Search Documents

By J. B. Clark, A. J. McEvily, R. L. Snyder

The nonuniform distribution of precipitate particles has been recognized as a leading factor contributing to the relatively low fatigue resistance of aluminum alloys. The structure of many of these alloys is characterized by narrow precipitate-free zones adjacent to the grain boundaries. Alloys with such zones exhibit a tendency for brittle inter crystalline fracture. The interrelation between this type of structure and mechanical properties was investigated in an Al-10 wt pct Mg alloy. It was found that deformation during fatigue occurs preferentially along these zones and cracks initiate there. In Al-10wt pct Mg, the zones were found to be supersaturated even after extensive general precipitation and are due to the absence of proper precipitate nuclei in the region near the grain boundaries. Cold working the alloy prior to aging improves the mechanical properties by inducing precipitation within the zones and also by jogging of grain boundaries. The mode of fracture is changed from brittle inter crystalline to more ductile trans granular fracture. THE process of fatigue is highly structure sensitive, with the strength of the whole often dependent upon some localized discontinuity, either geometrical or metallurgical in nature. Much has been learned about the role of geometrical discontinuities, e.g., notches, in fatigue, but with the exception of the effects of inclusions or the shapes of carbides, relatively little is known about the specific effects of discontinuities in metallurgical structure such as nonuniform precipitation. In most age-hardening aluminum alloys, metallo-graphic studies have shown that the extent of precipitation adjacent to grain boundaries is much less than that which occurs in the interior of the grains. The width of these almost precipitate-free regions, which are sometimes called denuded zones, and the extent of solute depletion within them, are dependent upon the particular alloy and its aging treatment. It has been observed1 that these zones are relatively soft with the result that plastic deformation takes place preferentially within them. It has also been shown 2-4 that there exists a tendency for intercrys- talline cracking in fatigue when such zones are present. It is of interest to note that Broom et al.2,3 were able to reduce the incidence of this type of failure in an A1-4 wt pct Cu alloy by stretching the material 10 pct prior to aging. In the present study, the effects of precipitate-free regions on the fatigue properties of an A1-10 wt pct Mg alloy were studied in detail, and the effects of deformation prior to aging on the nature of the precipitation process as well as on fatigue properties were also investigated. MATERIAL AND PROCESSING An A1-10 wt pct Mg alloy was selected for this study, because it was known that well-defined precipitate-free regions along the grain boundaries are readily obtained in this alloy after aging at 200oC.5 The starting materials were 99.998 pct A1 and singly sublimed magnesium of about 99.9 pct purity. The aluminum was induction melted in a graphite crucible, and then the magnesium addition was immersed until dissolved. Chlorine gas was then bubbled through the molten alloy for 4 min to degas the melt, after which the melt was cast at a pouring temperature of 730" to 760°C into a cold, graphite-coated, tapered steel mold. Since A1-Mg alloys are difficult to homogenize,5 special care was taken to obtain a uniform composition. Two-in. cubes were cut from the ingot and heated at 446°C for 30 min. These cubes were then hot forged approximately 35 pct in each of the three cube directions and homogenized for 16 hr at 446°C. Sheet specimens were then obtained by pressing 40 pct and rolling 35 pct per pass with reheating between reduction steps to a final thickness of approximately 0.10 in. The sheet was then solution treated for 16 hr at 446°C and water quenched. The age hardening behavior of this material at 200°C was then determined, and the results are shown in Fig. 1. The age hardening of this alloy when subjected to cold work prior to aging is also shown in this figure. Preliminary work indicated that extensive deformation after quenching was required to affect drastically the precipitate-free regions in this alloy, and a rolling reduction of 50 pct was chosen. For purposes of comparison the following three conditions were studied: a) Solution treated, quenched, and aged 20 hr at 200°C

Jan 1, 1963

By Rollin E. Hook, Attwell M. Adair

The yield and fracture behavior of recrystallized chromiztm have been studied in order to gain an understanding of the recrystallization embrittlement-rheotropic recovery phenomenon. The duc tile-brittle tensile transition temperature of recrystallized chromium was found to vary from 225 ° to -80°C as the ratio of specimen diameter/mean planar grain size decreased. The applicability of this observation to other bcc metals is briefly discussed. Brittle fracture in recrystallized chromium was found to be initiated predominately from intercrystalline rupture nuclei. Single crystals exhibit a high degree of ductility at 23 "C. The experimental data indicate that grain boundaries play a significant part in the yield and fracture behavim. Pres training significantly reduces the transition temperature, and high temperature aging does not result in re-embrittle-ment until complete recovery or re crystallization occurs. SEVERAL theories have been used to explain the recrystallization embrittlement-rheotropic recovery phenomenon in chromium. Wain, Henderson, and ohnstone' propose a Cottrell dislocation locking mechanism due to nitrogen, and Weaver and Gross and weaver 3 propose a strain-induced precipitation mechanism involving a chromium nitride precipitate. The researches of many investigators seem to leave little doubt that brittleness in chromium is promoted by interstitial impurities. Smith and seybolt4 have shown that the ductile-brittle transition temperature is lowered as the interstitial content is reduced. weaver3 provides evidence for a 150°C reduction in the tensile transition temperature as the nitrogen content is reduced from 20 to 5 ppm. Maykuth and affee' have suggested that the amount and orientation of grain boundary impurities may be responsible for the brittle nature of recrystallized chromium. This suggestion was based on their findings that wrought iodide chromium possessing a fiberous grain structure is ductile in tension at room temperature while a recrystallized grain structure is completely brittle. Previous investigations have not studied this phenomenon sufficiently to determine its cause. Perhaps this has been largely due to previous experimental findings that brittle fracture observed in chromium has been almost entirely transgranular.' In addition McNeil and imb,' who investigated coarse-grained as-cast chromium, describe the point of initiation of fracture as cleavage, close to the (100) plane, near or at grain bound-daries, or at included particles or voids. However, Hook, Adair, and ipsitt found that fracture in coarse-grained recrystallized chromium is initiated by intergranular rupture and then propagates nearly entirely by cleavage. This result, together with their finding that room temperature ductility in bending is observed for single crystals and coarse-grained polycrystalline specimens for which the number of grains through the specimen diameter is small, suggests that grain boundaries may play an important role in the brittle behavior of chromium. The recrystallization embrittlement problem can be resolved into two questions. These are, whether the controlling mechanism is predominately one of crack initiation or crack propagation, and whether brittle fracture is initiated predominately by slip or twin induced intergranular rupture or by cleavage. The following research was designed to study the role of grain boundaries on the brittle fracture of chromium in an effort to resolve these questions. EXPERIMENTAL PROCEDURES, RESULTS, AND DISCUSSION General Procedure. All tensile specimens (except 0.04 in. diam wires) were prepared from 1/4 in. diam swaged rod fabricated from one ingot. Details of fabrication are given elewhere. The results of impurity analysis for the swaged rod and for the

Jan 1, 1963

By T. G. Lovering

Nonparametric statistical tests have practical application to many geological problems because, unlike the standard statistical tests, they do not require assumptions regarding the form of the population frequency distribution. Many of them do not even require that the variable be susceptible to numerical measurement. These tests are useful either 1) for geological data representing population distributions of unknown form, or otherwise not meeting the requirements for standard parametric tests; or 2) for sample attributes that cannot be measured but only ranked in order of relative magnitude. The application of two such tests will be illustrated. The Kolmogorov-Smimov test will be used to test for significant differences in copper distribution in samples of two rock types; the Spearman rank-correlation method will be used to determine whether there is a significant association between intensity of rock alteration and copper content of the rock. Nonparametric tests are, in general, quicker and easier to perform than the corresponding parametric tests, and they may be applied to a greater variety of data. They are not as powerful as the standard tests if used on data to which the latter are applicable, and computer programs utilizing them are not generally available. The use of standard or parametric statistical procedures for the interpretation of measured geological data is familiar to most mining engineers and explora- tion geologists. The computation of product moment correlation coefficients on sample assays for more-than one component is already standard procedure in some company operations, as is the use of Student's F test and Snedecor's F test for the comparison of means and variances of sample groups. The application of these three tests is limited by the prerequisites that the samples must be random, independent representatives of normally distributed populations and that the variables involved must be measurable. If such tests are applied to data which fall short of fulfilling these prerequisites, then the probability level attributed to the computed statistic is invalid and the conclusions based thereon may be erroneous. A number of nonparametric, or distribution free, statistical tests have been developed for use with data which are not amenable to standard parametric procedures. These tests have been widely used in the behavioral sciences such as psychology and sociology for some time, but their application to geologic problems is relatively new. Nonparametric tests may be grouped into two main categories: those which test for significant differences in a single variable by comparing either sample values with a standard value or representative values between two or more groups of samples; and those which test for significant association between pairs of variables in a single group of samples. The versatile chi square test, the Kolmogorov-Smirnov one and two sample tests, the Wilcox matched pairs signed ranks test, the median test, the Mann-Whitney U test and the Cochran Q test are examples of nonparametric tests with demonstrated or potential geological applications that belong in the first category. The contingency coefficient, Spearman rank correlation coefficient, Kendall rank correlation coefficient and the

Jan 1, 1963

By F. J. Lecznar

An investigation of the problem of separating two minerals by a moving inclined plane has revealed that only minerals with considerable difference in sliding friction coefficients separate readily; that it is possible to separate two minerals when only one of them has a rolling friction coefficient; that difference in geometrical shape of the minerals is important to separation which improves as grain size increases; and that the thickness of the stream falling on the belt influences the grade of separation - the thinner the stream, the smaller the interference between the grains of the minerals to be separated and the higher the degree of separation. The problem of separating two minerals with different friction coefficients and shapes by a moving inclined plane has been investigated. Friction separators can be used to separate two materials that differ in shape and in friction coefficient. 1,3 Gladkov has presented some theoretical aspects of using a moving plane to clean corn and podded seeds; this method, however, did not result in practical application. When considering the separation of minerals, only those with considerable difference in sliding friction coefficients can be taken into account. Furthermore, it is possible to separate two minerals when only one of them has a rolling friction coefficient. The difference in geometrical shape of two minerals is also important to separation. It has been found that the separation of two minerals improves as the grain size increases. For example, coal and slate over 2 in. in grain size separate much better than materials of smaller grain size. As Table I indicates, the slate grains smaller than 25 mm gradually lose their flat shape with decreases in size and are more rounded than larger grains, which are almost flat. The smallest grains of slate in first and second fractions slide down the belt since their friction coefficients decrease and approach the value of the rolling coefficient. The percentage of separation of the first fraction is, therefore, low. Thickness of the coal stream falling on the belt also influences the grade of separation; the thinner the stream, the smaller the interference between the grains of the two minerals to be separated and the higher the degree of separation.

Jan 1, 1963

By F. Garofalo

It has been shown by various investigators that during constant stress creep the dependence of minimum creep rate, 6,, on stress, o, is given by em = A onat low stress levels, md by 6, = A' exp [ß s] at high stress levels. In these relations A, n, A', and ß are constant at constant temperature. A single relation has been found which satisfies conditions for both high and low stresses and agrees well with experimental results. This relation is 2, = AN (sinh a s)n, where A" is a constant at constunt temperature and a = p/n. This relation also satisfies the linear relation, 6, = k s, found at temperatures near the melting point at low stresses. EXPERIMENTAL creep results have led to a number of empirical relationships between minimum creep rate, 6,, and the applied stress, s. Under conditions of constant stress it is generally found that at low-stress levels,'-' the dependence of minimum creep rate is given by 6, = Asn. At high-stress levels'74 the experimental results fit the relation, 6, = A' exp [po l. In these relations A, n, A' and ß are constant at constant temperature. A model based on climb-of-edge dislocations from a pile-up array leads to a relation similar to that found experimentally at low stresses.' On the other hand. theories based on chemical-reaction rate,5 nonconservatively moving dislocation jogs, and jog migration and climb-of-edge dislocations4 lead to a relation similar to that found experimentally at high stresses. No difference in mechanism between low and high stress levels has been clearly defined; it is questionable whether such a difference really exists. In any event, the major argument usually given in substantiation of a change in mechanism from low to high stresses is that no single relation exists for defining the stress dependence of the creep rate over wide ranges in stress. However, such a relation has been found and is of the form, dm = A" (sinh a s)n, where A" is a constant at con- stant temperature and a = ß/n. This relation agrees well with experimental results over wide ranges in stress and temperature for copper, aluminum, an A1-3.1 pct Mg alloy, and an austenitic stainless steel. STRESS DEPENDENCE OF MINIMUM CREEP RATE At low stress levels the minimum creep rate, gm, depends on the stress, 0, under conditions of constant stress creep through the relation em=Aon [1] The quantities A and n have been defined previously. In the range in which this relation applies, a linear dependence is found in a log em,-log o plot. Above the stress range of application of relation [I], the minimum creep rate increases much mar; rapidly than predicted by relation [I]. This behavior is shown in Fig. 1 for a series of creep tests on copper4 at various temperatures ranging from 673 o to 973°K. Relation [I] is satisfied at the lower stresses, although the results at 973°K are quite limited. In all cases the transition beyond the applicable range of relation [I] is a gradual one indicating no abrupt change in mechanism. For all test temperatures, values of n have been determined from the results in Fig. 1. These are reported in Table I under Alog tm/hlog o. At high stress levels the stress dependence of minimum creep rate under conditions of constant stress is given by im = A' exp [ß a] 2] All factors have been defined previously. In a log em-0 plot, relation [2] predicts a linear function. Experimentally, a pronounced deviation from the prediction of relation [2] is found at lower stresses. As the stress is lowered, the minimum creep rate decreases more rapidly than predicted by relation [2]. This shown for copper in Fig. 2. Again it is found that the transition from high to low stresses is gradual and no really sharp change is found. Values of p determined from the high stress results are given in Table I. The dependence of minimum creep rate on stress at constant temperature for all ranges in stress

Jan 1, 1963

By R. C. Gifkins, A. Gittins

A. Gittins and R. C. Gifkins (University of Mel-houvne)— Evidence from somewhat similar experiments to those described in this paper has led us to the conclusion that possibility 2) of the four listed on p. 319 is a description of what happens in grain boundary sliding. That is, sliding is limited by accommodation deformation, it being necessary to the Continuance of sliding that this accommodation should take place. An example of such accommodation, though probably not the most important, is fold formation at triple points. The evidence and arguments for this point of view are set out in Refs. 14 to 16. In particular, an experiment with a lead bicrys-tall5 seems to support point 2) rather than 3). In creep of this specimen it was found that at least 90 pct of the extension of a zone 0.3 mm wide at the boundary was supplied by sliding at the boundary. The sliding was localized in a region at the boundary very much smaller than 0.3 mm, in fact, in a region too narrow to resolve optically. In poly crystalline Specimens&apos; there appears to be a surface effect, as found by Rachinger," but it appears that this may not manifest itself at low strains. "Low strains" means <10 pct for a Pb-TI alloy, <2.5 pct for magnesium or "Magnox" A12 alloy, or <40 pct for a 0 brass. Up to this extension the sliding, measured by the mean vertical step at boundaries, is proportional to extension. We are able to say this with some confidence, since the error bars on our points are statistically derived. Our results would lead us to think that the error bars on other work, including that of Mullendore and Grant, are much too large at low strains to show this

Jan 1, 1963

By S. R. Titley

The recent emphasis placed upon the use of molybdenum as a geochemical indicator has stimulated considerable inquiry into the behavior of molybdenum in the zone of oxidation. This paper represents a summary of recent and continuing experimental work and calculations in an investigation of a portion of the system Mo-Fe-S-H20 and treats the stability of various iron and molybdenum compounds with respect to oxidation potential and pH. Using experimentally determined free energy values for ferrimolybdite, diagrams of the stability fields of this mineral have been calculated with respect to ilsemannite (?), molybdenite, pyrite and limonite. This theoretical approach is the one utilized by Garrels3&apos;4 in studies of the relationships of various minerals in the zone of oxidation. Use of this approach is particularly suited to the study of geochemistry in the zone of supergene processes because of the low temperatures, low pressures and other relatively easily determined and duplicated factors that exist in the near surface environment. Such a study of molybdenum, however, is complicated by several factors. One of these is the fact that the strong amphoteric behavior of molybdenum makes possible a wide variety of substances and complexes whose exact chemical natures are unknown. The second factor is the dearth of thermodynamic data for many of the simplest of the possible compounds and ions important to the study. For the purpose of this study, the problem of complexes has been approached through considering only the simpler possibilities realizing that the complexes and systems so treated may be gross oversimplifications. All the complexes considered, however, have been suggested in the chemical 2,7,9,12 or geologi-cal literature. EXPERIMENTAL WORK Studies of the behavior of molybdenum ions and ferrimolybdite have been carried out in aqueous sys- tems, and reactions have been followed by simultaneous measurement and recording of pH and oxidation potential. A major problem involved in this study was the long periods of time necessary for equilibrium to be established in some of the systems considered; in a few instances, these periods exceeded four weeks, that is, four weeks during which measured conditions changed continuously. Another problem, in some instances, was the indeterminate nature of the reactions that have taken place. Because of the tendency of molybdenum to complex, reactions which take place under some of the more extreme imposed conditions of oxidation and acidity or alkalinity are unknown. However, useful data for the present work were obtained from studies of the molybdate ion and ferrimolybdite between pH 2 and 8 and under only moderately oxidizing to nonoxidizing conditions. Instruments used in the experiments were a Hewlitt-Packard Model 412A VTVM, a modified Beckman Zeromatic pH Meter and two Varian model 11 recorders. pH was measured between glass and calomel electrodes and Eh was measured between calomel and platinum electrodes. Conditions of pH were varied with 1N H2SO4 and 1N NH4OH and 1N NaOH. Eh was varied with gaseous nitrogen, oxygen or aqueous H202. Ambient temperature was essentially uniform at 24ºC and was increased in the reaction vessel by the positioning of a magnetic stirrer whose temperature was sufficient to bring the temperature to 2S°C in the reaction. The temperatures of reaction did not vary more than 1/2ºfrom 25°C. Two reactions were investigated, the ferrimolybdite-limonite-water reaction and the reduction of molybdate ion. Studies of both reactions yielded useful data that were used in preparation of the Eh-pH diagrams. The traces of the reactions are shown on Fig. 1. Ferrimolybdite-Limonite- Water: Studies of this reaction involved analysis of the solution of ferrimolybdite and simultaneous formation of ferric hydroxide. The data are summarized on Fig. 1. In this reaction, equilibrium was attained at various pH&apos;s when coexistence of the two solid phases (ferrimolybdite and ferric hydroxide) was attained. At the proper pH, controlled by the amount of ferrimolybdite and thus the dissolved content of molybdate species, both phases were observed and reaction times to complete the transition were on the order of weeks. The reac-

Jan 1, 1963

By R. G. Davies

Above 500°C, where dislocation climb is rate controlling, it is observed that the activation energy for creep is independent of the apblied stress, although it varies from 62 kcal per mol at 15 pct A1 to 73 kcals per mol at 20 pctA1 due to the change in short-range ovdw with composition. Below 500°C, when the nonconservative motion of jogs upon dislocations is thought to be rate controlling, the activation energy is stress dependent. A sudden increase in the creep rate, for which no explanation is availab2e, occurs at the highest stress at all temperatures. CREEP has been the subject of numerous research publications during the past decade with most of the attention focused upon high temperature steady state creep, i.e., above 0.5 TM, where TM is the absolute melting temperature, as it is the most amenable to theory. Many of the proposed theories involve the climb of dislocations out of their slip plane, which allows them to circumvent obstacles or to annihilate dislocations of opposite sign on other slip planes.&apos; For pure metals it is observed that the activation energy for this creep is equal to the activation energy for self diffusion.&apos; The position is much more complicated in alloys as there are many possible microcreep mehanisms, as well as dislocation climb, which could be rate controlling. There has been little systematic investi- gation of alloys, but Weertman (lead and indium base alloys)3 and Lawley, Coll, and Cahn (Fe-A1 alloys)4 suggest a microcreep process as the controlling factor. A Cottrell-Jaswon atmosphere mechanism is proposed for the lead and indium alloys, and a mechanism involving crystallographic order for the Fe-A1 alloys. Lawley et al. showed that the formation of a super-lattice, either of the FeAl or Fe3A1 type. decreased the rate and increased the activation energy for creep. From the measurements of lattice parameters,= resistivity,&apos; magnetic properties,7&apos;8 internal friction: and dilatometry10 the ordered region is thought to commence at -19 at. pct Al. However, a recent investigation by the author indicates a value

Jan 1, 1963

By T. H. Alden

T. H. Alden (General Electric Research Laboratory)—This paper as well as earlier ones of Dr. Wood represent an important contribution to the experimental description of fatigue fracture. The mechanism of fracture proposed by the authors, however, is not established by this data nor supported by other data existing in the literature. Although taper section metallography provides a rather detailed picture of fatigue crack geometry, photographs so obtained must be interpreted with care. The narrow bands revealed by etching, frequently associated with surface notches, are labeled by the authors "fissures". Measurement shows, taking into account the 20 to 1 taper magnification, that the depth of these structures is at most 2 to 3 times the width. This distinction is important in the conception of a mechanism of crack formation. It is difficult, for example, to imagine a deep, narrow fissure arising from a "ratchet slip" model. A surface notch, on the other hand, may form easily by this mechanism. The notches observed in the present work are the subsurface evidence of the surface slip bands or striations in which fatigue cracks are known to originate.4-6 It is clear that an understanding of the structure of these slip bands is of key importance in understanding the mechanism of fracture. The evidence presented shows that these regions etch preferentially, possibly because they contain a high density of lattice defects, or as the authors state equivalently, because they are "abnormally distorted." However, it is not possible to conclude that the distortion consists of a high density of vacant lattice sites. The fact of a high total shear strain in itself does not assure a predominance of point defects as opposed to other defects, for example, dislocations. Other evidence in the literature which suggests unusual densities of point defects formed by fatigue7-&apos; refers not to the striations or fissures, but to the material between fissures (the "matrix"). If a choice must be made, the preferential etching would seem to be evidence for a high dislocation density, since dislocations are known to encourage chemical attack in copper;g no such effect is known for the case of point defects. A third alternative is that the slip bands are actually cracked, but that near its tip the crack is too narrow to be detected by the authors&apos; metal-lographic technique. In this case the rapid etching can be readily understood in terms of the increased chemical activity of surface atoms. Unless a vacancy mechanism is operative, the motion of dislocations to-and-fro on single slip planes will not lead to crack growth. Point defect or dislocation loop generation are the principal non-reversible effects predicted by this model. In any case, the nonuniform roughening of the surface in a slip band6 requires a flexibility of dislocation motion which is not a part of the to-and-fro fine slip idea. The same is probably true of crack growth by a shear mechanism. Either some dislocations must change their slip planes near the end of the band and return on different planes,&apos;0 or dislocations of opposite sign annihilate." The mechanism by which these processes occur in copper at room temperature or below is that of cross slip. Thus cross slip appears to be essential to fatigue crack growth.6&apos;10"12 The fact that a tensile stress opens the slip bands into broad cracks does not indicate the structure of the bands or the mechanism by which cracks form. The charactersitic concentration of slip into bands during fatigue shows a low resistance to shear strain in these regions. (This fact in itself may be inconsistent with a high concentration of vacancies.) The authors contend also that continuing shear produces an additional mechanical weakening so that the bands fracture easily (are pulled apart) under the influence of the superimposed tensile stress. It is equally possible that the only weakness is a weakness in shear, that the crack propagates by a shear mechanism, and that subsequently the tensile stress pulls the crack apart. Even the direct observation of bands opened by a tensile stress would not be conclusive since, as argued above, they may be fine cracks. The same argument applies to internal cracks, their existence in the presence of a tensile stress not indicating the mechanism of formation. Internal cracks originating in regions of heavy shear have also been seen following tensile deformation of OFHC copper,13 so that this mode of fracture is not unique to combined tensile and fatigue straining. The authors point out in their companion report14 that 90 pct of the cracks formed during pure tor-sional strain were within 8 deg of the normal to the specimen axis. If the tensile stress were an important factor in crack propagation, it is surprising that the cracks cluster about the plane in which the normal stress vanishes. Similarly, a study of zinc single crystals showed that for various orientations the life correlated well with the resolved shear stress on the basal plane,&apos;= and was not dependent on the normal stress across this plane. W. A. Wood and H. M. Bendler (Authors&apos; reply) -Dr. Alden&apos;s discussion emphasizes the essential point in the relation of slip band structure to

Jan 1, 1963

By J. J. Kramer, G. W. Wiener, K. Foster

The effects of the primary grain size and sheet thickness on the secondary growth rates of grains with (100) surface planes were studied in 3 pct Si-Fe sheets. This secondary grain growth was carried out under conditions in which the {100} plane had the lowest surface free energy. The value of the surface energy difference, between the (100) plane and the random matrix planes was found to he only about 3 pct of the grain boundary energy, indicating that most of the driving force for secondary growth is supplzed by grain boundary energy. A method for obtaining cube texture in 3 pct Si-Fe sheets by a secondary recrystallization process has been reported.&apos; It has been demonstrated that a difference in surface free energy between the (100) plane and other planes is responsible for this growth, where the (100) plane has the lowest surface energy. A semiquantitative relation has been derived for secondary grain growth in sheets where primary grain growth has proceeded until the grains have grown through the sheet and the average grain diameter in the plane of the sheet is from one to two times the sheet thickness. Then, the linear growth rate, G, for the secondary grains can be given by where M is the grain boundary mobility, is the grain boundary free energy, is the average primary grain radius, A is the average difference in surface free energy between the (100) plane and planes of other orientations, t is the sheet thickness, and C is a term first proposed by zener6 which acts as a negative driving force, the effect of inclusions in holding up grain boundary motion. Since the problem of grain growth being discussed is under conditions where the matrix grain size is at least equal to the sheet thickness, the radius r is, of course, a function of the sheet thickness. The term B/r is the grain boundary driving force normally associated with inclusion inhibited or texture inhibited secondary recrystallization. Here, however, the term is modified to account for the fact that growth occurs in thin sheets. The term is the additional driving force available because of the difference in surface free energy. The purpose of the present work has been to evaluate the driving force terms of the growth equation by studying growth rates of (100) secondary grains as a function of sheet thickness and primary grain size. From these data it has been possible to determine the relative values of grain boundary energy, surface energy difference, and the inclusion term, C. EXPERIMENTAL PROCEDURE A 3 pct Si-Fe alloy was melted and cast under a helium atmosphere using electrolytic iron and a pure commercial grade of silicon as raw materials. The alloy was hot and cold rolled to a thickness of 0.6 mm, and samples approximately 5 in. long by 1 1/4 in. wide were cut from the cold-rolled sheets. Four different primary grain sizes were obtained

Jan 1, 1963

By E. T. Turkdogan, P. Grieveson, J. F. Beisler

Experimental results are given for the rate of reduction of silica from silicate melts by paphite-saturated iron in the presence of carbon monoxide. It is shown that, when gas bubbles are present at the slag-metal interface, the interfacial chemical reaction is the rate-controlling process in the transfer of silicon from slag to metal. From the silicon-concentration profiles in the metal sample, the diffusivity of silicon in paphite-saturated iron is found to be (5.4 * 0.4) x 10 sq cm per sec at 1600°C. Some consideration is given to the transport of oxygen from the slug-metal interface to the walls of the gvaphite metal container as becoming the rate-controlling process in the absence of gas bubbles at the slag-metal interface. ATTEMPTS were made by Fulton and chipman1 to investigate the kinetics of reduction of silica from calcium alumino-silicate melts by graphite-saturated iron in an atmosphere of carbon monoxide. In these experiments with 400 to 500 g of metal and slag, both phases were stirred in such a manner that the interface was essentially undisturbed. Initially, the graphite-saturated iron contained about 15 pct Si and, after about 8 or 10 hr of reaction time at 1600°C, the silicon content of the metal in- creased by about 0.2 pct. In a written discussion on this paper of Fulton and Chipman, schuhmann2 pointed out the possibility that this observed slowness of the reduction of silica could be due to the transport of oxygen across the liquid metal boundary layer at the slag-metal interface. Although this explanation of Schuhmann is quite feasible for the experimental data of Fulton and Chipman on silica reduction, such a reasoning should not be extended to other rate data on slag-metal reactions unless sufficient evidence is available satisfying the boundary conditions for such a transport process. The hydrodynamics of stirred liquids are very complex and even for a simple method of stirring there is no exact mathematical solution which will predict the mass transfer coefficient accurately. The best available solution of this problem is that given by Lewis&apos; and later modified by mayers~ for a two-liquid system where liquids are stirred by counter-rotating stirrer blades without disturbing the interface. In such a system the mass transfer coefficient is estimated from an empirical relationship using the data on viscosity and density of the two liquids, rate of revolution, the length of stirrer blades, and the diffusivity of the element which is transferred from one phase to the other. It was found in these investigations&apos; that the estimated and observed mass transfer coefficients, in systems containing two immiscible aqueous solutions, agreed within *40 pct. Although this is not considered to be an unreasonable error for the study of industrial processes, uncertainties of this magnitude cannot be tolerated in any theoretical study of reaction kinetics. In view of the complexity of the

Jan 1, 1963

By R. H. Richman, II Conard G. P.

AS part of a study of deformation twinning in bcc crystals, solid solutions of beryllium in iron have been found to twin profusely when strained slowly at room temperature. This note reports some crystal-lographic aspects of the operative mechanical twin modes in a quenched Fe-5 wt pct Be alloy. After melting twice under a partial pressure of helium, the alloy was homogenized, cut into pieces 7 to 10 mm on an edge, solution treated at 1130°C,1 and quenched. Debye-Scherrer patterns confirmed

Jan 1, 1963

By R. E. Hook, H. J. Garrett, A. M. Adair

X-ray diffraction and metallographic techniques were employed to examine the impurity phases occurring in as-swaged and recrystallized chromium produced fvom high-purity iodide chromium crystals. Chromium oxides are found to form in a consistent manner associated with the temperature of recrystallization. Microscopically observable second phase particles of Cr203 are observed to form intergranulurly beginning at 1300T. Beginning at 1500°C. Cr30, is formed intergvanulurly, atzd at 1600°C and above Cr30, is also formed in association with migrating grain boundaries. Geometrical traces formed by particles of this high temperature oxide uithin the grains rez)eal the histmy of grain boundary motion. X-ray evidence is presented which suggests the existence of a high temperature form of Cr203. Other impurity phases in addition to the oxides of chromium are also present. THE partitioning Of impurity elements between grain boundary aid lattice sires, either as solutes in solution or as second phases, may exercise an important influence on the mechanical behavior of metals and alloy systems. This partitioning of impurities could have a direct bearing on the iecrys-

Jan 1, 1963

By R. G. Ward, A. E. El-Mehairy

A technique was developed to calculate the density of liquid metals from the profile of a weighed levitated drop obtained by emitted light photography and calibration. The density of liquid copper was determined over the temperature range 1370" to 2100°K and expressed with the equation: The molar volumes and thermal coefficients of expansion are calculated. Copper, upon melting, was found to expand 3.34 pct. A knowledge of the density of liquid metals is essential as it is a parameter appearing in most theories of the liquid state. The liquid density is also necessary for the calculation of the commercially important contraction occurring on solidification. Unfortunately good density values are rarely available and the present work was undertaken to establish a new method of density determination free from the limitations of the existing methods. For liquid metals with high melting points, the variation and the limited number of density values reported in the literature have been caused primarily by difficulties in technique, a matter which has been adequately discussed by Urbain.&apos; In all the methods used for measuring the density of the high-melting liquid metals there are sources of error, and corrections have to be made. For example, in the method of balanced columns, the formation of gas bubbles in one or both arms was a great source of error. In the pyknometer3 and immersed-sinker methods, corrections of certain volumes for thermal expansion to the operating temperature exceed 4 pct. In the maximum-bubble method5 a correction for the thermal expansion of the bubble-tube is used (about 1.6 pct); a second correction takes into account the change in the volume of liquid displaced by the bubble-tube (about 3.5 pct). Another method used is based on the shape of a drop of liquid, at rest on an inert refractory plaque, obtained photographically or radiographi-cally. The volume is calculated from the shape, which is sensitive to the surface tension and supporting material, by some intricate empirical relation. The drop contour changes with 90 deg rotation of the support.7 The application of all methods is limited by reaction between liquid metal and the apparatus. The successful stabilization of a mass of liquid metal during levitation melting8 afforded a technique for measuring the density of liquid metals out of contact with solid materials. The technique was developed by using copper as the density of liquid copper has already been studied extensively, thus allowing comparison with the new method. It consists of calculating the volume from the dimensioned profile of the levitated drop obtained by emitted-light photography and careful calibration. The molten egg-shaped drop spins around the magnetic axis of the coil (lined vertically) at a speed which seems to increase with increasing temperature. The method does not involve corrections and the experimental setup allows the use of vacuum or special atmospheres. The maximum operating temperature is limited by excessive evaporation of metal from the droplet. EXPERIMENTAL Levitating System. A stabilizing ring was used in conjunction with a 3-turn levitating coil energized by a 450-kc per sec, 10-kw Toccotron with a high current-low voltage output stabilized to f 1 pet.&apos; APyrex glass tube with a Pyrex Optical flat window inserted through the coil and ring allowed the use of special atmospheres during levitation melting, as shown in Fig. 1. Optical System. Fig. 1 shows the optical arrangement used to photograph the profiles of the drop, parallel and perpendicular to the vertical spinning axis, simultaneously on the same frame of a high-speed 35-mm film. The optical system was adjusted to avoid any distortion of the profiles and to ensure that both profiles are in focus. The length of path from object to objective was about 30 cm, and a magnification factor close to 1 was obtained with the single-lens reflex-camera system. Temperature Measurement and Control. In operation, the specimen temperature was constant to 5 C for a fixed power input to the levitating coil and fixed rate of gas flow through the tube. The temperature was measured by an automatic two-color pyrometer (Coloratio) checked at the melting

Jan 1, 1963

By J. J. Kramer, W. A. Tiller, G. F. Bolling

The temperatures of solid-liquid interfaces have been measured during the growth of Sn-Pb alloy crystals which exhibited a cellulur substructure. A simple descriptive theory was used in combination with this data and the other measurable growth variables to predict some features of the solute segregation such as the intercell excess concentration and the depth of cell grooves during freezing. The effective distribution coefficient at the tip of a cell, a, was predicted and experimentally found to vary in the range, 1 >a > k0. At the lower limit, the severe segregation that must arise between dendritic cells was substantiated by various observations such as the existence of a second Phase even for alloys of only 0.2 wt pct Pb. ThE cellular morphology of a solid-liquid interface is known to coincide with the formation of a zone of constitutional supercooling ahead of the interface during growth.&apos; The occurrence of various cell structures, (two-dimensional cells, regular cells, dendritic cells), has been quantitatively investigated,&apos;2,4 but the segregation of solute which arises as a consequence of this mode of growth has only been examined qualitatively. An indirect method of obtaining the solute concentration at the tip of a cell cap during solidification has been de-

Jan 1, 1963

By Edna A. Dancy, Gerhard J. Derge

Jan 1, 1963

By K. E. Merklin, F. D. DeVaney

In the past most of the research in pelletizing has as its aim the production of a physically strong pellet. Now that this problem is solved, more attention is being paid to uniformity in grade and chemical composition. Consideration has been given in recent years to a study of additives which will make pellets partially or fully fluxed. Blast furnace tests show fluxed pellets to have advantages over both fluxed sinter and acid pellets. Lower slag volumes are used more frequently in the blast furnace and the uniformity of the charge, both physically and chemically, is becoming increasingly important. A considerable improvement can be secured by having all of the flux uniformly distributed through the pellets in a calcined state. A fluxed pellet, therefore, should be a superior product. The use of an ore burden that contains an appropriate quantity of fluxing materials is not new to the blast furnace industry. For years the Minette ores of France and Germany, the Clinton ores of the Birmingham, Ala., areas, as well as others, which are low in iron but contain calcium carbonate in sufficient quantities to give them self-fluxing properties, have proven economical. The adaption of the sintering process to the preparation of ore burdens provided a means for artifically associating the fluxstone with the iron ore prior to charging to the blast furnace. In 1944 Torpl discussed the effects of using lime in sinter made in Sweden on the sinter&apos;s metallurgical performance in a blast furnace. Hessle2 reported in 1945 on the advantages of using lime in sinter to increase the sintering plants&apos; capacity and reduce fuel requirements. Robinson3 discussed the use of lime in Swedish sintering practice at the 1950 Blast Furnace, Coke Oven, and Raw Materials meeting. One of the suggested advantages was the combining of lime with silica to prevent the formation of a fayalite slag, which is difficult to reduce in the blast furnace. Nilson4 at the same meeting commented briefly on experimental work that had just started in the U. S. on the use of lime in sinter. The literature tends to indicate that most of this early work was directed towards improvement in sinter physical quality, and the fact that the feed to the blast furnace was partially fluxed was incidental. It was not long, however, before blast furnace advantages were being claimed. The matter was forcibly called to the attention of North American operators by reports of remarkable production records established by the Russians using fluxed sinter. The leading step was taken on this continent by the Steel Company of Canada in 1957; McMahon5 reported on the marked improvement in performance of an 18-ft blast furnace using fluxed sinter at its Hamilton works. Other North American operators quickly followed theStelco lead.

Jan 1, 1963

By H. H. Dreissen, F. J. Fontein

The operating principle of the classifying cyclone has been described often but some special aspects of the cyclone receive further clarification here. In addition, several uses of the classifiers are summarized and made clear by means of practical examples. In coal and ore washeries, the washing cyclone is widely used for the separation by specific gravity of sizes up to about 12 mm. This application has lately been extended to the separation by specific gravity of particles up to about 50 mm. This use makes special demands on the construction and wear resistance of the cyclone. The sieve bend, a development of the Dutch State Mines which is less well known than the hydrocy-clone, is dealt with at some length. The sieve bend is an apparatus designed for wet screening. Attention is given to the variables influencing the performance of the sieve bend. Hydrocyclones are now so widely applied that it would be difficult to find an industry not, directly or indirectly, using them or knowing how to use them since publications on this subject abound.1-3,5,9-11,13-15 However, because of the large number of variables influencing the operation of classifying cyclones, it is not entirely possible to calculate the results of this type of cyclone. Therefore, a number of lesser known variables will be discussed here along with some special applications of the classifying cyclone. FACTORS INFLUENCING CYCLONE CLASSIFICATION Shape of Particles: The shape of particles may cause a much coarser classification than the one expected, considering the sieve analysis and the cyclone dimensions chosen. This was clearly shown by feeding a mixture of spheres and disks in water into a cyclone. The spheres were 0.36 mm diam and the disks, 3 mm diam and 0.03 mm thick. These dimensions were chosen in order that the disks and spheres might have the same rates of fall under gravity. Both spheres and disks had a specific gravity of 1.05. The results obtained with these particles in a cyclone are summarized in Table I. Test No. 2 shows that the catch of spheres was

Jan 1, 1963

By H. Biloni, G. F. Bolling

The microsegregation in tin specimens containing 0.2, 0.5. or 1 wt pct Pb has been studied m detail. The specimens were grown from the melt in a controlled fashion and exhibited a well-developed cellular substructure. In general, a severe microsegre-gation can be observed in all the photomicrographs presented. Some observations of special interest are: the distribution of a second phase, the presence of a eutectic in several forms, the evidence for dif- THE Smialowski or cellular substructure which forms during solidification has been extensively studied during the past decade.&apos;* Many of the im- *Most references to the cellular substructure except for the last 2 years are reviewed by Winegard.&apos; portant characteristics of this substructure, such as its occurrence and variability with growth conditions, its association with the phenomenon of constitutional supercooling, and its effect on dislocation production and mobility, have received quantitative fusion of solute in the solid, and the enhancement of segregation at a dislocation subboundary network ("striations"). On the basis of the metallographic observations, it can be deduced that the segregation observed is representative of the solute segregation produced during crystal growth. Some deductions are also made about the interplay of the two substructures, cells and striations, and about the mode of cellular growth. investigation both experimentally and theoretically. However, the concentration profile of the solute segregated during crystal growth has not been examined in detail. Tin crystals grown from melts containing various concentrations of lead have been most extensively studied. In a recent investigation of Sn-Pb alloys, Kramer et a1.2 were able to measure the effective solute distribution coefficient, k,, at the most advanced points of the solid-liquid interface, over a large range of growth conditions, finding it to approach the value of k,, the equilibrium distribution coefficient. As will be shown, this must lead to drastic segregation. Another recent study of several Sn-Pb alloys was made by Plaskett and winegard3 who investigated the variation of growth conditions

Jan 1, 1963

By d&apos, D. L. Guernsey, J. C. Entremont, John Chipman

The reaction of aluminum with oxygen in liquid iron has been studied at 1740° and 1910°C. Interaction coefficients are very much smaller than those previously published. The equilibrium constant falls within the approximate range of earlier experiments and calculations. It is now well established that the "deoxidation constant" for aluminum in liquid steel at normal steelmaking temperatures is extremely srnall. Thus at 1600 the product [% Al] is of the order of 10- 14. At this temperature, therefore, great experimental difficulties are encountered in any attempt to obtain a direct measure of the constant. If the aluminum content of the metal is high enough to permit accurate analytical determination, that of the oxygen is too low. If the oxygen content is high, the aluminum is below the limits of detection. Our best values of the constant have come, therefore, from indirect calculations or by extrapolation downward from observations at higher temperatures.. For their studies in the range 1695° to 1866°C, Gokcen and chipman&apos; restricted their aluminum concentrations to less than 0.06 pct. Oxygen analyses were made by the vacuum fusion method which appears to be dependable at these low aluminum concentrations. Experimental results, even at the highest temperature employed, are extremely sensitive to small errors in the analysis of either element. The slight change in the equilibrium constant with aluminum and oxygen concentration was utilized in the calculation of interaction coefficients expressing the effect of aluminum on the activity coefficient of oxygen and the reciprocal effect of oxygen on the activity coefficient of aluminum. The range of concentration, however, was so limited that only a rough value for the interaction coefficient could be obtained. This value was much more negative than that found for any other pair of reactants. In view of the very limited range of concentration employed, the reported values have always been regarded as very questionable. A redetermination was the object of the work reported here. EXPERIMENTAL METHOD At high aluminum concentrations, the vacuum fusion method is susceptible to grave error due to absorption of gas by aluminum distilled from the sample. Since it was desirable to extend the study to at least as much as 1 pct Al, another method for analysis was sought. In steels containing more than a few hundredths of a percent aluminum and no other deoxidizing element, it is safe to assume that

Jan 1, 1963