Search Documents

By R. Taggart, R. H. Ericksen, D. H. Polonis

MarTENSITIC a has been observed to form spontaneously from the retained ß phase during the preparation of thin foil specimens of metastable Ti-Cr alloys containing from 6.9 to 20 wt pet Cr. Similar spontaneous martensite reactions have been reported in thin foils of 0 brass' and other alloys.2 4 The absence of retained austenite in iron-based alloys has also been attributed to spontaneous decomposition during thinning.5 The present work indicates that a unique situation occurs during which the spontaneous transformation takes place when the transitional ? phase is present in the as-quenched or partially aged ß phase. Alloys containing 6.9, 7.0, 8.0, 12.0, and 20.0 wt pet Cr were prepared by the levitation melting of "io-chrome" chromium and iodide titanium. The 0.5-in.-diam ingots were vacuum-annealed at 1000oC, and wafers 0.020 in. thick were then cut from the ingots. These wafers were wrapped in molybdenum foil, encapsulated in Vycor under a reduced pressure of purified helium, and then held for 30 min at 1000°C. The specimens were quenched in water to retain the metastable high-temperature ß phase. The quenched alloys were thinned by mechanical grinding to about 0.005 in. thick, then electropolished in a solution of 300 ml methyl alcohol, 175 ml n-butyl alcohol, and 30 ml perchloric acid at a temperature of -40°C. Needle specimens 0.005 to 0.010 in. diam were taken from each ingot and heat-treated with the foil specimens. X-ray diffraction patterns indicated the structure of the 6.9, 7.0, and 8 wt pet alloys to be a

Jan 1, 1968

By M. H. Richman, L. M. Markowitz

MUCH of the available information on which the heat treatments of a particular steel are based is found in the isothermal-transformation diagram. Such is the case despite the fact that the usual thermal treatments involve continuous transformations, i.e., transformations which occur on cooling, rather than isothermal transformations. This rather unusual practice is continued because of the abundance of published isothermal-transformation diagrams1, 2 which are relatively easy to determine experimentally as compared to continuous-transformation diagrams.3, 4 However, this can easily lead to the production of improper micro-structures and may not allow full advantage to be taken of the inherent hardenability of a steel. A mathematical method of estimating the continuous-transformation diagram from the available isothermal data was developed by Grange and Kiefer;5 but, due to the mathematical approximations required in the calculations, the computation of continuous-transformation diagrams has been limited to individual cases and no large-scale compilation of these diagrams is available. With the present availability of high-speed digital computers, the problem of time-consuming approximations may be minimized and continuous-transformation diagrams may be constructed with the insertion of isothermal data into a basic program. It is the purpose of this report to announce the availability of such a program and compare the computed and experimental continuous-transformation diagrams for two different steels. Let us first review the method of Grange and Kiefer5 and then compare the results of the machine calculations with the experimentally determined diagrams. In order to apply this method, one requires the isothermal-transformation diagram and a set of curves representing cooling at constant rates. A section of an isothermal-transformation diagram is shown in Fig. 1 together with a cooling curve of M0F per sec. The cooling curve is constructed to start from the A1 (eutectoid) temperature if the transformation product is the eutectoid microconstituent (i.e., pearlite or bainite), from the A3 temperature if the product is primary ferrite, or from the 4, if it is proeutec-toid cementite. The cooling curve in Fig. 1 intersects the curve representing the start of isothermal transformation at a point A' corresponding to a temperature TX and a time fX. An arbitrary lower temperature To (on the

Jan 1, 1968

By J. R. Cahoon, W. V. Youdelis

FOR the diffusion of copper in silver-rich alloys, there have been only two previous investigations, that of Seith and Peretti in 1936, and that of Sawatzky and Jaumot2 in 1957. Seith and Peretti obtained a value of 24.8 keal per g-atom for the activation energy of diffusion whereas Sawatzky and Jaumot obtained a value of 46.1 keal per g-atom. Admittedly, Seith and Peretti used relatively high concentrations of 2 at. pet Cu and polycrystalline diffusion couples, while Sawatzky and Jaumot investigated tracer diffusion in single crystals of silver. The difference in the results, however, seems too large to attribute solely to solute concentration, as a change of over 20 keal per g-atom in the activation energy suggests a change in the mechanism of diffusion. Therefore, in the present investigation the diffusion coefficient for copper in polycrystalline Ag-2 at, pet Cu alloy is re-established using conventional sandwich-type diffusion couples and electron-probe microanalyses, and the results are compared with those of the previous studies. The diffusion couples were prepared from high-purity silver (99.999 pet) and high-purity copper (99.999 pet). The silver disks, 1/4 in. long and 4 in. in diam, were machined from an ingot which had been solidified in a dynamic vacuum of 10-3 mm of Hg to eliminate oxygen absorption which otherwise would result in copper oxide precipitation in the weld zone. The alloy disks of similar dimensions were machined from an alloy ingot (nominal composition 2 at. pet Cu) which had been unidirectionally solidified to obtain a predictable solute concentration,3 and homogenized for 1 week at 800°C. Sandwich diffusion couples were made by pressure welding an alloy disk between two silver disks in an induction heating coil. The couples were diffusion-annealed at temperatures of 800°, 756°, 699°, and 630°C in tube furnaces in which the temperature was controlled to better than ± l°C. The diffusion couples were examined metallographically for weld defects, and couples showing even minute defects at the weld interfaces were rejected. Concentration penetration curves were established using an A.R.L. Microprobe analyzer. CuKO radiation was counted and converted to concentration using 0.5 and 1.0 wt pet Cu chemically analyzed standards. Initially, four continuous profiles were obtained, and averaged to obtain the final diffusion curve. The initial analysis was performed commercially and the welds were not located with respect to the profile. Therefore, the results were rechecked by the authors using the Mellon Institute A.R.L. probe. In the latter case, concentrations were obtained at 10-µ steps for four different traverses, and then averaged to obtain

Jan 1, 1968

By R. B. Gavert, David J. Mack

ROOM- and elevated-temperature compression tests were made on the binary 3/2 electron compounds AuZn, CuAu, AgZn, AgMg, AuCd, and NiA1. These compounds possess either the B2 CsC1-type ordered structure or the A2 bee disordered structure represented commonly as ß and ß on equilibrium phase diagrams. In testing at room temperature, cylindrical specimens 0.508 cm by 0.330 cm diam were seated directly on an instron compression load cell (CF 0 to 10,000 lb). At elevated temperatures, testing included a special apparatus with a furnace heat source and a protective atmosphere.' The specimens were strained at a rate of 1 pet per min. The compounds were prepared from 99.99 pet or better high-purity elements. Vacuum and/or argon cover gas was used throughout the mixing, melting, and specimen drawing processes. The generally equi-axed specimen grains varied in size from 0.5 to 1.5 mm for the five different compounds. The stress-strain curves in compression showed

Jan 1, 1968

By Charles J. McMahon

In a recent paper, Floreen and Hayden report that additions of the noble metals iridium, rhodium, ruthenium, and platinum to Fe-C alloys containing 0.03 to 0.08 wt pct C act to significantly reduce the ductile-to-brittle transition temperature in a notched-bar impact test. They compare these results with the similar, well-known effects of nickel and manganese additions, and they note that this supports the hypothesis that addition of austenite stabilizers enhances notch toughness. Although not attempting to offer an explanation for this, they suggest that it might be found in an alteration of the bonding characteristics of the bcc lattice of iron by additions of elements which tend to stabilize the fcc lattice. While this suggestion appears to have some merit, it should perhaps be held in abeyance until two more likely possibilities have been explored. It is now fairly clear that the beneficial effect of manganese derives from its ability to promote a globular carbide morphology instead of the platelike form found in grain boundaries and pearlite patches in Fe-C al10~s.' Previous work14 has also shown that cleavage fracture in Fe-C alloys of the type tested by Floreen and Hayden originates almost exclusively in the cracking of cementite platelets, and that the ductile-brittle transition temperature is very sensitive to the volume density and size of these carbides. Treatments which lead to development of nonlamellar carbides or fine dispersions of carbides, such as quenching and tempering or appropriate alloy additions, promote toughness by removing the primary mode of cleavage initiation. The effectiveness of manganese in this regard has long been known. Nickel appears to behave in a similar manner, though the effect is perhaps not as dramatic as with manganese.15 However, nickel additions do lead to a definite refinement of ferrite grain size following a given transformation heat treatment, and this in itself acts to greatly enhance notch toughness.'' (The grain sizes of the various heats tested by Floreen and Hayden were not reported, so it is not known how much influence this factor exerted in their results.) In conclusion, it is suggested that one should consider the well-established effects of alloying elements on carbide morphology and ferrite grain size before resorting to more complex possibilities, such as modification of interatomic bonding.

Jan 1, 1969

By d&apos, C. Antonio, S. H. Reichman

In the course of our studies of the annealing behavior of vapor-deposited silver films we have determined the electrical resistivity as a function of annealing temperature for films in the thickness range of 630 to 3240A. Fig. 1 shows the resistivity of the silver films vs film thickness in as-deposited and annealed conditions. The films were prepared by depositing 99.999 pct pure silver from molybdenum boats onto Vycor substrates. Deposition was conducted at a residual pressure of 5 x 10"6 Torr at rates of 20 to 408, sec-'. Thicknesses were determined using a multiple-beam interferometer. During annealing, resistance was measured using the four-probe method. Measurements were made every 5°C during two consecutive anneals to 300°C under a continuous vacuum of 2 x X6 Torr. The heating and cooling rates during annealing were approximately l° and 2°C min-', respectively. Typical resistivity vs temperature curves are shown in Figs. 2 through 4. There is an inflection point evident at 120°C on Fig. 2, the first anneal of a 3240A film. This inflection point was found on all films above approximately 1!00A on both first and second anneals. Below 1400A the inflection occurred at approximately 180°C. Figs; 3 and 4 show the first and second anneals of a 1370A film. The inflection point occurs at 180°C on the first anneal and approximately 120°C on the second. Our interpretation of this phenomenon is based on oxygen liberation from the film surface. It is well-documented that the presence of oxygen on silver surfaces causes a reversible increase in resistivity.1"3 The liberation of this oxygen, for example during vacuum annealing, would therefore cause a decrease in resistivity as observed on Figs. 2, 3, and 4 as a change in slope. This is accounted for by the decomposition of Ag20 on the film surface. The free energy of AgzO becomes positive between 125" and 225°C.4 The thickness dependence of the inflection points can be explained as an orientation effect. The films used in this study were preferentially oriented with their (111) planes parallel to the Substrate.''~ In addition it has been sten that the thinner films, below approximately 1500A, possess a decided "hill and valley"

Jan 1, 1969

By D. J. Mottern, F. Wald

SEVERAL systems have been investigated by Bates, Wald, and weinstein&apos; to determine the compatibility of pure PbTe with various metals. This was accomplished by experimentally determining the direction of the reaction given by Eq. [I]: The direction of this reaction is very important if alloys of the PbTe system are to be used in thermoelectric generators because the presence of free lead in the matrix, followed by incipient melting in the vicinity of the hot junction, could cause a bonding failure. According to anssen,&apos; the Au-Te system contains only one stable compound, AuTez. The initial alloy which was reacted was therefore of the composition AuTez + 2Pb. Ten-gram samples of PbTe plus 1, 10, 20, 30, 40, 50, and 70 at. pct Au were encapsulated in carbon-coated quartz capsules in vacuo. All materials were of semiconductor (99.999+ pct) purity. Carbon coating was used to prevent reaction of the quartz with any of the alloying constituents (the solubility of carbon in the constituents of the alloys is negligible). Each capsule was then placed in a thermal analysis furnace, melted, and cooled through the transformation region. Thermal arrests were monitored on a strip chart recorder. Each sample was subsequently examined microscopically after standard grinding and polishing techniques were applied. For better contrast between the phases, the samples were etched in Murakami&apos;s etch; however, both phases could be identified without etching. Several of the samples were ground into powder, stress-relieved, and X-rayed in a Debye-Scherrer camera using nickel-filtered Cu-Ka radiation. The phase diagram shown in Fig. 1 indicates that the pseudobinary system exists as a simple eutectic with no measurable (<I pct) solid solubility at either limit. Figs. 2,3, and 4 are photomicrographs of the structures obtained with 20, 40, and 50 at. pct Au, respectively. At higher concentrations of gold (>40 pct) slight divorcement of the eutectic occurred, apparently due to a) the difference in density between gold and PbTe (19.6 and 8.25 g per cu cm, respectively) or b) the nonideality of the liquid as noted by the curvature of the liquidus lines. The eutectic temperature was determined to be 742° ± 5°C. The composition of the eutectic was determined by extrapola-

Jan 1, 1969

By G. S. Coe, G. F. Bolling

The columnar-to-equiaxed transition (CET) has often been studied as a function of concentration, C, temperature gradient in the melt, G, and rate of solidification, R. Although it may not be immediately obvious it is sure that the CET, expressed as some ratio G/R necessary to suppress equiaxed nucleation for a given C,&apos; cannot be a single-valued function of C. This restriction which is ignored in theoretical presentations is most trivially illustrated by the following simple reasoning. An ordinary casting is an unconstrained system where the thermal conditions change continuously with progressive solidification, and thus where the ratio G/R decreases from the outset. A CET occurring early in a casting (much equiaxed material) must then mean a higher value of G/R at transition, in comparison with the late CET which implies a more columnar ingot. Therefore, when we choose to experiment under constant superheat for all alloys across the simplest binary, and we find an inversion in the amount of equiaxed material, we shall find two values of C satisfying the G/R condition. However, in addition to this simple instance, the early experiments of ~orthcott&apos; seem to show that the amount of equiaxed material in cast ingots may be a multivalued function of composition in a given binary system. In the present investigation, the CET has been examined in Pb-Sb, Pb-Ag, Pb-Sn, and Sn-Pb under several types of casting conditions. First, we have used the "bidirectional" ingot section Simulation;~ and we have used .&apos;i~z situ" ingot solidification where a hot mold already filled with liquid is chilled so that there is no disturbance provoked by pouring. Differences in method are trivial to our purpose except for the fact that certain variations are more simply observed by one technique than by the other. Second, we have varied the mode of superheating from the simple extreme of casting all alloys from the same temperature regardless of composition to the careful casting at a standard superheat, calculated with respect to the alloy liquidus. Third, we have made provisions for decanting and interface observation, so that some "castings" were not completed. The results obtained with Pb-Sb are exemplary; some of these are shown in Fig. 1 along with explanatory captions. At 9 to 10 at. pct Sb, and thus above the solid-solubility limit Cs = 5.8 at. pct Sb and yet below the eutectic concentration Ceu = 17.5 at. pct Sb, there is a persistent anomaly in all the variations. Measurement was easy enough and standard enough so that we may forgo discussion of either method or possible error, in view of the forceful self-consistency shown by experiment. Solute distributions were measured along the bidirectional ingots in order to determine if a solute buildup had occurred at the CET. No anomalous changes in concentration were observed along the bars at either side of the solute anomaly of 10 at. pct Sb. The microstructures of the ingots were also examined across the range of composition investigated. At concentrations less than 9 at. pct Sb the micro- and macrostructures were typical of the cellular-dendritic growth usually reported for the columnar zone. Above the anomaly, both the columnar grain size and the amount of primary phase decreased. This microstruc-tural observation can be correlated with observations made of decanted interfaces. Below 9 at. pct Sb, the interface reveals the "normal" columnar dendritic structure; however above the anomalous concentration range, preferential growth of certain dendrites appears to form a "super dendrite" pattern. In Fig. 2 a microstructural view of the etched &apos;(super dendrites&apos;&apos; shows their constitution both in terms of relative phase proportion and in terms of the ordinary smaller dendrites.

Jan 1, 1969

By R. P. Laforce, ZJ. R. Low, A. M. Turkalo, D. F. Stein

The interaction of the crlloying eletnenls, nickel and chromium, with the impurity elements, antimony, pIzosphorus, tin, and arsenic, to producse reversible temper brittleness in a series of high-purity steels containing 0.40 wt pct C has been investigated. The alloyed steels contained approximately 3.5 pcl Ni, 1.7 pct Cr, and 0.05 to 0.08 pct of the particular irnpurity to be investigated. Susceptibility to teirlper embrittlement was measured by comparing the notched-bar transition temperature of each steel after quenching from the final temper and after very slow cooling (step cooling;) following the final temper. A plain carbon steel without alloying elements, bu/ ud/h 0.08 pel Sh, does not embrittle when step-cooled through the emzbrittling range of temperatures. The same embrittling treatment, applied to a steel with about the same antinzony content but with nickel and chvonziunz added, causes a 700°C increase in transition temperature. If chromium or nickel is the only alloying element, the increase in transition temperature is only 50%, again with antimony present. A carbon-free iron containing nickel, chromium, and antimony shou~s a 200°C shift in transition temperature for the same thermal treatment. Specific alloy-impurily interactions are also observed for the other impurity elements, phosphorus, tin, and arsenic. Additional investigations involving electron microscopy, trzicrohard-ness tests of vain boundaries, minor additions of zirconiutn and the rare earth and noble metals, nzainly with negative results, are also described. HE particular type of embrittlement investigated is that which is encountered in alloy steels tempered in the temperature range from about 350" to 525&apos;C or slowly cooled through this range of temperatures when tempered above this range. This type of embrittlement is sometimes called reversible temper brittleness to distinguish it from the embrittlement indicated by a minimum in the room-temperature V -notch Charpy energy vs tempering-temperature curve encountered in the range 28 0" to 350°C. Temper brittle-ness seriously restricts the use of many alloy steels since it precludes tempering or use in the embrittling range of temperatures and may significantly raise the ductile-brittle transition temperature of heavy-section forgings and castings tempered above the embrittling range, since such sections cannot be sufficiently rapidly cooled after tempering to avoid embrittlement. The very voluminous literature of temper brittle-ness up to about 1960 has been reviewed by woodfine&apos; and LOW.&apos; Of particular significance to the present investigation was the demonstration by Balajiva, Cook, and worn3 that high-purity Ni-Cr steel does not exhibit temper brittleness and the subsequent detailed and systematic study by Steven and Balajiva~ of the effect of impurity additions on the susceptibility to embrittlement of Ni-Cr steels. Steven and Balajiva showed that, of the impurities which may be found in commercial steels, Sb, As, P, Sn, Mn, and Si could all produce temper brittleness in a high-purity Ni-Cr steel. The principal purpose of the present investigation was to study the effects of particular alloy-impurity combinations on susceptibility to temper embrittlement. The steels used were high-purity 0.30 to 0.40 wt pct C steels containing 3.5 wt pct Ni and 1.7 wt pct Cr, separately or in combination. The susceptibility of these steels was then determined when approximately 500 ppm by weight of antimony, arsenic, phosphorus, or tin were added as an impurity. The melting, casting, and forging practices used in the preparation of the materials investigated are described in Appendix A. Table A-I in this appendix shows the analysis of all steels to be discussed. The steels were produced as 20- or 2-lb heats. The smaller heats were used after it had been demonstrated (see Appendix B) that a small, round, notched test specimen could be used to measure the shift in the ductile-brittle transition temperature caused by temper brittleness with about the same result as that obtained by Charpy testing. HEAT TREATMENT Unless otherwise noted, all steels were tested for embrittlement in the tempered martensitic condition. A typical heat treatment for a 0.40 C, 3.5 Ni, 1.7 Cr steel was: 1 hr at 870"C, in argon, quench into oil at 100"C, quench into liquid nitrogen, temper 1 hr at 625"C, and water-quench. The warm oil quench was used where quench-cracking was encountered; otherwise the initial quench was into room-temperature oil or water. For other compositions austenitizing temperatures were 50°C above Acs with the remainder of the thermal cycle the same. Steels in this condition, with no further heat treatment, are designated as non-embrittled. The above quenching and tempering cycle for the 0.40 pct C steels resulted in as-quenched hardnesses of 48 to 53 RC and as-tempered hardnesses of 24 to 31 Rc except in the case of the plain nickel or plain carbon steels. In these, the as-tempered hardness was as low as 80 to 90 Rg. No attempt was made to adjust the tempering temperature to obtain the same hardness in ali steels since it was felt that a uniform thermal cycle was more important than exactly equivalent hardness values. Pro- the standard quench and temper described above, the standard embrittling treatment was "step-cooling". For this the thermal cycle was: 593"C, 1 hr; furnace-cool to 538"C, hold 15 hr; cool to 524"C, hold 24 hr; cool to 496"C, hold 48 hr; cool to 468&apos;C, hold 72

Jan 1, 1969

By D. H. Forbes, H. M. Manasevit, F. L. Morritz

Single-crystal thin films of molybdenum have been grown. by the pyrolytic decomposition of molybdenum hexafluoride in a hydrogen ambient on various insulating substrates at temperatures front 650° to 900° C. Single-crystal discs of various orientations of magnesium aluminate spinel, a aluminum oxide, magnesium oxide, and the natural facets of beryllium oxide were all observed to exert epitaxial control on the depositing film. Determinations of the crystallographic relationships existing between film and substrates have been made, using X-ray diffraction three-circle goniometry and Laue back-reflection techniques. Using these data, crystallographic overlays have been constructed which suggest a possible interfacial geometry on the basis of a match between metal deposit atoms and metal ion sites in the substrate. A continuing program has been in effect in the Auto-netics Laboratories over the past several years aimed at the development of electronic materials in new and useful forms using chemical vapor deposition techniques. During the course of these investigations, the epitaxy of silicon1"5 and tungstenep7 on single-crystal metal oxide insulators was studied extensively. The deposition of molybdenum is a logical extension of the tungsten work since the crystal structure is identical and the lattice dimensions differ by less than 1 pct. The pyrolytic decomposition of molybdenum hexafluoride in hydrogen was used to grow molybdenum films on the oxide substrates AlzOs, MgA1~04, MgO, and Be0 at from 650" to 900°C, depending on the substrate material. Under the experimental conditions to be described, monocrystalline films of controlled orientation were obtained. The materials investigated and their crystal structure are listed in Table I. EXPERIMENTAL The molybdenum depositions were effected by the hydrogen reduction of molybdenum hexafluoride (MoFe) in a flowing system of hydrogen at atmospheric pressure. The schematic of the apparatus is shown in Fig. 1. The hydrogen, purified by passage through a heated palladium thimble, was mixed with the vapor of the liquid MoFe maintained in a steel bottle at room temperature (25°C). The mixture was passed over the substrate, which was heated by rf induction of a sup-

Jan 1, 1969

By P. T. Chiang

Chemical etching methods for the simultaneous revealing of the tellurium or selenium Phase and the chalcogenide grain boundaries of the alloy systems are given. A tellurium eutectic was found Present in zone-melted ingots. Similarly, a selenium monotectic was present in ingots. In general, the second phase (tellurium or seleniumn) occubies three different sites; viz., along the chalcogenide grain boundaries, as inclusions within the chalcogenide grain, and on the undersurface of the ingot. The detection limit for the tellurium phase is about 1 u in width. THERMOELECTRIC materials based on Group V (bismuth, antimony) and Group VI (selenium, tellurium) elements have aroused considerable interest in recent years in the practical application of thermoelectric cooling. In many cases, a small amount of excess tellurium (or selenium) was added to the material to optimize its thermoelectric properties. Then the question immediately arises as to the number of phases present in the resultant alloy. In the binary systems of Bi-Te, Sb-Te, and Bi-Se, the congruent melting compositions have been reported to be non-stoichiometric and are represented by Bi~Te respectively. It is to beexpected and known that Bi2Te3 and SbzTe3 crystallize from the melt with an excess of bismuth and antimony in the lattice and that tellurium forms a eutectic.~&apos; The same could be assumed to take place in the pseudo binary systems of (Bi,Sb)zTe3 and Bi2(Se,Te)3 as well as in the system studiedby puotinen5 and other workers. Likewise, BiaSe3 crystallizes from the melt with an excess of bismuth in the lattice and selenium forms a monotectic.~ Therefore, in practice, alloys solidified from the melt often contain a second phase (tellurium or selenium) in one region or another of the solid mass even without the addition of excess tellurium (or selenium). ~u~~recht&apos; studied the thermoelectric properties of (Bi,Sb)2Te3 alloys with excess tellurium and simultaneous additions of selenium. He mentioned that the materials show two phases because of the considerable excess of tellurium or selenium. However, he did not report as to how the tellurium or selenium phase was identified. It is generally believed that the presence of an excessive amount of tellurium or selenium phase in the alloy would adversely affect its thermoelectric properties and its uniformity. Consequently, there is a need for a simple method for the identification of the tellurium and selenium phase. The quantity of the second phase present is usually too small to be detected either by chemical analysis or by normal X-ray techniques. This investigation was therefore carried out, first, to devise a simple metallographic method for the identification of the tellurium or selenium phase coexisting with the chalcogenides and, second, to determine the distribution and specific location of the tellurium or selenium phase in the ingots. EXPERIMENTAL PROCEDURE The starting materials used for the alloy preparations were 99.999 pct pure bismuth, antimony, and tellurium and 99.997 pct pure selenium. The bismuth and antimony were obtained from Consolidated Mining and Smelting Co. of Canada Ltd., while the selenium and tellurium were obtained from Canadian Copper Refiners Ltd. The tellurium was purified further in the laboratory by zone refining. The elements were pulverized in a stainless-steel pestle and mortar. The amounts for the desired composition were weighed out each time on an analytical balance to make up a 100-g sample. Then the sample was introduced into a Vycor ampule (19 by 150 mm), pumped down to a vacuum of 10"5 Torr for 15 min, and sealed off. The ampule was then heated in a horizontal resistance furnace at 800" to 900°C for about 20 hr. During this period the assembly was rocked back and forth several times to ensure good mixing. At the end of the heating period, the ampule was quenched in cold water and then transferred to the zone-melting apparatus described in a previous publications to grow large-size aligned polycrystals. The background and ring-heater temperatures were adjusted to make the freezing solid-liquid interface slightly convex to the liquid. The recorded temperature gradient in the vicinity of the freezing solid-liquid interface was around 15°C per cm. The ampule was moved horizontally at a speed varying from 0.4 to 2 cm per hr so that the ring heater would cover the whole ingot length from end to end. A single zone-melting pass was used for the Bi-Te, Sb-Te, and Bi-Sb-Te ingots. Two passes in the forward and reverse directions were carried out for the Bi-Se and Bi-Se-Te ingots. Six passes in the forward and reverse directions were performed for the Bi-Sb-Se-Te ingot. The zone-melted ingots were found to contain several large crystals, with their basal planes (0001) approximately parallel to the growth axis. Samples of bismuth and antimony tellurides coated with a layer of tellurium, and bismuth selenide coated with a layer of selenium, were prepared for comparison in phase identification. These coatings were made by dropping a piece of the zone-melted ingot into some molten tellurium or selenium under argon atmosphere and allowing them to cool slowly to room temperature. The metallographic specimens were prepared by

Jan 1, 1969

By M. C. Flemings, R. Mehrabian, G. E. Nereo

Analytical expressions derived previously are used to describe quantitatively effects on macrosegregation of some solidification and mold design variables. Al-4.5 pct Cu alloy is used as example. It is shown that abrupt changes in heat transfer rate result in solute-rich or solute-poor "bands". Severe solute-rich bands accowzpany "reheating" that can result, for example, from abrupt formation of an air gap at a mold-metal interface. With planar isotherms, solute-poor regions also result from reduction in cross section through which fluid flow to feed shrinkage takes place. It is suggested that the negative "cone" of segregation often found at the base of large ingots results from this reduction in cross section for feeding. Consideration is given to the positive and negative segregation to be expected in laboratory ingots in association with changes in ingot cross section. Under-riser positive segregation is considered and it is concluded this segregation results when vertical temperature gradient is low in the region of the riser; the segregation is accentuated by the temperature gradient reversal which can occur in inadequately insulated risers ("&apos;hot tops"). Finally it is shown for Al-4.5 pct Cu alloy that the assumption of constant k and p in analytic solutions gives calculated results close to those obtained for the case of ziariable k and 0. A basic "local solute redistribution equation" was derived earlier&apos; for calculation of macrosegregation resulting from flow of solute-rich liquid to feed solidification shrinkage, thermal contraction, and/or other contractions occurring during solidification. This expression relates fraction liquid, g~, to liquid composition, CL, at a given position x, y, z, in a solidifying Eq. [I] is for the general case of three-dimensional heat and fluid flow assuming constant solid (but not necessarily liquid) density during solidification, negligible net solute change from diffusion, and no pore formation. In this paper we first demonstrate application of this equation for calculation of macrosegregation in several examples involving one-dimensional heat and fluid flow. Specifically considered are examples in which segregation results from sudden changes in heat flow conditions. Next, Eq. [I] is used to describe segregation in examples where isotherms remain planar but where fluid flow is no longer one-dimensional. Application of this expression illustrates the type of segregation to be expected in unidirectionally solidified ingots of gradually or abruptly varying cross section. For segregation calculations, A1-4.5 pct Cu alloy is employed with constant k and p Vz = 0.172, 4 = 0.055), eutectic composition (CE = 33 pct Cu). In all examples, fraction eutectic is neglected in velocity calculations; this was previously shown to introduce little error.&apos; A linear distribution of fraction liquid in the liquid-solid zone is assumed. In a concluding section it is shown that the assumption employed of constant k and /3 gives results close to those obtained when the actual (variable) values of k and 0 are employed.

Jan 1, 1969

By M. C. Flemings, R. Mehrabian, G. E. Nereo

Analytic expressions were developed and applied in two preuious papers to predict effects of solidification variables on macrosegvegation. In this paper, experiments are reported to test the analyses quantitatively and qualitatively. Good quantitative agreement is obtained between predicted and measured compositions for the following types of segregation (all experiments were on laboratory-size ingots of Al-4.5 pct Cu alloy): (a) inuerse segregation, (b) composition distribution in ingots solidified with unidirectional heat and fluid flow, (c) simulated centerline segregation, and (d) segregation resulting fro, change in solidification cross section. In addition, qualitative agreement of prediction of theory with experiment is obtained for (a) segregation resulting from sudden change in heat extraction during solidification, (b) centerline segregation in a bidirectionally solidified ingot, and (c) under-riser positive segregation. IN two previous papers,&apos;&apos;2 it was shown that a variety of apparently different types of macrosegregation result from the same basic mechanism. This mechanism is the flow of solute -rich liquid to feed solidification and thermal contractions. Analytic expressions were derived and examples given of their application to predict macrosegregation. Calculations were for Al-4.5 pct Cu alloy. In this paper, we describe a series of experiments designed to test results of the foregoing calculations. All experiments were on laboratory-size ingots (weighing 10 lb or less), of A1-4.5 pct Cu alloy, nominal composition. APPARATUS AND PROCEDURE A typical mold and chilling arrangement used for unidirectional solidification is shown in Fig. 1. Before casting, this assembly was placed in an electrically heated air furnace, and heated to 680" to 700°C. When the plaster reached the desired temperature (usually 1 to 2 hr after the furnace temperature), the mold was filled through a funnel extending from the outside of the furnace into the cavity. The assembly, with liquid, was held for several minutes to allow convection to subside. The coolant, either air or water (depending on the desired solidification rate), was turned on and the ingot solidified. With the four sides and top insulated, heat removal for solidification was through the bottom surface, resulting in a fully columnar structure. Design of the casting cavity was varied, as discussed below, to produce desired macrosegregation effects. However, for castings solidified with unidirectional heat flow, the top was always open to the furnace atmosphere and temperature. The bottom sur- face rested on a chill through which the heat was extracted, and all other surfaces were insulated with plaster. In one case to be discussed, the mold design was changed to obtain bidirectional solidification. In this case, two vertical chills were used (on two faces of the mold) and the bottom was insulated. In a last case discussed, that dealing with under-riser segregation, experiments were conducted in bottom-chilled sand molds at ambient temperature. Except for this one case, all experiments were conducted in the electrically heated air furnace as described above. In many ingots, thermocouples were used to obtain continuous records of solidification. In all cases, these were chromel-alumel couples with output continuously monitored on a twelve-channel recorder. Where thermocouples were employed, these were embedded in the refractory mold walls during mold-making so their heads extended to the vertical centerline. Six to nine couples were employed, placed at intervals from the chill. METAL-MEL TING PROCEDURE Aim analysis was A1-4.5 pct Cu; alloy was prepared from high-purity aluminum (99.9 pct Al) and A1-Cu master alloy (50 pctOFHC Copper, 50 pct high-purity Al). The A1-4.5 pct Cu alloy was cast in 5- to 7-lb ingots for subsequent remelting for the macrosegregation experiments. Melting for ingot casting was as follows. After preheating the melting crucible for an hour or more, the charge was melted and superheated to a temperature of 690" to 720°C. After stirring, the melt was degassed with chlorine for 10 min. After stirring again for 1 min, a gas sample, under reduced pressure, was taken. When the sample was gas-free, the crucible

Jan 1, 1969

By Koichi Kashida, Mayumi Someno, Kazuhiro Goto

The present work has two objectives: first to develop an experimental method which would give some new information on the micromechanism of the reduction of oxides by gases at elevated temperature; second to clari,fy the intermediate phases of the W-0 system, formed during the reduction. A Galvanic cell with the solid electrolyte , ZrO2 . CaO, was imbedded in powders of W03, CuO, or Fe20s with two measuring electvodes and with a reference electrode. The change of the intraparticle gas composition was measured at An oxygen concentration cell consists of two electrodes separated by an electrolyte through which the transference of electrical charge is accomplished predominantly by the migration of oxygen ions. Many works have been reported on the oxygen con- 600" to loo@ C during the reduction by the gas mixture of hydrogen and argon. The intraparticle gas composition changes stepwise with time. The electromotive force of such temporary interruption corresponds to the equilibrium electromotive force of two solid phases formed during the reduction. The intermediate phases of W-0 system, which are formed in a large quantity during the reduction by hydrogen, are W0 2 W02.72 and WOz. The micromechanism of the reduction of oxides by gases is discussed. centration cells with the solid electrolyte of solid solution of zirconia and lime, since Kuikkola and wagnerl demonstrated the possibility of determination of standard free energy of formation of oxides by an electromotive-force method with the Galvanic cell. One of the present authors (K.G.) and his coworkers have applied the cells with the solid electrolyte of zirconia and lime to the following measurements:

Jan 1, 1969

By S. D. Dahlgren

The plasticity of unalloyed P plutonium was studied by evaluating the strain-rate-hardening exponent, m, in the equation Values of m up to 0.33 and tensile elongations up to 680 pct were obtained for ß plutonium formed directly from the a phase. Casl bars that were cycled and tested had lower flow stresses and higher values for than uncycled bars , hut their total plastic elongations were not as great as expected because of premature failure at voids formed during cycling. Flow stresses were higher and m was lower (m = 0.15) for 9 plutonzutn formed from the y phase. ThE strain-rate-hardening exponent for plastic flow of p plutonium was determined as a function of strain rate, test temperature, and heat-treating history, and it was related to the rather large plastic elongation found previously for some plutonium tensile samples. Gardner and ~ann&apos; obtained tensile elongations of up to 620 pct for P plutonium samples which had been completely transformed from the low-temperature a phase to the P phase before testing. Such a large percentage elongation is unusual for unalloyed metals or single-phase alloys.&apos; The a phase is stable below approximately 112°C and (3 is stable between approximately 112" and 190"c.~ Both phases have mono-clinic structures. A temperature of 160°C is 0.58 of the effective absolute melting point, 475°C) of the p phase. The effective melting point is obtained by considering the enthalpy and entropy changes associated with the higher-temperature phase transformations and it is not the same as the experimental melting point of the highest-temperature allotrope of unalloyed plutonium. Nelson, Bowman, and Bierlein observed a striking difference between the ductility, as well as other mechanical properties, of p plutonium formed from the next higher-temperature phase (y) and the ductility of /3 plutonium formed from the a phase. The total elongation of /3 plutonium formed from the y phase was less than 5 pct. Earlier, the creep rate of 13 plutonium formed from the a phase was shown to be 200 to 500 times faster than the rate for 0 formed from y for the same test temperature and stress.

Jan 1, 1969

By Edwin F. Brush, John Chipman

Published data on the activity of carbon in binary and ternary austenite at 1000°C are reviewed. The composition variables are . The coefficient, Vj, has the value for substitutional elements but may be assigned a ualue -1 for silicon to effect a simplification on the mathematical treatment. The activity coefficient PC = is obtained from gas equilibrium data. For the ternary systems studied the Henrian activity coefficient is given by the expression: The principle advantage in cornparison with the analogous expression in terms of atom fractions is found in the fact that for most systems the final term and in many the last two terms are zero. Values of the interaction parameters are tabulated and the graphical comparison of the equation with experirt~ental data is sho,wn in several figures. The thermodynamic properties of the solid solution, austenite, are of considerable importance in the heat-treatment of alloy steels and in other aspects of ferrous metallurgy. Early studies embodied in the Fe-C diagram have been extended to include other binary and some ternary diagrams. Our best hope of obtaining a useful thermodynamic treatment of reactions in complex alloy steels will not be found in quaternary and quinternary phase diagrams but in determinations of activities and interaction relations among the several solute elements. The foundation for such a development was laid by smith&apos; in his early studies of the activity of carbon in Fe-C alloys and his subsequent work on Fe-C-Si, Fe-C-Mn, Fe-C-Ni, and Fe-C-Co solid solutions.2"4 Other observers have also made noteworthy contributions and these will be discussed and correlated here in the light of a somewhat simplified mathematical treatment of solute interactions. Fe-C AT 1000°C In a recent review of the thermodynamic properties of binary Fe-C austenite it was shown that the simple mathematical treatment employed by smith&apos; was capable of describing all the published results on this solution. The treatment utilizes the atom ratio, yc = nC/nF,, rather than the atom fraction. The activity of carbon is proportional to the equilibrium gas ratio, either measured on a scalk in which the activity of graphite is taken as unity or, alternatively, aC may be set equal to yc at infinite dilution. In either case the ratio aC/y = (PC, the activity coefficient. It was found that the experimental data of smith1 and those of Diin-wald and wagner; Bungardt, Preisendanz, and Lehn-ert,&apos; and Scheil, Schmidt, and Wiinnings at temperatures of 800" to 1200° C as well as reliable data on the solubilities of graphite and cementite could be represented by the equation for which graphite is the standard state: The data on alloyed austenite have been obtained at 1000°C. For this temperature, the "Henrian" activity coefficient (which is unity at infinite dilution) is: FOr an interstitial solute such as carbon in austenite it has been shown by statistical mechanicsg that the activity in the infinitely dilute solution is proportional to the ratio of the number of lattice sites occupied by carbon to the number of yet unfilled lattice sites. A variable zz proportional to this ratio serves as a useful measure of concentration:" in which b is the number of lattice sites provided by each atom of solvent; in austenite b = 1. The Henrian activity approaches zz at infinite dilution and the ratio a is the activity coefficient. The effect of a change in concentration is then given by: It has been shown further that in some systems the derivative in Eq. [4] remains constant over a wide range of concentrations. In such a case second-order terms become negligible and the Henrian activity coefficients is:

Jan 1, 1969

By David L. Holt

The contribution of grain boundary shear to total elongation, CS/E&apos;, has been measured in an Al-Zn eu-tectoid alloy that was quenched from above the invariant temperature, then annealed at 250° C to a grain size of&apos; 1.8 p. At 250°C, ks/E&apos; is low at both high and low strain rates, but reaches a maximum, estimated as 60 pct at an intermediate rate of 5 X 10 per rnin. Rate sensitivity, as measured by the index m = a log a/a log E&apos;, follows the same trend, and furthermore the maximum values of m and -cur at approximately the same strain rate. This result, combined with the metallographic observation that boundary migration enhances boundary shearing, is interpreted as supporting a previous suggestion that the high rate sensitivity characterizing super-plasticity is the result of combined boundary shearing and migration. It is suggested that the latter event relieves stress concentrations at triple points, and smoothes boundaries so that stress is governed largely by a viscous boundary shear. GrAIN boundary shear has been considered in relation to superplasticity in several recent papers.&apos; The problem has been to explain the high strain rate sensitivity of flow stress, and the variation of rate sensitivity with strain rate (E&apos;) and grain size (L). The requirements for superplasticity, small L and high T, suggest the reasonableness of an approach to high rate sensitivity involving grain boundary shear. Further support came from experiments on the A1-Cu eutectic alloy,&apos; where it was found that strain rate sensitivity of cast material annealed to produce an equiaxed, micron-size grain is always low; taking as an index of rate sensitivity m = a log a/a log <, m < 0.3. However, m in hot-worked alloy of comparable grain size can be as high as 0.7. In the cast and annealed material, each phase is a single crystal, the only boundaries are interphase boundaries, and it is, consequently, geometrically impossible for boundary shear to contribute to deformation in any major way. Other observations (for hot-worked material) were a-L at constant (low) strain rates and indications that the rate of recrystallization was enhanced as strain rate increased. As a result of this work, it was proposed that high rate sensitivity arises from a deformation mode of boundary shear associated with boundary migration. Migration serves to relieve stress concentrations at triple points, and smoothes boundaries so that they assume properties of fluid films. On the other hand, the low rate sensitivity observed at high and low strain rates reflects deformation of bulk material. Measurement of the variation of grain boundary shear with strain rate and m have not yet been made. Such measurements are important, especially in view of a proposal, differing in detail from the above, that high m arises merely from a transition between a grain boundary shear mode of deformation at low rates to a transgranular mode at high rates.2&apos;4 In the present work, the contribution of boundary shear to total deformation is measured and in addition metallographic observations are made on surfaces of deformed specimens to look at the interaction between boundary shear and migration. The Al-Zn eutectoid alloy was chosen for its homogeneous, fine-grained structure, which is obtained readily without hot-working. It has also been the subject of a previous phenom-enologically directed study. EXPERIMENTAL Material. Compression specimens, cross section 4 by + in., length \ in., were machined from a sand-cast ingot of composition 77.5 wt pct Zn, 22.5 wt pct Al. (The melt was prepared from 99.9 pct Zn and 99.99 pct Al.) After homogenization at 375°C for 50 hr, the specimens were quenched in brine and removed before the heat evolution that accompanies de -composition of the high-temperature phase.5&apos;6 The resulting microstructure, see Fig. l(a), was too fine for grain boundary sliding to be easily studied; coarser structures were obtained by annealing for various times at 2 50°C. Annealing was terminated by a brine quench. Final average intercept lengths between all grain boundaries (both interphase and those lying in a phase), L, were: 0.5 p [annealed for 15 min, Fig. (a)], 0.8, 1.1, and 1.8 p [Fig. l(b)l. Testing Procedure. An Instron machine was used for most of the compressive deformation. Tests were of two types: those in which crosshead velocity was changed in steps to measure m as a function of strain rate15 and tests at constant velocity to a fixed (engineering) strain of -0.2 (20 pct). Stress reached a steady-state value (a) which was plotted, on a logarithmic scale, against log strain rate (E&apos;). An alternate and equivalent evaluation of m was to take the slope of the log o vs log 6 curve. Time at temperature before testing was 15 min. Strain rates covered by the Instron (4 x lo-&apos; to 4 x 10&apos; per min) were insufficient; at a higher rate of 5 x lo2 per min a gas-operated testing machine was used, the gas driving a piston to compress the specimen at a controlled velocity.&apos; To obtain points on the log a vs log E&apos; curve at low rates, specimens were compressed by a dead weight. strain rate was an average value computed by dividing strain at the end of test by loading time. In some tests strain was measured at fractions of the loading time; creep rate was found to be reasonably constant.

Jan 1, 1969

By P. C. J. Gallagher

A number of recmt determinations for the ratio of extrinsic to intrinsic stacking fault energy in fcc solid solutions are examined. Some of these arise from incomplete analyses which can yield only approxi?nate values for the ratio. Reliable results, on the other hand, obtained using extrinsic-intrinsic fault pairs, show that the extrinsic and intrinsic fault energies are essentially equal in several materials. There is some reason to believe that this finding is of general applicability to fcc elements and alloys. A wide range of values has been obtained for the relative magnitudes of the extrinsic and intrinsic stacking fault energies (yext and yint, respectively) in recently published studies in a variety of materials. In contrast, Hirth and Lothe&apos; using a central force model have shown that out to tenth nearest-neighbor interactions the perturbation in energy caused by both types of fault is the same. Although the model used is not completely valid in metals, there is nevertheless some indication that marked variations of yext/nnt should not be observed from one material to another. In early work in Cu-A1, Cu-Ge, Ni-Co, and stainless steel all the deformation faults observed in the electron microscope were found to be intrinsic in nature, which led to an attitude that the extrinsic fault energy must be considerably greater than the intrinsic. Extrinsic faulting arising from deformation has, however, more recently been directly observed in Au-4.8 at. pct n;~ Ag-6 at. pct Sn and Ag-8 at. pct sn; Ag-7.5 at. pct In and Ag-11.8 at. pct 1n;"&apos; pure silwer and Ag-0.5 at. pct ~n;&apos; and Cu-22 at. pct Zn, Cu-30 at. pct Zn, and Cu-7.5 at. pct ~1.l&apos; Multilayer loops containing extrinsic faulting have also been observed in quenched aluminum." While peak asymmetries in X-ray faulting probability studies were generally attributed to the presence of twins,Lelel2 has recently reinterpreted earlier X-ray data in Ag-Sb alloysU in terms of the presence of extrinsic faulting. The determinations of yeXt/yint made from the above studies are shown in Table I, with a brief description of the techniques employed. A number of the methods utilized are deficient in one or more respects, and the reliability of the values listed will be discussed. ~ele&apos;~ recognizes that his approximate determinations of yext/yint assumes equal numbers of extrin-sically and intrinsically faulted dislocations. It is well-known, however, that such an assumption is not at all justified since extrinsic faulting has but rarely been observed in samples studied in the electron microscope. The only conclusion that should be drawn from the X-ray results at present is that the total intrinsic scattering cross section (i.e., the product of the width of the intrinsically faulted dislocations with their density) is approximately ten times greater than for extrinsic faulting in these particular samples. An important point is that the relative magnitudes of the energies cannot be inferred from results of this type, unless the intrinsic and extrinsic faults form with equal ease. One must recognize that, although a formation barrier may restrict the amount of extrinsic faulting which occurs, this in no way implies that the extrinsic and intrinsic energies should be different. It is unlikely that a worthwhile estimate of the relative densities of extrinsically and intrinsically faulted dislocations can be made at the high deformations present in X-ray samples. ~oretto,&apos;~ from a statistical argument applied to the nonobservation of extrinsically faulted tetrahedra out of a large sample, concluded that yeXt/yint could not be less than -4.5. However, the present author feels that a high-energy formation barrier as just supposed also explains this finding satisfactorily and that no conclusion can possibly be drawn concerning the actual extrinsic stacking fault energy. The same argument also serves to explain the fact that extrinsic faulting has been relatively little observed in the electron microscope. Extrinsic-intrinsic node pairs and isolated extrinsic nodes were first reported by Loretto~ and subsequently by Ives and Ruff,&apos; Gallaher,&apos; and Gallagher and Wash-burn.&apos; Ives and Ruff&apos; found a wide spread in the ratio of extrinsically to intrinsically faulted area in the node pairs they observed, and drew the very tentative conclusion that yeXt/yint 2 2. They recognized that a straightforward comparison of the size of the faulted areas could provide no more than a qualitative result without a theoretical analysis of the dislocation geometry associated with extrinsic faulting. A theoretical

Jan 1, 1969

By R. W. Smith, M. E. Brookes

The temperature at which th: quenched 0&apos; (ordered bcc) phase transfo.rms to the ( (complex hexagonal) phase on heating has been examined for certain Ag-Zn and Ag-Au-Zn alloys using an electrical resistivity method. The transition temperature decreases by approxinzately 7°C as the zinc content is reduced by 1 at. pct. The critical temperature, Tc, for long-range order has also been measured for a number of Ag-Au-Zn alloys. Extrapolation to zero gold content suggests that Tc decreases by approximately 5°C per 1 at. pct reduction in zinc content. The 0 phase of the Ag-Zn system exists over a narrow range near the composition Ag zn,&apos; Fig. 1, and has a bcc structure which is disordered above 272°C. Slow cooling to temperatures below the range 274"/258"C allows the complex hexagonal phase, r° to form.2 This is the stable structure at room temperature.3 Upon quenching 0 -phase alloys from high temperatures, it is found that the 0 — (&apos; transformation is suppressed; but ordering takes place (i.e., 0 -0&apos;). This ordered structure can also be obtained by the room-temperature aging of cold-worked (0.4 Deformation of the 0&apos; phase gives rise to a fur- ther hexagonal phase ( (of nearly ideal axial ratio) which reverts to b&apos; within a few hours.= In view of this, it has been suggested that the equilibrium struc-

Jan 1, 1969

By Carl Altstetter, Roy Adams

Measurements of enthalpy changes and transformation temperatures me reported for the fcc -hcp marten-sitic transformation in pure cobalt single and multi-variant crystals andpolycrystalline specimens. From these measurements the free-energy difference between the two phases was calculated for both heating and cooling and for repeated cycles through the transformation. For heating of single crystals the average value of enthalpy change was 113 cal per mole but on cooling it was 84 cal per mole. This difference is interpreted in terms of lattice imperfections which are introduced during transformation. A calorimeter suitable for measurement of enthalpy changes of athertnal transformations is described. COBALT exhibits an allotropic phase transformation at around 417°C&apos; where an hcp phase (a = 2.507, c = 4.0686, c/a = 1.6228)~ stable at low temperatures transforms to an fcc phase stable up to 1495°C. The transformation has a hysteresis of about 30°C— the exact magnitude of the hysteresis depending on the number of transformation cycles, grain size, impurity content, and prior treatment.&apos;-6 The athermal character of the transformation and surface relief due to transformation shear indicate that the transformation is martensitic. Pole mechanisms for the transformation have been discussed by Sebilleau and Bibrin, Bilby, Seeger,&apos; and Basinski and Christian. These authors have proposed that the cooling transformation is effected by rotation of a Shockley partial dislocation around a pole dislocation with a screw component of 2a/3(111). Such a rotation would change the stacking of close-packed planes from the abcabc fcc stacking to the ababab hcp stacking sequence. This mechanism would predict a lattice correspondence of {111}fcc~~(0002~cp, have been observed. One would expect the heating transformation to occur similarly, involving dislocation movements on the (0002)hcp. However, a high density of pole dislocations and high dislocation ve- locity are required to account for microscopic observations. Furthermore, the high back stress developed on the Shockley partial dislocation as it rotates toward the sessile dislocation associated with it makes it doubtful that this is in fact the transformation mechanism. A mechanism for spontaneous nucleation of partial dislocations1*l2 seems more reasonable. If a single crystal of cobalt is cycled through the transformation but never heated above 600°C, it will remain single."13 However, if the specimen is heated through the transformation and held for a short time at or above 1000°C before cooling, the cooling transformation usually involves all four {lll}fcc habits, resulting in neighboring orientations of hcp differing by 70.5 deg. This condition is termed multivariance and has been observed by Bibring and sebilleau5 and Nelson and ~ltstetter.&apos; The latter authors used single crystals of cobalt grown in an electron beam zone refiner. They proposed that the strongly directional cooling caused the initial cooling transformation to occur on the {lll)fcc most nearly perpendicular to the temperature gradient. Transformation shear on one set of {lll}fcc planes immobilized dislocations on the other {lll}fcc planes involved in the initial transformation. Since there is only one (0002)hcp plane orientation, the product of subsequent heating transformations could have no more than two variants—the original fcc orientation or its twin. If, however, the specimen was annealed at a high temperature, such as 1000°C, there would be relaxation of dislocations into low-energy configurations, and in the absence of a sharp temperature gradient transformation on all four (111 Ifcc planes could occur. If a dislocation mechanism is responsible for the transformation, the perfection of the lattice should affect the reversibility of the transformation. Mutual constraints of neighboring grains in polycrystalline material would inhibit dislocation movement resulting in increasingly sluggish transformation. Indeed, Nelson and Altstetter found the M, to decrease and the A, to increase in going from single crystals to multi-variant specimens to polycrystals. The effect is enhanced by decreasing the grain size of a solid specimen or a powder.6&apos;14 Data on the free-energy difference necessary to initiate the transformation gives information about the driving force of the transformation. This is an important consideration in the understanding of the transformation mechanism. The free-energy change associated with the transformation of the hcp phase to the fcc phase at some temperature T different from the equi-

Jan 1, 1969