Search Documents

By G. R. Love, M. L. Picklesirner

THE use of color adds a new beauty, power, and versatility to metallography. This has been amply demonstrated in a number of public exhibits and on the walls of corporate, government, and university metallographic laboratories around the world. Indeed, in a number of cases, the color not only is of esthetic value but is scientifically essential to the competent examination of particular samples and systems. Optical strain analysis,' determination of true grain size in U-A1 alloys: phase identification in Nb-base alloys,3 analysis of transformation rates in heat-treated Nb-Zr alloys,4 measurement of solidification segregation in cast Nb-Zr alloys,' study of the intermediate phases in Nb/Sn diffusion couples,6 investigation of the oxidation kinetics at various surface orientations in single crystals,7'8 and a number of other research efforts9'10 would have been vastly more difficult without the extra clarity and distinction introduced by color—either the colors resulting from particular polarized light techniques or the interference colors obtained by oxidation. It is unfortunate that the beauty of colored micro-structures cannot be shared more widely, but the economics of publishing a technical magazine apparently prohibits color plates except under extraordinary circumstances. It is doubly unfortunate that no convenient, general method exists for communicating the information available in colored microstructures to the scientific community. True, in some instances the color changes observed are dramatic enough that direct black and white photography is adequate. In photoelastic stress analysis, several orders of interference fringes may be produced and it is frequently desirable to suppress the color effects (use monochromatic light) to simplify interpretation. On the other hand, relatively minute color variations may be readily detectable by eye, be of considerable assistance in interpreting a structure, and yet be enormously difficult to reduce to an unambiguous black and white print. In order to illustrate a possible compromise or solution to the above problems, we wish to show three black and white photographs of the same region of an Nb/Sn diffusion couple whose interpretation depends heavily upon use of color techniques. These micro-structures were obtained from a section through a rod consisting of commercial-purity niobium tubing filled with 99.995 pct Sn (melt-filled in 10"6 Torr vacuum). The composite rod, measuring 500 mm long, was bent into a hairpin shape and inserted into water-cooled copper grips in a vacuum high-temperature furnace. The rod was outgassed for 2 hr at a furnace temperature of 800°C and was further heated by self-resistance using a 2.2-kva ac power supply. The rod was raised to 1910°C over a 4+-hr period, held for 15 min, slow-cooled to 1500°C over a -hr period, and free-cooled to room temperature (all temperatures were measured with W5Re/W26Re thermocouple). Throughout the heat treatment the ambient pressure was kept below 8 x lo-' Torr. Metallographic preparation of the sample was completely conventional except that it was necessary to extend the grinding steps to remove damage in the brittle NbsSn phase introduced by sectioning. No satisfactory chemical or electrochemical etch has been found for these diffusion couples; of those which have been tried, any that attack the niobium-rich phases at all attack the tin-rich phases disastrously. However, the final polishing steps did produce a slight relief effect and this, coupled with anodizing, yielded excellent clarity in the microstructures for direct observation. The anodizing voltage was chosen to give maximum change in color per unit change in composition. In this case, indeed normally, this maximum sensitivity occurred in the blue-to-red color range. The pure niobium tube wall anodized to a relatively bright blue

Jan 1, 1967

By E. Lifshin, R. E. Hanneman

QUANTITATIVE applications of the electron micro-beam probe frequently involve the evaluation of complex mathematical expressions and/or the analysis of large amounts of experimental data. The purpose of this communication is to describe briefly a versatile and useful computer program system that is applicable to analyze rapidly a wide variety of practical microprobe problems. This system consists of a group of ten FORTRAN programs that can be stored on tape, cards, or in the memory disc of the computer. These programs, or links, can be run individually or in any prespecified sequence without interrupting the operation of the computer or without destroying information which is being transfered from one link to another. For the program system described here a GE-235 computer with disc storage was used, so that the DCHAIN method of program linking was employed. Included in the library are programs to: 1) initiate analysis of a new set of data and transfer control between all other programs in any predetermined manner; 2) generate theoretical calibration curves of composition vs relative intensity; 3) generate empirical deviation parameters from least-square fits of experimental calibration data from standards of known composition; 4) convert raw X-ray data to corrected composition; 5) determine inter diffusion coefficients by Matano analysis of con centrat ion -distance data on a uniaxial diffusion couple; 6) determine activation energies and frequency factors of temperature-activated processes such as diffusion; and 7) generate calibration curves for determination of the thickness of thin films using microanalysis. A detailed description of these computer programs and their underlying principles is available on request from the authors."' The first program to generate theoretical calibration curves of corrected relative intensities vs composition uses the Poole and Thomas atomic number correction' and the Philibert absorption factor' with a voltage-dependent mass absorption coefficient for electrons in the alloy. A modified Castaing fluores- cence correction is also used which includes the effects of both Ka and KO radiation and over voltage.' Once the theoretical curves have been calculated in 1 wt pct intervals, these results are least-squares fit to obtain Ziebold deviation parameters' which are stored in COMMON in the computer memory. The net discrepancies between the original theoretical calibration curve and the regenerated curve using the Ziebold parameter are computed. Although this link is explicitly written for K-K fluorescent interactions, it can be applied to K-L, L-K, and L-L interactions as previously disc~ssed."~ Similar programs have also been written to utilize the Wittry fluorescence correction, Birks combined corrections, and various other corrections.' These modified programs have proved to be quite useful for quantitative comparisons of the results of the various theories. The program for conversion of raw X-ray data to corrected composition includes corrections for drift, backround, and instrumental dead time. The corrected intensities are converted to composition by use of the Ziebold equation"2 and parameter obtained from the program system. The results can be obtained for either atom or weight fractions. In addition to accurately computing interdiffusion coefficients the Matano analysis program calculates least-square smoothed values of concentration, concentration gradient, and curvature for each point on the raw input concentration profile. In order to obtain high accuracies a unique method of performing a least-square polynomial fit to incrementally advancing profile segments which overlap is used.' This program has been successfully modified for use in ternary diffusion problems3 and can readily be modified to handle analysis of diffusion profiles which include phase boundary discontinuities. This link is generally applicable to analysis of interdiffusion data obtained by other techniques as well as by the microprobe. The primary function of the next program is to least-squares fit experimental diffusion data to the normal Arrhenius function: D = Doexp(-Q/RT), to obtain values of Do and Q. In addition the probable error and one, two, and three u statistical confidence limits of DO, Q, and log D are evaluated. This program is also directly useful for analysis of any other simple temperature-activated processes including conductivity, and certain deformation and chemical processes. The program to generate calibration curves for film-thickness determination using the microprobe is based on a numerical integration of the equation derived by Cockett and ~avis.' Values of film thickness obtained by this program for copper on various substrates are in good agreement with measurements made by other techniques. Versions of the above program system have been prepared for use with or without a remote teletype connection to the computer for processing on either a real-time or time-share basis. The instrumentation coupling a microprobe to a teletype for automatic data collection and analysis by the presently described program system has been reported elsewhere by the authors.' If teletype equipment is not used to communicate with the computer, standard methods of card reading and tape reading can be used. In either

Jan 1, 1967

By E. R. Gilbert

Klopp et al. have reported data on tensile and creep properties of are-melted and electron-beam-melted tungsten. We would like to point out some similarities between their creep results and ours on are-melted and powder-metallurgy tungstenand offer some alternate interpretations of the data which lead to different rate-controlling mechanisms. Klopp et al. observed extensive transient creep in the 2250" to 3500°F temperature range where the acthetivation energy is 106,400 ± 3300 cal mole-1. They suggest cross slip as the possible rate-controlling mechanism since "cross slip may occur only by the movement of screw dislocations, and not edge dislocations, thus allowing only partial recovery of the strain hardening and resulting in a continuously decreasing creep rate, characteristic of transient creep.'' Our tests on are-melted tungsten and powder-metallurgy tungsten displayed an extensive transient creep stage in the same temperature range but they also displayed a steady-state creep stage. The stress levels and strain rates of the two investigations, 1.4 to 7 kg mm-2 and 10-7 to 10-4 sec-1 for that of Klopp et al. and 4.5 to 10 kg mm-2 and 10-8 to 10-4 sec-1 for those of Gilbert et al., are overlapping. The activation energy for creep of our powder-metallurgy tungsten is 105,000 cal mole-1. This close agreement with the value found by Klopp et al. suggests that the same mechanism may be common to both types of materials. It may be possible to gain some additional insight into the rate-controlling mechanism by applying the analysis made by conrad22 in terms of the activation volume. The value for the activation volume depends upon the dislocation mechanism and may be used to help differentiate between various dislocation mechanisms. The activation volume for cross slip is 10 to 100 b3. Values determined from our data were 130 to 220 b3 which are in better agreement with the range of 102 to 104 b3 for nonconservative motion of jogs. Values determined from data of Klopp et al. are 240 to 1040 b3. We suggest that the rate-controlling mechanism for creep in this temperature range may be nonconservative motion of jogs assisted by enhanced diffusion along grain boundaries or dislocations. This model requires an activation energy lower than that for self-diffusion as is observed. If this is the controlling mechanism, a higher activation energy which is closer to that for self-diffusion might be expected for tests conducted on specimens with reduced grain boundary area, i.e., larger grain size. In support of this premise we have conducted tests on speci- mens having one to three grains per cross section for which a higher activation energy of 130,000 cal mole-1 was observed. Also, in agreement with results of Klopp et al., a faster creep rate resulted from the increased grain size and probable decreased dislocation density. Further argument against the cross slip mechanism is based on the constant activation energy we obsserved over the range of 5.0 to 10 kg mm-2,21 whereas the activation energy for cross slip displays a significant stress dependence.23 In addition to this, the powe on the stress term for the strain-rate dependence appears too high for cross slip. The values for tungsten fall in the range of 4.5 to 7 (Klopp et al. and Ref. 21) whereas values of approximately 2 are observed in creep by cross slip for hep metals.24 At temperatures from 3300° to 4000°F, Klopp et al. observed an activation energy of 141,000 ± 4000 cal mole-' which is less than that for self-diffusion, 153,000 cal mole-1, 25 and less than that reported for high-temperature creep of powder-metallurgy tungsten by Green:' 160,000 cal mole-1. Klopp et al. suggested that recovery of strain hardening by dislocation climb is the rate-controlling mechanism in this temperature region. Actually an activation energy larger than that for self-diffusion is predicted by this model if the temperature dependence of the elastic modulus27 is taken into consideration. Klopp et al. attributed this lack of correlation to the compositional differences between powder-metallurgy tungsten used by Green and their own arc- and EB-melted tungsten. It is our position that the value of 141,000 ± 4000 cal mole-' represents a transitional value between the higher and lower activation energies and does not represent a discrete mechanism of deformation. If the tests had been extended to higher temperatures the higher value of activation energy would have been observed as has been demonstrated by both our analy sis'' of isothermal test data by GE-NMPO28 and differential temperature tests28 on are-melted tungsten. The activation energies determined (170,000 and 160,000 cal mole-1) were greater than that for self-diffusion and are in better agreement with the value found by Green. The model of dislocation climb is associated with ; activation volume of 1 b3 according to Conrad.22 The activation volume calculated for the creep data of fou independent investigators in the temperature range where the activation energy for creep is near that for self-diffusion (Klopp et al. and Refs. 26, 28, and 30) is 110 to 2800 b3 and is in the range suggested for nonconservative motion of jogs. In the case of tungsten this mechanism may be preferred over that of dislocation climb. Sincere appreciation is expressed to Messrs. J. E. Flinn and F. L. Yaggee for the helpful discussions and suggestions in the preparation of this communication.

Jan 1, 1967

By J. E. Flinn, E. R. Gilbert

Recently, Green reported some high-temperature, 0.60 to 0.89 T3, creep results on unalloyed tantalum. From his study he determined an activation energy of 114 kcal per mole for steady-state creep using the T, differential-temperature method25 for testing. He accounted for the difference between this value and 110 kcal per mole for self-diffusion by the temperature dependence of the elastic modulus, 4 kcal per mole. This reported activation energy is in apparent disagreement with that which can be determined from Green's isothermal data, which we have plotted as log creep rate vs 1/T in Fig. 15. From this figure, it appears that there are two activation energies of -185 kcal per mole at the higher temperatures and -75 kcal per mole at the lower temperatures. It is not obvious that the data can be fitted to a straight line corresponding to an activation energy of 114 kcal per mole. One is led to two possible conclusions regarding Green's study: there are two discrete processes controlling the creep rate from 0.60 to 0.89 Tm or the apparent activation energy is not discrete but continuously increases with increasing temperature. The former speculation would be in closer agreement with the high-temperature creep behavior of tungsten.28 The apparent disagreement between the activation energies determined by the AT and isothermal methods certainly suggests that further study on the high-temperature creep of tantalum is needed.

Jan 1, 1967

By C. G. Dunn

James, Jones, and Leak (JJL) conclude that growth-rate and orientation-distribution data obtained in a steep thermal gradient should be used with caution to account for isothermally produced recrystallization textures. I agree. However, it is not clear whether they conclude a) that the orientation dependence of rate of growth of secondaries in (commercial) silicon iron depends on both temperature and a temperature gradient or b) that, in applying rate data obtained in a steep thermal gradient, the isothermal annealing temperature must be close to the temperature for which the rate data apply. Also, it is not clear what tem-

Jan 1, 1967

By H. M. Strong, R. A. Oriani

THE technological interest attaching to light, age-hardening alloys for structural uses has been pointed out to one of us.1 A potentially interesting alloy would be titanium-rich Ti-Mg alloy, provided that sufficient solid solubility exists. or could be induced at high pressure. and provided also that age hardening can be induced by proper temperature and/or pressure annealing. It was decided therefore to explore the feasibility of producing a Ti-Mg alloy in the high-pressure apparatus. Experimental. Three high-pressure runs have been carried out. In each case the geometry of the sample was that of a cylinder of titanium of commercial purity, into which a narrow hole had been drilled along the axis; this hole was filled with magnesium wire,

Jan 1, 1967

By C. S. Yust, C. S. Morgan

ThIS note reports the observation of masses of dislocation etch pits around the weld necks of small single-crystal particles of calcium fluoride sintered to a cleaved face of a larger CaFz crystal. Crystal particles in the -50 +I00 mesh size range were placed on freshly cleaved CaFz surfaces and heated at 1250°C for 30 min in a tantalum container in an argon atmosphere. The sintered specimens were then etched at room temperature for 40 to 50 min in a 2 pct sulfamic acid solution which is a selective etch for dislocations in caF2.' Figs. 1 through 3 show the masses of dislocation etch pits observed around weld necks after etching. In some cases, particles were gone after removal of the specimens from the furnace although weld necks were readily visible. In other instances, no etch pits formed around a weld neck. A possible cause of concentrations of dislocations on the crystal surface is mechanical deformation of the base crystal when particles are dropped on it. If such a deformation does occur, it would be much smaller than that caused by the light punching of a crystal with a sharp instrument which results in large concentrations of etch pits. However, crystals deformed by punching with the sharp instrument and annealed at the conditions of these experiments (1250°C for 30 min) before etching showed no concentrations of dislocation etch pits. It is obvious that dislocations from the much smaller deformations, possibly created when particles are placed on the surface, would also anneal out. Etching of large crystals which had particles placed on them in the usual manner but which were not heated showed no dislocation concentrations such as shown in Fig. 1. The magnitude of dislocation etch-pit concentrations did not vary with extensive variation of the cooling rate. The appearance of many of the etch-pit clusters is typical of a well-annealed condition, indicating that the dislocation concentrations formed before the specimen temperature was lowered. Therefore, it is evident that thermal stresses in the neck area did not create the deformation indicated by the presence of the dislocations. The possibility exists that the etch-pit concentrations at places where particles had been removed may be the result of deformation accompanying the fracture of the sintered bond. That this is not so was demonstrated by breaking off particles around which no etch pits occurred. Re-etching never resulted in the large concentrations observed at sintering sites. The stress available to effect material movement during sintering of two particles will depend on the geometry at the contact point but can easily exceed the macroscopic yield stress in CaFz systems. It appears that the plastic deformation observed results from sintering or simply from the weight of the small particles. It has been previously argued that plastic flow makes an important contribution to the initial densification of oxides having the fluorite crystal structure.2p3 If plastic deformation occurs, it would begin as soon as the temperature is sufficiently high

Jan 1, 1967

By J. W. Downey, A. E. Dwight, R. A. Conner

In a previous paper' we have reported the existence of a number of AB2 Laves phases in which A is a member of the scandium group (including the 1an-

Jan 1, 1967

By R. N. Orava, L. A. Rice

One of the difficulties encountered in the preparation of titanium specimens for transmission electron microscopy is the formation of a surface hydride phase during the thinning operation at ambient temperatures. This necessarily destroys contrast, obscures features of interest, and often leads to misleading results. By cooling the polishing solution to temperatures below 0°C this problem has been overcome in the past.' However, since the cooling procedure is somewhat inconvenient, a technique was developed at this laboratory for thinning titanium at room temperature. The material examined was A-70 grade commercial-purity titanium purchased from the Crucible Steel Co. of America. The mechanical and thermal treatments, described elsewhere,* yielded a recrystallized grain size of 4 to 5 p. A slice, 0.5 mm thick and 1.5 mm wide, was cut from a bulk specimen2 using a diamond cut-off wheel. The sample was then coated with paraffin wax. Two concentric holes, about 0.03 in. in diam, one at each broad face, were scribed through the wax to the surface of the metal. Subsequent polishing for several minutes resulted in a preferentially thinned or "dimpled" region. Thereafter, the wax was removed and the entire specimen polished to the point of breakthrough at the base of the dimple. Polishing was effected chemically in a solution comprised of 30 pct HF and 70 pct HN03 by volume, kept at room temperature (20" to 25OC). The specimens were mounted in a stereo holder and examined at 100 kv in a Philips l00 B Electron Microscope. A micrograph from a foil prepared in this way is shown in Fig. 1. Several points are noteworthy. This electron micrograph shows neither evidence for the formation of a hydride phase due to specimen preparation nor, moreover, any obvious surface contamination in electron-transparent areas of a freshly prepared foil. Localized variations in foil thickness observed in some specimens seem to be a characteristic of this chemical polish but do not seriously detract from foil quality. With care, the lifetime of a good specimen exceeded 2 hr in the beam. On the other hand, the exposure of a sample to air at atmospheric pressure, when still warm from the beam, clearly resulted in the deterioration of a formerly clean area just previously in the field of view. An example in Fig. 2 of a contaminated region reveals features, and "spots", not unlike those observed in much purer material by Yamane and Ueda. Consequently, one can conclude that: 1) titanium can be thinned satisfactorily in a solution at room temperature for transmission electron microscopy; 2) large-scale precipitates, >50A in diam, are absent in commercial-purity, recrystallized, furnace-cooled, A-70 titanium containing about 3800 wt ppm interstitials and the same amount of iron; 3) care must be taken not to attribute surface contaminants to precipitates. The authors are grateful to the Franklin Institute for financial support and permission to publish.

Jan 1, 1967

By Edward J. Ryan, E. A. Peretti

iV preliminary X-ray and metallographic survey of the ternary system Bi-In-Te revealed that the following isopleths could form quasibinary systems: InTe-Bi, InTe-InBi, InTe-BizTe3, and InzTe3-Bi2Te3. The nature of the InTe-Bi phase diagram was established by Bologna and Peretti.' This article presents the results of an investigation of the InTe-InBi system by thermal analysis, X-ray methods, and metallographic examination. The bismuth and tellurium were bought from the American Smelting and Refining Co. The tellurium had the following spectrographic analysis: 0.0001 pct Mg, 0.0001 pct Si, 0.0001 pct Fe, 0.0002 pct Al, and 99.999+ pct Te (by difference). The bismuth assayed: 0.0001 pct Mg, 0.0001 pct Pb, 0.0001 pct Fe, 0.0001 pct Cu, and 99.999+ pct Bi (by difference). The indium was obtained from the Indium Corp. of America and contained 0.002 pct Cu, 0.006 pct Pb, 0.01 pct Sn, 0.01 pct Zn, and 99.97+pct In. The alloys were made by melting the required proPortions of InTe and in evacuated, sealed quartz

Jan 1, 1967

By G. P. Conard, E. H. Rennhack

The creep behavior of hot-rolled, hypoeutectic Zn-Ti alloys was investigated in the temperature range from 0.43 to 0.53 TM. Secondary flow was found to originate primarily from strain-induced gvain growth where grain boundary )nigvation served to relieve the strain energy of distortion introduced by slip, grain boundary sliding, and subgvain formation. The extent to which this recovery mechanism operated was determined by the ratio of grain width to the spacing between planar fibers of TiZn,, compound particles generated in these alloys during rolling. When this ratio was unity, creep resistance demonstrated a marked improvement. In this condition, which was fulfilled by annealing following rolling, structural stability was enhanced with decreasing grain size below the equicohesive temperature (-0.5Tm), while the reverse was true above this temperature. TITANIUM concentrations approaching the eutectic composition of 0.23 wt pctl have been shown to promote a significant increase in the creep resistance of rolled zinc,2 The alloying effect created with titanium is somewhat unique; a structure closely resembling that of a fiber-reinforced metal composite can be developed which selectively modifies creep strength in preference to other mechanical properties. In an earlier investigation,~ the present authors found that, while the fiber network, composed of individual TiZn,, compound particles, had a distinct influence on rolled texture, the crystallographic variations produced were of minor importance with respect to creep. Rather, creep resistance seemingly increased when the grain size appeared to coincide with the in-terfiber spacing. The work described here was undertaken to explore this effect in greater detail. EXPERIMENTAL PROCEDURE Three zinc-base alloys containing 0.05, 0.12, and 0.16 wt pct Ti were prepared from CP zinc and iodide titanium in the form of 4 by 2 by f in. chill-cast ingots. The melting and casting procedures for these alloys have been detailed el~ewhere.~ Individual ingots of each alloy were hot-rolled at 200°C (392°F) to total reductions of 10, 25, 50, 75, and 90 pct in from one to five passes, respectively, employing a 10-min reheat prior to each rolling pass. With grain, tensile-type creep specimens with a 1-in.-long, -in.-wide gage section were machined from the rolled strips for test purposes. Annealing studies to explore the influence of grain size on secondary creep flow were carried out at 400°C (752°F) in argon for times extending up to 60 min. The grain-size effect was evaluated in terms of average grain width and length values statistically derived from lineal intersection measurements.4 A similar method was applied in establishing the average interfiber spacing, i.e., average perpendicular distance between adjacent planar fibers. The creep characteristics of the alloys were investigated by means of constant-load and constant-stress creep tests. The former tests were conducted at 25°C (77°F) under an initial stress of 10,000 psi, while the latter were performed in the range from 25°C (77°F) to 90°C (194°F) at stress levels varying from 8000 to 22,000 psi. Total specimen strain, as determined with Budd HE-1161-B strain gages, was in excess of 0.10. Maintenance of constant stress was achieved through periodic load reductions made at 0.01 strain intervals to compensate for the attendant incremental reduction in specimen cross-sectional area. The maximum indicated error in the applied stress at these strain intervals was less than 3.0 pct. RESULTS AND DISCUSSION Constant-Load Creep. In an effort to clarify the in-terrelation between interfiber spacing and grain size with respect to the creep resistance of the Zn-Ti alloys, their separate effects on secondary creep rate were determined as a function of titanium content and rolling reduction. These results are set forth in Figs. 1 and 2, respectively. The average grain diameter plotted in Fig. 2 was resolved from average grain width and length values. No data are presented for reductions of less than 50 pct because of the inability to obtain consistent measurements on these strips. The curves of Fig. 1 indicated that, for a given titanium content, a decrease in interfiber spacing, as produced with increasing reduction, promoted a decrease in creep rate. Depending on titanium content, however, wide variations in creep rate occurred at the same interfiber spacing suggesting that interfiber spacing, by itself, has little or no influence on creep resistance. Grain size, on the other hand, decreased progressively with both increasing rolling reduction and titanium content, the effect of which led to a pronounced decrease in creep rate, particularly when the average grain diameter became smaller than 3.0 x 10"4 in., Fig. 2. The continuity of this relationship tended to support the view that grain size rather than interfiber spacing was predominant in controlling secondary creep. Annealing Effect. The observed dependence of creep flow on grain size suggested that a further contribution to creep resistance would result when the alloys were annealed to effect a coincidence between grain width and interfiber spacing, see Fig. 3(b). ~eiides creating an immediate barrier to grain boundary movement, annealing offered the possibility of providing increased structural stability by eliminating many high-energy, mobile grain boundaries.= To test this hypothesis, specimens from the Zn-0.16 Ti strips reduced 75 and

Jan 1, 1967

By P. Wynblatt, J. E. Dorn

The effect of strain rate, temperature, and interstitial impurity concentration on the flow stress was investigated in a poly crystalline Fe-2 pct Mn alloy. The temperature dependence of the flow stress was found to be independent of the interstitial impurity concentration, removal of interstitials merely decreasing the atherma1 stress level. Below 160%, the large temperature dependence of the flow stress was interpreted in terms of the Dorn-Rajnak theory of the Peierls mechanism of plastic deformation. Above 160"K, some other thermally activated intrinsic meckanisln seems to be operating. The strong temperature dependence of the flow stress of bcc iron at low temperatures has been studied by numerous investigators and interpreted in terms of several different thermally activated dislocation mechanisms. The following mechanisms have been proposed: a) interaction of dislocations with interstitial impurity atoms1 or with solute atoms in general; b) interaction of dislocations with clusters of impurity atoms;3 c) resistance to the motion of dislocations due to jogs on screw dislocations;* d) resistance to the motion of dislocations due to the Snoeck effect;' and e) interaction of dislocations with the intrinsic resistance of the bcc lattice or Peierls "hills". In previous work: it was found that the plastic behavior of polycrystalline iron containing 2 wt pct Mn (and 100 ppm C + N) is in good agreement with predictions based on the Peierls mechanismlo from 77" to 160°K whereas from 160" to about 370°K another as yet unidentified thermally activated mechanism is operative. The purpose of the present investigation was to determine the effect of interstitial impurities on the temperature dependence of the flow stress, by purifying the previously investigated alloy with respect to interstitial impurities. The investigation revealed that removal of the interstitial impurities to a level that eliminated Cottrell locking and the Portevin-LeChatelier effect (i.e., serrations in the stress-strain curve due to dynamic strain aging) affects neither the strong temperature dependence of the flow stress at low temperatures nor the general characteristics of the higher-temperature behavior. Such purification, however, lowered the athermal yield stress. I) EXPERIMENTAL PROCEDURE AND RESULTS The material used in this investigation consisted of an Fe-2 wt pct Mn alloy having the following additional elements present: 0.004 pct C, 0.006 pct N, 0.05 pct 0, 0.004 pct S, 0.003 pct P, and 0.001 pct Si. This was the same material as that studied in the previous investigation and will henceforth be referred to as the "impure material". The as received -jr by 2 in. hot-rolled bars were cold-rolled to 0.08-in. thickness, recrys-tallized under argon for 30 min at 80O0C, and further cold-rolled to 0.053-in. thickness. Flat tensile specimens 3 in. wide having a 1.625-in.-long gage section were machined from the sheet and then purified. The first stage of purification consisted of holding the specimens at 850°C for 24 hr in a stream of hydrogen saturated with water vapor at room temperature. Material prepared in this fashion will subsequently be referred to as "wet hydrogen purified material". It has been estimated that this type of treatment reduces the carbon content of iron to less than 5 ppm.' The second stage of purification consisted of holding the specimens in a closed system through which hydrogen was circulated past the heated specimens as well as a zirconium hydride getter. The specimens were held at 850°C and the getter at 800°C. The process was continued for 212 hr. The starting material for this process was wet hydrogen purified material. Material prepared in this manner will be referred to as "ZrHz purified material". The technique used was substantially the same as that described by Stein et al.' who have shown that the carbon level at the end of the process is reduced to about 5 parts per billion. The mean grain size of the three materials was found to be -60 C( for the impure material and -160p for the wet hydrogen and ZrHz purified materials. All tensile testing was carried out on an Instron Testing Machine either in controlled-temperature baths or at fixed-point baths such as boiling liquid nitrogen, and so forth. When fixed points were used the temperature was kept constant to better than *1°C whereas in all other tests the temperature variation was less than *2°C. Test of Purity. It was not possible to determine the purity achieved at each stage of the purification because the resulting level of interstitial impurities was near or below the limit of sensitivity of standard analytical techniques. It was possible, however, to obtain a qualitative measure of the extent of purification from the appearance of the stress-strain curves for the three materials. Fig. 1 shows the stress-strain curves for the three different materials tested at 300°K. It can be seen that the effect of wet hydrogen

Jan 1, 1967

By George A. Colligan, John M. Lo

Electrical resistivity as a function of temperature was investigated in undercooled gallium and under-cooled Ga-In alloy of eutectic compositian to examine structural changes in undercooled melts. A continuous linear extension of the resistivity was determined for 99.999 pct pure gallium when undercooled below the stable Ga(l)-liquid, and the metastable Gafl)-liquid equilibria. In icndercooled Ga-In, a significant continuous negative departure from linear resistivity behavior was observed below the eutectic temperature. All resistivity measurements were made by the four-point probe technique. ThE temperature range for investigations into the physical properties of liquids before freezing has been greatly extended with the undercooling of massive melts of pure metals and alloys. Experimental determinations of thermal and transport properties and diffraction patterns in undercooled systems should throw further light on the structural behavior of liquids. Many measurements showing prefreezing anomalies in viscosity1'5 have been regarded677 as evidence for structural groupings in the liquid state. However, the majority of viscosity results,~)~-~~ conforming to the law 17 =A exp(~/RT), do not support theories of atomic aggregation. For the system Cd-Sb, which exhibits a metastable solid phase, Fisher and Phillipsg reported that, in undercooled alloys, the activation energies for viscous flow below the stable liquidus are much higher than those above it. Similarly, Miller et a1." observed a doubling of the activation energies for electrical conductivity in undercooled Cd-ib alloys. These effects were suggested as indicative of atomic association and clustering in the liquids at lower temperatures. With the recent establishment of metastable phases in pure gallium22 and its alloys,23 attempts were made during this investigation to undercool bulk samples and to determine the temperature dependence of electrical resistivitv. Pure gallium, when undercooled below -17°C at atmospheric pressure, can crystallize to either a stable Ga(1) phase (mp, 29.78"C) or a metastable phase identi-

Jan 1, 1967

By A. W. H. Morris, G. H. Laurie, J. N. Pratt

Using- galvanic cells of the form Sn(liq)/Sn" (LiCl-KC1-SnCl,)/Sn-Ag (alloy), measurements have been made of relative thermodynamic properties of the a, C, E, and liquid phases of the Ag-Sn alloy system. Partial heats of solution of the components in the liquid alloys lzave also been observed by direct cal-orimetric measurement in an isoperibol calorimeter. The thermodynanzic quantities are disczlssed in relation to structural and other properties and the existence of anomalous minor fluctuations in the partial heats and entropies of solution in liquid alloys is tentatively suggested. In the course of a recent electro motive-force study of the thermodynamic properties of Sn-Ag-Pd liquids,' some measurements were also performed on the Ag-Sn binary system. Most previous thermodynamic studies of this system have been concerned with the liquid state. Measurements confined to the limiting heat of solution of silver in liquid tin have been made by many calorimetric workers2 while high-temperature calorimetric measurements of the heats of formation of the full range of liquid alloys are reported in the early work of Kawakami~ (1323°K) and more recently by Wittig and Gehrin~(1248°K). Electromotive-force studies on liquid alloys have been made by Yanko, Drake, and Hovorka' (606" to 686°K; 86 to 99.4 at. pct Sn) and by Frantik and Mc Donald' (900°K; 30 to 90 at. pct Sn). The only known measurements on the solid state are of heats of formation of the a, £, and c phases; these values were obtained using tin-solution calorimetry, at 723"K, by Kleppa,~ whose investigation also yielded heats of formation of liquid alloys containing more than 64 at. pct Sn. The present experiments on the Ag-Sn alloys include a re-examination of the liquid phase and, because of the dearth of free-energy data for the solid state, attempted measurements on the a, c, and E phases. The principal new feature of electromotive-force results obtained for the liquid phase was an indication of anomalous fluctuations in the partial heats and entropies of solution of tin at certain compositions. However, since the values for these thermodynamic quantities were determined from the temperature coefficients of cell potentials, which are commonly subject to considerable error, confirmation by calorimetric measurements was considered desirable. A detailed investigation of the partial heats of solution of the components in the binary liquids was made using a liquid metal solution calorimeter. I) GALVANIC CELL STUDIES a) Experimental Details. Measurements were made, as a function of alloy composition and temperature, of the potentials of reversible galvanic cells of the form: ~n(liq)/~n++/~n-Ag (solid or liquid alloy) Details of the apparatus and experimental techniques have been given elsewhere.' so that a brief account will suffice here. Molten salt, 58 mole pct LiC1-42 mole pct KC1, containing small amounts (1 to 2 mole pct) of stannous chloride was used as the electrolyte. The salts were dehydrated by pre-electrolysis and vacuum -drying techniques. Cells were established under an argon atmosphere by immersing tin and alloy electrodes in the molten salt contained in a large silica tube, heated in a vertical resistance furnace. The tube was sealed at the top by a head plate provided with openings permitting the simultaneous insertion of six electrodes, a central thermocouple sheath, and connections to vacuum and argon lines. Temperatures were controlled to *0.2"C over prolonged periods, with maximum variation over the electrodes at any time of 0.l°C. Temperatures were measured with a standardized Pt/13 pct Rh-Pt couple. The electromotive force of this and the cell potentials were measured on a Cambridge Vernier potentiometer and short-period galvanometer. Alloys were prepared from Pass "S" tin (99.999 pct) and Johnson-Matthey high-purity silver (99.999 pct) by melting in evacuated silica capsules and quenching in cold water. For liquid phase experiments, pieces of the resulting alloys were remelted into prepared silica electrode units, while solid electrodes were prepared by remelting into 3-mm bore tubing, inserting a cleaned molybdenum lead wire, and quenching to produce uniform rods about 3 cm in length, with soundly attached leads. In all cases remelting was done under an argon atmosphere. The solid electrodes were subsequently annealed in evacu ated silica tubes for 14 days at about 20°C below the solidus and quenched. Analyses showed that these techniques produced uniform electrodes with no significant change from weighed out compositions. b) Results and Discussion. Measurements were made on about forty alloys in the solid and liquid states, over varying ranges of temperature between 550" and 1050°K. Stable, mutually consistent, and reproducible electromotive-force data were obtained with most liquid alloys and these are shown in Fig. 1. Investigations were extended below the liquidus tem-

Jan 1, 1967

By R. E. Mussler, F. E. Block, T. T. Campbell

An exploratovy study was made to deternzine the feasibility of preparing beryllium by the metallother-. mic reduction of beryllium oxide. The procedure involved heating a relatively nonvolatile metal reductant in contact with beryllium oxide in a vacuum. Beryllium vapors formed in the reaction were condensed as a coherent deposit on a heated surface. Reductants tested included tanthanun, yttrium, thorium, Th-Mg, and Zr- Ti alloys, and the hydrides of lanthaniir~z, yttriunz, and zirconiur?7,. Test ten7pevatures ranged from 1350'to 175O' C, and reaction times ranged frot 30 to 100 lzr. In most cases metal recoveries were lola, although yields close to 90 pct of theoretical were obtained with latzthanum, yttrium, and zirconium reductants. In most cases metal deposits were contanzinated by beryllides, /ls/cally of the Beln M or Bel3 M type. BERYLLIUM has a variety of unique properties including light weight (sp gr 1.85), low thermal neutron absorption cross section, high scattering cross section, excellent dimensional stability, high strength to weight ratio, high melting point (1285"C), high specific heat, and high thermal conductivity. Interest in beryllium as a structural material in aerospace and nuclear-reactor applications is based on these properties. In general, the potential of beryllium has been only partially realized, largely because of the brittleness of poly crystalline forms of beryllium, difficulties associated in its fabrication, and joining problems encountered in its use in structural applications. Several excellent books and monographs are available to those interested in the occurrence, extractive metallurgy, and utilization of beryllium.1"* Although an extensive amount of research on beryllium has been in progress during the past 10 years, much of the work has been concerned either with the so-called "ductility problem" or with the development of methods of fabrication and evaluation of properties of various commercial or purified grades of metal. A recent survey of the literature indicates that only limited efforts are presently being directed toward developing new methods for reducing beryllium compounds to metal. Two major commercial methods now used to produce beryllium are the magnesium reduction of beryllium fluoride and e lec trow inn ing of beryllium from fused sodium chloride-beryllium chloride salt baths. Only a limited amount of work has been done on the direct reduction of beryllium oxide. Major factors which complicate this method are the great stability of beryllium oxide and the reactive nature of beryllium metal. The stability of the oxide necessitates extremely high reduction temperatures and vacuum operation to liberate beryllium metal vapor and cause reactions of the following type to proceed: XBeO(s) +YM(s)-MYOX(s) + XBe(& As early as 1939, roll' used this reaction for the direct reduction of beryllia by zirconium and titanium, and later in 1952 ~a~el~ studied the reduction of beryl-

Jan 1, 1967

By S. A. Mersol, C. T. Lynch, F. W. Vahldiek

The room-temperature microhardness of as-grown and annealed MoaC single crystals was measured on the (0001), {2110), and1012) planes using Knoop and Vickevs indenters at loads ranging front 25 to 1000 g. The orientatimz dependence of hardness with respect to crystal axes was also studied. The average random hardness of as-grown crystals was determined to be 1520 kg per sq nm. Annealing to 2000°C decreased. the average hardness by 150 units. An increase in hardness after annealing at 2200 ;C was noted. Optical and electron microscopy revealed slip and twin traces on all planes studied, as produced by mi-cvohavdness indentations. Basal (0001)(2i10) slip was determined to be the primary slip system and was substantiated by electron transmission microscopy. A secondary {1010)(2110) slip was produced by mi-crohardness indentations. The lattev also produced twinning- of the {10i2)[0001] type, as proven by electron diffraction. Electrical resistivity and elastic-modulus anisotropy were found and correlated with hardness anisotropy and Mo2C crystal structure. Elastic-modulus values were obtained by microhard-uess and ullrasonic methods. Bonding mechanism of Mo2C is discussed. ROOM-TEMPERATURE microhardness indentations are useful for studying hardness anisotropy, slip, and twinning in brittle materials. Slip has previously been produced in this manner in and WS~~.~ Recently, the authors5 reported slip of the {10i0)(11~0) type produced by high pressure and microhardness indentations on hexagonal TiBz (c/a = 1.066) single and polycrystals. This slip system was also reported by French and ~homas' and Taka-hashi and ~reise~ for hexagonal WC (c/a = 0.976) crystals. These results suggest that prismatic rather than basal slip is favored in hexagonal nonmetallic materials having a c/a ratio considerably less than the ideal (1.633). Buerger precession and cylindrical X-ray rotation patterns were previously' taken on cleaved sections of the Mo2C single crystals studied in this work. Th~y were found to be hexagonal MoaC with a. = 3.0233A, co = 4.7344A, and C/O = 1.5660. The latter ratio is close to that of the beryllium metal (c/a = 1.57), which slips primarily on the (0001) plane: but also slips on the (1010) planes.'0 ~irconium" (c/u = 1.59) and titanium12 (c/o = 1.59) deform mainly by slip on nonbasal planes which contain a close-packed direction. This is due to the fact that for these two metals the initial resolved shear stress for slip on the (10i0) prismatic planes is lower than that on the (0001) plane. The prominence of basal rather than prismatic slip in metals of high c/o ratios is shown by cadmium (c/a = 1.89), zinc (c/a = 1.86), and magnesium (c/a = 1.62) which deform mainly by basal slip. However, in case of the latter, by stressing magnesium crystals in tension or compression parallel to the basal plane, slip on (10i0) planes can also be produced.13 Several hardness values for polycrystalline Mo2C are reported in the literature: Biickle ' and Samsonov'~ give a value of 1800 and 1479 kg per sq mm, respectively, at a 100-g load; and Kieffer and Benesovsk~'~ report a value of 1950 kg per sq mm at a 50-g load. ~ott'~ reports a Vickers hardness value of 2000 kg per sq mm, with the load unspecified. A Rockwell A hardness value of RA = 88 has also been reported.'' In the present work, for comparison with single-crystal Mo2C hardness values, a Khnloo value of 1600 It 150 kg per sq mm was found on 99.6 pct pure and 99 pct dense hot-pressed Mo2C. This work was undertaken partly to explain the considerable differences in hardness values reported for polycrystalline Mo2C. EXPERIMENTAL The Mo2C single crystals investigated were prepared by a Verneuil-type process using an electric arc by the Linde Co. of the Union Carbide Corp.lg The largest specimens grown were boules 7 mm in diam by 40 mm in length. The crystals had an average density of 9.04 g per cu cm, with a Mo + C content of 99.8 wt pct. The major impurities were: 100 ppm each Na, Zr, and Ca; 85 ppm 0, 55 ppm Fe, and 10 ppm each Cr and Ta. The crystals were found to be carbon-poor, the average carbon content being 5.73 wt pct (stoichiometric value is 5.89 pct). The molybdenum content was found to be 94.08 pct, which is nearly stoichiometric. Electron-microprobe traverses of selected specimens were done with a Phillips-AMR microanalyzer. Thin-film and carbon replicas were used to prepare electron micrographs. This work was done with a JEM-6A electron microscope. Prior to optical and electron-optical studies, specimens were mounted in Lucite and polished on a vibratory polisher using diamond-paste grades ranging from 9-1 p and Linde A powder for up to 48 hr. Dilute nitric acid was used for thin-section polishing and chemical etching for 1-15 min. Electrical-resistivity measurements at room temperature were taken with a Rubicon bridge, using gold contacts. For hardness measurements, a Tukon Microhardness Tester Type FB with Knoop and Vickers indenters was used. Measurements were taken at loads ranging from 25 to 1000 g; however, the 100-g load was chosen as the standard load. All measurements were taken at room temperature. Indentations of cracking classes 1 and 2 only were considered for hardness determinations.20)21 (There are six cracking classes, ranging from "class 1" for a perfect inden-

Jan 1, 1967

By R. Ebeling, M. F. Ashby

The paper is in two parts. The first develops the formulae and method needed to calculate the size, nu)nber, spacing, and volume fraction of hard or inert particles in the interior of a specimen from measurernents made on an extraction replica; throughout, the spread of- particle sizes in a single specimen is taken into account, and the extraction efficiency of the replica is considered. A method for testing whether the particles are randotnly distributed in space is described, and the spacing of randomly distributed particles is discussed. The second Part describes the application of the method to copper containing spherical particles of silica; the description includes an estimate of the errors involved, and a comparison of this with alternative methods. All parameters except uolume fraction can be determined satisfactorily, and the method offers advantages over other electron-,rzicroscopic techniques when dealing with hard particles in a softer matrix, or inert particles in a chem-ically less inert matrix. 1 HE experimental measurement of sizes and numbers of discrete particles of a second phase embedded in a matrix material from opaque, plane sections has been considered by many authors, notably by Fullman.' It is assumed that, when the specimen is sectioned, the particles are also sectioned, so that the diameters of the circles of intersection of particles with the plane of the cross section ("surface diameters") are measured. It is difficult to derive the distribution of true particle diameters ("volume diameters") from measurements of the distribution of "surface diameters" on a plane section. The measurement of size and numbers of small second-phase particles by transmission electron microscopy through thin foils of matrix containing the particles has been described by Cahn and ~uttin~' and by ~illiard.~ Because of overlap and the fact that particles which intersect the foil surface may be sectioned whereas those in the bulk of the foil are not, it is again difficult to measure the true distribution of particle sizes. The use of replicas for this sort of measurement has not been analyzed in detail. A replica made from an opaque cross section, thus replicating the circles of intersection of particles with the plane of section, is analyzed by Fullman's method and suffers from the same difficulties. Under certain circumstances, however, extraction replicas can be used, and offer a special advantage in that they make the determination of the true distribution of particle diameters particularly simple. It is the purpose of this paper to describe this extraction replica method. Consider the example of an alloy consisting of SiOz particles, whose mean diameter is about 1000A, dispersed in a copper matrix. When the alloy is sectioned and mechanically polished, the particles are not sectioned, because they are much harder than the matrix, and so will be either left in the surface or pulled out whole, and because the particle size is smaller than the abrasive size of even the finest abrasive. The particles are chemically inert in solutions which chemically etch or electropolish the matrix, so that chemical treatments do not section the particles either. (Experimental measurements of particle shape in this alloy confirm this—truncated particles are never seen.) Small, hard or inert, particles are therefore not sectioned when the specimen is sectioned and polished. An extraction replica taken from this polished surface shows the true "volume" diameter of each extracted particle; the distribution of diameters seen on the replica is the surface distribution of true "volume" diameters, and the volume distribution of true "volume" diameters is easily derived from it. The extraction replica method is not subject to the difficulties associated with sectioning of particles, and avoids the problem of overlap from which the transmission method suffers. The method is well-suited for studying hard particles in a soft matrix. Examples are provided by internally oxidized alloys, oxide-dispersion strengthened metals like T.D. Nickel, and carbides, nitrides, and strong intermetallic compounds in metals. It should be possible to adapt polishing techniques in alloys containing softer but inert particles to avoid sectioning the particles (say by electropolishing or etching). The method is tested and illustrated in Sections 8 and 9 by applying it to a dispersion of spherical SiO2 particles in a copper matrix. I) SYMBOLS Symbols are defined where they first appear in the text. Volume fractions of particles are denoted by the symbol f, particle diameters by x, center-to-center particle spacings by D, and numbers of particles by N. Mean particle diameters and standard deviations about these means are denoted by x and a. Both arithmetic (subscript A) and geometric (subscript G) means are employed. In this paper we measure the distribution of true diameters of particles which touch or pass through a random-plane section of the specimen. We may want to know the distribution of particle diameters in the

Jan 1, 1967

By H. Majima, E. Peters

The oxidation rates of pyrite, pyrrhotite, chalcopy-rite, chalcocite, covellite, bortzite, galem, sphalerite, and stibnite have beet2 carefully compared at 120 oC, using aqueous phosphate solutions buffered at pH 2.7, 7.1, and 11.2 as zvell as in 1.0 M NaOH and in am~onia-containing solutions. Under conditions of these studies, variations in rates of more than a factor of two per unit of mineral surface area were not observed for different minerals in phosphate buffered solutions, except that pyrrhotite showed unusually rapid oxidation rates In acid solutions, and stibnite showed very rapid oxidation rates in basic solutions. In caustic solutions, oxidation vates varied considerably, the most rapid being for pyrite and the slowest for sphalerite. In the presence of ammonia, only copper minerals showed more vapid oxidation rates in neutral solutions. Oxidation rates were measured on the basis of oxygen-consunzption rates per unit surface area of exposed mineral, using equivalent surface areas, mineral mesh sizes, pulp densities, agitation, and oxygen pressure. OXIDIZING pressure leaching of sulfide minerals has been a commercial method of treating concentrates of nickel, cobalt, and copper for over a decade, and recently the treatment of lead and zinc concentrates,' and even the recovery of sulfur from pyrrhotite by such processes, have been shown to be technically feasible.' Furthermore, the leaching of uranium ores by solutions acidified through the oxidation of pyrite and other sulfides present in the ore is commercially practicable.3 5 The mechanisms by which sulfide minerals oxidize in aqueous environments are not yet understood. However, certain generalities can be made for systems containing no unusual reagents, in which oxygen is the oxidizing agent. In acid solutions the oxidation of many sulfide minerals leads to the formation of elemental sulfur.8'7 The attack of some sulfides, notably galena (PbS), sphalerite (ZnS), and pyrrhotite (FeS), at temperatures below about 110°C leads to an almost quantitative conversion of the sulfide mineral to elemental sulfur according to the equation: On the other hand pyrite (FeS2) and copper minerals do not yield elemental sulfur quantitatively. Not more than 50 pct of the sulfur in pyrite is converted to the elemental form in acid oxidation, and this percentage is lowered by higher temperatures and possibly higher oxygen pressures.8'9 Copper minerals sometimes yield elemental sulfurlo'" but, since sulfur can reduce cupric ions with the precipitation of CuS or Cu2S, it is not likely that elemental sulfur is obtained in good yield. In acid solutions sulfur that is not converted to the elemental form always appears as sulfate salts or sulfuric acid. Sulfide minerals that yield elemental sulfur nearly quantitatively on oxidation in acid solutions tend to generate H2S when the conditions are not oxidizing. However, although it has been suggested that H2S is an essential intermediate in the mechanism of oxidation," this is doubtful because it is oxidized very slowly by oxygen. In basic solutions elemental sulfur is never formed, as predicted by thermodynamical considerations, summarized in the form of a potential-pH diagram in Fig. 1. However in neutral and basic solutions, unstable forms of sulfur not shown in the diagram appear during oxidation, notably thiosulfate (SOT-) and polythi-onates (S,O;-, x = 2 to 6).l3,l4 On prolonged oxidation only hexavalent sulfur as sulfate (SOi-) or sulfamate (NHzSO;) is found to remain, the latter appearing only in the presence of ammonia. The oxidation kinetics for thiosulfate by oxygen are the subject of a separate paper,'5 and were studied on the presumption that thiosulfate may, in fact, be a primary oxidation product for sulfide minerals, at least in neutral and basic solutions. A survey of the literature indicated that comparative oxidation rates for different sulfide minerals were needed to establish the role of the mineral structure and its thermodynamic properties on the oxidation rates and mechanisms.

Jan 1, 1967

By W. M. Robertson

The rate of crack Propagation thvough a solid meta1 in the presence of a liquid metal has been calculated. The crack is assumed to propagate by solution of the solid in the liquid under the influence of an applied stress, with volume diffusion of the dissolved solute through the liquid controlling the propagation. The calculated velocities can be tens of centimeters per second or higher. THE embrittlement of normally ductile solid metals by liquid metals has been the subject of extensive study, as summarized in the book by Rostoker, McCaughey, and arkuus.' In experimental studies of liquid-metal embrittlement it is usually concluded that the observed crack-propagation rates are too high to be accounted for by diffusion processes. Diffusion is considered to be an unimportant factor and various other mechanisms are invoked to account for the experimental observations. The similarities of liquid-metal embrittlement to stress-corrosion cracking, to hydrogen embrittlement of steel, and to low-temperature embrittlement of bcc metals are frequently used as analogies for these mechanisms.1"3 The present paper assesses how rapidly a crack could propagate by a mechanism of volume diffusion through the liquid metal. We assume that the crack propagates by solution of the solid in the liquid at the crack tip under the influence of the applied stress and diffusion of the dissolved solute away from the crack tip. The model uses only macroscopic thermodynamic, kinetic, and elastic concepts. The calculations indicate that this dissolution-diffusion mechanism can give high crack velocities. In some previous treatments4'' of crack propagation in a liquid-metal environment the fundamental step has been the breaking of atomic bonds at the crack tip. The liquid greatly decreases the energy necessary to break these bonds. The present model assumes the crack is filled with liquid metal; the presence of the liquid is essential because diffusion through the liquid is assumed to control the rate of propagation. With proper modification this model could also apply to stress corrosion in a liquid environment. THEORY Consider a crack in a homogeneous elastic continuum, ignoring all the structural details of the solid. This neglects a host of metallurgical details which can be significant in the problem of liquid-metal embrittlement;' but this simple assumption is a useful approach in evaluating the role of diffusion through the liquid. The stress distribution around the crack tip establishes a gradient in the chemical potential at the tip, which causes a diffusion flux through the liquid away from the tip. The solid-liquid interfacial tension works against crack propagation, since propagation requires creation of interfacial area. From these two considerations we derive the diffusion flux, which then yields an expression for the velocity of the crack tip. The concentration of the solute in the liquid, C, is equal to the equilibrium concentration in the presence of an unstressed, flat surface, Co, plus the excess concentration due to the stress, AC(a), plus the excess concentration due to capillarity, AC(y): We calculate AC(o) and AC(y) and then obtain the diffusion flux. Consider (Fig. 1) a long flat crack of length 2L, thickness 2r, and tip radius r, where r remains constant as the crack propagates in the positive x direction. There is no stress ay in the y direction across the flat surfaces of the crack. The magnitude of the stress at the crack tip due to the applied stress, a,, is given by the relation6 ^=^[l-2^] 12] which, for long narrow cracks, becomes The stress increases the chemical potential in proportion to the strain energy.? The excess chemical potential due to the stress, Ap(a), is given by where 52 is the atomic volume of the solid and E the Young's modulus. Inserting Eq. [3] gives Now consider the crack to be filled with a liquid metal. The excess solubility due to stress, hC(a), is given by the excess chemical potential due to capillarity, Ap(y), is given by8 where y is the solid-liquid interfacial energy, and v1 and YZ are the two principal radii of curvature of any surface point. At the crack tip, r1 = -r and vz = m, so TRANSACTIONS OF THE METALLURGICAL SOCIETY OF AIME

Jan 1, 1967

By J. B. Conway

Stress-vuptuve data for several of- the refractory metals are frequently found to yield a linear relationship between the Larson-Miller parameter and the logarithm of the applied stress. In such cases linear stress-rupture isotherms result with slopes bearing a definite relationship to the temperature. It also follows that the stress to produce rupture in a certain period of time will be linear in temperature. Data for several refractory metals have been reviewed and excellent linearity is shown in this type of isochronal plot. In addition, the af ore - mentioned lineavity leads to a linear relation between the log of the stress to produce rupture in a certain time and the homologous temperature. This has been illustrated for the Group VI-A metals, tungsten and molybdenum. EXTENSIVE use has been made of the Larson-Miller' parameter in the interpolation and extrapolation of stress-rupture and creep data. In those cases where this particular parametric approach is applicable a convenient and fairly straightforward procedure is made available for the correlation of experimental stress-rupture data. It is quite common to employ this parameter in the form of a master rupture plot in which the parameter, T(C + log tr), is expressed as a function of log stress. In many cases this functional relationship in log stress is linear within acceptable accuracy and hence the following relation results: where P is the parameter, C is the Larson-Miller constant, T is the absolute temperature, t~ is the rupture time, a is the stress, and a and b are constants. Examples of such a relationship are contained in the work of Green, Smith, and 01son2 dealing with high-temperature rupture behavior of molybdenum and in the work of Green' dealing with the high-temperature behavior of tungsten. In addition, pugh4 has shown a similar linearity for some fairly low-temperature data for molybdenum. It can be shown that when the relationship in Eq. [I] is exhibited certain generalizations can be made concerning the form of the stress-rupture isotherms. For example, rearranging yields: For a given material (constant C) at a given temperature the first term on the right-hand side of Eq. [2] is a constant and hence this equation defines a straight line when log stress is plotted as a function of log-rupture time. This is recognized as the standard form usually employed in this type of data presentation. Such linearity then suggests the linear form of the Larson-Miller parameter. Or, in other words, the linear parametric relationship in Eq. [2] results only when the stress-rupture data are linear on a log-log plot of stress vs rupture time. Another interesting observation can be made in regard to Eq. [2]. It can be noted that the slope of the stress-rupture isotherms is given by - T/b and hence a direct calculation of the constant b is available. It also follows that since the value of b is the same for all temperatures the slopes of the various isotherms on the log-log stress-rupture plot cannot be the same. Indeed, the existence of the relationship in Eq. [2] precludes a system of parallel lines on this common stress-rupture plot. As a matter of fact it further specifies that in addition to being nonparallel the slope of these isotherms must decrease (i.e., become more negative) with increasing temperature. Such a condition is indeed found to exist in the case of the stress-rupture data reported for molybdenum.' As a corollary to the above, it may be stated that stress-rupture data which do not lead to a linear log-log stress-rupture plot or whose isotherms do not exhibit a decrease in slope as the temperature increases will not yield the linear relationship of Eq. [I]. Applying Eq. [2] to two different temperatures and solving for C yields: Eq. [3] affords a simple and rapid method for calculating the Larson-Miller constant from the log-log stress-rupture plot. The slope of a given linear isotherm is measured and the value of "b" calculated based on Eq. [2] as: slope = - -Tb Then at an abscissa value of 1.0 hr (making log tr in Eq. [3] equal to zero) read the stress corresponding to rupture for two different temperatures. Substitution in [3] yields: A value of the Larson-Miller constant can thus be calculated from a few simple mathematical procedures employing data read directly from the log-log plot of the stress-rupture data. Of course, it is not to be overlooked that the above reasoning has been based on the linear relationship of Eq. [I] being applicable. However, if as mentioned above the log-log plot is

Jan 1, 1967