Search Documents

By R. Frenkel, O. D. Sherby, J. E. Dorn, J. Nadeau

A method is shown for the study of the creep rate dependence of metals on the applied stress under the condition of constant structure. The method was applied to pure aluminum and to dilute solid solution alloys of magnesium, copper, germanium, zinc, and silver in aluminum. A linear relationship was found to exist between the true stress and the logarithm of the creep rate. ALTHOUGH extensive creep data are now avail-Aable, the fundamentals of the creep phenomena are as yet only vaguely understood. Many attempts have been made to formulate theories for creep. Eut their agreement with fact has been disappointing. It appears that merely pyramiding more of the same type of data that is now available will not provide the essential aid toward formulating a better understanding of the complicated process of creep. Perhaps attempts to construct better models for the mechanism of creep might have to be tentatively abandoned until new types of definitive creep data are uncovered. The question therefore arises regarding what type of data might be needed to provide the essential knowledge for formulating better and more realistic theories of creep. Perhaps one source of the failure to uncover an adequate theory arises from the fact that practically all theories for creep and nearly all analyses of experimental data on creep disregard the well-known fact that the structure of metals changes during creep. For example, it is customary in evaluating the effect of stress on the creep rate to correlate the secondary creep rate with the applied stress. But, as has been demonstrated recently,' the structures obtained during secondary creep tend to exhibit deformation banding at high stresses and polygoniza- tion at the lower stresses. Consequently the stress-secondary creep rate relationships found in this way are not fundamental, inasmuch as they reveal not only the effect of stress but also the effect of differences in structure on the secondary creep rates. It is the purpose of the study reported here to isolate the effect of stress alone on the creep rate in an attempt to provide new data for possible formulation of a better theory of creep. In addition an attempt will also be made to ascertain how structure, temperature, and alloying might affect the stress-creep rate relationship. The procedure that was adopted was simple: A specimen was precrept at a given engineering stress U, (initial engineering stress during creep under constant load, namely the load divided by the initial area of the specimen) and temperature T to a selected engineering strain, e (engineering strain, namely, the instantaneous gage length minus the initial gage length divided by the initial gage length), at which time the true stress was reduced to some lower value of the true stress, ^ (the true instantaneous stress, namely, the load divided by the instantaneous area). The instantaneous true creep rate, i, (the true strain rate; with E, the true strain, namely, the natural logarithm of the instantaneous over the initial gage length) following reduction of the true stress to U, was then determined. Second, third, etc., tests were conducted under identical conditions except the stress was reduced to yet lower values, us, u,, yielding yet lower instantaneous creep rates, ;, and i,. Inasmuch as the precreep conditions were identical in each series of tests, the instantaneous structures obtained immediately after reducing the stress were presumed to be identical.

Jan 1, 1955

By W. R. McMillan, E. A. Gulbransen, K. F. Andrew

Annealed and electrolytically polished pure iron was oxidized between 650° and 850°C at oxygen pressures of 0.1 to 2 microns Hg. Electron optical studies showed that oxidation occurs discontinuously over the surface and orients crystallographically with the substructure of the metal grains. IN previous papers', " the authors described the formation of oxide films on high purity iron at norrnal pressures and at temperatures below 300°C. Electron optical studies showed that the oxide film consisted of many individual oxide crystallites having nearly random orientation on each metal grain. The size of the oxide crystallites depended upon the temperature and upon the extent of oxidation. In these studies the crystallite size varied from 250 to 1800Å. Recently, Bardolle and Benard3 studied the crystal habit of the oxide crystallites formed on Armco iron between 650" and 850°C in what might be defined as a "poor vacuum" atmosphere in which the pressure could be varied between 10-1 to 10-3 mm of Hg. Their results showed that, at 850°C and a vacuum of l0-2 to 10-3 mm of Hg, a few well oriented nuclei of oxide were formed on the metal grains. At pressures between l0-1 to 10-2 mm of Hg, many oriented nuclei of oxide were formed. At 750°C the metal surface was bright although an oxide was present. Unfortunately, the extent of oxidation, the oxygen pressure, and the influence of carbon in the metal were not determined, and it was, therefore, impossible to give a theoretical interpretation to the very beautiful light micrographs. Bardolle and Benard used the light microscope to determine the crystal habit of the oxide and X-ray diffraction to determine the orientation relationships. It was believed that the use of controlled amounts of oxidation at high temperature is an important method for the study of the initial stages of oxidation of metals at high temperatures. This paper presents the first results of an electron optical study of this problem for iron. Since the electron microscope and electron-diffraction methods using stripped oxide films and surface replicas yielded finer details of the crystal habit and structure of oxide film, these methods were used in this study. Two types of iron having different carbon contents were used: Puron and Armco* iron. A com- parison of the results for these materials made possible a study of the influence of the surface oxide-carbon reaction on the initial stages of oxidation. Before discussing the experimental results, the thermodynamic equilibria of 1—the oxidation reaction, 2—the surface oxide-carbon reaction, and 3—the solubility of oxygen in iron, will be considered. In addition, it is important to consider the kinetic-theory predictions on the rate of collision of oxygen with the iron surface. Thermodynamic and Kinetic-Theory Calculations Only one type of reaction is usually considered in the oxidation of iron at normal pressures at high temperatures: Fe(a) + 1/2O2(g) = FeO(s) 3Fe(a) + 2O2(g) = Fe3O4(s) 2Fe(a) +3/2O2(g) = a-Fe2O3(s). [11 However, two other reactions may occur: FeO(s) + C (solid solution in Fe) + Fe(a) + CO(g) [2] and Fe(a) + O2(g) e Fe(a) (solid solution of 0,). [ 3 ] In addition, iron will react with the higher oxide at certain temperatures to form FeO. For a complete interpretation of the oxidation of iron at low pres-

Jan 1, 1955

By L. E. Murr, P. J. Smith

A study has been made by transmission electron microscopy of thin foils of 304 stainless steel fatigued external to the electron microscope in reversed bending, and of thin foils fatigued directly within the microscope in alternating tension. The build-up of stacking faults in the thin foils during fatigue zoas correlated with the dislocatirm structures found in thin films prepared fror fatigued bulk specimens. The performance of the special devices designed for fatigue of thin foils so outlined and the importance of alternative methods of preparation of more uniform fatigue specimens bY vapor deposition are emphasized. INVESTIGATIONS on fatigued bulk aluminum1,' and stainless stee13j4 have revealed the existence of dislocation substructure on examination by transmission electron microscopy of electrolytically thinned foils representative of these bulk specimens. While this technique has proved extremely valuable, it has several shortcomings. First, the method of fatiguing bulk specimens and then thinning to foil for electron-transmission observation allows only one observation of internal structure at any one portion of the fatigue life. Second, thinning the bulk fatigued specimens to foil results in the loss of at least one original surface. Thus, what one sees in the remaining electron-transparent sections is an internal dislocation or fault structure which in many cases cannot be correlated with the original surface markings. This is an undesirable feature, since it is well-known that metal surfaces play an important part in the fatigue process.5 An obvious third undesirable feature of the thinning-from-bulk technique is the fact that static observations have difficulty (in the case of fatigue) explaining the mechanism of a dynamic process. What is required then is a method whereby a selected thin area can be continuously observed either while undergoing cyclic deformation or at various fixed stages of fatigue deformation. While Murr and wilkov6 have reported some success with an apparatus designed to fatigue thin metal foils directly within the electron microscope, the nature of the specimen-mounting procedure and the mechanical features involved in the adaptation of the fatigue device to the electron microscope make this method difficult to operate. It was not possible, for example, to make frequent observations of a selected area because of difficulty in maintaining a chosen area in a stable viewing position inside the electron microscope. Except for the build-up of a dislocation substructure and what are commonly referred to as "slip striations", little else has been reported from observations on thin-foil sections prepared from bulk fatigue specimens. segal17 has found "slip striations" in stainless-steel fatigued specimens which were electropolished from both sides, but gave no explanation as to their origin or identity in terms of lattice imperfections. The research to be reported in this paper was undertaken with the following objectives in mind. First, an attempt was made to devise a technique or techniques whereby thin metal foils could be fatigued and repeatedly observed by transmission electron microscopy. Second, it was hoped that some correlation could be made between deformation striations found in fatigued thin transmission specimens and thin foils prepared from bulk fatigued specimens. These investigations illustrate quite convincingly that a feasible method is available for the direct study of the fatigue mechanism and similar dynamic phenomena in thin transmission specimens inside the electron microscope. I) EXPERIMENTAL METHODS Three modes of specimen fatigue and observation of fatigue damage were used. These involved fatigue of thin-foil specimens in a special arrangement external to the electron microscope and observation of a selected area at various stages in the fatigue life, the fatigue of thin-foil transmission specimens inside the electron microscope, and the fatigue of bulk specimens external to the electron microscope followed by thin-foil preparation for direct observation at predetermined stages. Design of the External Fatigue Clip and Specimen Holder for the Hitachi H.U.11 Electron Micro-scope. The purpose of the fatigue clip was to provide a reversed bending fatigue stress to a thin-foil transmission specimen outside the electron microscope. In order to accomplish this, a flat-bottomed, U-shaped brass clip was made as shown

Jan 1, 1965

By C. S. Barrett

THE crystal imperfections known as faults in stacking (stacking disorder) are of importance to both fundamental and applied science and are receiving increasing attention. On the theoretical side there have been recent advances in the analysis of diffraction patterns in terms of stacking faults, and on the experimental side it has been found that certain metals develop a faulted structure on cooling through a phase transformation. Beyond this, little is known of what may be called the physical and mechanical metallurgy of faulting, although there has been no lack of discussion and conjecture as to the presence of faults in metals and the effect of faults on mechanical properties and microstructures. In the close-packed metals the nature of faults can be described very simply. Face-centered cubic (FCC) crystals consist of close-packed layers of atoms, (111) planes, stacked above each other in a definite manner. If one layer is designated as the A layer, the next must be a B or a C layer; that is, every atom of the second layer lies not above the atoms of the A layer, but nestles in the hollows between these, using the hollows designated as the B positions, or the equivalent hollows designated as C. If the sequence of layers is ABCABC...., so that layers three spacings apart are directly over each other, the crystal is FCC; if CBACBA ... it is an FCC crystal that is a twin of the former; if ABAB........or BCBC..it is hexagonal close-packed (HCP). A fault is a break in the sequence, such as ABCABABC .. . which has the effect of inserting an HCP lamella into an FCC crystal. Regularly spaced faults spaced two layers apart would convert an FCC crystal entirely to HCP, or an HCP to FCC. A fault in an FCC crystal is also equivalent to creating an FCC twin of minimum thickness; in fact, the entire crystal would be converted to its twin if faults were introduced regularly. (In HCP crystals, however, twinning does not occur on the basal plane and a fault on this plane is not identical with a thin twin). Faults are one of the possible causes for annealing twins,' for the strain markings that are brought out by etching deformed metal,'," for the altered X ray reflecting power of metal at slip bands that is revealed in X ray reflection micrographs,' for the extinction of Kikuchi lines in electron diffraction patterns, and for the hardening of latent slip planes.' Since other explanations of these effects are also available," there has been no direct proof of faults from deformation; yet it has appeared likely that they would occur. Faults resulting from phase transformations in metals, however, have been clearly revealed by diffraction in cobalt" and lithium. It is to be expected that the faulting tendency is greatest

Jan 1, 1951

By R. H. Tien

Temperature distribution as well as position of the solidified front is solved by means of "heat balance integral", for the case of freezing a slab with time-dependent surface temperature. Numerical solutions for the cases of linear variation and exponential decay for surface temperature are given. Application also includes solidification of casting. An increased interest in the heat-transfer problem involving a change of phase has been developed in recent years. Practical problems include solidification of a casting, freezing and thawing of soil, ice formation, and others. Despite the importance of the topic, a literature search reveals relatively few solutions which may be extended to practical problems. Due to mathematical complexity, which comes from the nonlinearity of governing differential equations, exact analytical solutions which have been published are limited to a few cases with simple boundary conditions. Although numerical solutions of these problems have become available after highspeed digital computers were introduced, lack of simple and general methods of programming for computation makes this approach impractical. The case of the semi-infinite slab with constant surface temperature was solved in the 19th century. In honor of Neumann, who first found an analytical solution, this kind of problem is generally referred to as the Neumann solution. As has been shown in Fig. 1, the problem to be analyzed is the freezing of a semi-infinite region (x > 0) which is initially liquid at the melting temperature with the surface (x = 0) maintained at time-dependent temperature. This analysis can be applied to the study of many casting processes where most of the heat removal is essentially unidirectional. "Heat balance integral" which has been used by Goodman1 is employed in this analysis. The concept of "heat balance integral" is to define a quantity, d(t), called the thermal layer. Beyond this thermal layer, for all practical purposes, there is no heat transfer, and the region is at the equilibrium temperature. "Heat balance integral" is obtained by integrating the conduction equation with respect to the space variable over this thermal layer d(t). Choosing some plausible temperature distribution, usually a polynomial or a finite trigonometric series in the space variable, this integral equation becomes an ordinary differential equation, with time as the independent variable, which can be solved by classical methods. The thermal layer and "heat balance integral" are equivalent, respectively, to the boundary-layer thickness and the momentum integral in hydrodynamics. A modern account of Prandtl's concept of the boundary layer and von Karman's momentum integral may be found in Ref. 2. Although the problem considered here has been limited initially to uniform melting temperature, the present method is equally applicable to the case with arbitrary temperature distribution in the liquid. STATEMENT OF THE PROBLEM The physical situation and coordinate system used in the problem to be analyzed are shown in Fig. 1. Notations used are listed in the table of nomenclature. Let us assume a semi-infinite region extending over positive x, initially at the uniform melting temperature. At the surface x = 0, a time-dependent temperature, which has a numerical value

Jan 1, 1965

By C. S. Roberts, S. L. Couling

PLASTIC strain in polycrystalline metal as a result of bulk movement of one grain with respect to another along grain boundaries is not new. Rosenhain and Humphrey observed such effects shortly after the turn of the century.' Since then, observations of grain boundary deformation have been made on several metals. In recent years, intensive studies of the deformation at grain boundaries have been performed on polycrystalline metal and bicrystals.14-20 Aluminum or its alloys have been used in these latter studies. The subjects of controversy at present appear to be not whether grain boundary deformation exists, but 1) how quantitatively important it can be in polycrystalline metal, and 2) what its detailed mechanism is. Some investigators have concluded that such de- formation is minor in importance while others have concluded that it is a major consideration. Some have concluded that it involves shear of one grain over another at the interface while others have concluded that it operates by crystallographic deformation within the grains near ,the boundary. A categorical statement for electrolytic magnesium cannot be made on either of -these points. When polycrystalline magnesium is deformed in creep under conditions of high strain rate and low temperature, grain boundary deformation is relatively slight." Under these conditions, much deformation occurs within the grains and boundary contours change to a higher-energy jagged form. The result is shown in Fig. 1, where boundary jaggedness is pronounced upon polishing and etching after a plastic strain of approximately 6 .pct at 300°F and 3000 psi. No migration of the boundaries has been detected in this process. When the same metal is deformed the same amount at 600°F and 300 psi, resulting in a much lower strain rate, the boundaries migrate toward a lower energy configuration, where

Jan 1, 1958

By S. Yukawa, M. J. Sinnott

A high resolution autoradiographic study of the diffusion of a nickel isotope (NP3) into copper in the temperature range of 650° to 925°C, with particular emphasis on grain boundary diffusion, has been made. The extent of grain boundary diffusion is a function of the grain boundary angle and the diffusion temperature. The ratios of the grain boundary diffusion coefficient to the lattice diffusion coefficient ranges from 104 to 105. The activation energy for grain boundary diffusion decreases with increasing grain boundary angle. STUDIES and reviews&apos;-8 of metallic diffusion have indicated that grain boundary diffusion is a common phenomenon but quantitative studies of the many variables involved are not numerous. The present work was concerned with determining the influence of the crystallographic orientation of the grain boundaries and the temperature on the extent of boundary diffusion of nickel into copper. The study involved essentially three steps: 1—the production of grain boundaries of known crystallographic orientation, 2—the determination of the extent of diffusion into the boundaries, and 3—the analysis of the resulting data in terms of the usual diffusion concepts. The grain boundaries investigated approximated the simple tilt variety having l° of freedom. They were produced by controlling the solidification from a melt of Cuprovac (a high purity vacuum-melted copper obtained from Vacuum Metals Corp.) of a pair of seed crystals so that the bicrystals of copper produced had a common <100> axis. By rotating one seed crystal relative to the other, a simple tilt boundary was produced. The angle of rotation between the two crystals is the grain boundary angle 8, or what is sometimes termed the misfit angle. In all, some 16 boundaries ranging from 2" to 72" in 5" to 10" steps were investigated. The crystallographic orientations were measured with the use of both X-ray and optical goniometric techniques. The <100> axes in the crystals were found to be parallel within 2" to 3" with the exception of one crystal (8 = 12") in which the axes were 6" from being parallel. All bicrystals showed evidences of sub-grain structures. Laue analyses showed that these subgrains differed in orientation up to a maximum of 2". Because of these variations, and some curvature in the boundaries, the grain boundary angles listed are actually average values and deviate from the 1 o of freedom boundaries. The stripping film technique of high resolution autoradiography was used to detect the diffusion of the nickel isotope into the grain boundaries of the bicrystals of copper. Kodak experimental permeable stripping film with an emulsion layer 5 u thick

Jan 1, 1956

By G. R. Kotler, W. A. Tiller, H. V. Ashcom, S. O&apos, Hara, W. C. Johnston

The grain-refinement effect of electromagnetic stirring of Sn-Pb alloys during solidification is investigated at higher field strength and in larger-sized samples. It is found that the grain-refining effect is due to a crystal multiplication phenomenon rather than a nucleation phenomenon. The effect increases with sample size for a fixed field strength and a fixed supercooling bath temperatures. The effect in other alloys was investigated. IN a recent publication1 it was reported that significant grain refinement could be produced during the solidification of a supercooled melt of Sn-Pb alloy when a critical degree of electromagnetic stirring was present in the melt. Thus, for a particular supercooling, AT, a critical magnetic field, Hc, existed at which the number of grains per unit volume, n, increased rather abruptly by a factor of 103 to 104 as the field, H, was increased by a small amount. It was further found that the undercooling at which the melt spontaneously nucleated, ?Tn, was independent of the magnetic field. These authors matched their experimental results against the qualitative expectations of existing theories and found each theory somewhat unsatisfactory. One important question that was not answered in this previous work was "should the effect be classed as a nucleation phenomenon or as a crystal-multiplication phenomenon?" Other valid questions that need answering are: i) do higher magnetic field strengths alter the nature of the results?; ii) what is the effect of higher lead content in the alloy?; iii) what changes occur when one extends the experiments to larger-diameter samples?; and iv) is this a universal effect applicable to the grain refinement of all materials? The present work directs its attention to answering these five questions. The

Jan 1, 1965

By M. Cohen, D. Caplan

The scaling characteristics of three Fe-Cr alloys have been investigated by determining their weight gain vs. time curves at 1600° to 2000° F. The scales formed thereby have been examined using the techniques of X-ray diffraction and spectrographic and metal-lographic analyses in an attempt to explain the discontinuities in the curves and to elucidate the mechanism of scaling. DESPITE the considerable number of investigations that have been carried out on heat resistant alloys, the characteristics of the scales formed at high temperatures are not fully known. The research reported here was undertaken in an attempt to ascertain the mechanism of scaling of the stainless steels. Scaling experiments were carried out first, the weight increase of the specimens being followed continuously with time. It was observed that, as well as showing the expected decrease in oxidation rate with time, the oxidation curves showed breaks corresponding to intermediate periods of accelerated oxidation, after which protectiveness again increased. This phenomenon was observed with austenitic stainless steels (types 302, 309, and 330) and with Fe-Cr alloys (types 410, 430, and 446), but only the latter are treated in this report. An examination of the scales was made using the techniques of X-ray diffraction and spectrographic and metallographic analyses in an attempt to obtain a correlation between the nature of the scales and the oxidation curves. A search through the literature revealed only a very few previous reports of such periods of accelerated oxidation. Dunn&apos; found breaks in the oxidation-time curves of some Cu-Si alloys but saw no rational explanation of the phenomenon. Heindlhofer and Larsen2 attributed a discontinuity in the weight gain-time curve of iron at 1290°F to the formation of blisters, the subsequent cracking of which exposed an unprotected surface and permitted rapid oxidation until a new protective scale had been reestablished. They advanced no explanation, however, for what they termed the peculiar behavior of a 27 pct Fe-Cr alloy at 2000°F which gained weight very rapidly in between two periods of very slow weight gain. Portevin, Pretet, and Jolivet3 in describing breaks in the weight gain-time curves of Fe-A1 alloys suggested that they might be associated with the occurrence of localized and deeply oxidized areas on the specimens. Bandel4 in a general discussion of oxidation curves of heat resistant alloys considered that the discontinuities were due to a local disruption of the protective layer by the growth of iron-rich oxides. Day and Smith" in their report on the scaling of a large number of iron alloys noted but did not explain occasional relatively rapid changes in oxidation rate at higher temperatures. Chevenard and Wache6 found breaks, often two per specimen, in the oxidation curves of an 18-8 type alloy. They suggested that the cause might be a depletion in chromium of the surface layer of metal due to its selective oxidation, the resultant high concentration of iron and nickel in the scale leading to a poorly protective scale. McCullough, Fontana, and Beck&apos; explained the breaks in the oxidation curves of types 304, 430, and 410 alloys as due to mechanical ruptures. Experimental Work Table I lists the chemical compositions of the materials used. Cylindrical specimens 1/4 in. in diam and 11/2 in. long were machined from cold rolled % in. rod. After a fine finish cut with a sharp tool, the specimens were abraded while still mounted on the lathe with Nos. 2, 1, 0, and 00 metallographic grade emery papers. A 3/64 in. hole was drilled at a distance of 1/8 in. from one end to permit suspension in the furnace. Specimen Nos. 1, 2, and 3 were tested with this surface preparation. All others, after being similarly prepared, were electropolished in a perchloric-acetic electrolyte, electrical contact being made by pressing a tapered platinum hook into the drilled hole. The specimens were then washed in hot water, rinsed with distilled water, rinsed with methanol, dried at 120°F, and weighed. Thereafter,

Jan 1, 1953

By Leonard B. Gulbransen, John R. Lewis, W. Martin Fassell, J. Hugh Hamilton

A simple reproducible method was developed for determining the ignition temperatures of magnesium and magnesium alloys and by this method magnesium and over 100 magnesium alloys were measured. The ignition temperature of magnesium was determined in O2-SO2, O2-N2 mixtures and in O2 from 0.166 to 10 atm pressure. The ignition temperature of magnesium is generally lowered by alloying and increased by an increase in oxygen pressure. WITH the expanded use of magnesium alloys in industry, ignition temperatures are of considerable importance, especially in heat treating and for service at elevated temperatures. Magnesium alloys exhibit a relatively slow linear oxidation rate up to temperatures near the melting point.&apos; At some critical temperature the rate of oxidation becomes extremely rapid. This high rate of oxidation is accompanied by the emission of light and additional characteristics suggesting a flame. The literature concerning the ignition temperatures of magnesium and magnesium alloys is relatively incomplete and the values reported by various authors are not in agreement. The most probable reasons for this are: 1—The absence of a suitable definition of ignition, and 2—the need of a standardized method of determining the ignition temperatures of inflammable metals and alloys. The exact date of the discovery of the fact that magnesium will ignite and burn is unknown. It is likely that H. Davey encountered this property of the metal during its preparation in 1808. The first reference in the literature specifically on ignition of magnesium is by Lenze, Metz, and Rubens.&apos; They investigated the inflammability of certain magnesium alloys. Brown3-&apos; reported that magnesium ribbon will ignite in air at 507 °C by what was called the rising temperature method. Hartman, Nagy, and Brown" reported that magnesium powder ignites from 475" to 560°C depending on particle size. According to Guise, Mars, and Wilson prolonged heating in air at 427°C of common casting magnesium alloy causes it to ignite. Samples of magnesium alloys were placed directly in the flame of a torch by Carapella and Shaw.&apos; These investigators obtained some evidence of melting prior to ignition. For commercial magnesium, they reported an ignition temperature of 650°C. Their values varied from 450" to 800°C for magnesium alloys. Willmore and Peterson" studied the effect of air velocity, humidity, rate of heating, and foreign metal contact on the ignition temperature of magnesium. They conclude that melting is not a prerequisite for ignition and that the conditions which influence ignition are complex and difficult to analyze. The only quantitative theoretical treatment of the ignition temperatures of magnesium and magnesium

Jan 1, 1952

By P. J. Soltis

Direct as well as indirect evidence of a-phase ordering was found in a "near-a" type titanium alloy, Ti-8Al-1Mo-1V. The annealed alloy showed an anomalous decrease in second-stage creep rate at 950°F (510°C) and 47,500 psi. Anomalous increases were noted in initial yield stress after 500 hr creep exposure and in tensile strength after 1100 hr creep exposure. Debye-Scherrer X-ray diffraction patterns obtained on fine-grained polycrystalline speci-ment showed much line broadening and development of superlattice lines after 500 hr at 510° C. Examination with the light and electron microscopes failed to reveal phase changes in thermally exposed material, but detailed metallography is presented illustrating the plastic and fracture characteristics of the alloy. Electron-metallographic evidence is presented, however, which shows the nature of the nucleation and growth of ordered regions from the disordered a in annealed material. In the original phase diagram of the binary Ti-Al system constructed by Ogden et al.1 aluminum was shown to be soluble in a titanium over a wide temperature range at concentrations up to about 30 pct. Subsequently, a number of papers2-7 have appeared which indicate the existence of either intermetallic compounds and/or ordered phases in the a field up to 25 pct Al. In the most recent of these, Goldak and Gordon paw7 have determined unequivocally the ordered hexagonal D019 structure for Ti3Al in a 25 pct Al alloy. No superlattice lines were observed above 1112°F (600°C). Previously, the embrittle-ment encountered by Crossley and Carew4 in low- temperature heat treatments of binary Ti-Al alloys low in aluminum had been attributed to the appearance of a new phase in light micrographs. In the present work, direct as well as indirect evidence is presented showing the existence of a-phase ordering in what is essentially the binary Ti-8A1 alloy containing small percentages of the ß stabilizers molybdenum and vanadium, i.e., the "near-a" Ti-8Al-1Mo-1V alloy. Although Bohanek and Kessler8 have shown that this alloy has good stability at temperatures up to 1100°F (593°C), the present work shows instability at 950°F (510°C) where long exposure times were used. Light and electron microscopy proved very useful in revealing the mechanical behavior of ordered a and led to the conviction that ordered "near-a" alloys may have remarkably good properties and useful engineering applications.

Jan 1, 1965

By R. M. Waterstrat

The phases occurring in the V-Mn system were studied by means of X-yay diffraction and metallo-paphic techniques, using are-melted alloy specimens annealed in the temperature range 800° to 1150°C and quenched. The bcc solid solution extends at 1250°C all the way from vanadium to 6-manganese. Below 1050°C the a-phase is formed, and the terminal a-manganese phase is stabilized up to about 900°C by vanadium in solid solution. IN the only previous general survey of the V-Mn system Cornelius, Bungardt and Schiedtl reported the existence of three intermediate phases corresponding to the approximate compositions VMn,, VMn, and V5Mn. The phase VMn8 has recently been identified as a o phase2 but the alloy VMn was found to have a bcc structure2 corresponding apparently to the vanadium solid solution rather than to the large cubic unit cell reported by Cornelius et al. 1 Subsequent work by Rostoker and Yamamoto3 has shown that the vanadium-base bcc solid solution extends to at least 15 pct Mn at 900°C. An alloy corresponding to the composition VMn, was examined by Elliott,4 who reported that the as-cast sample as well as samples annealed at 1200o and 1300°C had bcc structures, but that annealing at 1000°, 800") and 600°C produced two phases. One of these phases was apparently the bcc solid solution and the other resembled the o phase structure. Hellawell and Hume-Rothery5 established the phase relationships in manganese-rich alloys above 1000°C, and showed that the o phase in this system is replaced by the 6 Mn (bcc) solid solution at temperatures above 1050°C. These results suggest that a continuous bcc solid solution may exist above 1050°C between vanadium and 6 Mn. The present investigation was undertaken in order to develop more complete information in regard to this system. EXPERIMENTAL METHODS The alloys used in the present work were prepared by arc-melting electrolytic manganese having a minimum purity of 99.9 pct and vanadium lumps with a purity of 99.7 pct. The major impurities present in these metals were carbon, nitrogen, and oxygen and this would account for the small percentage of nonmetallic inclusions observed metal-lographically. The arc-melting was at first performed under a helium atmosphere and it was necessary to keep the melting times as short as possible in order to minimize the loss of manganese by vaporization. It was later found that the evaporation of manganese was considerably reduced when the melting was done under argon atmosphere. The final composition of each alloy was calculated by assuming that the total weight loss during melting was due to evaporation of manganese. Compositions which were calculated in this manner agreed reasonably well with the results of chemical analysis, as shown in Table I. Spectrographic analysis revealed the presence of contamination by tungsten, but in no case was the percentage of tungsten greater then 0.4 at. pct. The specimens were in each case broken in half and the fractured section was examined visually and microscopically for evidence of inhomogeneity. Each specimen was homogenized at temperatures near l100°C, as shown in Table I. After this treatment most specimens consisted of large columnar grains of the bcc vanadium solid solution. The etchant used in most of the metallographic work consisted of 20 pct nitric acid, 20 pct hydro-flouric acid, and 60 pct glycerine. It was found that this etchant would clearly delineate the phases present in these alloys although it does not produce any striking contrast between the phases. For certain manganese-rich alloys, a 1 pct aqueous solution of nitric acid was used. This etchant gave a brown color to the a-manganese phase, whereas the o phase was virtually unattacked and appeared very light as shown in Fig. 1. The etchants used by Cornelius et a1.l were found to produce spurious effects in some of these alloys. In particular, the vanadium-rich alloys etched in hot sulfuric acid often appeared to consist of two phases when both X-ray diffraction and etching with the glycerine-acid mixture indicated the presence of single phase bcc solid solution. A few percent of what appears to be an oxide or nitride phase was found at the grain boundaries and in the interior of the grains, especially in the vanadium-rich alloys. All alloys were annealed in sealed silica tubes containing 1 atm of pure argon and these tubes were then quenched in cold water. Although some manganese loss occurred during annealing, the loss seemed to be confined to the surface of the speci-

Jan 1, 1962

By G. Bhat, J. F. Libsch

lsoembrittlement curves depicting the influence of time and temperature in the range 800&apos; to 1260°F (425&apos; to 680°C) on the development of embrittlement in a commercial chromium alloy steel and a commercial molybdenum alloy steel are presented. Two distinct regions of embrittlement occur in the chromium alloy steel: I—at 800&apos; to 1000°F (425&apos; to 540°C) and 2—in the region just below the lower critical temperature. Embrittlement is most pronounced at 800&apos; to 1000°F, decreasing very rapidly with increasing temperature above this region, only to increase again as the lower critical temperature is approached. The data suggest two distinct modes of embrittlement with possible superposition of the two modes at extended embrittling times in the temperature range 1100° to 1150°F (590&apos; to 620°C). While the molybdenum alloy steel shows little susceptibility to embrittlement at 800&apos; to 1000°F (425&apos; to 540°C), considerable embrittlement may occur just below the lower critical temperature. THE subject of temper embrittlement in alloy steels has received considerable attention in the last few years. Points of view on the mechanism of embrittlement differ, however, resulting in part from the incompleteness of the data developed and in part from the speculation regarding the susceptibility of plain carbon steel to temper embrittlement. Libsch, Powers, and Bhat1 carried out short-time embrittling treatments on an AISI 1050 steel and demonstrated that hardened plain carbon steels are quite susceptible to embrittlement when tempered in the range from 850°F (455°C) to the lower critical temperature. The isoembrittlement diagram,&apos; representing the embrittling characteristics of this steel, is reproduced in Fig. 1. It is evident from the shape of the curves shown that embrittlement in plain carbon steel increases progressively with both temperature and time in the embrittling range. A comparison of the isoembrittlement diagram for AISI 1050 steel with that presented by Jaffe and Buffum&apos; for an SAE 3140 steel shows that up to 930°F (500°C) the isoembrittlement characteristics of the plain carbon steel are similar to those of SAE 3140 steel, although the embrittlement is much more severe in the latter steel. Above 930°F (500°C), the rate of embrittlement in the plain carbon steel increases continuously with increasing temperature; whereas, in the SAE 3140 steel, the embrittlement rapidly decreases. The influence of alloying elements upon embrittlement during tempering thus appears to cause a decrease in embrittlement above the region of maximum embrittlement, i.e., 850" to 1000°F. The question naturally arises as to what effect individual alloying elements have upon the embrittling characteristics of the plain carbon steel. Current knowledge on the influence of alloying elements on temper brittleness may be found in the review papers of Hollomon" and Woodfine. Hollo-mon," from the results of other investigators, has shown that, in general, the amount of embrittlement increases with increasing alloy content (except for molybdenum and possibly tungsten and columbium). Jaffe and Buffum," by a comparison of the embrittlement in a plain carbon steel with that of a SAE 3140 steel postulated that the presence of alloying elements in moderate amounts tends to retard the development of temper brittleness. It is difficult to determine what effect chromium has upon temper brittleness, since most of the information available has been based on the combined effect of other elements with chromium, particularly nickel and manganese. However, Wilten, and recently Jolivet and Vidal,&apos; Vida1, and Woodfine have reported that chromium steels are temper brittle, that the embrittlement is reversible with a maximum rate of embrittlement at approximately 975°F (525"C)," and that the susceptibility increases with increasing amounts of chromium. Taber, Thorlin, and Wallacel" have found a large embrittling effect with increasing chromium content in a medium C-Mn-Ni steel. But Hultgren and Chang," from their experiments conducted on synthetically prepared ternary Fe-C-Cr alloys, could not conclude that these alloys are susceptible to temper embrittlement. However, on addition of manganese or phosphorus, these Fe-C-Cr alloys became susceptible, from which fact they concluded that the embrittlement developed in chromium-bearing Fe-C alloys is due chiefly to the presence of these elements. Considerable data are available to show that molybdenum decreases the susceptibility of steel to temper embrittlement. However, its effectiveness in preventing or decreasing embrittlement appears limited to its presence in small amounts. Vidal" has shown that a plain 2 pct Mo steel was susceptible. Hultgren and Chang" also have shown that molybdenum additions in excess of 2 pct to synthetically prepared Ni-Cr steels did not prevent embrittlement. Jolivet and Vidal&apos; and Lea and Arnold found that molybdenum reduced temper brittleness. Lea and Arnold further stated that molybdenum decreased the rate of embrittlement rather than the total amount of embrittlement, whereas Preece and Carter" have shown that the presence of molybdenum greatly reduces the equilibrium extent of the change at a given temperature but does not appear to influence the rate of embrittlement. There appears to be very little information as to how molybdenum by itself affects the temper brittleness susceptibility of a plain carbon steel.

Jan 1, 1956

By R. Ruka, E. A. Gulbransen

ONE of the interesting questions in the understanding of the reaction of iron with oxygen is the kinetics and the mechanism of the crystal structure changes occurring in the formation and breakdown of the oxide film. In a recent work&apos; the electron diffraction method was applied to the study of the solid phase reactions occurring internally in the oxide film on iron above 570°C and under vacuum conditions. The higher oxides which form under oxidizing conditions are reduced internally by the diffusion of iron into the oxide leaving the bulk of the oxide FeO (wüstite). This paper will ask and consider three questions. These are: (1) What is the nature of the depressed transformation temperature for the Fe3O4 + Fe ? 4Fe0 reaction in very thin oxide films?; (2) What is the mechanism of the forward transformation of Fe3O4 to FeO at or near 570°C?; and (3) What is the nature and mechanism of the slow reverse transformation of FeO to Fe3O4 and Fe below 570°C?. The composition of the oxide as well as its electrical, magnetic, mechanical, and chemical properties depends upon the kinetics of these reactions. Literature Survey The composition of the oxide film on pure iron at temperatures up to 570°C has been studied extensively by the electron diffraction method. Nelson,&apos; Jackson and Quarrell,3 and Gulbransen and Hick-man4 have shown that Fe3O4 and a-Fe2O8 are formed in the temperature range of 225" to 570°C, while y-Fe,O, is most likely formed below 225°C for long oxidation times. In very thin films FeO (wüstite) is found to form at temperatures as low as 400°C, and its existence depends essentially upon the extent of initial oxidation. FeO is not found in the formation of thick films below 570°C. On the other hand X-ray5 and micrographic8 studies show that Fe3O4 may exist as the main com-ponent in thick films up to 650°C, well above the equilibrium temperature for the reaction Fe3O4 + Fe ? 4FeO. The kinetics of the forward and reverse trans-formations of the reaction Fe3O4 + Fe ? 4FeO have not been studied in oxide films. However, analysis of the data of Gulbransen and Hickman4 on the heat-ing and cooling of iron oxide films of known thick-nesses in high vacua shows that the forward reac-tion proceeds rapidly even for a 4000 A film at temperatures of 575" to 600°C. The reverse reaction or decomposition of wüstite (FeO) occurs at tem-peratures below 450°C. Thus, there appears to be a slow process in the decomposition which is in marked contrast to the rapid forward reaction. The decomposition of a bulk sample of FeO free from the metal has been studied extensively by Chaudron, Benard and coworkers using the micro-graphic, X-ray diffraction, and thermomagnetic methods. Chaudron7,8 first observed that the de-composition of FeO (wüstite) below 570°C occurs by the reaction 4FeO ? Fe3O4 + Fe. Later work by Chaudron and Forestier5 showed that the rate of de-

Jan 1, 1951

By L. S. Darken, H. A. Wriedt

In a previous paper,1 an experimental study of nitrogen dissolved in a cold-rolled ad heat-treated, low-carbon steel at 300° to 450°C yielded the equilibrium solubility relations and the concentrations of the two types of deject sites involved. In this paper, the probable identities of the defects that absorb nitrogen are restricted from the available evidence to screw dislocatioms for B sites, and to either micro crack surfaces or edge In a previous paper.&apos; we reported an experimental study of the excess nitrogen solubility in a low-carbon steel produced by a combination of plastic deformation and nonrecrystallizing heat treatment. The concentrations of the defect sites that absorb the excess nitrogen were measured in terms of their capacities for nitrogen. Also measured were the equilibria of nitrogen distribution among the sites. Several problems and questions arise from these measurements. We should like a physical identifi- dislocations for A sites. The concentrations of these defects in experimental steel specimens are estimated from the concentrations of defect sites. finally, the influence of chemical composition upon the nitrogen-trapping defects—the relative numbers generated in deformation, the relative numbers surviving as temperature is raised, the apparent numbers observed, and the distribution equilibria-is discussed. cation of the defects involved and of the two kinds of nitrogen-absorbing sites, called A and B types, upon those defects; the concentrations of the defects themselves could then be computed. We should also like to know how exclusively the findings apply to the "X steel" used in the experiments and to what extent they apply to cold-rolled steel given no subsequent treatment. By considering the data already given, together with results from a few specific tests which will be reported as they are invoked, we are able to eliminate some interpretations altogether and to secure an acceptable though limited interpretation from among the remaining possibilities.

Jan 1, 1965

By L. Zwell, H. A. Wriedt

X-ray Lattice parameter and bulk density measurements were made at room temperature on solid solutions of nitrogen in purified a, iron, containing from 0 to 0.053 wt pct N. When nitrogen is dissolved in a, iron, the fractional increase in unit cell volume (X-ray diffraction) and the fractional increase in volume per gram of iron (bulk density measurement) are the same within experimental error. The two sets of results when averaged show that the unit cell edge of a, iron increases linearly by 3.2 X 10-2 A per wt pct N dissolved and that the partial gram-atomic volume of nitrogen at infinite dilution in a, iron is the same as it is in nitrogen-martensite, that is, 5.90 cm3. RECENT reviews17&apos; of literature dealing with the lattice parameter of ferritic iron-nitrogen alloys indicate that the dilating effect of dissolved nitrogen on the bcc unit cell of a iron is not known with confidence. Five investigations3-7 either revealed no effect of nitrogen or were inconclusive. Among other investigations, where quantitative data were obtained, the results of Osawa and Iwaizumi8 indicate too large an effect, and those of Eisenhut and Kaupp9 show the lattice parameter of ferrite varying with the nitrogen content at concentrations far beyond the known solubility limit. It is not clear in Burns&apos;10 data what value should be used for the lattice parameter of nitrogen-free ferrite in calculating the dilation. Burdese11 shares with Burns the shortcoming that the iron-nitrogen alloys were solution-treated and quenched with undissolved nitride present. The consequence is that the un- certainties in the solubility limit1 are added to those in the lattice parameter measurements when the variation of lattice parameter with nitrogen content is computed. For a system of such widespread interest, we felt that a new investigation was merited. Both X-ray diffraction and bulk density measurements were used to determine the dilation of the , a iron lattice by dissolved nitrogen. EXPERIMENTAL PROCEDURE Material. The bulk density and X-ray diffraction experiments were both made with specimens of a purified iron (denoted as "B iron")- from Battelle Memorial Institute. The impurity contents were as follows: As, C, Cu and Pb—50 ppm; Ag, Al, Co, Ga, 0, Ti-11 to 50 ppm; B, Be, Ca, Cd, Cr, Mg, Mn, Mo, N, Ni, S, Sb, Si, Sn, V, W-10 ppm or less. The vacuum-cast ingot was hot-forged and annealed. Pieces of the forged stock were machined and cold-rolled to give strip stock 0.08 and 0.002 in. thick for density and for X-ray measurements, respectively. Preparation of Specimens. Starting with a piece of the cold-rolled iron stock, the general sequence of operations, after an initial recrystallizing anneal, was to measure the density or lattice parameter, nitrogenize, remeasure the property, age or denitrogenize, and again measure the property. Duplicate analyses for nitrogen were made by a modified Kjeldahl method on portions of the actual density specimens and on pieces nitrogenized to equilibrium alongside the X-ray specimens. Annealing of the density specimens was performed by sealing the pieces with the as-rolled surfaces removed by filing into evacuated "Vycor" capsules and treating for 1 hr at 750°C followed by furnace cooling. (TO check the effect, if any, of annealing procedure, a specimen also was

Jan 1, 1962

By I. A. Lesk

The various materials utilized in the construction of integrated circuits, and the resultant materials interfaces, are discussed with emphasis on a materials system that is compatible with all types of integrated circuits and minimizes the number of different interfaces. Problems with interfaces may originate in the silicon, on its surface, and between other materials within the integrated-circuit package independent of the silicon. ThE difference between integrated-circuit theory and practice may generally be spelled out in terms of the difficulties in fabricating a desired structure. There are, of course, problems in forming the requisite materials, of the necessary composition and geometry, where needed. A distinct class of problem, however, is associated with interfaces— generally between different materials, but occasionally between regions of the same material. We require that each interfacial system perform its proper electrical, mechanical, and thermal function, and do so in an unchanging manner with time under the most severe operating conditions. In a practical sense, we deal in nonideal systems, such that it is generally not possible to form an interface exactly the way it was intended, even if we allow for a knowledge of the practical limitations of the fabrication system. Another way of looking at this situation is to realize that, when we form a group of interfaces under apparently the same experimental conditions, there are often differences of important consequence between them. Very high-magnification pictures, such as are obtained by an electron microscope, reveal fine structure that often differs over the area of a given sample, as well as from sample to sample. Low-magnification pictures look better, but don&apos;t eliminate the problem. The trend towards integrated circuits, and away from circuits fabricated utilizing separate components, is very strong in many areas of electronics. A typical silicon monolithic integrated circuit, still in place in the wafer on which it was fabricated, is shown in Fig. 1. The increased reliability which is being experienced through the use of integrated circuits can, to a large extent, be credited to the reduced complexity of the interface system that is present in the electronic system. A glance at Fig. 1, however, with all of its epitaxial and diffused p-n junctions, and interconnect system, shows that interfaces have been far from eliminated. Fig. 2, depicting a schematic cross-sectional view of a (partial) typical integrated circuit, illustrates in detail the numerous interfaces that occur in this type of structure. Besides the individual circuit elements, there are interfaces for isolation, and, Fig. 3, even a crossover requires a series of interfaces. What has been accomplished, however, irrespective of whether or not the total number of interfaces in the circuit is increased or decreased, is that the variety of interfaces is very much reduced. This is because there are fewer materials involved. Since the expense incurred in an R & D program to improve reliability is pretty much directly proportional to the variety of interfaces involved, it is economically advisable to work with integrated circuits for highest reliability. That is not to say that circuits built using standard components could not be made as reliable as integrated circuits, but that the expense involved because of the more complex over-all interface system would be much greater. Although electronic performance utilizing integrated circuits with all silicon elements (i.e., transistors, diodes, resistors, capacitors) is adequate for many applications, there are numerous instances where thin-film passive elements must

Jan 1, 1965

By J. P. Nielsen

When two grains in a polycrystalline specimen meet at a point in the course of grain-boundary movements, and the new boundary created at the point is one of relatively low specific free energy, a nonequilibrium boundary condition occurs. The nonequilibrium is enhanced if the other grain boundaries involved at the point of meeting are relatively high in specific free energy. The nonequilibrium results in a correction by growth of the complex grain (two subgrains) to a large size, sufficient in many cases for it to become a self-propagating unit, i.e., a recrystallization grain. This mechanism, herein called "geometrical coalescence," is proposed as a logical origin for recrystallization nuclei, particularly for secondary recrystallization. RECENT publications1-&apos; indicate that considerable progress has been made toward a complete understanding of the three microstructural processes in metals: recovery, recrystallization, and grain growth. However, a major problem persists, namely, the origin of recrystallization nuclei. It is generally accepted that once a recrystallization nucleus, a secondary recrystallization nucleus particularly, has reached a certain size, it will continue to grow at the expense of its neighbors, simply as a consequence of grain-boundary free-energy considerations. What has remained obscure, however, is the mechanism whereby such a nucleus comes into being and grows to the self-propagating size. Recrystallization Nucleation in a Two-Dimensional Grain-Boundary System In the left portion of Fig. 1 is shown schematically the meeting of grains 1 and 4 on disappearance of boundary 2-3, i.e., the boundary that existed between grains 2 and 3, as might happen in the course of normal grain-boundary migration. If, by chance —and the probability is not necessarily small— grains 1 and 4 are close together in their mutual orientations, or in some other special way mutually oriented so that the new common boundary has relatively a very low specific free energy (hereinafter referred to by s), then the grain boundary arrangement as depicted in the figure is highly unstable. This is particularly so if boundaries 1-3, 3-4, 4-2, and 2-1 are high in their s values. Assuming, for the sake of argument, that boundary 1-4 has a s equal to zero, and the s&apos;s of all the other boundaries present are equal to each other and constant, then the boundary arrangement on the right in Fig. 1 will be

Jan 1, 1955

By W. A. Wood, W. H. Reimann, Maria Ronay

The basic mechanism of fatigue is studied in annealed a brass subjectecl to alternating torsion at room temperature, 100°, 200°, 300°, and 400°C, and in air. It is shown that the slip-zone micro-cracking which characterizes fatigue damage produced by small amplitudes at room temperature is progressively replaced by grain boundary cracking at elevated temperatures, replacement being complete at 400°C. Replacement occurs not because slip activity decreases hut because slip movements at elevated temperatures cease to concentrate in narrow zones and instead disperse. Decrease of amplitude at 400°C through permitting longer lives of specimens before fracture actually causes in- MOST studies of fatigue at elevated temperature have been concerned with data for design. Observations on the basic mechanism of fatigue have been largely incidental, though they have led to the interesting inference that this mechanism has features in common with that of creep.&apos; This paper is concerned primarily with the mechanism itself. Advantage has been taken of mater- crease in pain boundary damage, an anomaly attributed to the greater difficulty at elevated tempwature of starting and propagating a crack. Surface corrosion was most pronounced along slip bands but since at elevated temperatures the slip hands showed no cracking the surface corrosion did not appear to influence onset and early spread of fatigue damage. Similarly the type of fatigue damage produced at room temperature by cyclic strain at large amplitudes, characterized by irregular mi-crocracks at zones between disoriented regions in a grain, was also replaced by mainly grain boundary cracking at elevated temperature, though not so completely. ials and procedures known to show readily how fatigue affects the microstructure of a metal at room temperature, so that direct comparison might be made with what the same procedures show when the material is subjected to fatigue deformation at higher temperatures. 1) MATERIAL This was 70/30 a brass, used extensively in previous studies at room temperature.&apos; Specimens were turned from extruded rod to have parallel test portions 1-1/4 in. long and 7/32 in. diameter, annealed to a grain size -1/10 mm, and electro-

Jan 1, 1965

By B. Chalmers

THE practical importance of the phenomena of melting and freezing must have been recognized for a very long time. The difference between ice and water, for example, has had a profound influence on the history of mankind and the evolution of society. The possibility of melting a metal and allowing it to freeze in a mold of chosen shape has been an essential ingredient in our mastery of the art of shaping metals, and therefore in the evolution of the: machine age in which we find ourselves. The importance of melting and freezing, as applied to metals and alloys, has been so great, in fact, that empirical solutions have been found for the multitude of practical problems that have arisen. This approach has been so successful that relatively little attention has been directed to arriving at an understanding of the fundamentals of the processes. But metallurgy has come to a stage at which we may expect that some, at least, of the more complex problems that have not yet been solved (or perhaps even recognized) may be handled more effectively by scientific study, theoretical understanding, and logical experimentation than by trial and error. In this lecture, therefore, I propose to describe in outline what I think really happens when a metal freezes. In doing so I hope to explain many of the phenomena which have been observed, and in particular to account for the structures that are obtained in actual ingots and castings. The basic problem, to which this lecture represents a tentative partial answer, is this: a mass of metal, containing known proportions of various elements, is melted, heated to a given temperature, and then allowed to freeze under specified conditions. What will be the "structure" of the resulting metal? The term structure includes: 1—crystal size, shape, and orientations, 2—distribution of chemical elements, and 3-—shape, including cracks, cavities, pores, etc. The Solid-Liquid Interface We will first consider what takes place if a single crystal of a metal in the form of a rod is heated, not uniformly, but so that one end is hotter than the other. If this heating process is continued long enough, the hotter end will eventually melt; we will suppose that the rod is in a containing vessel so that the molten metal does not run away, Fig. 1. When some of the metal has melted, we have some solid, some liquid, and an interface or surface of contact between them. If the source of heat is now removed, the interface will move so that some of the liquid freezes, and if the supply of heat is suitably adjusted the interface will remain at rest. This very simple arrangement allows us to study the basic processes of melting and freezing, and if we fully understand this simple case, we may be able to account for what takes place under practical conditions where the heat does not all flow in the same direction, and where the heat flow is determined not by a controllable source of heat but by the heat capacity and temperature of metal and mold, and by the heat loss from the mold surface. The solid-liquid interface is evidently the region of the greatest interest to us; on one side of it there is crystalline solid, and on the other, liquid. In the solid, each atom has a well defined position, around which it vibrates as a result of thermal agitation. It only leaves this position in the relatively rare event of a "diffusion jump." The liquid is much less systematically organized. The atoms are about as far from their neighbors as in the solid, but the arrangement is much less systematic and is continuously changing. The solid and the liquid are represented diagrammatically in Fig. 2. The average energy of the atoms in the liquid is greater than in the solid by an amount that corresponds to the latent heat of fusion, i.e., the amount of heat that has to be supplied to convert unit mass of solid into liquid at the same temperature. The Two Processes As has recently been shown by Jackson and Chalmers,3 many of the features of the processes of freezing and melting can be understood if it is assumed that a continuous and rapid interchange of atoms between solid and liquid always takes place at a solid-liquid interface." It is necessary to con- sider two distinct processes, that of melting, in which atoms leave the surface of the solid and become part of the liquid, and the converse process,

Jan 1, 1955