attributed to the diminution of the velocity of the oxidation boundary. The smaller this velocity is, the more slowly the free-oxygen concentration rises towards the concentration in equilibrium with the oxygen gas. And a small 0 concentration favors dissociation and thus conglomeration of the oxide. For instance, the hardness decreases more rapidly during long annealings in nitrogen than in air.&apos; Measurements by Gregory and Smith3 on sections through internally oxidized Al-silver wire showed a smaller variation in hardness, probably because the wire diameter was only 0.9 mm. Wood4 found a drop in hardness with depth in internally oxidized Al-copper strip, and also—by electron microscopy—the concomitant increase in Al2O3 particle size. Under certain circumstances internal oxidation leads to formation of rather irregular inner oxide films in the midst of dispersed oxide. This special sort of conglomeration—discussed in Ref. 5—is also favored by greater depth below the surface.3,5 In this paper hardness measurements in large diameter cylindrical and spherical specimens are described. Here the oxidation velocity* goes through ness initially decreases with depth but increases again towards the center. Normally, the diffusion coefficient D1 of oxygen is very large compared to D2, that of the dissolved metal. If, furthermore, the atomic fraction c2 of the latter is much larger than the O-solubility cl- but c2D2 << c1D1- the velocities in question are easily calculated2 to be: for the cylinder and for the sphere. Here p is the radial coordinate of the oxidation boundary and R the specimen radius, which in dilute silver alloys is constant, as no external oxidation occurs. These equations have been derived for a divalent solute; for nondivalent solutes c2 has to be multiplied by half the valency. The equations are not valid near the center. For which can be considered as a standard case,2 the Eqs [I] and [2] begin to yield U-values which are appreciably too high for p/R = 0.05 and 0.15, respectively. But at the v-minima they are still very good approximations. These minima lie at p = emlR= 0.37R in the cylinder and p = 1/2 R in the sphere. Other factors besides the velocity v will also influence the hardness. The oxide concentration cox is constant in a flat strip, apart from a narrow region in the middle,2 but in cylindrical and spherical specimens cox decreases steadily with depth. When C1 << c2 and c1D1 >> c2D2 one can calculate

Jan 1, 1961

By H. D. Merchant

NUMEROUS publications have examined hot hardness of metals and alloys. Some have studied creep in long-time hardness tests, few of which, however, were tested under a spherical indentor. 1-3 The results of Mulhearn and Tabor&apos; indicate a parabolic-type creep for indium and lead, whereas Moore and Tabor2 how a logarithmic-type creep for indium. Similar semilog relationships between hardness and time have been found by Goffard and wheeler3 for magnesium, aluminum, and their alloys. In this note, the elevated-temperature hardness of some complex alloys, shown in Table I, was examined to discover whether any basic information about the deformation behavior of the alloys could be obtained. One of the alloys, H-13 alloy steel, was tested for short-time creep tests to examine the hardness-time relationship at various temperatures. A standard Brinell Hardness Tester, mounted with an appropriate furnace and 99 pct A12O3 poly-crystalline indentor, was employed for the purpose. The alloys were tested at temperatures from room to 1400°F for 30 sec under 1500-kg load. The creep studies were conducted for about 8500 sec at 7l°, 500°, 1000°, and 1200°F. For constants A and B, a relationship of the type H = Aexp(-BT) [ll was obtained for all alloys between hardness, H, and temperature, T. In the semilogarithmic plot, the data points for each alloy were arranged into two groups, each group on a straight line with a specific set of A and B values. These results are in agreement with many such observations on hot hardness, notably those of Westbrook for pure metals,4 The thermal softening coefficient, B, and transition temperature, Tt, for various alloys are shown in Table I. The change in slope around the transition temperature suggests an apparent change in the mechanism of deformation, probably from the shear (below Tt) to the diffusion type of mechanism. At the temperatures above Tt, for constants A&apos; and B, a relationship of the type was obtained between H and T, Fig. 1. The nature of the equation, where R is gas constant, suggests the possibility of a thermally activated mechanism of deformation above Tt. The constant B&apos; may hence be defined as the apparent activation energy of indentation. The B&apos; values of various alloys are shown in Table I. A plot of B vs B&apos; gives a curvilinear relation shown in Fig. 2. The indentation-creep data for H-13 alloy steel is shown in Fig. 3. At 71" and 500°F, the creep initially follows the logarithmic time law. However, the indentation size stabilizes after 7 sec at 71°F, and after 10 sec at 500°F. Further increase in loading time has no effect on indentation size. The creep of 1000° and 1200°F follows the logarithmic time law for the first 1000 sec, after which the

Jan 1, 1964

By David A. Thomas, David N. French

The lrnrdness of WC crystals has been measured with the Knoop indenter at loads of 100 and 500 g on the (0001) and (1070) planes. The hardness as tneasitred on the basal plane is 2400 kg per sq mm and shows little anisotropy. The hardness on the prism plane, however, shows a marked orientation dependence, with a low value of 1000 kg -per sq mm when the long axis of the Knoop indenter is oriented parallel to the c axis and a high value of 2400 kg per sq mm when the indenter is perpendicular to the c axis. Slip lines (Ire observed surrounding the microhardness indentations and they show slip on (1010) planes, probably in [0001] and (1120) directions. This slip behavior can be explained by the crystal structure of TVC, which is simple hexagonal with a c/a ralio of 0.976. The hardness anisotropy call be explained by [0001]{1010} and (1130) {10l0] slii) and the resolved shear-stress analysis of Daniels and Dunn. HARDNESS anisotropy is a well-known phenomenon and has been reported for many metals, with both cubic and hexagonal structure.1-6 For hexagonal tungsten carbide, WC, a wide range of hardness values is reported in the literature. For example, Schwarzkopf and Kieffer7 give a hardness of 2400 kg per sq mm and report a value of 2500 kg per sq mm by Hinnüber. Foster and coworkerss give the average Knoop microhardness as 1307 kg per sq mm with a maximum value of 2105 kg per sq mm. Although these values and the structure of WC suggest the likelihood of hardness anisotropy, no such measurements have been made. We first detected a large apparent hardness anisotropy in WC crystals about 75 p large, in over-sintered cemented tungsten carbide. Prominent slip lines also occurred around many indentations. This report presents further observations and interpretations of hardness anisotropy and slip in WC crystals obtained from Kennametal, Inc. Both Knoop and diamond pyramid indenters were used on a Wilson microhardness tester with loads of 100 and 500 g. EXPERIMENTAL RESULTS The carbide crystals tended to be triangular plates parallel to the (0001) basal plane of the hexagonal structure. The side faces were parallel to the ( 1010) prism planes. Specimens were mounted approximately parallel to these two types of faces and metallographically polished. Laue back-reflection X-ray patterns were used to orient the specimens, which werethen ground to within ±1 deg of the (0001) and (1010) planes. The Knoop hardness values measured on the basal plane are plotted in Fig. 1. There is only a small anisotropy, with a hardness range of 2240 to 2510 kg per sq mm. The additional points at angles from 52.5 to 67.5 deg confirm the sharp minimum hardness at 60-deg intervals, consistent with the sixfold hexagonal symmetry. The average hardness of all values obtained on the basal plane is 2400 kg per sq mm. While the basal plane shows only slight anisotropy, the (1010) plane exhibits marked hardness anisotropy, from 1000 to 2400 kg per sq mm. Fig. 2 shows the hardness as a function of the angle between the long axis of the indenter and the hexagonal c axis, the [0001] direction. The minimum and maximum occur when the indenter is oriented parallel and perpendicular to the [0001] direction, respectively. The anisotropy of the prism plane is contrary to that reported for hexagonal zinc and hard- However, the basal-plane anisotropy is similar to these two metals.1&apos;2 To check the accuracy and reproducibility of the measurements, a series of fifteen impressions was made at 100-g load in the same orientation in the same area of the specimen surface. The average for all was 2040 kg per sq mm, with a range of 1950 to 2130 kg per sq mm, giving an accuracy of about ± 5 pct. Thus the slight anisotropy on the basal plane is almost within experimental error. Fig. 3 shows slip lines around the Knoop indentations on the basal plane. The slip traces are in directions of the type (1130). The presence of slip steps on the basal plane indicates that the slip direction lies out of the (0001) plane. Because WC has a c/a ratio of 0.976,&apos; the shortest slip vector is [0001], which suggests slip systems of the type [0001] (1010). Fig. 4 shows slip lines around the Knoop intentations on the (1010) plane. These slip lines are inconsistent with [0001] slip but can be

Jan 1, 1965

By M. Schwartz, S. K. Nash, R. Zeman

Knoop hardness in the rolling plane and in the longitudinal plane of hot-rolled and cold-rolled sheets of sublimed magnesiu?w was measured as a function of the angle between the long axis of the indenter and the rolling direction. These measurements were correlated with similar data taken on the (0001) and (1010) planes of a single crystal of magnesium where the hardness was measured as a function of the angle between the long axis of the indenter and the [1120] direction. The results were analyzed for compliance with the hypothesis of Daniels and Dunm to account for slip, and with a similar hypothesis to account for twinning. Some hardness anisotropy data are also presented for magnesium-indium and magnesium-lithium solid solution alloys. It is well known that the hardness of a crystalline specimen is different for its different surfaces, and also that the hardness is a function of direction within a single surface. Variations in hardness for single crystals have been found to be much larger than those for polycrystalline materials. Also, materials having low crystal symmetry were found to have a greater anisotropy of hardness than those of high symmetry. 0&apos;Neill1 and Pfeil,2 using a 1-mm Brine11 ball, studied single crystals of aluminum and iron, respectively; and they found a variation of hardness of about 10 pct between readings taken along the principal crystallographic faces. Daniels and Dunn3 found that the Knoop hardness number varied about 25 pct as the long axis of the indenter rotated on the basal plane of a zinc single crystal. The variation on the (1450) plane was about 100 pct, and the average hardness on this plane was about twice that of the basal plane. They also studied the variation of hardness within the (loo), (110), and (111) faces of a single crystal of silicon ferrite and found variations of about 25 pct although the average values for these planes were almost identical. Single crystals of zinc were also studied by Meincke.4 He found that the Vickers hardness numbers varied about 30 pct depending on whether the axis of the indenter was parallel or perpendicular to the (1010) and (1110) planes. Mott and Ford,5 using a Knoop indenter, found a 25 pct variation in hardness on the basal plane of zinc. Crow and Hinsley6 studied heavily cold-rolled bronze, steel, brass, copper, and other metals. They found that the difference in hardness numbers based on the difference in the length of the diagonals of the Vickers indenter was from 5 to 12 pct. Some minerals and synthetic stones show a very large anisotropy of hardness. Robertson and Van Meter7 found the Vickers hardness of arsenopyrite to vary from 633 to 1148 kg per mm2. stern8 using the double-cone method on synthetic corundum found the hardness number to vary from 950 to 2070. And winchell9 reported a variation of hardness number from 184 to 1205 in kyanite. The variation of hardness as a function of direction in a given crystallographic plane in single crystals possesses a periodicity which is related to the symmetry of the lattice. Daniels and Dunn3 found a six-fold periodicity of hardness in the (0001) plane of zinc. They found that the hardness curves of silicon ferrite had a four-fold symmetry in the (100) plane, a two-fold symmetry in the (110) plane, and a six-fold symmetry in the (111) plane. Mott and Ford5 also reported a six-fold symmetry of hardness in the basal plane of zinc. And vacher10 found two-, four-, and six-fold periodicities of hardness in copper on the (110), (100), and (111) planes, respectively. The purpose of this paper is to report the results of an investigation on the anisotropy of hardness as a function of orientation in single crystals of mannes-ium, and samples of rolled magnesium, magnesium-indium, and magnesium-lithium solid solution alloys. The anisotropy of hardness of pure magnesium which had been hot rolled, and then cold rolled various amounts to fracture, was studied by means of Knoop indentation hardness numbers; and the results were correlated with the preferred orientation as determined by quantitative X-ray pole-figure data. A comparison was made of the hardness data obtained from the rolled sheets and those of single crystals of magnesium. In order to obtain a more fundamental understanding of the variation of hardness and of Knoop hardness testing, the data were analyzed by

Jan 1, 1962

By H. J. Booss

THERE is a considerable lack of data on thermody-namic properties of hard-metal alloys. Only two papers 1,2 give mean values of specific heat in an unknown temperature range; more recently the author3 gave more detailed information on the mean specific heat in the temperature range from 20" to -190° C.. EXPERIMENTAL 1) Method—Experimental procedure followed a method suggested by Oelsen and coworkers.4 It requires samples of some 100 g of weight. The time of measurement is prolonged (30 min) but one run allows the measurement of a lot of positions on the heat-content curve. Results4 were found to be in good agreement with "best values" given by kelley,5 Kubaschewsky and Evans,6 etal.

Jan 1, 1960

By A. A. Watts, R. I. Jaffee, F. C. Holden, H. R. Ogden

Hypoeutectoid Ti-Cu alloys are responsive to heat treatment, and considerable variation of mechanical properties may be produced by transformation of the ß phase. Control of cooling rate, isothermal transformation, or quench tempering determine the transformation structure and properties. The transformation products are similar to those of carbon steel. THE equilibrium diagram for the Ti-Cu system has been studied by various investigators, and the most recent diagram is that determined by Jou-kainen, Grant, and Floe.&apos; The titanium-rich portion of this diagram is reproduced in Fig. 1. The alloy compositions and annealing temperatures used in this work are shown. The Ti-Cu alloys containing up to 17 wt pct Cu are representative of an active eutectoid alloy system. Joukainen et a1.l have stated that ß phase cannot be retained on quenching and this has been confirmed in the present work. The similarity between the Ti-Cu system and the Fe-C system is evident and, as will be shown later, many of the transformation products appear to have similar microstructures. Heat treatments in this program were designed to produce the structures of primary interest for study and determination of mechanical properties. Experimental Procedures Preparation of Alloys: Five ½ lb ingots, ranging in composition from 0.8 to 6.83 pct Cu, were prepared by double arc melting in an atmosphere of high purity argon. Titanium melting stock was obtained from high purity iodide titanium, and copper additions were made from copper wire clippings. Segregation was reduced by inversion melting, in which the once-melted ingots were inverted in the crucible and remelted. The double-melted ingots were forged at 1600°F to ¾ in. diam rods, with both heating and forging operations being carried out in air. Surface scale was removed by grit blasting and grinding, after which the forgings were vacuum annealed at 900°C. The alloys then were swaged to ¾ in. diam rod at 750°C and test specimens were cut from this material. The vacuum annealing operation was to remove dissolved hydrogen, which is commonly present in the iodide titanium as-received. Although vacuum fusion analyses were not obtained for all these alloys, past experience indicates that the 6 hr treatment with a limiting pressure of 10- to 10-5 mm of Hg is sufficient to reduce the hydrogen level to 10 to 30 ppm. Vacuum fusion analysis of the 0.8 pct Cu alloy showed 19 ppm of hydrogen. In addition to the five hypoeutectoid alloys, three 15 g ingots were prepared with nominal copper additions of 9, 10, and 11 pct. Fabrication procedures were similar, except that these alloys were rolled to 0.040 in. sheet. Microstructure specimens only were obtained. Table I shows analyzed compositions and fabrication temperatures for these alloys. Heat Treatments: All heat treatments were made in potentiometer-controlled, resistance-wound tube

Jan 1, 1956

By R. I. Jaffee, F. C. Holden, H. R. Ogden

The properties of quenched Ti-Fe alloys have been correlated with their microstruc-tures. For specimens quenched from equilibrium in the a-ß field, the dominant micro-structural variable is the a-ß ratio. A comparison of specimens having equiaxed a-ß structures with those having acicular a-ß structures shows that the equiaxed specimens have better tensile ductility, but lower impact resistance. There is evidence to show that specimens with acicular structures reach equilibrium in the a-ß field more rapidly than specimens with equiaxed structures. Both strength and ductility are lowered by heat treatments below 700°C. IN previous work,1, 2 certain principles of the physical metallurgy of a-ß titanium alloys have been evolved. Most of these principles have been based on data obtained on alloy systems which undergo no eutectoid reaction, or in which the eutec-toid reaction is so sluggish as to be inoperative. The alloy systems previously studied included the Ti-Mn alloys, with and without a-stabilizing additions, and the binary Ti-Mo alloys. The most important of these principles are: I) The compositional factors which affect mechanical properties are solid solution strengthening, the martensite transformation, and instability of

Jan 1, 1957

By R. I. Jaffee, F. C. Holden, H. R. Ogden

IN a previous series of papers, the results of studies made on various types of high purity base titanium alloys were presented. Included in this work were the Ti-Mn&apos; and Ti-Cu&apos; alloys, representative of sluggish and active eutectoid systems, respectively. The present paper gives the results of a simi- lar study made on Ti-Mo alloys as an example of a ß-isomorphous alloy system. Experimental Procedures Alloy Preparation—Six ½-lb ingots were prepared from iodide titanium and high purity molybdenum stock. The ingots were arc melted in a water-cooled copper crucible under an argon atmosphere. Each ingot was melted at least twice to insure homogeneity and complete solution of the molybdenum. This was done by inverting the ingot in the crucible and remelting completely.

Jan 1, 1957

By R. I. Jaffee, F. C. Holden, H. R. Ogden

Ti-Mn alloys were studied in order to determine the factors affecting the mechanical properties of &stabilized titanium alloys. The principal compositional factors have been found to be solid-solution strengthening, the martensitic transformation, and instability of the P phase. Structural factors, such as grain size and shape, were found to have more influence on ductility and toughness than on strength. THE alloys of titanium with the &stabilizing elements offer the chief hope for developing useful heat treatments. Conversely, the heat-treatment response possible with the p-stabilizing elements causes difficulties such as weld embrittlement and thermal instability during service at elevated temperature. Therefore, it is of considerable technical importance to understand the factors that govern the heat-treatment responses in these alloys, so as to be able to soften the alloys if they are embrittled by an adventitious heat treatment, to render them stable in service, or to harden them while maintaining adequate ductility and toughness. The Ti-Mn system is a good example with which to demonstrate the factors that govern strength and heat-treatment response in. a P-stabilized system. The region of the a-j3 transformation, after the work of Maykuth, Ogden, and Jaffee,&apos; is shown in Fig. 1. The a solubilities are low. This means that, in the two-phase a-p field, partition of manganese between the two phases occurs practically only to the P phase. The solid-solution effects are therefore chiefly the result of solution in the ,8 phase. The eutectoid occurs at 20 pct Mn and 550°C, and proceeds very sluggishly. It can be ignored in heat treatments in hypo-eutectoid alloys, except for long-time storage at low temperatures below the eutectoid temperature. Even here, there is some question that diffusion proceeds far enough to involve eutectoid decomposition. Three compositional factors bear on the mechanical properties and structure of titanium a-P alloys. These are: 1—solid-solution strengthening, 2—mar-tensite transformation, and 3—P instability. Combined with the purely structural factors of grain size and shape, these govern the properties of the alloys. Perhaps the most important factor is solid-solution strengthening. The facts that manganese dissolves in a titanium to a maximum of only 0.5 pct and most of the manganese partitions to the p phase dominate the structure and properties of the a-/3 alloys. Because of the overwhelming preference of manganese for the ,3 phase, the P phase is harder and the a phase is softer. The relative amounts of both phases may be modified by heat treatment in the a-P field as application of the lever rule in the two-phase field clearly shows. A critical temperature in the Ti-Mn diagram is 800°C. This is the P transus temperature of the 6.4 pct Mn alloy, which has the lowest manganese content which will form all retained-P phase on quenching to room temperature. Any quench performed from below 800°C will produce a structure consisting of only a, a plus retained P, or retained p, for alloys of the compositions used in this work. For alloys containing less than 6.4 pct Mn, any quench from above 800°C will always result in structures

Jan 1, 1955

By P. D. Frost, H. A. Robinson, J. R. Doig, M. W. Mote

The mechanism of hardening in heat-treatable zirconium alloys was foUNd to be analogous to that for titanium alloys. Zirconium containing a relatively large addition of a ß -stabilizing element such as molybdenum or columbium can be hardened by the following transformation: Lean alloys, having insufficient alloy content to retain ß at room temperature are hardened by the following transformation: Application of these findings in Zr-Mo-Sn and Zr-Nb-Sn alloys produced strengths as high as 190,000 psi with reasonable ductility. PRIOR research on zirconium alloys has been concerned with development of greater strengths and creep resistance by solid-solution hardening or by dispersion hardening. Although solution and aging heat treatments had been applied earlier to zirconium alloys, quenching frequently produced low ductility. The possibility that the ß-to-w transformation, discovered earlier in titanium alloys,&apos; might also occur in zirconium alloys and cause low ductility was considered, and, indeed, was found as predicted2 to be the case. In the earliest experiments, the heat-treatment response of titanium, too, was disappointing. However, by avoiding the formation of the embrittling w phase, it became practical to obtain room-temperature strengths higher than 180,000 psi, with elongation values of 10 pct.3 Based on the present concept of titanium heat treatment, the principal requirement for a heat-treatable zirconium alloy is that it contains a certain minimum amount of one or more ß -stabilizing elements. Only molybdenum, columbium, and tantalum, when present in sufficient amounts, are known to stabilize ß zirconium to room temperature during rapid cooling from the ß field. Tantalum was undesirable because of its high thermal-neutron cross section. Aluminum and tin are effective a stabilizers, but aluminum has been shown to seriously decrease the corrosion resistance of simple zirconium alloys in hot water. Therefore, for the present study, tin was used as the stabilizer addition. The alloys selected for this evaluation are presented in Table 1 along with their analyses and ß-transus temperatures. The analyses indicated that intended compositions were usually realized. EXPERIMENTAL PROCEDURES All of the alloys were melted from sponge zirconium and fabricated to sheet for heat treatment

Jan 1, 1960

By L. A. Bromley, R. L. McKisson

A high temperature calorimeter designed for use up to 1500°K is described and the theory of its operation presented. This calorimeter was used to measure the heats of formation of Na-Sn alloys ranging in composition from tin to Na2Sn. The values tend to support those reported by Kubaschewski and Seith. ONE method of obtaining the heat of formation of an alloy is to measure the heat of mixing of the elements at high temperature. The principal advantage of this method is that the desired quantity is measured directly, while in conventional room temperature methods the differences of the heats of solution of the elements and the alloy in a suitable solvent are measured. A method of improving the accuracy of the high temperature measurements, when the precision of the measurement is lower than that of the heat capacity values for the elements, is to add one component at room temperature to the other at the high temperature, so that part of the heat of reaction is dissipated in heating the cold component to the high temperature. Thus, only a fraction of the heat of reaction is measured, and a result of higher precision is obtained. The limitations of the high temperature technique are that volatile systems cannot be studied, and that many alloys melt at temperatures higher than those at which the calorimetric apparatus operates readily. Nevertheless, a calorimeter operating at 1500°K would be useful in investigating many of the lower melting alloy systems and the low melting mixed oxide systems. A unit designed for these applications was built and is described here. The Na-Sn system was chosen for investigation because the temperature involved, 880°K, would adequately test the operation of the calorimeter. Further, the data reported in the literature for the various alloys of sodium and tin differ markedly, and it was hoped that this investigation would help to fix the values of the heats of formation in this system. Design of Apparatus A detailed discussion of the considerations involved in the calorimeter design is given by Mc-Kisson and Bromley,&apos; and a brief summary follows. Fig. 1 shows a sectional view of the calorimeter. The inner graphite chamber is maintained at constant temperature by means of the circulating molten tin bath in which it is submerged. The tin is contained in a graphite vat. This unit comprises a thermostat whose heat capacity is about 7000 cal per "C. The main heating element consists of a machined graphite spiral with a room temperature resistance of about 1/2 ohm. These parts are separated from the powdered carbon external insulation by the cage, a cylindrical graphite chamber. The electrical insulators and separators, the auxiliary heater spool, and the loading tube liner are made of sintered alumina. The tin vat stirrers, the sample stir-rer, the melt thermocouple well, and the power leads are made of molybdenum. The auxiliary heater is wound with molybdenum wire and serves to compensate for the heat loss up the loading tube. All of the individual components in the high temperature region of the calorimeter will withstand at least 2000°K, and the various contacting materials should easily withstand 1500°K. The external container for the insulating carbon powder is a copper shell upon which copper tubes are soldered to serve as cooling coils.

Jan 1, 1953

By M. J. Pool, J. R. Guadagno

THE relative partial molar heats of solution in liquid tin have been determined for phosphorous, arsenic, and antimony at 750°K. This work was carried out as part of an over-all program to determine the heats of formation of the III-V semiconducting compounds. A differential twin-type liquid tin solution calorimeter was used in the investigation. All three elements are endothermic upon solution in tin. The relative partial molar heats of solution go through a minimum value at arsenic. No concentration dependence of the heat of solution was found in the dilute (0 to 2 at. pct solute) solution range investigated. The measured heat effects and the calculated relative partial molar heats of solution at infinite dilution are listed in Table I along with the relative heat contents of the solutes calculated from the data of kelley.1 All of the values are referred to 750°K even though the calorimeter temperature fluctuated by *1°K during the course of the experiments. In this range of temperature change no variations were found which were outside the experimental error. This would be expected since the partial molar heat of solution is in general a very insensitive function of temperature. The limits of error given in Table I represent the standard deviation from the mean. The reported error estimates only apply to the estimated precision of the data and not the accuracy. Systematic errors could be present which would affect the absolute values given. This is especially true in the case of phosphorous and arsenic where vaporization could occur. If this were true, the experimental values reported here would be too high and the true partial molar heat of solution would be less than that given in Table I. Moreover, the heat-content data could be in error and this would also affect the calculated values. The relative partial molar heat of solution of phosphorous at infinite dilution in liquid tin is given in Table I. This value is based on eight drops at

Jan 1, 1965

By Irving Cadoff, Richard Nolder

An analysis oj a series of samples of single -crystal silicon grown on sapphire shows that ,four distinct orientation relations exist. There are at least thirteen crystallographic planes which serve as substrates ,for epitaxial growth The model which assumes that silicon substitutes for aluminium and bonds with the oxygen atoms of the sapphire to form the first later for subsequent growth can account for the epitaxial geometry in all four cases. Multiplici-ties 01 specific orientations are found which are explained by the symmetry of both the sapphire and silicon crystal structures. It was reported in the previous paper&apos; that single-crystal silicon films can be deposited on single-crystal a A1,03 (sapphire) with a wide variety of orientations. The silicon films, however, were found to deposit in only four different orientation relationships with respect to the sapphire substrate. The X-ray data have been analyzed to determine the plane and direction of atomic fit and an atomic model to account for the epitaxial relationships at the silicon-sapphire interface has been developed. These results are described in the present paper. EXPERIMENTAL The Laue X-ray data described in the previous paper were reduced to stereographic projections in a standard manner. These data were supplemented with X-ray reflection data obtained with -& E and A Full-Circle Goniometer (Electronics and Alloys, Inc.) mounted on a Phillips diffractometer unit. The silicon films were thin enough to result in characteristic reflections from the substrate as well as the silicon. The superimposed stereographic projections were analyzed to determine the orientation relationships. Ambiguities were resolved by referring to constructed models of various lattice planes for both silicon and sapphire. These lattice models were also used to develop the model for atom fit at the interface. DATA AND ANALYSIS 1) Crystal Structures. The crystallography of silico-n, which is of the diamond cubic class with a. = 5.4~, is straightforward. The sapphire structure is more complex and generally difficult to represent in pictorial form. It is of the rhombohedra1 class but usually referred to hexagonal axes. The transformation of indices between these two systems can lead to ambiguities which will have bearing on the analysis presented herein. Therefore, the basis used in this paper is described in the appendix. Referred to the hexagonal system, the parameters of sapphire are co = 12.95.&, a, = 4.75~, c/a = 2.73. All parameters are for room temperature. Figs. 1 and 2 illustrate schematically the hexagonal unit cell model of sapphire used for this analysis. The oxygen atoms are assumed to occupy positions in an idealized hexagonal array on basal planes. Their actual positions are in a slightly distorted hexagonal array. The aluminum-atom positions are scaled to their actual positions in the lattice with respect to the basal layers of oxygen atoms. It was found that an idealized model where the aluminum atoms are assumed to lie exactly at the octahedral interstices of the oxygen lattice, similar to that used by ~ronberg,&apos; was not adequate for this study.

Jan 1, 1965

By D. Turnbull, R. E. Cech

FISHER, Hollomon, and Turnbull have developed a theory for the nucleation of martensite. They first tested the theory on Fe-C alloys and low alloy steels. The major factor influencing nucleation of martensite was considered to be statistical composition fluctuations occurring in small regions at high temperature and frozen-in on quenching. These local regions of varying size and composition serve as nucleation centers. They become supercritical, one by one, as temperature is progressively lowered, resulting in temperature-dependent or athermal transformation. Fisher next applied nucleation theory to substi-tutional solid solution alloys. Detailed predictions were made for Fe-Ni alloys because of the availability of free energy data on this system. It was shown that composition fluctuations that were significant energy-wise did not occur, and nucleation frequencies could be calculated from average properties of the system. Nucleation was predicted to occur as time-dependent and having the functional relationship to give a C curve of nucleation frequency vs temperature. The analysis further predicted that the nucleation frequency was extremely sensitive to composition. Experimentally, it would be found that the transformation in some compositions is so slow that measurable amounts will not form in a reasonable length of time. With other compositions, only slightly different, the nucleation frequency becomes so great that the material becomes transformed while still distant in temperature from the maximum nucleation frequency. On quenching an alloy of such composition, the observed transformation kinetics would be similar to those found in Fe-C alloys. Cech and Hollomon repeated experiments of Kurdjumov n which the kinetics of transformation were similar to those predicted by Fisher for Fe-Ni alloys. The alloy studied in this investigation contained 73.3 pct Fe, 23.0 pct Ni, and 3.7 pct Mn. Fisher,&apos; using an idealized model for the extent of transformation as a function of the number of martensite crystals per grain of parent phase, derived nucleation frequencies from the transformation curves of Cech and Hollomon. Complicating influences such as coupling effects between grains in the polycrystalline specimens were neglected. Nevertheless, excellent agreement was found between the theoretically and experimentally derived nucleation frequencies. These experiments, however, could not provide a critical test of the theory. Experimental nucleation frequencies could vary widely from those calculated, depending upon the extent of deviation from ideal partitioning and the extent of coupling effects. Further, since the compositions of material theoretically analyzed and experimentally determined were different, the free energy changes involved in the experimental work could only be estimated. Also, the effect of heterogeneities on the transformation kinetics was not considered. For these reasons, it was decided that experiments designed to test the validity of the Fisher analysis must be conducted on binary Fe-Ni alloys, which were the ones considered theoretically by Fisher. The martensite transformation in Fe-Ni alloys has been the subject of considerable study. Machlin and Cohen have shown that transformation proceeds in a manner quite unlike that in any other ferrous alloy system. They found that single crystals would transform to a large extent in a single burst. In large grain polycrystalline specimens, frequently more than one grain and sometimes the whole specimen would transform at the same instant in this manner. Results on filings indicated that different particles would undergo the burst transformation at widely different temperatures. These results support the conclusion that the transformation behavior could not be described by a single nucleation frequency as would be the case if the nucleation were homogeneous. It appeared that further work was necessary to define the factors responsible for burst-type transformation, so that the conditions could be altered to favor homogeneous nucleation of martensite if such could be accomplished. This report summarizes the results of some experiments conducted with powdered Fe-Ni alloys for this purpose and the re-

Jan 1, 1957

By M. J. Saarivirta

A high-purity copper-zirconium alloy system was imesti-gated. The zirconium content of the alloys studied varied from 0.003 to 0.23 pet. The solid solubility of zirconium in copper and some physical and mechanical properties of the alloys have been determined. By proper heat treatment, Cu-Zr alloy can develop a high electrical conductivity and resistance to softening at temperatures up to 500°C (930°F). This combination of desirable properties makes this alloy superior, to other com-merzcrrl copper base alloys for- use in the electrical conductor field. DEMAND for high conductivity copper alloys that retain their mechanical properties at elevated temperatures is growing rapidly. At present, copper-base alloys such as copper-silver and copper-chromium are used where these characteristics are required. As an example, commutators in electric motors are often made of these alloys. However, the life at elevated temperature of commutators made of copper-silver alloys is much shorter than the life of the same part made of copper-chromium and copper-zirconium alloys. Copper-silver alloys are subject to creep and subsequent failure when the service temperature exceeds several hundred degrees centigrade. Use of copper-chromium alloys has been known for the past several years, but little information is available in the literature concerning the properties of alloys in the copper-zirconium alloy system. Some work had been directed toward determining the solubility of zirconium in copper as well as determining properties of the Cu-Zr alloys. These investigations have revealed the superior high-temperature and electrical conductivity properties of copper-zirconium alloys over copper-chromium alloys. The United States Metals Refining Co., Carteret, N. J. has recently produced OFHC R copper-zirconium alloys for experimentation and evaluation. Because OFHC R copper contains no appreciable oxygen, no harmful oxides are formed and the full benefit of Zr as an alloying addition is obtained. This paper describes the results of this work. It contains room-temperature and some high-temperature data which have been obtained after various processing techniaues. Actual elevated temperature properties are being determined and will be reported in a separate paper. Preliminary results obtained at elevated temperatures suggest that the Cu-Zr alloy is superior to other high conductivity copper-base alloys. WORK PROCEDURE The high-purity oxygen free (OFHC R) copper-zirconium alloys with zirconium content ranging from 0.003 to 0.23 pct were cast into 4-by 4-in. wirebars by both continuous casting and conventional casting methods. Zr was added in the form of a master alloy containing 30 pct Zr and 70 pct Cu. Castings were analyzed for zirconium and nine bars containing various amounts of zirconium were selected for this study. The results of analyses and the sample numbers are given in Table I. Typical analyses of OFHC R copper are also shown. The following properties were studied: A) Solid solubility of zirconium in copper. B) Effect of zirconium on properties of copper. C) Effect of combined working and heat treating on properties of Cu-Zr alloys. The solid solubility of zirconium in copper was determined by metallographic techniques. Specimens for this phase of the investigation were hot rolled at 900oC (1650oF) to 0.250-in. rod. The rolled rods were cut into 2/3-in. long specimens and solution annealed at 950 °C (1740°F), 980°C (1795 °F). and 1020 °C (1870 °F) for 6 hr and quenched in water. In order to find the solid solubility limits at lower temperatures, the homogenized specimens were reheated at various temperatures for 1 hr and quenched. The physical and mechanical properties were determined on 0.250 in. rod, 0.132- and 0.081-in. wires. The wires were cold drawn from the 0.250-in. rods after they were solution annealed at 980 °C (1795°F) and quenched in cold water. Mechanical and electrical properties were determined on cold drawn and

Jan 1, 1961

By M. E. Wadsworth, R. C. Peterson, W. M. Fassell

The temperature and pressure dependence of the reaction of tantalum in oxygen were investigated from 500° to 1000°C at pressures from 10 mm Hg to 600 psi total oxygen pressure. Tantalum was found to oxidize linearly under the above conditions. Three distinct regions of temperature dependence were found with different energies of activation. From 500° to 600°C the rate of oxidation of tantalum was found to be essentially independent of the oxygen pressure at the pressure investigated. The oxidation rate increases rapidly with an increase in pressure from 600&apos; to 800°C. The dependence of the oxidation rate on the bulk concentration may be expressed by V = k&apos;?, where k&apos; is the specific rate constant and ?=k1Co2/(l + K1Co2), where k, is the equilibrium constant for the adsorption of oxygen on tantalum. ONLY a limited amount of data has been presented in the literature on the oxidation behavior of tantalum metal. Some values of its oxidation rate were reported in the Corroston Handbook&apos; and by McAdams and Geil.&apos; The first extensive investigation of tantalum was the work of Gulbransen and Andrews." They determined the reaction rates of tantalum in O2 and N2. Their results indicate that this metal follows the parabolic law in the temperature range of 250" to 450°C in O2 at pressures of 7.6 cm Hg with an activation energy of 27,400 cal per mol. In nitrogen, the metal follows the parabolic law from 500" to 800°C at 7.6 cm pressure. All experimental determinations were of 2 hr or less duration and deviations from the parabolic rate conceivably could occur at longer times. A spectrophotometric study of the oxidation of tantalum by Waber, Sturdy, Wise, and Tipton&apos; indicated good agreement between this method and the microbalance technique. The experiments were conducted in air from 220" to 350°C. The metal was found to follow the logarithmic law up to 320°C and the parabolic law at higher temperatures. The activation energies in the logarithmic region were reported as 12.57 kcal per mol and 12.0 kcal per mol, respectively, for the microgravimetric and spectrophotometric methods. In the parabolic region the activation energy was found to be 27.2 kcal per mol, in good agreement with the value of Gulbransen and Andrews." The present paper is concerned with the behavior of tantalum from 500" to 1000°C in oxygen at pressures ranging from 1/2 to 40.8 atm pressure. Earlier preliminary data by McKewan5,6 indicated the following facts concerning the behavior of tantalum: 1—At temperatures from 500" to 600°C, the metal followed the linear law. 2—At 575°C the rate of oxidation was found to increase with O2 pressure. 3—The Arrhenius plot in this temperature region was nonlinear. The nonlinearity of the Arrhenius plot was indicative of a more complex behavior than originally suspected for this metal.

Jan 1, 1955

By W. Mckewan, W. M. Fassell

The oxidation rates of copper have been determined at temperatures from 600" to 900°C in oxygen from 14.7 to 400 psi total oxygen pressure. The oxidation rate of copper is unchanged by oxygen pressures within the range studied. The observed data extend Feit-knecht&apos;s conclusion concerning the pressure independence of copper from atmospheric pressure to 400 psi. All samples studied in the above-mentioned temperature and pressure range have both Cu2O and CuO present in the oxide film. NUMEROUS papers have been published on the oxidation of copper, some of which have noted the effect of oxygen pressure on the oxidation rate of copper. It must be noted, however, that no data on the oxidation rates of copper at pressure in excess of 1 atm have been reported in the literature. (Le Chatelier¹ reported that a black oxide of silver was formed at 300°C and 15 atm pressure.) This is likewise true for the other metals. This investigation was initiated to obtain experimental oxidation rate constants for pure metals at elevated temperatures and high oxygen pressure. Copper was selected as the first metal since probably more reliable data are available on its oxidation rates and related physical and chemical properties than for any other metal or alloy. As a general summary, the following conclusions appear well established for the oxidation of copper: 1—The oxide coating consists of Cu2O with a superficial coating of CuO providing the ambient oxygen pressure exceeds the equilibrium pressure for the coexistence of Cu2O and CuO.² The following sequence of phases is then Cu/Cu2O/CuO/O2 (gas). The Cu2O consists of large crystals that have no relation to the orientation of the metal crystals, except for a very thin layer adjacent to the metal." The CuO consists of very fine crystals randomly oriented. If the oxygen pressure is below the equilibrium pressure for the coexistence of Cu2O and CuO, the coating consists of Cu,O only. 2—The rate-determining factor in the parabolic oxidation of copper is the diffusion of Cu+ ions through the Cu,O layer.4,6,7 The Cu+ ions are accompanied by electrons to maintain electrical neutrality. The mechanism of the reaction at the Cu2O/CuO interface and the method of growth of the CuO is in some doubt. 3—On copper, there are three reported pressure effects: (a) At pressures below 0.3 mm Hg oxygen pressure, the oxidation rate increases directly with pressure.&apos;, &apos; .Vxygen starvation at the surface may account for this effect. (b) At oxygen pressures from 0.3 to 63 mm, the oxidation rate increases as the 7th root of the oxygen pressure providing the temperature is such that only Cu2O is present in the film. (c) At pressures from above the equilibrium pressure for the coexistence of Cu2O and CuO (varies with temperature) to atmospheric pressure, the oxidation rate is independent of the oxygen pressure.&apos; Equipment In order to determine the oxidation rates of metals in an oxidizing atmosphere at pressures up to 400 psi, the equipment is somewhat different than for low pressure work as is shown in Figs. 1 and 2. A Leeds and Northrup Micromax recording controller (A) is used to control the temperature of the Ni-chrome furnace winding in conjunction with a Micromax electric drive unit. The drive unit adjusts a 20 amp Variac (variable transformer) to control the power input into the Nichrome furnace winding. Due to the "self heating" of the sample by oxidation, frequently encountered in oxidation rate studies, it is essential that the sample temperature be measured independently. This is done by means of a series of six chromel-alumel calibrated thermocouples located spirally around the oxidizing metal sample. Each thermocouple is partially shielded from the direct radiation of the furnace wall by

Jan 1, 1954

By John Brownson

Ge-Si N-N heterodiodes hare been built recently which show promise as high-speed logic devices. Low-resistivity germanium is deposited on silicon substrates held at temperatures above the germanium melting point and an alloyed heterojunction formed. The diodes are fabricated from this material using conventional mesa techniques. Over-all device qualify has been greatly improved by the use of epitaxial silicon substrates, that is, a subsirale consisting of a film of thin, high -resistivity silicon on a thicker Piece of low-resistirily silicon crystal. A semimpirical device design equation has been devised which predicts switching performance fairly well within the range of resistivitics and geometries employed. Switching times 01 0.9 ns and PIV &apos;s of —20 v are typical of the better 10-ma diodes. Switching times as low as 0.5 ns have been observed for 10-ma diodes and as low as 2.8 ns for 200-ma diodes. With further development it should be possible to improve switching speed by a factor of four. IN our laboratory, Ge-Si heterodiodes have been built recently which show promise as high-speed logic devices. The diodes have been built using a process reported in some detail in an earlier publication.&apos; The process in short involves alloying a thin film of low-resistivity germanium into a silicon substrate. the germanium having been transported and deposited by the well-known Theuerer halide reduction process.2 This process contrasts with that reported earlier by Oldham who used an essentially conventional direct epitaxial deposition technique.3 Oldham&apos;s process, unlike ours, requires cleavage of the silicon sample inside the reactor furnace the instant prior to deposition. Our process has the advantage of yielding large-area macro-scopically plane heterojunctions. Heterodiodes in general have several interesting properties. Without attempting to summarize the theory of heterojunctions.4 two of these properties which have direct relevance should be mentioned. First, N-.V and P-P heterojunction diodes rectify alternating current. Provided that either the band gap or the electron affinity of the two materials is different (which in general is the case), electronic barriers will exist in the junction band structure. When the device is sufficiently forward-biased, the barriers shift so as to permit current flow, and when it is reverse-biased. the barriers block current flow. Second, for N-N and P-P heterodiodes. the foward-conduction process involves majority carriers only. Consequently. when such a device is turned off. there is no minority-carrier storage and hence the diodes switch on and off quite rapidly. As we have reported earlier. N-P and N-N Ge-Si heterodiodes have been built in this laboratory. As one would predict from theory, the N-P heterodiodes were rather slow in turning off while the N-N&apos;s were quite fast. Similarly, the turn-off time of the N-P&apos;s increased rapidly with forward current while that of the N-N&apos;s was independent of forward current. Both N-N&apos;s and N-P&apos;s were moderately photosensitive. Photocurrents resulting from normal room illumination were as high as 1 µa at 1 v for small-area devices. The most disappointing parameter of the early N-N devices was reverse leakage.&apos; The reverse characteristic was invariably "soft" and leakage at only several volts reverse bias was intolerable. The reverse characteristic was greatly improved by using higher-resistivity silicon substrates. Leakage becomes large at voltages about one third of the avalanche voltage one would expect of a P-N homo-junction formed on silicon of the same resistivity. Thus leakage could be reduced to reasonable limits by using unconventionally high-resistivity silicon. Medium-power diodes formed on 30 ohm-cm substrates leaked typically 15 to 50 µa at 100 v reverse bias. The use of high-resistivity silicon substrate material, however, created a new problem. In order to have an acceptable forward-conductance characteristic. the area of the diodes had to be increased. With this increase in area, the capacitance increased. Switching time for N-N heterodiodes is essentially a linear function of capacitance. Thus when the resistivity was raised. so was the switching time. A SEMIEMPIRICAL HETERODIODE DESIGN EQUATION As a result of our experiments with a limited range of diode geometries and resistivities. a semi-

Jan 1, 1965