Search Documents

By Akira Yazawa, Alžbeta Gubcova

RECENT lead-smelting techniques have tended toward use of high-grade concentrates with minimum amounts of flu, accompanied by profitable employment of roast reaction.1-5 In order to better understand these smelting processes, thermodynamical calculations have been carried out on the Pb-S-0 system, and the results obtained have been presented in isothermal reaction diagrams in which log Ps, and logPo, were the axes. The diagram thus constructed seems to be more convenient to grasp the equilibrium aspect than those published hitherto.6-8

Jan 1, 1968

By R. C. Hall

We are pleased to confirm some of the mechanical data presented by Dr. Detert on the 18 pct Ni marag-ing steel.15 Bar 3 in. in diam and containing 0.02 wt pct C, 18.5 Ni, 8.9 Co, 5.1 MO, 0.55 Ti, 0.05 Al, and balance iron has been examined in our laboratory. Cylinders 0.5 in. high and 0.4 to 0.5 in. in diam were cold-compressed from 3 to 40 pct reduction in height. Table VI shows some typical data; deformation in-

Jan 1, 1968

By J. Weertman

The paper of Gifkins and Snowden considers the interesting but difficult problem of determining the stress dependence of secondary (steady-state) creep at low stresses. These authors have concluded that at stresses below 250 psi (2 x 107 dynes per sq cm) the secondary creep rate of lead is proportional to the stress (viscous creep) and is not proportional to the stress raised to about a fifth power. The experimental data considered by them were obtained on tests conducted at room temperature and at 50°C. The lowest stress employed was 50 psi (3.5 X 106 dynes per sq cm). The authors pointed out the main difficulty in determining the stress dependence of creep at low stresses. The creep tests must be run for very long lengths of time. However they made no estimates of when an experimental creep rate determination must be rejected because it does not represent a true steady-state or minimum creep rate. In order to be certain that a creep curve is within the steady-state region, the total creep strain should be of the order of 0.1 to 0.2. For a creep test of a year's duration this requirement implies that a secondary creep rate smaller than about 10-3 per hr cannot be measured reliably. The corresponding creep rate for a 10-year test is 10-6 per hr. The creep rates of the tests that were considered by the authors to prove the existence of viscous creep were of the order of or less than 10-6 per hr. One can conclude reasonably that this data does not prove unambiguously that large strain steady-state creep rate of lead is proportional to the stress in the stress range of 50 to 250 psi (3.5 x 106 to 2 X 107 dynes per sq cm). Another technique can be used to obtain the stress dependence at low stress levels. The creep rate is a very sensitive function of temperature. The creep rate can be increased by very large amounts merely by increasing the temperature. We carried out steady-state creep tests on lead single crystals25 at temperatures up to 320°C. We were able to obtain creep rate data down to stresses as low as 35 psi (2.5 x 10' dynes per sq cm). Our smallest creep rate was 8 x 10-5 per hr. Thus we obtained large strain, steady-state creep rates to even lower stresses than were considered by Gifkins and Snowden. No evidence was seen for viscous creep. The creep rate was proportional to the stress raised to about a 4.5 power down to the lowest stresses. Since there is no reason to believe that changing the temperature should change the stress dependence of steady-state creep, we feel that large strain viscous creep does not occur in the stress range quoted by the authors for lead single crystals or large-grain polycrystalline samples of lead. This conclusion does not imply that viscous creep may nat occur in a lower stress range or in the same stress range for fine grain material or at creep strains very much smaller than 0.1. Support by the U.S. Office of Naval Research is acknowledged. Authors' Reply R. C. Gifkins and K. U. Snowden We thank Dr. Weertman for his discussion and although, as we hope to show, we do not agree with his reservations, we do concur in stressing the importance of ensuring that creep rates are reliably obtained. Dr. Weertman appears to be content to accept n = 1 for low stresses with fine-grained material but not for single crystals. We believe our results show that the former result cannot be accepted without also accepting the latter. We will also show that the probable errors in our minimum creep rates are insufficient to alter our conclusions, that the criterion proposed by Dr. Weertman is arbitrarily restrictive and his alternative experimental approach possibly invalid. 1) A principal result of our Fig. 1(a) is that n = 1 for polycrystalline specimens at room temperature and 50°C for stresses below -250 psi. There was evidence that crystal slip and grain boundary sliding contributed approximately equaily to the overall strain in this low-stress regime. This implies that either a) grain boundary sliding controls slip within the grains or b) both grain boundary sliding and crystal slip independently occur according to mechanisms which give n = 1. Alternative a does not seem acceptable, so we were forced to consider b. This led us to reexamine work on bicrystals by Strutt and Gifkins and plot curves Fig. l(b). Previously Strutt et al. (loc. cit.), had merged these points with others using the Zener-Holloman parameter and thus, we now believe, had been led to overlook the behavior where n = 1. Curves C and D in Fig. l(b) did appear to confirm the hypotheses that n = 1 for crystal slip at these low stresses and the sliding curve F was similarly of the expected form. It was comparatively easy to find a quantitative theory to account for n = 1 for sliding and the similarity of curves C and D to curve F (all obtained from the same set of specimens) led us to feel that the single-crystal curves were valid. 2) We believe the secondary creep rates for both the polycrystalline and single-crystal specimens to be in error by factors c2. In Fig. 6 creep curves for polycrystalline specimens of lead(1) and lead(II) are reproduced as curves a and b, respectively, and curve c is for a single crystal at 100 psi. It is clear that, although the attainment of secondary creep rate takes 2 years for a and 150 days for b, thereafter the curve is linear for periods of 7 and -1 year, respectively. The single crystal has a linear portion commencing after 20 and extending to 90 days. Creep extension was measured directly using a traveling microscope reading to 0.01 mm on gage lengths marked on the specimens; the gage lengths

Jan 1, 1968

By Ronald C. Koo

SMALL additions of potassium, aluminum, and silicon as oxides (referred to as dope) to tungsten have been known for several decades to have a very large effect in raising the recrystallization temperature and altering the recrystallized grain morphology of heavily drawn wire. Previous investigations of this doping effect1 6 have led to various speculations but all failed to deduce unequivocally the physical state or the mechanistic influence of the dopant. Recently, walter7 has observed, by transmission electron microscopy, strings of "fine particles" in heavily rolled and annealed doped tungsten sheet and concluded that they were particles of mullite (Al6Si2013). His conclusion was inferred from the electron diffraction patterns of large particles ex-trated by carbon replication of fractured surfaces of slabs rolled from sintered ingots. That the strings of "fine particles" in heavily rolled and annealed sheet were also mullite was an assumption for which no evidence was presented. A similar investigation has been carried out in this laboratory over the past 2 years on a variety of tungsten wire: doped tungsten, and undoped tungsten, as well as 218 doped tungsten purchased from the Lamp Metals and Components Department of the G.E. Lamp Division. Although the electron microscope observations are similar to those reported by Walter, the writer has reached a different conclusion concerning the physical state of the dopant in annealed tungsten. The purpose of this communication is to present evidence that the strings of "fine particles" observed in transmission specimens are not mullite, but are voids. In the present investigation, tungsten wire, 0.88 mm in diam, was rolled into ribbon 0.2 mm by 125 µ thick. The ribbons were annealed in vacuo at a pressure of -1 x 10-5 Torr and then thinned electrolytically in a 2 pct KOH solution. Fig. 1 shows a typical transmissicjn electron micrograph of a sample annealed at 1700°C for 30 min which reveals rows of closely spaced voids having a light contrast (referred to as "fine particles" by Walter) aligned parallel to the working direction. Stereo micrographs (by D. Moon, Westinghouse Research and Development Center) have shown that these voids do not exist on the surfaces of the specimens as a result of thinning but are within the thin-foil specimens. The following experiments have been performed to verify that these are indeed voids. I) Effect of Annealing. Voids are absent in unannealed specimens and appear only after annealing. This behavior cannot be explained if they were solid particles such as mullite. Precipitation (in an unquenched specimen) upon annealing requires solute elements having decreasing solubilities with increasing temperature. Although the solubility of aluminum in tungsten does decrease with increasing temperature, precipitation requires a concentration in excess of 1.4 wt pct,8 which is at least two orders of magnitude higher than the aluminum content in doped tungsten. It is also unreasonable to assume that particles aligned in rows initially present in unannealed specimens were too

Jan 1, 1968

By V. K. Grotzky, F. J. Fraikor

A number of investigators have noted the importance of various precipitation reactions on the properties of commercial-purity beryllium.1-5 Carrabine, for example, has demonstrated the interaction of iron and aluminum on the mechanical properties of hot-pressed polycrystalline beryllium.6 He considered the determining factor on elevated-temperature properties to be a reversible AlFeBe, precipitation reaction which could be directly related to the Fe/Al impurity ratio of the beryllium. Any aluminum in excess of the stoichio-metric amount required to form A1FeBe4 was presumed to remain primarily in the grain boundaries. We have observed, by transmission electron microscopy, such a grain boundary effect after various heat treatments of beryllium sheet. The starting material in this investigation was bare-rolled beryllium ingot sheet (0.001 to 0.005 in. thick), in contrast to the hot-pressed powder utilized by Carra-bine. It should be noted from the ingot sheet analysis shown in Table I that the Fe/Al weight ratio was near unity which would provide some free aluminum although the total aluminum and iron content was still small. Fig. l(a) shows some aluminum or aluminum-rich precipitates in the grain boundaries of a beryllium foil which had been quenched from 1200°C into brine at -TPC. Similar grain boundary precipitates were found, particularly at grain boundary intersections, in other samples quenched from 1100° and 1000°C. The length of annealing time at these elevated temperatures did not appear to substantially alter the amount of grain boundary precipitates, thus attesting to their relative insolubility. No matrix precipitates were observed other than the normal BeO and Be2C inclusions. A microprobe analysis of the larger precipitates at

Jan 1, 1968

By W. A. Nordland

BISMUTH can be readily grown into single-crystal form using the so-called "quick freeze" growth method.' Some advantages of this method are that very large crystals can be grown in a relatively short time, and with comparatively simple apparatus. We have found the general quality of crystals grown by this method to be not as good as that of "pulled" crystals.' This communication describes a "quick-freeze"

Jan 1, 1968

By J. Warner, M. J. Marcinkowski, B. Cullen, T. Prevender, J. Mcllwain, T. Carnahan, J. Pahlman, J. Demel, J. Munford

The following analysis was an outgrowth of a midterm examination given by one of the authors (M.J.M) in a course entitled Metallurgy 541 "Applications of Dislocation Theory'' at Iowa State University. Fig, l(a) shows the three-layer sequence of atoms lying in the (111) plane which characterizes the fully ordered AB3 fcc crystal, i.e., LI2 type. If any two-layer sequence of close-packed planes is constructed from Fig. 16)' the structure would be hcp of the DO,, type. Consider for convenience that the circled atoms lie in the plane of the drawing while those shown square and triangular lie immediately above and below the plane of the drawing, respectively. It has already been shown' that if a Shockley partial dislocation is passed over the plane of the paper, then the square atoms in Fig. 1(a) are displaced from their original positions as shown by the arrows in Fig. l(b). Not only is an intrinsic fault produced but the crystal is simultaneously disordered across the slip plane. In particular, 2AB bonds are converted to 1AA and 1BB bonds when referred to the area outlined by the solid line in Fig. 1. If, on the other hand, a Shockley partial is allowed to pass immediately above and immediately below the plane of the paper so as to displace the atoms in opposite directions as shown in Fig. l(c), an extrinsic fault is producedZ and twice the amount of disorder is introduced into the crystal as compared to the dislocation-produced intrinsic fault, if only broken first-neighbor bonds are counted. If an intrinsic stacking fault is produced by removing a close-packed plane of atoms in Fig. l(a) and displacing the two adjacent planes by 1/3a0[111], no wrong nearest-neighbor bonds are produced. Similarly, if an extra half plane of atoms with the same arrangement as any of those shown in Fig. l(a) is inserted between any two planes in the same figure, then an extrinsic stacking fault is produced in which once again no wrong nearest-neighbor bonds are produced. The latter two faults are bounded by Frank partial dislocations. Since, at least in the alloy CU3AU,3 nearly all of the ordering energy resides in the first nearest-neighbor interactions, intrinsic and extrinsic stacking faults and associated APB's of the Frank type are expected to have significantly lower energy than those of the Shockley type in CusAu in particular and quite likely in L1, and DO19 type alloys in general. Also, because of the higher degree of disorder plus loop interaction

Jan 1, 1968

By R. DiBella, H. Biloni, G. F. Bolling

A study of the detailed relationship between solidification substructure and microsegregation in dilute tin alloys has been continued. New observations reveal that depressions form at the solid-liquid interface not only as an initial form of segregation but throughout the range of increasing constitutional supercooling. In order to show the importance of this node formation it proves convenient to make two simplifications. Although substructure changes continuously as constitutional- supercooling increases,&apos; the following discrete stages may be used for discussion: 1) the planar solid-liquid interface; 2) ordered nodes, or depressions at the interface;&apos; 3) two-dimensional or elongated cells;3 4) regular or hexagonal cells;4 5) distorted cells, or hexagonal cells with branches;4 6) dendrites. It is also convenient to ignore observations made near grain boundaries which interact with the cellular substructure.2-4 And, it should be clearly pointed out that all singular depressions at the solid-liquid interface will be called nodes. Heretofore, the term "node" has been used without clear justification to describe both the truly isolated depression and the junction of cell walls appearing in a later stage of segregation. This usage is continued because it seems proved by the observations. Standard systematic growth of tin crystals was undertaken over a broad range of constitutional supercooling using various added solutes. Except for one or two chance observations, this research failed to reveal any appreciable range of growth conditions over which a disordered array of isolated nodes exists. Even though &apos;some finite range of random instability must exist, it is now concluded that stage one passes very rapidly to stage two. Next, after confirming that stage two passes to stage three by the elongation of isolated nodes: a reexamina-tion of stages three, four, and five followed. Standard metallography showed that pockets of higher solute concentration (for ko < 1) became more numerous in cell walls the more severe the growth conditions. These high-concentration regions must arise from depressions; however, in stage three their segregation is not distinguishable from the segregation of the nodes which join cell walls together. It seems, therefore, that the presence of nodes is causal in the formation of new cell walls, and that the mechanism by which stage three passes to stage four is initiated by nodes. If a similar reasoning is followed, the separation of stage four from stage five appears to be just a convenience that has arisen from the decanted interface observations made in the past. The present observations again show that more severe growth conditions relate directly to an increase in the number of nodes in the hexagonal cell walls. These nodes apparently serve as the origin for lines of segregation extending within the body of a cell, and can thus be assumed to represent the earliest form of branching,* as shown here in Fig. 1. The examination of stage six involved decanted-interface structures as in Fig. 2, which are similar to those considered by Rutter and Chalmers,5 Plaskett and winegard,6 and Hellawell and Herbert,7 and where distinct protrusions are each surrounded by a depressed boundary. The corresponding microsegregation behind the interface, but not far enough behind to allow a smearing of the segregation because of diffusion,4 is shown in Fig. 3. Both the shape of the decanted interface and the segregation pattern behind it

Jan 1, 1968

By N. A. D. Parlee, E. M. Sacris

OXYGEN solubility in liquid silver has been well studied.1-6 It has been established that the solubility follows Sieverts&apos; law reasonably well5&apos;6 up to po2

Jan 1, 1968

By Olav Idelgeland

FROM indirect evidence point defects have been believed to influence the fatigue behavior of metals and alloys.1,2 Measurements of changes in resistivity made after cyclicstressing at low temperatures indicate that a very high density of defects is present, part of which may be removed by room-temperature annealing.3&apos;4 These results are consistent with the generation of a high number of point defects at low temperatures. Corresponding measurements made after room-temperature stressing did not show any recovery effects.3 This was attributed to a stabilization of the dislocation network after a few thousand cycles without generation of vacancies. In this laboratory resistance measurements of copper single crystals, shaped as normal fatigue specimens ,5 have been made with a Kelvin double bridge at room temperature during and after cyclic stressing at 80 cps by comparing with a dummy at a close and constant temperature. Temperature differences were measured to ±0.025°C with fine thermocouples taped onto the gage lengths. Potential leads were tin-soldered into small holes in the gage length to give stable values. Corrections for differences in temperature and changes in mean length of the specimen were made according to the formula:

Jan 1, 1968

By T. L. Berger, M. C. lnman, M. F. Rose

AFTER shock loading, mechanical twinning has been observed by several authors1-4 for a variety of fcc metals and alloys. It has been shown2,3 that low stacking fault energy materials deform primarily by the formation of mechanical twins and stacking faults during dynamic loading. Further, Inman, Murr, and Rose2 have shown that a correlation exists between the average hardness and the volume percent of mechanically twinned material in stainless steel. It has been observed in nickel&apos; that mechanical twins are produced during shock deformation if the peak pressure associated with the shock wave is greater than 350kbar. It is the purpose of this work to discuss shock-hardening mechanisms in nickel and to determine to what extent the initial condition influences the final response. The "flying plate technique" was used to generate planar shock waves with peak pressures as high as 710 kbar. The details of the technique5" have been discussed elsewhere and will not be repeated here. The material used in this study was polycrystalline nickel of 99.98 pct purity and had an initial grain size of 0.18 mm. Prior to any deformation, all samples were given a stress-relief anneal of 6 hr at 600°C in a vacuum furnace. After annealing, the initial condition of the samples was changed by cold rolling individual samples from 0 pct to a maximum of 80 pct reduction in thickness. This provided samples with controlled initial histories for shock deformation. Immediately after shock loading, the specimens were slowed and quenched by suitable arrangement of buffer material and water bath. This technique minimized specimen damage and the effects of shock heating on the samples. Since deformation twins in high stacking fault energy fcc metals are only produced at high stress levels, the initial defect configuration prior to mechanical twinning should greatly influence the number of twins formed. In fact, Zukas and Fowler7 have found that prestraining in iron and steel greatly suppresses the formation of mechanical twins during the shock deformation process. Following the methods of Warren and warekois8 and Cohen and Wagner,9 X-ray diffraction profile analysis was performed to determine the incidence of stacking faults and mechanical twins in the deformed nickel. Analysis was made on samples in the annealed, cold-rolled, shock-loaded, and cold-rolled—shock-loaded states. The estimated accuracy of the peak position measurements is *0.01 deg in 2?. Therefore the accuracy in the measurement of the relative peak shifts between the (400) and (222) diffraction line profiles is on the order of *0.04 deg in 2?. None of our measurements indicated peak shifts greater than 0.04 in 28. This would correspond to a fault density of -1 pct and sets the limit on the sensitivity of our instrument. Similarly, analysis of the shift in the center of gravity of the diffraction line profiles yields the minimum measurable value of the twin fault probabilitya to be -0.002. Fig. 1 shows mechanical twinning in nickel shock-loaded to 400 kbar. In this photograph, fault intersections are clearly visible. The thickness of the individual fault lamellae was determined from the image widths of their lines of intersections by the

Jan 1, 1968

By N. A. D. Parlee

THE gas-liquid metal diffusion cell technique1-3 has been used to study the diffusion of nitrogen in Fe-C alloys. The actual apparatus and technique were the same as employed by the authors to study the diffusion of hydrogen in liquid iron.&apos; The diffusion cell&apos; was formed by immersing an alumina tube in a stagnant melt contained in an alumina crucible. The tube (hereafter called absorption tube) contained nitrogen gas at 1 atm and was connected to a sensitive gas-measuring system. The change in the average concentration of nitrogen in the melt in the absorption tube was determined by measuring the rate of absorption of nitrogen in the diffusion cell. Eq. [I], 1-3 derived from the appropriate solution of Fick&apos;s second law, was used to examine the data and to calculate chemical diffusion coefficients (D) in situations where the absorption appeared to be controlled by diffusion in the liquid: where V is the stp volume in cu cm, d is the diffusion cell diameter in cm, p, is the density of the melt in g per cu cm, pg is the stp density of nitrogen in g per cu cm, C, is expressed in wt pct, D is in sq cm per sec, and t is in min. According to Eq. [I], a plot of v vs vt should yield a straight line passing through the origin if absorption is controlled by diffusion in the liquid. The slope of this line could be used to determine the value of D under appropriate conditions. As discussed in earlier reports,1,2 the initial thermal expansion of the gas upon immersing the absorption tube distorts the line in the initial stages of a diffusion experiment. However, this is not a serious problem since the slope of V vs 6 plots can be determined from the data points obtained after the initial expansion ceases. The volumes reported on the data plot presented here represent absorption after this thermal expansion ceased. Several preliminary diffusion experiments were attempted with melts of pure iron and with Fe-C alloys containing up to 0.6 pct C. In these cases, no or very little absorption was observed. The only data obtained were for an Fe-C melt which was in addition thoroughly deoxidized by successive hydrogen and vacuum-degassing treatments. These treatments introduced an appreciable amount of aluminum into the melt and reduced its omgen content. The original charge material for this melt contained 0.23 pct s and approximately 0.001 pct &. After repeated hydrogen and vacuum-degassing treatments two diffusion experiments were performed. The analysis of the melt

Jan 1, 1968

By G. R. Leverant

RECENT investigations1,2 of the low-cycle fatigue behavior of pure copper under strain cycling conditions have shown that a unique saturation stress level is eventually attained for each value of applied plastic strain amplitude. The dislocation substructure of copper at the saturation stress level is in the form of cells whose diameter decreases with increasing saturation stress level, i.e., plastic strain amplitude. In addition, both the saturation stress level and the cellular substructure are completely reversible. That is, the saturation stress level and accompanying cell size are determined solely by the present conditions of testing and are, therefore, independent of any previous cyclic history. The mechanical and substructural reversibility observed in pure copper, which has a relatively high stacking fault energy, is not observedz for Cu-7.5A1 which has a much lower stacking fault energy. This implies that easy cross slip (wavy slip) is a prerequisite for reversible fatigue behavior. A dispersion of fine, nondeformable particles of a second phase in a metal matrix tends to stabilize substruc- tures introduced by forming operations during subsequent elevated-temperature heat treatments.3 This raises the question as to whether a finely dispersed second phase can induce an element of irre-versibility into the cyclic deformation behavior of a wavy slip material at room temperature, due to partial stabilization of the initially introduced substructure. Since nickel has a higher stacking fault energy than copper,6 pure nickel should be characterized by cyclic reversibility. Therefore, TD-nickel* (Ni-2ThO.r) was

Jan 1, 1968

By John R. Hutchins

1ms is a report of nickel (wire) oxidation studies designed to test the validity of Valensi&apos;s1 equation developed for the kinetics of a reaction controlled by diffusion in a continuously growing product layer on cylindrical surfaces. The equation is:

Jan 1, 1968

By S. F. Bartram, G. A. Chase

COMPUTED X-ray diffraction powder patterns of YNi3, YNi5, and CbBe3 are available on request from either author.

Jan 1, 1968

By Arnulf Muan, R. E. Johnson

Equilibrium CO2 /CO ratios of a gas phase coexisting with selected oxide phase assemblages of the system CaO-"Fe0"-SiOz and vnetallic iron have been determined at 1080°C. The data obtained are used to calculate activity-conzposition relations for metasili-cute, orthosilicate, and CaO- "FeO" solid solutions, and to determine a new value for the free energy of formation of C&S&0,. The metasilicate solution is practically ideal, whereas large deviations from ideality exist in the othe,r two solid solutions. BECAUSE of the geochemical and metallurgical importance of the oxides CaO, "FeO", and SiOz, the system CaO-"FeO"-SiO2 has been the subject of several experimental studies during the past 30 years. Following the delineation of phase relations at liquidus temperatures by a group of investigators at the Geophysical Laboratory of the Carnegie Institution of Washington in 1933,1 Chipman and associates in the early forties2,3 determined activities of FeO in liquids of this system. Since that time, many attempts have been made to interpret the data of Chipman et al. in terms of the constitution of the silicate liquids and the interaction parameters among the various constituents of the system (see for instance papers by Richardson4 and by Flood and coworkers5). One of the assumptions made in some of these interpretations4 is that the binary silicate systems CaSiO3-FeSiO3 and CazSi04-Fe2SiO4 mix ideally in the solid state. The objective of the present study is to provide experimental data on activity-com position relations in solid solutions of the system CaO-"FeO"-SiO2. Such data will contribute toward an understanding of the thermodynamics of silicate solid-solution phases in general, and will serve specifically to confirm or refute some of the assumptions made in interpreting liquid slag behavior. PREVIOUS WORK Phase relations in the system CaO-"FeO"-SiO2 in contact with metallic iron were determined by Bowen, Schairer, and Posnjak,1 and additional phase-equilibrium data for this system have been presented by Allen and snow.&apos; A reproduction of the most up-to-date phase diagram for this system is shown in Fig. 1.7&apos;8 Of particular interest in the present investigation are the relations among the various crystalline phases slightly below the lowest solidus temperature in the system, as shown in Fig. 2. The present work will be concerned primarily with some of the composition

Jan 1, 1968

By A. G. Guy, T. B. Lindemer

A second intepation of the Fick law has been developed and applied to three-place data generated from the Gosting and Fujita solution of the diffusion equation. This integration provides new mathematical techniques for the analysis of experimental data. Through use of the synthetic data it is shown that for concentration-independent interdifiusion coefficients these techniques permit a more accurate determination of the coefficients than does the usual Boltzmann-Alatano analysis. Another technique, applicable to the most general concentration-dependent diffusion coefficients , can be used to determine whether given sets of coefficients are consistent with the experimental ternary concentration-distance curves. DURING the past 10 years the subject of three-component diffusion has advanced from the stage of a laboratory curiosity to the status of an accepted portion of physical metallurgy.1 In the course of this development the underlying treatment for two-component sys- tems was extended and generalized to accommodate the added component, but fundamentally the analyses currently employed for ternary systems are analogous to those previously developed for binary systems. There are several indications that an improved analysis is needed for the ternary case. First, the two cross coefficients, D312 and D321 are typically an order of magnitude smaller than the direct coefficients and consequently are difficult to determine with adequate precision. Second, instead of a single concentration-penetration curve (as in binary diffusion), there are four such curves in the usual determination of ternary diffusion coefficients. These curves have definite theoretical interrelationships, but actual experimental curves exhibit significant departures from these relationships. Finally, the amount of data needed to adequately describe the diffusion behavior of a ternary system is so large that radically new methods are needed to facilitate the gathering of these data. ICirkaldy2j4 has proposed several possible analyses of ternary diffusion data, including some of the methods in current use. A "second-integration" procedure proposed by him4 differed in important respects from that presented here and was not carried to the point of application. Two other methods have been explored by the authors and will be published separately; these are a numerical analysis technique and a method for the analysis of a single ternary diffusion couple. Still other analyses are being explored by various investigators. At present it seems likely that the final analysis of ternary diffusion that evolves will be a composite

Jan 1, 1968

By Victor A. Phillips

A series of strips of a Ni-12.7 at. pct A1 alloy were Prepared containing cubical y&apos;(NisAl) precipitates with edge lengths from 60 to 500A. A particle-free solution-tveated strip was included for cornparison. They weve cold-rolled 95 pct and the effects of particle size on the isochronal (1/2 hr) annealing behavior between 300° and 950°C studied (by hardness and light and electron microscopy). It ulas inferred that the particles deformed with the lnatvix becoming lamellae which remained coherent. Comparison with published data fov pure nickel showed that aluminum greatly re-tavded softening and recrystallization, but it made little difference whether or not particles were present. The presence of pakticles led to a heterogeneous distribution of precipitates after annealing at 700" to 750°C. Recovery was not detected. Recrystallization occurred by the growth of new grains into unrecrys-tallized material. In a previous study by the author,&apos; the growth of Ll2-type ordered yl(Ni3Al) precipitates was followed in Ni-12.7 at. pct A1 alloy as a function of aging at 600" and 700°C. The particles were showo to be cubical in shape in all sizes from 50 to 3000A and remained coherent. This work was used as a guide in preparing the starting structures for the present study of the effect of these particles on the annealing behavior of heavily cold-rolled strip. Another question of present interest was whether dislocation and particle hardening were additive, since the structures before rolling ranged from solution-treated to peak-aged to overaged. Also, precipitation might occur on annealing after cold-rolling. Reference may be made to other papers2"5 for previous work in this relatively unexplored field and only some recent work will be mentioned her:. phillips2 studied the effect of deformable 0 to 590A-diam cobalt particles on a Cu-3.23 pct Co alloy rolled 95 pct and found that the particles, which rolled out into thin lamellae, impeded softening and recrystallization. Tanner and servi3 likewise studied the annealing of cold-swaged Cu-2 pct Co alloy containing 150A-diam particles and found impeding effects. Haessner et a1.,4 on the other hand, found that incoherent 2-p-diam non-deformable particles of B4C (0.04 vol pct) tended to increase the rate of recrystallization of copper rolled up to 95 pct reduction. They attributed this to the formation of new grains at the particle interfaces. Humphreys and artin&apos; found that nondeformable silica particles in copper rolled to 30 pct reduction accelerated recrystallization if the particle spacing was large and retarded it if the spacing was smaller. Haessner et a1 4 also studied a rolled Ni-Cr-A1 alloy; however, the particles of y&apos;(Ni3Al)-type precipitate were not put in before rolling, but separated during the isothermal annealing at 750°C. No previous work appears to have been carried out on the effect of y&apos; (Ni3A1) particles on the annealing of Ni-A1 alloy. Hornbogen and ICreye7 redetermined the solubility c of aluminum in nickel as a function of temperature T and showed that it was given by c = 32.6 exp(-1940/RT). This relation gives aluminum solubilities of 15.1, 14.2, 12.0, and 10.7 at. pct at 1000°, 900°, 700°, and 600°C, respectively. The phase precipitated from the nickel-rich solid solution is fcc y1 (Ni3A1) which has a Cu3Au -type ordered structure8 and remains ordered up to 1000°C.B EXPERIMENTAL PROCEDURE The alloy used was identical with that used before. Chemical analysis showed 6.27 wt pct (12.71 at. pct) Al, the principle impurities being 0.065 pct Fe, 0.022 pct Co, 0.020 pct Cu, and 0.004 pct C. Bar stock of 1 in. diam was cold-swaged to % in. diam, cold-rolled to 0.300-in.-thick strip, and annealed at 900°C in dry hydrogen. It was cold-rolled to 0.100-in. thickness and solution-treated for 1 hr at 1000°C while sealed in a quartz tube in argon, quenching in iced brine with the aid of a device to snap off the nose of the tube. Lineal analysis gave an average grain size of 0.055 mm. Pieces of strip were aged at 700°C in vacuo for 30 min, 51/4 hr, and 1 week to produce nominal average particle widths of 60, 150, and 500A, respectively, as known from the previous work.&apos; The average diamond pyramid hardness was determined. The heat-treated strips were rolled from 0.100 to 0.005 in., a reduction of 95 pct, and the rolled strips stored at about -5°C. Small pieces were annealed within 1 week for 30 min at temperatures from 300° to 950° ±2°C in a horizontal vacuum furnace. Strips were withdrawn into a cooling zone, giving an estimated initial cooling rate from 950°C of about 50°C per sec. Average diamond pyramid hardnesses were determined on a lightly electropolished spot on the surface of each strip using 300-g load. Each point on the softening curves represents a separate annealed specimen. Sections containing the rolling direction were examined by optical metallography. Selected specimens were electrothinned to the center plane&apos; and examined by transmission at 100 kv in a Siemens Elmiskop I electron microscope. It is well-known that changes in the structure tend to occur when a deformed strip is electrothinned below a thickness of a few hundred angstroms, although this is less serious with a material such as nickel which has a high melting point, and also is apt to be less serious when particles are present. Observations were nevertheless confined to thicker regions of the foils with estimated thicknesses over 1000A. No changes were observed due to beam exposure.

Jan 1, 1968

By John L. Walter

A sintered compact of iron containing 0.007 vol fraction of glass , prepared 6). powder rnetullurgtcal lechniques, was rolled lo foil, 0.001 in. /hick. The foil sullrples were anneuled a1 650° to 850°C in hydrogen. Pri,narj recryslullization prodtnced large grains which Lc3ere highly elongnled in the rolling direclion und which had an unusual prgferred orientation. The glass parlicles assumed a stringerlike dispersiorz on working; llie slrinyers were parallel lo the rolling direction and to the long axis of the grains. The same unusual grain structure, preferred orientation, and stringerlike dispersion of purticles are found in heal-ily rolled and annealed foil of tungsten which has heen doped wilh small amounts of aluminum, silicon, and polassiutvt. Both the proper dispersion of the particles and severe working (reductions in excess of 98 pct) are necessary to produce the elongzcted grain structure. RECENTLY, the author reported on the origin of the elongated or "nonsag" grain structure in tungsten foil which was doped with aluminum, silicon, and potassium.&apos; It was observed that particles, identified as mullite (Al6Si2Ol3), were present in the doped ingot (as sintered) and at all stages of processing from ingot to sheet or wire. It was shown that the particles, being plastic at tungsten working temperatures (up to 1550"C), underwent a "stringing out" parallel to the working di-direction. It was concluded that these strings of particles were responsible for the formation of the highly elongated interlocking grains observed in doped tungsten wire or sheet. To give further credence to this explanation of a long-observed but not understood phenomenon, it was decided to design a new system, one which was analogous to the doped-tungsten system. Thus, a bcc metal, iron, was chosen to be doped with glass which would be plastic at reasonable working temperatures for iron. The glass chosen was a #0080 Corning lime glass which melts at about 810°C. This is approximately the same fraction of the melting point of iron (0.6) as the melting temperature of mullite is of tungsten. It will be shown that all the observed unusual features of the doped tungsten are duplicated in the glass. doped iron, even to the preferred orientation of the grains in the recrystallized sheet. In the tungsten sheet the preferred orientation consists of two components, (1i3)(691) and (013)(362), neither of which has been observed as a component in textures of primary recrystallized bcc metals. EXPERIMENTAL PROCEDURE The glass, in the form of fine powder, was added to high-purity electrolytic iron powder and mixed dry. The particular composition chosen was 0.007 vol fraction of glass. After mixing, the powder was mechanically pressed to a disk 0.5 in. diam and 0.15 in. thick at a pressure of 5000 psi. The pellet was sintered, in pure dry hydrogen, for 2 hr at 8 50°C. The sintered pellet was then heated to 750°C and rolled to 50 pct reduction of thickness in two passes, with reheating between passes. The rolling temperature was then reduced to 600°C and the slab was rolled to a thickness of 0.050 in. Rolling to 0.010 in. was accomplished in several passes at 300°C. Final reductions to 0.003 and 0.001 in. were taken at room temperature. This rolling temperature schedule was similar, in terms of the fraction of the melting point, to that used in the rolling of the doped tungsten. The total reduction of thickness was in excess of 99 pct. Samples of the rolled foil were lightly electro-polished and annealed for 1 to 6 hr at temperatures of 650" to 85O°C in hydrogen. Samples for transmission electron microscopy were thinned on a Disa Electropol using AC-1 electrolyte.

Jan 1, 1968

By L. C. Brown, E. B. Hawbolt

The formation of grain boundary precipitates has been studied in Ag-5.64 wt pcl Al alloys at 688°C. Large-grain specinzerzs were used and the boundary misorientation measured by X-ray diffraction. At low-angle grain boundrrries Primary sideplates form. u)hile abore a 17-deg misorientation lenticular precipitates predominate. Growth of the lenticular precipitates was studied on individual grain boundaries using a statistical technique. The thickening and lengthening rates appeared to be independent of &apos;pain boundary misorientation, indicating that grain boundary diffusion was not significant in this system. The precipitates grew urith constant shape, with both the length and thickness increasing parabolically with time. The parabolic lengthening rates were unexpected since previous theory and experimerztal results had indicated linear growth. By approximating the shape of the precipitate to that of an oblate spheroid grouling with constant shape an equivalent diffusion coefficienl was calculated. The zlalue obtained was in good agreement with nleasurements obtained fron, diffusiorz couples. IF a single-phase alloy is heated or cooled to a temperature such that the phase becomes supersaturated, there is a tendency for a second phase to precipitate from it in order to remove the supersaturation. Precipitation takes place initially on grain boundaries due to preferred heterogeneous nucleation, but subsequently Widmans kitten plates may form in the grain interiors if the grain size and supersaturation are sufficient. The type of grain boundary precipitate formed depends on the nature of the boundary. Aaronson1 and clark2 have shown that grain boundary misorientation can be characterized by the simple parameter R, where and X, Y, and Z represent the minimum rotations about three orthogonal axes required to bring the two lattices forming the grain boundary into coincidence. The maximum value for R is 60 deg. In using this parameter the orientation of the grain boundary itself is, of course, neglected. However it was found in both Fe-si3 and A1-Ag2 that the precipitate shape depended only on the parameter R. Clark found that the vast majority of precipitates formed were allo-triomorphs— lens-shaped precipitates that grew along the grain boundaries. At low-angle boundaries, however, (R < 15 deg) Widmanstatten sideplates nucleated at the grain boundaries and grew into the adjacent grains. It has been reported that the rate of growth of grain boundary precipitates is controlled by diffusion within the supersaturated phase.&apos; The theories which have been developed for growth rates have assumed the diffusion coefficient constant over the composition range of the supersaturated phase. This is a fairly good approximation if the composition range involved is not large. Thickening rates have been calculated using the theory developed by Zener4 and by Frank5 for the growth of planar precipitates. A parabolic growth law

Jan 1, 1968