Search Documents

By Hans-Jürgen Schaller, Horst A. Brodowsky

Activity coefficients of boron in palladium were determined at concentrations up to PdB0.23 by reducing B2O3 between 870" and 1050°C in a controlled H2-H2stream and measuring the resulting weight gain. The deviations from ideal behavior closely resemble those of the system Pd-H and are interpreted in terms of three principles: 1) The solute atoms occupy octahedral interstitial positions. 2) They donate their valence electrons to the 4 d and 5s bands of palladium, raising its Fermi energy. 3) The lattice strain energy is lower for two nearesl -neighbor interstitial particles than for two farther separate ones. SOLID solutions of hydrogen in palladium are a useful subject for studying thermodynamic aspects of the formation of alloys and of nonstoichiometric systems.1-3 The activity of hydrogen is readily measurable to a high degree of accuracy,4'5 even at low temperatures where the deviations from ideal behavior are more pronounced, and its simple structure facilitates an interpretation of these deviations in terms of a detailed model. Two effects are discussed to account for the non-ideal properties:3 An "electronic" effect, connected with the rise of the Fermi energy, as electrons of the interstitial hydrogen atoms enter the electron gas of the metal, and an "elastic" effect, due to an interaction of the regions of strain around each interstitial atom. The electronic effect is based on the idea that the lowest energy levels of the dissolved hydrogen atoms are higher than the Fermi energy, so that the electron will not occupy a localized state but enter into the electron band of the metal.6 The elastic effect is based on the observation that dissolved hydrogen distorts and expands the palladium lattice. The hypothesis is put forward that the elastic strain energy is lower for two adjacent dilatational centers than for two separate ones; i.e., they attract each other. The resulting pair interaction can be used to calculate an elastic contribution to the thermodynamic excess functions by means of one of the statistical methods. This model permitted a detailed description of the solution properties of hydrogen in palladium3 and in palladium alloys.798 An extension of the approach to describe the excess functions of substitutional palladium alloys is possible.9 In order to further test and refine the model, an investigation of other interstitial alloys was started. Palladium dissolves considerable amounts of boron in homogeneous solid solution.10 The palladium lattice expands linearly up to nB = 0.23 (nB = B/Pd atomic ratio), the highest concentration studied." The expan- sion, extrapolated for 1 mole of interstitial per mole of palladium, is 17 pct of the lattice constant of pure palladium vs 5.7 pct in the case of hydrogen.12 The fact that the lattice expands rather than contracts is a strong indication that interstitial positions are occupied. According to neutron diffraction experiments, hydrogen occupies the octahedral sites of the fcc lattice.13 Unfortunately, this direct evidence is not available for the Pb-B system, mainly because of the high-reaction cross section of boron with thermal neutrons. However, by way of analogy and on the grounds of the rather close similarities between the two systems to be reported here, it seems safe to attribute octahedral positions to the dissolved boron, too. At higher boron contents, compounds of stoichiomet-ric compositions are reported such as Pd3B, which has the structure of cementite,14 so that a close structural relationship seems to exist with the system r Fe-C. In their study of hydrogen absorption in Pb-B alloys, Sieverts and Briining noted that alloys with an atomic ratio of about nB = 0.16 are no longer homogeneous15 This observation was confirmed in an extensive X-ray investigation.11,16 The phase boundaries of two miscibility gaps were established. One two-phase region was stable below a transition temperature of about 315°C and extended from nB = 0.015 to 0.178. The other one extended from nB = 0.021 to 0.114 slightly above the transition temperature and had an apex at nB = 0.065 and 410°C. All phases involved have the fcc structure of pure palladium with lattice expansions proportional to their boron contents. The occurrence of miscibility gaps, i.e., the coexistence of dilute and concentrated phases, points to an energy of attraction between the dissolved particles, in the Pb-B system as well as in the Pd-H system. The filling up of the electron bands seems to be analogous, too, in the two systems, as indicated by the hydrogen absorption capacit15,17,18 and by the suscepti bility of Pd-B alloys.l8 In both types of experiments, boron acts as an electron donor. A chemical method was used to measure the activity of boron in palladium. Boron trioxide was reduced in a moist hydrogen stream: B2O3 + 3H2 = 2B + 3H3O [l] At known activities or partial pressures of boron trioxide, hydrogen, and water, the activity of boron could be calculated from the law of mass action. The equilibrium concentration of boron corresponding to this activity was determined as the weight gain of the sample. EXPERIMENTAL The samples consisted of small pieces of foil of 0.1 mm thickness and about 100 mg weight. The palladium was supplied by DEGUSSA, Germany, and stated to be

Jan 1, 1970

By M. J. Mullikin, J. B. Conway

for use in describing combined first- and second-stage creep data. In the above ?o represents the instantaneous strain on loading, ?t the limit of transient creep, ?S the steady-state creep rate, t the time, and ? a constant leading to the exhaustion rate of transient creep. It was noted that this equation was similar in form to that proposed originally2 by McVetty. This type of equation was also suggested by Andrade and Aboav3 in studies of the shear strain of cadmium. Calculation of the constants ?o, ?t, ?, and ?s, in Eq. [ l] is somewhat involved due to the slight complexity of this relationship. In the paper by Garofalo a computer approach based on a least-squares analysis was recommended. Such an approach involves a trial and error solution but when a computer facility is available the difficulties in this calculation are of little consequence. It does mean however that the evaluation of this type of equation would become quite tedious if computer facilities were not available. Some recent studies have identified a fairly straightforward method for calculating the constants in the Garofalo equation. While not leading to a least-squares evaluation the approach is accurate and is recom-

Jan 1, 1967

By R. A. Pasternak, B. Evans

A dynamic technique for the determination of solubilities of gases in metals has been explored, using the N-Nb system as a test case. An initially clean sample, maintained at constant high temperature, is exposed to the gas at low pressure: the rate of sorp-tion and the corresponding change in sample resistance are measured continuously. A pronounced decrease in the slope of the resistance concentration curve is observed at the terminal solubility. The rate of gas uptake by the sample was determined in an ultrahigh-vacuum flow system equipped with a servo-control which maintains constant pressure in the sample chamber; the system and the technique have been described in detail elsewhere. 1,2 The sample was held at constant temperature during sorption by maintaining approximately constant power input into it; this results in constant temperature, since the emissivity of a metal is independent of moderate concentrations of dissolved gas. The simple electrical circuit used both for heating and for resistance measurement is shown in Fig. 1. The power input into the filament is where R is the filament resistance, RL is an adjustable resistance in series with the filament, and VT is the constant dc potential of the power source. If RL is set equal to R at the beginning of an absorption run, the power input becomes insensitive to subsequent resistance changes; for example, an increase of 20 pct in sample resistance is accompanied by a decrease in power input of less than 0.8 pct; i.e., the

Jan 1, 1967

By V. Leroy, A. G. Guy

Serious disagreements are often found between experimentally determined intrinsic diffusion coefficients and those calculated employing the usual form of the vacancy theory. In the new theory it is proposed that the total intrinsic flux, Ji, of component i, is the sum of a part, f, due to the usual random exclzanges of component i with the vacancies, and a second part, Ji, due to exchanges with the uacancies composing the net vacancy flux. The present treatment, while less powerful than that of Manning, has the advantage of easy uisualization and of facilitating the application of the vacancy-flux effect to complex systems. IT is becoming increasingly evident that there are serious deficiencies in the version of the vacancy theory of diffusion that has been widely used for the past 20 years. One type of evidence is the frequent lack of agreement between intrinsic diffusion coefficients and tracer diffusion coefficients, even taking account of the thermodynamic factor. A second kind of evidence is the observation of a Kirkendall shift larger than theoretically possible, that is, larger than can be accounted for without assigning a negative value to one of the two intrinsic diffusion coefficient.'- The thermodynamic factor could conceivably make both coefficients negative, but not just one. It is clear that a cause of these anomalies, apart from any inadequacy of the usual vacancy theory, might lie in an oversimplified treatment of the data. Adequate experimental techniques, including the use of moderate pressure during the diffusion anneal, are now available to insure that porosity, lateral expansion, and so forth, can be kept negligibly small in most cases. The effect of differences in atomic volume can be of major importance, and it is essential that one of the available methods4 be used to account for this factor. In the present treatment this is accomplished by the consistent use of moles per cubic centimeter as the unit of concentration. Of the various possible inadequacies of the vacancy theory, attention will be given here only to effects of the net vacancy flux. annin' has previously considered this question, beginning with an analysis of atomic jumping of tracer atoms. When he added the effect of a concentration gradient, new terms arose that could be associated with the flow of vacancies. The present treatment uses quite a different approach. The usual vacancy flux, J,, is introduced explicitly, and a simple analysis predicts major changes in the intrinsic diffusion coefficients from this cause. The usual assumptions are made that only a vacancy mechanism is operative, that the formation of voids can be neglected, and that changes in the partial molal volumes, vl and v2, are negligible. The significant diffusion coefficients for the present topic are Dl and D,, the intrinsic coefficients, which enter in the equations, where the flux Ji, moles per sq cm per sec, is that crossing the Kirkendall interface. The concentration, ci, is in units of moles per cu cm, and the concentration gradient, aci/ax, is evaluated at the Kirkendall interface. It will be recalled' that the calculation of Dl and D2 involves the measurement of areas on the diffusion curve with respect to the positions of the Kirkendall and Matano interfaces. In the case of the anomalies mentioned earlier, the Kirkendall shift is too large to be accounted for by the diiferetzce in fluxes (J2 -J1), given by Eqs. [I] and [2]. The logical inference is that the flux of the solvent atoms, J1, is actually in the same direction as the flux of the solute atoms, Jz. In terms of Eq. [I] this requires that Dl have a negative value. However, it would be somewhat misleading to state that the solvent atoms are diffusing up their own concentration gradient. The explanation that will be advanced here pictures competing processes producing the net flux of solvent atoms: 1) diffusion of the solvent atoms down their own gradient by random exchanges with vacancies; and 2) diffusion of solvent atoms in the opposite direction by exchanges with the net vacancy flux. ACTION OF THE NET VACANCY FLUX Theories of vacancy diffusion can be formulated with varying degrees of refinement, and the present theory has purposely been kept as simple as appeared adequate to explain the phenomenon in question. In particular the following aspects have been neglected: 1) the gradient of vacancy concentration in comparison to the gradient of the atomic species; 2) departure of vacancy concentration from the local equilibrium value; 3) variation of the jump frequency, LO, with the specific surroundings of the atom-vacancy pair being considered; 4) correlation effects. These and other refinements can be considered once the essential mechanism has been established. The essential idea of the present analysis is to calculate the total intrinsic flux, Ji, of component i as the sum, JlJ?±j{ [3] where J; is attributable to the usual random atomic jumping, and J{ is a contribution arising from the net vacancy flux, J,. The latter quantity, of course,

Jan 1, 1967

By K. A. Jackson, J. D. Hunt

A new classification of eutectics is proposed, based on tlze entvopies of wzelting of the tuio eutectic phases. The clnssification was used to predict suitable tvansparent analogs of the metallic systems. Experimental confir?nation loas obtained for the theovetical shape of the lamellar solid-liquid interface, fov the fault mechanisms of lanzellar spacing changes, and for the development of low-energy solid-solid boundaries between the lamellae. An explanation is presented to account jov the irvegular and coinplex regular structures zrhich are found in some eritectic systems. FrOM experimental observations, single-phase materials can be divided into two groups according to their solidification characteristics, those that grow as faceted crystals and those that do not. acksonl' showed from thermodynamic reasoning that the type of growth depended on a factor a which was almost thg entropy of melting. Most nonmetals have high entropies of melting (a greater than 2) and grow with crystalline facets. Most metals have low entropies of melting (CY less than 2) and grow almost isotropically with no facets. The authors propose that eutectics may be classified in a similar manner. There are three groups of eutectics, those in which both phases have low entropies of melting, those in which one phase has a high and the other phase has a low entropy of melting, and those in which both phases have high entropies of melting. Lamellar or rodlike structures are formed in systems in which both phases have low entropies of melting. In these alloys dendrites of either phase may be formed, when the alloy is rich in the relevant component. Examples are Pb-Sn, Sn-Cd, Pb-Cd, Sn-Zn, Al-Zn. Irregular, Fig. 14((), or complex regular, Fig. 14(b), structures are formed in alloys in which one phase has a high entropy of melting and the other has a low entropy of melting. Examples are A1-Si, Zn-MgzZnll, Pb-Bi, Sn-Bi. When the alloys are rich in the low entropy of melting phase, dendrites are formed; when the alloys are rich in the high entropy of melting phase, faceted primary crystals are produced. These crystals are sometimes called hoppers or pseudodendrites. In this work the term dendrite will only be used to describe nonfaceted primary crystals. Dendrites are not formed during solidification in high entropy of melting single-phase materials. The third group of eutectics includes alloys in which both phases have high entropies of melting. Each phase grows with a faceted solid-liquid interface. Since most metals do not have high entropies of melting, metallic examples in this eutectic group are rare. However they may occur between some intermetallics and semiconductors or semimetals such as silicon, germanium, and bismuth. Attempts have been made to study eutectic solidification visually by watching the growth process.374 Since metals are not transparent, the observations had to be made on external surfaces. This difficulty can be overcome by using transparent analogs of the metallic systems. As was mentioned earlier, most single-phase compounds have entropies of melting greater than 2 and so grow as faceted crystals. Recently organic materials with entropies of melting less than 2 were investigated.' These materials grow in exactly the same way as the low entropy of melting metals. When the materials are pure, they grow with a solid-liquid interface parallel to an isotherm; when they are impure, cells or dendrites are formed. Since these materials are transparent, have low melting points, and even have cubic structures, they should be ideal for making up transparent analogs of the metallic eutectics. The purpose of the present work was to investigate these organic eutectics and to see whether this quite different series of eutectics could be classified in the same way as the metallic systems. The observations made on the organic alloys are also discussed with reference to the current theories of lamellar growth. Explanations are proposed to account for the structures formed in the other eutectic groups. EXPERIMENTAL Thin cells containing the organic alloys were uni-direction ally solidified on a specially constructed microscope stage.' Uniform growth rates were obtained by moving the cells, with a motor drive, through a fixed temperature gradient, so that the solid-liquid interface remained stationary with respect to the microscope objective lens. The cells were made by fusing two microscope cover slides 7/8 by 7/8 by 1/100 in. together on three sides, leaving a gap of 1 to 3 mils between the slides, and these were filled by surface tension. A preliminary investigation of the phase diagram between two components could be made very rapidly. One side of the cell was filled with component A and the other side with component B. Since only a small amount of mixing could occur every composition from pure A to pure B was present in the cell. When the cell was placed in the temperature gradient a pictorial representation of the phase diagram was obtained. Eutectics, peritectics, "interorganics", and solid -solid transformations could be readily detected. Fig. 1 shows part of a eutectic phase diagram. The Sample was first grown slowly then stopped. The two

Jan 1, 1967

By A. J. Opinsky

ThE superior performance of nonsag tungsten wire in lamps has been ascribed to the large-grained micro-structure that can be developed in it. The orientations of these large grains have been studied several times; most published data1-' appear to be similar to those plotted in Fig. 1. These new data were obtained by X-ray back-reflection Laue technique from all of the large grains occurring in thirteen different wires that had been heated at the lowest possible temperature for the formation of large grains. Here the wire axis is plotted in terms of the principal directions of the cube. Rosi decided that the preferred orientation of the large crystals could be described by a single ideal orientation, (531), being parallel to the wire axis.' Rieck2,3 and Mannerkoski' have essentially confirmed this finding. Opinsky, Orehotsky, and Seigle argued that the data suggest that (211) is parallel to the wire axis instead of a single ideal orientation.5 This {211} contains the (531) ideal orientation and would appear to be a more general case. Ahlborn and Wassermann report that their orientation data stretch from ( 110) through (531) to (311) 8 This is part of the locus of (211). All these investigators have determined the orientations of the large grains by back-reflection Laue techniques. They have all obtained essentially similar results. On the other hand, Swalin and Geisler, who obtained their data by neutron diffraction, indicated that the preferred orientation of the large grains in doped wire is (110) .7 These findings cannot presently be reconciled with the pole density data described in the previous paragraph. Rosi1 and Rieck2 have advanced arguments about how their orientations could occur. Nevertheless, the crystallographic link between the (110) (cold-worked) fiber texture and the preferred orientation of the large grains has remained missing. This link can be provided through the concept of coincidence-type boundaries. Such boundaries separate two grains that have a certain number of common lattice sites. Coincidence relationships that result from rotations about low index directions are given by Brandon et a1 .' What is desired is a calculation of those orientations that bear a coincidence relationship with the (110) fibers of the cold-worked matrix. This situation arises in the following fashion. Heating the tungsten wire for a combination of time and temperature falling short of the formation of the nonsag structure yields a structure that is still fibrous in nature and the preferred orientation remains (110). During grain growth, the growing grain is in contact with these fibers. It is hence possible that the entire process may begin by the passage of a coincidence boundary into a (110) fiber. The calculations were carried out analytically and checked graphically. Following Brandon et a1.,8 coincidence relationships from 1 in 5 to 1 in 19 have been considered. The results of the calculations are given in Fig. 2. Of the coincidences examined, some of those occurring upon rotation about ( 110) come closest to the experimental points in Fig. 1. Nineteen of the twenty-three data points in Fig. 1 can be attributed to the ideal orientations A, B, C, and D in Fig. 2. As determined analytically, three crystals lie within 4 deg of point A, four lie within 21/2 deg of point B, eleven lie within 5 deg of point C, and one lies within 1 deg of point D-a total of nineteen. The off-lying points are near the |110| point of the triangle.

Jan 1, 1968

By A. Gittens

In two recent papers stevens13,14 has compared the methods used by various investigators to calculate the contribution of grain boundary sliding to the total creep strain. In previous work Gittins15 has shown that the strain due to grain boundary sliding Egb may be estimated from the lateral, d, and vertical, v, displacements of a longitudinal marker line as it crosses a grain boundary. He obtained the formula, where nL is the number of grains per unit length measured on the specimens before creep, and \p and ?2 are, respectively, the angle between the grain boundary and the stress axis in the plane of the surface and normal to the surface (see Ref. 13, Fig. 1). The subscript L indicates that observations are made in a longitudinal direction parallel to the stress axis. It was generally found more convenient to measure the grain size of a specimen before creep; this was necessary to characterize the specimen. Stevens13,14 however prefers to define Egb in terms of NL, the number of grains per unit length after creep. Since nL - NL(l + €1) where cr is the total strain, Eq. 1 11 becomes The quantities d and ?, can be determined directly from the marker lines, while v and ?2 are found by sectioning normal to the surface. In the interior when symmetry considerations suggest that v = d and ?1 - ?2 the strain due to grain boundary sliding may be estimated from a suitable internal longitudinal marker line by the equation To determine cgb at the surface using either Eqs. |1| or [2] requires a large number of observations, particularly if many estimates of cgb are required, and also it is impossible to determine the variation of cgh with creep strain on the same specimen because the specimen must be sectioned to determine ?2. Gittins15 found that for a number of materials the sliding vectors d and v were related to each other so that provided one was measured the other could be estimated. Also during creep, grain boundaries tend to migrate to meet the surface at angles close to 90 deg,16 so that the term nL(v/tan?2)L was comparatively small when 4,- 90 deg. By determining the frequency distributions of ?2 on control specimens this term can be estimated without sectioning the creep specimen. In this way an empirical equation based on Eqs. |1| and |2| enables cgb to be determined with reasonable accuracy with fewer observations of the angles and sliding components. For B brass, the relationship between the sliding vectors indicated that ceb = 0.85nLVL . In Table I of Stevens paper this equation is stated to be incorrectly derived. However it should be pointed out that this is an empirical equation derived from Eq. |1|, based on a knowledge of the relationship between d and v and on the frequency distributions of ?1 and ?2 for a particular material under given experimental conditions. It was used only when these relationships were known, to reduce the number of observations required to estimate cb. Therefore the comment "derivation incorrect does not apply. Gittins did in fact make use of Eqs. |1| and 131. Author's Reply I must apologize to Dr. Gittins for the inadvertent misrepresentation of his work. Other slight errors in Table I have been pointed out to me. Langdon (No. 12 Table I), Gifkins (No. 7, Table I), and possibly other workers did not measure the number of boundaries per unit length after creep (A') but the corresponding quantity bejove creep (n). As Dr. Gittins points out, a completely analogous set of equations to those given in the paper can be derived by the same method: using the number of boundaries per unit length before instead of after creep. The basic equations corresponding to Eqs. [2] and 171 of the paper, from which all the others can be derived, are

Jan 1, 1968

By J. C. Scully

In June, 1965, Sanderson and scully5 reported at the Conference on Environment Sensitive Mechanical Properties of Materials at the Research Institute for Advanced Studies at Baltimore that thin foils of titanium could be prepared by chemical polishing in a mixture of 44 pct HF and 73 pct HNO3 in the volumetric ratio 2:3. The technique is simpler than that since reported by Rice and Orava. As-received annealed titanium (containing 0.8 pct Fe and 30 ppm H) in the form of sheet 0.05 cm thick is placed in the solution and moved up and down. The reaction is violent but suitable foils are obtained very easily and consistently. A typical micrograph is shown in Fig. 3. The small precipitates were not identified. Yamane and ueda6 observed many spotlike features which they referred to as small dislocation loops. While their spots may be due to the reaction of the warm foil with the air as suggested by Rice and Orava, the precipitates in Fig. 3 are also found when foils of titanium are prepared by electro-polishing at low temperatures and subsequently examined in the electron microscope without any inter- mediate warming. Furthermore, they are not a surface effect. It is necessary to add that alloys of titanium cannot be so easily thinned by the chemical polishing solution described above since it contaminates the a, phase with hydrogen and produces y hydride. These matters are described at length in a paper7 and their relevance to stress corrosion is considered elsewhere.'-* Author's Reply The authors are grateful to Dr. Scully for bringing to our attention his hitherto unpublished technique for thinning unalloyed titanium, and deeply regret the unintended apparent duplication. The chemical thinning solution (7 parts HNO3 and 3 parts HF, by volume) reported by us is the same as that with which, for some period in the past, we had experienced success in the preparation of foils from single and polycrystalline niobium. Its subsequent effective application to titanium was a logical step, which, clearly independently of Scully, was conceived by Mr. Rice, to whom full credit is due. A few comments are warranted concerning some of the details of the preceding Discussion. It is evident to the discerning reader that the two thinning techniques under consideration are chemically almost identical, and in that sense are equally "simple". The chemical aspect is, of course, the one of primary importance in both communications. Thereafter, the relative simplicity is dictated by the mechanics of the actual thinning operation selected by the individual investigator. Our standard dimpling, or dishing, procedure was chosen with a view toward providing, for the electron transparent areas of the sample, the maximum protection from extraneous deformation during handling. That the long, straight dislocations visible in Scully's electron micrograph of a seemingly recovered structure closely resemble handling dislocations would tend to indicate the merit of our adopted technique. Features similar to those which Scully has attributed to precipitates, without any apparent diffraction contrast confirmation, were not observed in any of our clean foils taken from annealed or deformed specimens. This was admittedly somewhat surprising since the 3700 ppm (nominal) Fe in the A-70 commercial-grade titanium examined should have given rise to TiFe precipitates. Presumably these existed but were un-resolvable. Curiously, Scully's "precipitate1' density is fairly high although the iron content quoted was less than one-fourth of that in our material. One is forced to conclude from the evidence that prior thermal and mechanical treatments are responsible for the two different types of structure observed.

Jan 1, 1968

By Paul Cherin

ThE angle of intersection, 0, between h2k2l2 planes and the (loo), (110), (101), and (111) planes was calculated on an IBM 7040 using the well-known relationship

Jan 1, 1968

By W. C. Winegard, C. J. Beinjessner

PREVIOUS investigations 1,2 of Fe3P described by pearson3 and schubert4 show that Fe3P has a bct unit cell having four Fe3P tetrahedra centered around each lattice point. The space group is 14, No. 82, and all atoms occupy general positions 8(g). The cell dimensions3 are: a = 9.090, c = 4.446 kX, with an axial ratio of c/o = 0.4891. We have measured the lattice parameters of Fe3P, using a Fe as an internal temperature standard, at 23°, 414°, and 678°C and computed the thermal expansion over this temperature range. The Fe,P was prepared by melting under vacuum (after flushing with argon) a commercially pure iron and "ferrophosphor" in a high-frequency induction furnace. Part of the ingot was crushed for X-ray examination and for chemical analysis. X-ray diffraction and light microscopy indicated iron phosphide (Fe3P) present as the major constituent and a small amount (about 5 pct) of a Fe. Chemical analysis showed 0.05 wt pct Si, 0.008 wt pct N, 0.018 wt pet C, and less than 0.02 wt pct each of other metallic elements. X-ray diffraction powder patterns were obtained from a mixture of iron and Fe3P sealed in an evacuated quartz capsule using a Unicam high-temperature

Jan 1, 1968

By H. Biloni

PORE formation during solidification has recently been studied theoretically by Piwonka and Flemings.1 For the ordinary conditions of cellular solidification, these authors state that there is a difficulty in explaining the formation of pores. The difficulty arises because the conditions necessary for pore formation are not those obtained during cellular growth. Therefore Piwonka and Flemings1 had to examine various unusual possibilities in order to explain the apparent presence of pores in a cellular substructure of aluminum, taken from Doherty.2 For aluminum, Doherty and Davis3 have developed a pitting technique to reveal substructure, which depends upon the condensation of vacancies from supersaturated solution after quenching. This technique as further developed4 was used to study solidification substructures. In the case of striations, delineation of the substructure is produced by the absence of surface pitting after quenching: in the vicinity of the striations the dislocations which constitute this substructure are sinks for vacancies. In the case of the cellular substructure the experimentally observed effect is the same. However, in this latter case Doherty and Davis4 suggested that micropores associated with the cellular substructure were the agent responsible for providing sinks for vacancies. This interpretation seemed necessary for the cellular substructure because no misorientation arising from a dislocation array could be observed. It is noted especially that Fig. 1 presented by Piwonka and Flemings1 and Fig. 5 by Doherty and Davis,4 as

Jan 1, 1968

By R. Kossowsky

USING etch-pit techniques, Johnston and Gilman' have shown that the stress dependence of dislocation velocity can be represented by the expression where r, is the dislocation velocity, a is the applied stress, and oa, is the stress to produce a unit velocity. Noble and Hul12 showed that the exponent m is strain-dependent and increases rapidly in the work-hardening region. This was explained by arguing that in the work-hardened material the applied stress required for any given velocity will be greater than in the annealed material by an amount equal to a back stress due to work hardening. schadler3 has stressed the importance of substructure as a possible source of error in dislocation velocity measurements, and took care to make velocity measurements in which dislocations moved such short distances that interactions with boundaries were avoided. schadler3 also pointed out that, if standard testing equipment is used, measurements of m from flow stress data could be seriously in error in materials with a high initial work-hardening rate. Noble and Hu112 have further argued that, due to the strain dependence of m , the extrapolation to zero strain4 to obtain the true value of m can be justified only in the preyield region. Microstrain experiments in high-purity iron5,6 have led to the conclusion that aiA (Ref. 5) (defined as the stress to first produce a permanent set of 10-6 in. per in.) and oflow are not controlled by the same deformation mechanism. It has been proposed5,6 that aA is governed by the lattice resistance to the motion of screw dislocations in a given slip plane and that oflow is controlled by the lattice resistance to a full-scale cross slip. The purpose of this note is to extend Noble and Hull's argument further and to show that measurements of m from flow stress data can be in error due to the rapid increase in stress in the microstrain region6 when stress is increased beyond aA. Double z-one-refined single crystals of molybdenum (with the (101)[ 111] slip system operating and a Schmidt factor of 0.475) were used in an experimen-

Jan 1, 1968

By David P. Shoemaker, Clara Brink Shoemaker

The I phase was discovered by Bardos, Malik, Spiegel, and Beck1 in the V-Ni-Si system at 1100°C and in the Mn-Co-Si and Mn-Ni-Si systems at 1000°C. Kuzma and Hladyshevskii2 and Kuzma, Hladyshevskii, and cherkashin3 have reported a new phase, which they named the S phase, in the Mn-Co-Si and the Mn-Ni-Si systems at 800°C at approximately the I phase compositions. The powder diffraction diagrams of the I and S phases, compared by Bardos el al.,1 are related but show distinct differences in relative intensities. Professor Paul A. Beck kindly sent us a specimen of the (v-Ni-Si) I phase for single-crystal study. Precession and Weissenberg diagrams taken with a tiny single-crystal fragment showed that the structure is monoclinic with a large unit cell. The cell dimensions were refined from a powder diffraction diagram taken with CrKa radiation in a 57.3-mm camera with our computer program LSCELP4 using only the indices of those reflections that were observed on Weissenberg diagrams (hnl, with n = 0 through 9 and n = 12) and precession diagrams (Okl, lkl, Izk0, hkl, hk2). Table I shows the agreement obtained between the l/d values calculated with the final values for the cell dimensions and the l/d values observed by Bardos et al .1 for I(V42Ni33Si25), by US for I(V41Ni36Si23), and by Kuzma and Hladyshevskii2 for S(MnpCoSi). The powder patterns contain many broad and diffuse lines. In fact practically all powder lines observed are near-superpositions of several reflections. The refinement program is not set up to handle several reflections contributing to one observed line: the position of each line is refined with the aid of one set of indices only. The final values for the cell dimensions were obtained from the lines of strong or medium intensity with the indices indicated by asterisks in Table I: a0 = 13.403(6), bo = 23.336(15), co = 9.129(6)A ß = 99.11(6) deg The standard deviations (given in parentheses) are those calculated by the refinement program from least-squares theory and do not take into account that the lines included in the refinement may not have been correctly indexed in every case or that the position of a line may also be partly determined by the overlapping reflections. The agreement between the calculated and observed I phase patterns is fairly good, although some discrepancies occur. Not all observed lines could be indexed with the incomplete set of sin-

Jan 1, 1968

By O. D. Gonzalez

ThIS communication presents the results of a determination of the permeabilities of hydrogen and deuterium in a iron from 360° to 560°C. Recently Heu-mann and primas' have given values of the diffusivity and solubility in a iron of the two isotopes from which the permeability can be calculated. The work reported in the present communication, however, is a direct measurement of the rate of permeation of the isotopes through an identical iron specimen under the same conditions, so that systematic errors can be minimized in the calculation of the ratio of the permeabilities—the value which is usually of main importance in establishing limits for isotopic differences in diffusive processes. Experimentally the procedure used was the usual one: a steady-state flow of gas produced by a pressure difference across an iron membrane was observed as the pressure rise with time in a previously evacuated, calibrated volume connected to the outlet side of the membrane, while a constant pressure of gas was maintained on the opposite inlet side. The membrane was a cylindrical tube 9.38 cm long, 0.730 cm inside diam, 0.187 cm wall thickness, fabricated of 99.95 pct vacuum-melted National Research Corp. iron (less than 0.003 pct C). One end was sealed by a heliarc-welded iron (99.95 pctj lid of 0.147 cm thickness; the other end of the tube was welded to a 304 stainless-steel tube (0.318 cm ID and 0.318 cm wall) which connects to a vacuum system and to the calibrated volume including a dibutylphthalate manometer. The temperature of the iron membrane was maintained constant to ±0.l°C and was measured with an iron-constantan thermocouple which was calibrated against a standard Pt/Pt-Rh thermocouple: the temperature variation along the length of the membrane did not exceed ±0.5°C. Hydrogen was allowed to pass through the membrane assembly for 72 hr at 615°C before permeability measurements were begun. The system was tested at temperature before and after the experiments by a substitution of helium for hydrogen in the inlet side; under these conditions the pressure rise observed in the outlet side was negligible at all times. The gases used for the permeability measurements were 99.95 pct Matheson prepurified H2 and 99.7 pct D2 (0.3 pct HD) from the Isomet Corp. The gases were dried with liquid N2 before use. In Fig. 1 the logarithms of observed flows, jt, for a pressure difference of 73.4 cm Hg are plotted against T-1 Each point in Fig. 1 represents the slope determined by a least-square evaluation of a plot of the pressure in the outlet side of the membrane against time, including at least two separate sets of observations. Each point in the figure has an uncertainty of ±0.6 pct arising from the experimental uncertainty in the pressure and time measurements. The standard deviation of the observed pressures from the straight lines in all cases is less than the experimental precision of the measurements. The observed flow, jt, includes a small but not negligible flow through the stainless-steel connection.

Jan 1, 1968

By S. S. Lau

The authors of the previous comment stated that Dorn and Rajnak33 and others34-40 concluded that "agreement with this function [(t'/tp) -f(Un/2Un) = T/Tc} may be used as a criterion for that (the Peierls) mechanism of deformation". In contrast, however, Dorn and Rajnak33 and others34-40 never made such a statement. In order to clarify the issues involved it is appropriate to recapitulate the conditions that must be satisfied in order to suggest that some deformation process might obey the Peierls mechanism: 1) Within permissible variations due to changes in the slope of the Peierls hill the T*/Tp - T/Tc relationship must be obeyed. Although some mechanisms, e.g.. solute atom stress field, recombination of dissociated partials in bcc metals, intersection of dissociated dislocations, and so forth, give about the same t* - T relationship, other mechanisms, such as cross slip, disruption of attractive junctions, climb, motion of jogged screw dislocations, and so forth, give distinctly different t*-T relationships. On this basis some distinction can be made of the various mechanisms. 2) An important feature of the Peierls mechanism concerns its physical origin based on the nucleation of pairs of kinks. The value of T* at 0°K increases with the density of dislocation (i.e. , cold work) for the intersection mechanism and it increases with the square root of the atomic fraction of solute atoms in the solute atom interaction mechanism. In contrast the value of t* at 0°k, for the Peierls mechanism which depends only on the line energy and the shape of the Peierls hills. is independent of cold working and the square root of the solute atom concentration. These differences have been employed by various investigators for elimination of a few other possible mechanisms. On the other hand it is not possible to eliminate all mechanisms, e.g., the recombination of dissociated dislocations in bcc metals, and so forth, on this basis alone. 3) Further discrimination between possible mechanisms is based on the activation volume. For the in-

Jan 1, 1968

By E. P. Whelan

A recent investigation1 into the origin of "ghost boundaries" in dilute Cd-Mg alloys has indicated the persistence of a segregated solidification substructure in spite of a homogenizing treatment designed to remove such segregation. Although no grain growth experiments have been reported on these segregated alloys, the homogenizing treatment used in grain growth studies on other dilute alloys2-7 is similar in

Jan 1, 1968

By A. J. Sedriks

ALTHOUGH the occurrence of a readily visible tarnish on the stress-corrosion fracture surfaces of titanium alloys tested in oxygenated nitrogen tetrox-ide solutions has been reported,' the possibility that the formation of this tarnish is a requirement for cracking to proceed does not appear to have been considered. It has been proposed that failure occurs in this system as a result of some electrochemical-mechanical process.2 However, recent studies3-6 of the stress-corrosion cracking of a brass in certain ammoniacal solutions, which also cause visible tarnishing, have led to the view that failure occurs by the repeated formation and fracture of a brittle surface film of cuprous oxide. The purpose of this note is to point out the similarities which exist between the tarnish which forms on a brass and that found on titanium alloys failed in nitrogen tetroxide, and to suggest, by analogy, that a similar tarnish-rupture mechanism may be operative in the latter case. The titanium alloy selected for the present study contained 5.2 pct Al, 2.5 pct Sn, 0.025 pct C, 0.16 pct Fe, 0.014 pct N, 0.012 pct H, 0.07 pct 0, and 0.002 pct Mn by weight. Vacuum-annealed alloy strips measuring 4 by 0.25 by 0.04 in. were used to prepare self-stressed test specimens. Each specimen was made from two strips by bending the end portions and spot welding them together to produce curved sections of uniform bending stress. (The exact details of preparing such specimens have been described by Lisagor el al.7) The curvature in the strips determines the ma-magnitude of the outer fiber stress, which in the present specimens was 100,000 psi (yield stress = 120,000 psi). The specimen surface in the spot-welded end regions was shot-peened to induce compressive stresses and thereby eliminate the possibility of stress-corrosion cracking in these regions. The test environment was nitrogen tetroxide, initially containing 99.5 pct N2O4, 0.1 pct H2O, and 0.08 pct C1, through which oxygen had been bubbled for 30 min. During testing the tetroxide was subject to an oxygen pressure of 250 psi. The duration of the test was 96 hr, and during this time stress-corrosion cracks propagated through some 50 pct of the thickness of the specimens. After testing, the tarnish was liberated from the fracture surface by immersion in a methanol-5 pct bromine solution, and then examined by electron dif- fraction and transmission electron microscopy using a JEM-7 electron microscope. The possibility that the bromine solution had affected the structure of the tarnish was investigated by an "in silu" examination of the stress-corrosion fracture surface using a reflection electron diffraction attachment designed for handling bulk specimens. Typical areas of the tarnish liberated by the me-thanol-bromine solution are shown in Fig. 1. Fig. l(a) reveals that the tarnish consists of individual platelets, and from Fig. l(b) it can be seen that the platelets are joined together to form a semicontinuous film, the boundaries of the titanium substrate grains being retained in the tarnish. Electron-diffraction patterns taken from these areas exhibited diffraction spots typical of a monocrystal. The diffraction pattern shown in Fig. 2 was taken from the central area in Fig. l(n). Monocrystal patterns obtained from areas containing numerous individual platelets can only be explained by assuming that their growth is governed by the titanium substrate (i.e., their growth is epitaxial). In other areas, the tarnish had folded over or became mechanically damaged, and these areas gave diffraction patterns consisting of "spotted" rings, enabling interplanar spacings of the tarnish lattice to be determined. These spacings, measured using a

Jan 1, 1968

By Y. C. Liu, G. A. Alers, S. S. Choi

In searching for an explanation for the obvious difference between the copper and the brass type of rolling texture, it is common practice to look for correlations with other properties. For example, there are obvious differences between copper and brass in the critical resolved shear stress, the extent of easy glide or stage I, and the amount of "overshooting)) during single-crystal deformation.' Also, the slip lines in copper are fine and wavy while those in brass are more coarse and better described as slip bands.' Since neutron irradiation is known to make these properties in copper change in such a way as to resemble brass,2'3 an excellent test for these possible correlations is to see whether neutron irradiation produces a corresponding change in the texture, i.e., whether irradiated copper develops a brass-type rolling texture. The experiment would also answer the more important question of whether or not lattice defects like those introduced by neutron irradiation have any effect on the development of rolling textures in fcc metals. Experimental studies using conventional X-ray pole figure techniques are inconvenient because the X-ray detection system must be modified to discriminate against the radioactivity which is induced in the specimen by the necessarily long exposure to neutrons. Nevertheless, some texture determinations have been

Jan 1, 1968

By L. H. Cadoff, L. M. Adelsberg

The layer growth method has been used to investigate the reaction kinetics of liquid titanium and hafnium with carbon and the [liq] - [liq + TiC] and [liq] - [liq + HfC] phase boundaries at temperatures above 1800°C. The equipment and experimental procedures used in this investigation are basically identical to those described in an earlier paper' on the Zr-C reaction. Temperature control, however, was improved significantly by connecting a Latronics Colo-Ratio automatic pyrometer and a controller to the power source. The subsequent temperature variations were ±10°C; previously ±20°C was obtained using manual controls. The absolute accuracy of the pyrometer, however, must be considered to be only ±40°C.* The spectrographic analyses of the materials used in this study are listed in Table I. The isothermal reactions of titanium (1800° to 2700°C) and hafnium (2300" to 3000°C) with carbon (graphite crucibles) resulted in the formation of carbon-saturated liquid metal and layers of Tic and HfC, respectively. The thicknesses of the carbide layers, determined from metallographic measurements at 100 times, as a function of carburization time and temperature are given in Figs. 1 and 2. At each temperature, parabolic growth behavior for these reactions was observed and could be expressed as x2 =Kpt, where X is scale thickness (corrected for heating and coolingL), Kp the parabolic growth constant, and t the isothermal carburization time. Values of Kp as a function of temperature for the carbides are shown in Fig. 3 and listed in Table 11. The growth constants are expressible as: kpTiC = 0.2 exp(-61,800/RT) sq cm per sec for Tic and Kp HfC = 2.62 exp(-95,300/RT) sq cm per sec for HfC. The liquidus phase boundaries were determined from metal samples treated in graphite crucibles for different times ranging from 4 to 100 min at each of the temperatures indicated in Table 11. Metallographic examination revealed that the carbide precipitates were distributed uniformly throughout the frozen metal matrices. This is attributed to the rapid cooling rates (800°C per min) and the small difference in density between the precipitated carbide and the liquid metal.

Jan 1, 1968

By H. A. Davies

DURING a current program of research on the structure and properties of liquid sodium alloys, the need arose for information on the solubilities as a function of temperature of several B subgroup metals in liquid sodium. In the absence of reliable information in the literature for several of these systems, par- ticularly at temperatures in the range 100° to 300°C, an apparatus was constructed to determine solubilities by a direct melt-sampling technique, followed by wet-chemical analysis for both sodium and the solute metal. The details of the technique have been described previously.' The solubility of indium in liquid sodium was determined at eight temperatures in the range 100° to 270°C. The solubility as a function of temperature is shown on Fig. 1; this represents the liquidus line over this composition range. A cooling curve was plotted after the completion of the sampling and an arrest was observed at 94.9°C which must be interpreted as being the result of a eutectic reaction. By extrapolation of the liquidus to this temperature, the eutectic composition was estimated to be about 0.25 at. pct In. The present estimate of the eutectic temperature and composition are in good agreement with the values of 96.0° to 96.2°C and between 0.28 and 0.45 at. pct In, respectively, which were reported by Heycock and Neville.2 However, no further information on the liquidus of this system at the sodium-rich end could be found in the literature for comparison. One compound: NaIn, has been reported in this sys-

Jan 1, 1968