Search Documents

By F. Keller, A. H. Geisler

Although the subject of precipitation from solid solution appears to be one of the more profitable fields in metallurgy for study with the electron Microscope, few comprehensive studies have yet been made. Occasionally electron micrographs have been published that illustrated alleged precipitation in various alloys, but frequently the apparent size, shape and distribution of the particles do not fully agree with those expected from theory or from observations of the Precipitate when the particles have grown to a size resolvable with the light microscope. The presence of a Widmanstatten pattern for submicroscopic precipitate particles has been demonstrated for only a few aged a1loys.l The initial problem has been the development of suitable techniques for applying the transmission-type instrument to a field that has been inherently associated with reflection-type microscopes. The various techniques have been the subjects of numerous publications, and instead of describing them individually here it will sufficc to say that the oxide film method has proved to be the most satisfactory method yet found for studying the microstructure of aluminum alloys with the electron microscope. This method has the definite characteristic advantage that the actual surface layer of the specimen is examined and not a plastic or silica replica. Doubtless suitable methods for forming and removing the oxide film could be developed for alloys of other metals. The purpose of this report is to point out the characteristics of the oxide film method that have been observed during the studies of numerous aged aluminum alloys and to present the results of these studies. Preparation OF Specimens The oxide film method for preparing specimens of aluminum alloys for examination in the electron microscope has been described in detail previously,¹,² Briefly, the procedure consists of preparing the metallographic specimen, forming the film by anodic oxidation and removing the film from the prepared surface of the oxidized specimen, The specimen is polished accordillg to the usual procedure for microscopic examination.³ he specimen may then be etched to remove flowed metal but etching to attack the alloy constituents or to leave them in relief as in the replica processes is not necessary. Electrolytic polishing is not generally recommended, Deep-etched specimens are used frequently, however, since they provide information that is not revealed by polished specimens and frequently present the same information more clearly than do the polished specimens, The surface preparation of specimens to be deep-etched is not important, since specimens that have been subjected to the first wet PoliShing operation are generally used, but frequently the as-rolled surface is Suitable. These specimens are then etched using a suitable macroetching reagent.

Jan 1, 1947

By Yves Dardel

The literature concerning the remelting of alunlinum and its alloys is very rich. Unfortunately the majority of the papers on this subject have no scientific value, for the theories or explanations are too often in opposition to the laws of physical chemistry. Furthermore only a few of the methods of purification are considered in the literature. Hence it has seemed necessary to the author to sketch a general theory of the purification of secondary aluminum. The purpose of the purification is to eliminate the impurities that the charges, and the atmosphere of the furnaces bring into the bath during the successive remelt-ings. Indeed scrap as well as metal can be used when soiled with oil, greases, water, or others, if it has been sufficiently dried and contains oxides. At the melting point of the metal, the oils and greases are decomposed into their constituent elements H², O2, H²O, co², CO, CH,, and others. These gases are added to the corresponding gases in the atmosphere of the furnace. The impurities introduced into the metal during the remelting will accordingly be the products resulting from the action of these gases on the liquid bath. The quantity of hydrogen that is soluble in aluminum and its alloys is given by the relation of sieverts3 as a function of its partial pressure PH2 in the atmosphere m = K vPH² [I] where K is a constant, which depends only upon the temperature and increases with it. The experiments of Rijntgen and Braun4, RÖntgen and M611er5, Bircum-shaw6 and Winterhagerl have rendered it possible to determine the value of K, when the temperature is less than 1.000°C. One obtains: log K = - 4,702.5T + o 238 [2] Besides its usual significance, the term Ph~ represents also the internal pressure of dissolved hydrogen, since in the equilibrium, that is1 when no gas exchange Occurs between the two phases, the Pressures of hydrogen in the metal and in the atmosphere must be equal. Eq I shows that when the partial pressure of H2 increases the amount of dissolved hydrogen in equilibrium increases, and that, if the contrary Occurs, the quantity of H² should diminish. However, if for any reason the diffusion of hydrogen is hindered and the total amount of dissolved gas remains constant while the temperature decreases, the internal pressure of H² increases until it reaches the value of the hydrostatic pressure. At this point, bubbles of hydrogen appear in the bath and escape out of it. The amount of hydrogen in the metal is then limited by a a depending On the hydro- static pressure, which is practically equal to one atmosphere. Indeed hydrogen does not accumulate in the form of a hydride appearing as a separate solid phase. However1 even the mechanism of the solubility is not yet elucidated. One does not know exactly as yet, whether the

Jan 1, 1949

By H. L. Talbot, A. J. Hall, R. S. Young

A method for the chemical differentiation of oxidized cobalt from sulphide cobalt in complex mine and concentrator products is of considerable interest to a number of mining companies throughout the world. At Rhokana Corporation, in Northern Rhodesia, for instance, cobalt is found chiefly in the sulphide form, as carrollite, but a cobalt oxide mineral also occurs irregularly underground, and relatively small quantities in the mill feed from time to time are difficult to detect visually at the concentrator. The oxide cobalt is not amenable to the sulphide flotation process used at Rhokana and nearly all of it passes into the tailing. The occasional period when high tailing losses occur with no appreciable change in the cobalt content of the mill feed has naturally been disturbing to the concentrator staff, who had no means of determining whether an influx of cobalt oxide ore or laxity on the part of the mill flotation operators as responsible for these results. A procedure for determining oxidized copper minerals, based on their solubility in dilute sulphuric acid saturated with SO2, has been used for many years in the copper industry. In the presence of cuprite the procedure fails, but otherwise it gives results quite satisfactory for control pur- poses. Empirical methods for determining oxidized forms of zinc and lead have been proposed, but they are generally unsatisfactory except in isolated cases of a very simple ore. No published procedure for determining oxidized forms of cobalt or siinilar elements has come to the authors'notice. Experimental Procedure There are obviously two alternatives in attacking a problem of this nature. One is to employ a solvent for oxide cobalt that will not dissolve cobalt sulphide. The other is to find a substance capable of oxidizing the cobalt sulphide to sulphate, which of course is water-soluble and can then be leached from the water-insoluble oxide cobalt. A representative sample of Rhokana cobalt oxide ore was secured, ground to minus 200-mesh, and carefully rolled to furnish a standard sample for this work. The analysis of this material gave the following percentages: Co, 0.55; Cu, 1.64; Fe, 2.98; S, 0.08; SiO2, 41.42; Al2o3, 11.20; CaO, 13.82; MgO, 3.80. As a source of cobalt sulphide a similar large sample of carrollite was prepared, analyzing Co, 23.36 per cent; Cu, 27.32; and S, 31.13. This sample of course was not pure carrollite, but was the best obtainable in quantities sufficient for a standard intended for considerable work. It was a pure sulphide sample, microscopic examination showing only carrollite, a little bornite, and gangue, with no visible sign of oxidized ore present. A wide variety of reagents as used in an effort to selectively dissolve oxide cobalt or to oxidize sulphide cobalt to

Jan 1, 1949

By A. A. Collins

OF widespread interest to all lead metallurgists is a dross smelting process that will consistently give mattes and speiss of low lead and high copper contents. It is a problem that has intrigued operators through the years. In the past few years, with increased amounts of copper and overcrowded metallurgical furnaces, it has been almost imperative to evolve some method. Experience at El Paso has shown that high copper-to-lead ratios can be consistently attained by use of the new soda process. † Older Methods Three older methods, i.e., 1. Dross plus siliceous ore and limestone. 2. Dross plus one fourth blast-furnace speiss and siliceous ore, 3. Dross plus sulphides and siliceous ore and scrap iron, do not give consistent results and the products are higher in lead and lower in copper than those produced with the soda process. Any radical changes in the blast-furnace drosses will throw any of these older methods into a tailspin. For almost ten years, El Paso used the third method, with varying results. It was noted that as the arsenic and anti- mony contents of the dross increased, the speiss and matte steadily increased in lead and decreased in copper. A dross containing four parts of copper to one part of arsenic was considered to be ideal. At times this balance of copper to arsenic was almost impossible to obtain, especially when handling copper Cottrell dusts, which vary so greatly in copper and arsenic. Custom smelters like El Paso, handling a great variety of dusts, concentrates and ores, are apt to have anywhere from I to 15 pct copper on charge, with a variable in arsenic of from 0.5 to 2 pct. With such conditions, and drosses varying from 12 to 40 pct copper, an operator using the older dross reverberatory practices would find it just about impossible to control the reverberatory operations and to produce a good product low in lead and high in copper. These conditions prevailed at El Paso from the year 1940 to the present. In May 1942 it was decided to experiment with the soda process. The Soda Process After considerable work the soda method was finally evolved to the present state. It has never failed to produce a good product even under the most adverse conditions. It might be well to remember that many men of thought in the smelting industry have at one time or other sugge~ted the removal of all copper, arsenic and antimony from the lead blast-furnace charge as an aid to better metallurgy.

Jan 1, 1949

By F. E. Buch

The smelter is designed for a production capacity of 10,000 short tons of blister copper per month, when operating on the present concentrate grade. The :smelter lay-out is shown in Fig I. The major units of equipment comprise two reverberatory furnaces, three 13 by 30 ft Peirce-Smith converters, two straight line casting machines, one being served by a 13 by 30 ft holding furnace and the other by a ladle tilting quadrant, two converter aisle cranes and two converter blowers. The plant is compact and the general arrangement of standard gauge railway tracks and conveyor belts provides for a smooth flow of incoming materials and metallurgical products. Concentrate is received by conveyor belts and fluxes in standard railway trucks. The local track system is covered by an electric trolley locomotive. The Smelter treats a wet (10 pet moisture) flotation concentrate containing approximately 53 pct total copper and 0.9 pet oxide copper. The predominating minerals are chalco-cite, chalcopyrite and bornite accompanied by small quantities of malachite and cuprite. The gangue is highly siliceous and requires limerock as a flux. Metallurgy would be straightforward were it not for the necessity for eliminating bismuth which occurs as an impurity in the ore along with the copper. Directly or indirectly most of the Smelter problems can be traced to the necessity for bismuth removal. Copper recovery, based on new copper treated, averages 97.5 pct. General Description Flotation concentrate, containing 10 Pet moisture, is conveyed from the filter plant to the smelter storage bins by conveyor belts, the material being weighed in transit by a revolving drum type of weigh-hopper manufactured by Samuel Denison and Son, Ltd. The weigher consists of a steel drum with dished ends carried on a horizontal shaft and rotated by an 8 hp squirrel-cage motor through a worm and Spur reduction gear. A feed mouth in the drum permits of filling and discharging. The whole unit is supported on a framework resting on weighing levers, which in turn connect to transfer levers, thence to an indicator of the steelyard type situated 16 ft from the centre line of the drum. The indicator is of the "Denison" patent equilibrium controlled ticket stamping type, insuring mistake-proof records by a combination of electrical and mechanical movements which Prevent the operator from printing until the cover enclosing the steelyard is closed and until the poise-weights have been Set correctly to the weight of the hopper and contents. The machine is capable of weighing 30 short tons at a time but the usual load is 28 tons. After obtaining the gross weight the drum is rotated and discharged into a receiving bin situated below it. The empty drum is weighed and the net weight re-

Jan 1, 1949

By J. H. Hamilton, J. R. Lewis, J. C. Nixon, C. L. Graverson

Recently there has been much speculation concerning the advantages of using oxygen enriched air or pure oxygen in pyrometallurgical processes. The advantage of using oxygen in the iron blast furnace and elsewhere in ferrous production metallurgy has been given careful consideration and, as a result, several experiments have been tried and a number of papers have recently appeared on this subject.1,2 The use of oxygen in nonferrous smelting likewise poses interesting problems, therefore any experimental data obtained in this area should likewise be of general interest.3 Ashcroft4 in 1933 claimed to have obtained improved results when copper concentrates were roasted in pure oxygen. However, his paper gives neither data nor experimental details. A few years ago a research program was started in this laboratory which had for its objective the determination of the mechanism of oxidation of metallic sulphides under varying conditions. To date we have studied the oxidation of galena,5 sphalerite, and chalcocite. It is a well known fact that copper sul-phde minerals are oxidized slowly by weathering agents, so that the sulphides of copper which find their way into a tailing pond are gradually changed into water-soluble and acid-soluble compounds. Commercially, the oxidation of copper sulphide minerals is accomplished by roasting them at elevated temperatures in air using a suitable (usually multiple hearth) roaster. Most copper smelters are equipped with such roasters. In this phase of the study we were especially interested in determining the effect of temperature on: (I) the air oxidation of chalcocite and (2) the oxidation of chalcocite using commercial oxygen. The amount of sulphur removed as sulphur dioxide, the percentages of water-soluble copper, and of acid-soluble copper were determined for the air oxidation tests. In the tests using commercial oxygen, the percentages of sulphur dioxide, water-soluble copper, acid-soluble copper, water-soluble sulphur and acid-soluble sulphur were determined. For these experiments the temperature was varied from 250 to 700°C. Experimental Preparalion of Chalcocite (Cu2S) Natural chalcocite is not pure, Minerals such as bornite, covellite, and chalcopy-rite may be present. It was decided, therefore, to prepare some chalcocite in our own laboratory. An intimate mixture of sulphur and hool divided metallic copper in the ratio of 4 g of copper to I g of sulphur was prepared. TO this was added a slight excess of

Jan 1, 1949

By R. E. Rick-Secker, W. M. Baldwin, R. McC. Baker

The development of residual stress in strip during rolling has not been systematically studied. A few scattered papers1 -3 mention the existence of residual stresses in rolled strip or touch upon some of their consequences,4 but there is no report of experimental studies in the literature describing the effect of roll diameter, strip thickness, reduction per pass, metal or metal temper on the magnitude of residual stresses developed in strip. A knowledge of the effect of these factors is of utmost importance since the present trend in some mills to roll thicker castings brings the rolling operation directly into that set of conditions which, as will be brought out in the present paper, can develop residual stresses of a high order of magnitude. Fundamentally, residual stresses are developed when a metal is inhomogeneously unloaded from a plastic stress state. A quantitative treatment of the problem requires a reasonably complete understanding of both the plastic stress state existing in the metal during deformation and the elastic stress state that would exist in the metal if the metal deformed as a perfectly elastic body. Our knowledge of either state during rolling is so meager that we can make no more than certain limited predictions. The external stress state* operating between the roll and the metal consists of Fig i—External stress state operating on metal being rolled (after siebel and LUEG5) compressive stresses acting almost perpendicularly to the strip surface but varying in intensity both across the width of the strip and along the length of contact between the roll and the strip. An example of such a stress distribution is shown in Fig I. The nature of the internal stress state, produced by this distribution, is not known. Various investigators,5-7 however, have felt that a qualitative similarity must exist between the internal stress state so produced and that produced in a wide strip that is compressed between two flat-nosed punches. In the latter case the external stress state is represented by a uniformly distributed compressive stress acting per-pendicularly over the face of the tool. The

Jan 1, 1949

By Frances Jane Cromwell, D. J. McAdam, G. W. Geil

In a series of papers, the authors and their associates have shown that the technical cohesion limit is affected by the same four factors that affect the flow stress, namely, the stress system, plastic deformation, temperature, and the strain rate.16-28 Moreover, each of these factors affects the technical cohesion limit in the same way that it affects the flow stress,t The influence of any one of these factors may be represented by a curve of cohesion limits, although only one point on the curve may represent actual fracture. As the term ilfracture stress' has come into use with the same significance as 'ctechnical cohesion limit," the authors are here using the shorter term with the understanding that it does not always refer to actual fracture. This paper presents the results of an investigation of the flow of notched and unnotched specimens betqrcen yield and fracture, A study has thus been made of the flow and fracture of metals as affected by the stress system, and of the ductility of metals as affected by the stress system throughout plastic deformation. Since two of the principal stresses were equal (S2 = S3), the stress systems can be represented in terms of the radial stress ratio, the ratio of the transverse radial stress s3 to the longitudinal stress S1.* The results of the experiments are represented by a series of their flow-stress curves for each Of the metals investigated, and these diagrams have been used for the development Of Other diagrams Of various types. Ludwik was the first to suggest that fracture occurs when the flow stress becomes equal to a technical cohesion limit. Although Ludwik was incorrect in his view that the technical cohesion limit depends only on the amount of plastic deformation and not On the stress system, he was 'Orrect in his view that the flow-stress curve for any stress system terminates at a point representing a technical cohesion limit. The authors have seen some recent references to assertions by other investigators that a point representing the fracture stress need not be on the flow-stress curve. The Ludwik principle, however, "auld Seem to be axiomatic, provided that the flow Stress and fracture stress are properly defined and that the fracture stress is properly determined. Disregard Of this principle has led to some untenable theories Of criteria for fracture. When a metal is very ductile, the conventional method of determining the ductility and fracture stress by dividing the breaking load by the sectional area after fracture, generally gives values that are much too high. With some metals and in some conditions, the fracture stress thus determined may be two or more times the

Jan 1, 1949

By E. R. Parker, C. F. Riisness

That grain size has a great effect on the mechanical properties of metals has been recognized for a long time. Bassett and Davis1 in 1919 did excellent work in determining the effect of grain size on the hardness of brass. About the same time Jeffries2 and Sykes3 ascertained the effect of grain size on the strength and ductility of several metals at temperatures ranging from 80°K. to their melting points. When creep was recognized as an important property of metals, those engaged in the study of this property tried to determine the effect of grain size4-9 on creep rates. Unfortunately, most of the investigations were conducted on only two grain sizes, often designated simply as "fine" and "coarse." It is well known that at high temperatures coarse-grained test bars creep less than fine-grained bars. The relation between grain size and size of test bar, however, has been frequently overlooked. In his early work, Jeffries2 showed that the number of grains in the cross section has a more important influence on the properties of the material than the actual grain size. More recently Hanson and Sandfordl0 and Pell-Walpole,ll in their measurements of the creep and tensile properties of tin, tested a wide range of grain sizes and correlated the grain size with the strength. Hanson9 later established the relation between the number of grains in the cross section and the creep rate. At high temperatures, generally a coarsegrained structure is considered superior in creep to a fine-grained structure, but it is evident that for a given bar size continuously increasing the grain size will

Jan 1, 1944

By C. F. Riisness, E. R. Parker

That grain size has a great effect on the mechanical properties of metals has been recognized for a long time. Bassett and Davis1 in 1919 did excellent work in determining the effect of grain size on the hardness of brass. About the same time Jeffries2 and Sykes3 ascertained the effect of grain size on the strength and ductility of several metals at temperatures ranging from 80°K. to their melting points. When creep was recognized as an important property of metals, those engaged in the study of this property tried to determine the effect of grain size4-9 on creep rates. Unfortunately, most of the investigations were conducted on only two grain sizes, often designated simply as "fine" and "coarse." It is well known that at high temperatures coarse-grained test bars creep less than fine-grained bars. The relation between grain size and size of test bar, however, has been frequently overlooked. In his early work, Jeffries2 showed that the number of grains in the cross section has a more important influence on the properties of the material than the actual grain size. More recently Hanson and Sandfordl0 and Pell-Walpole,ll in their measurements of the creep and tensile properties of tin, tested a wide range of grain sizes and correlated the grain size with the strength. Hanson9 later established the relation between the number of grains in the cross section and the creep rate. At high temperatures, generally a coarsegrained structure is considered superior in creep to a fine-grained structure, but it is evident that for a given bar size continuously increasing the grain size will

Jan 1, 1944

By William M. Baldwin

Jan 1, 1946

By William M. Baldwin

Jan 1, 1946

By Lester Guttman

The connection between short-range order and thermodynamic activities in binary solid solutions has been pointed out by Birchenall1 who calculated approximate values of the energy of interaction of Zn and Cu and of the degree of short-range order in alpha-brass at one composition and two temperatures. The purpose of this note is to call attention to a somewhat more accurate theory than that used by Birchenall and to extend the calculations to a wider range of temperatures and compositions as a means of testing its applicability. This is an example of the well-known general method of comparing measured thermodynamic functions of a system with those calculated from a microscopic model by means of statistical mechanics. In this case, mathematical approximations must be used which make difficult any decision about the validity of the model. Birchenall's statistical method resembles closely that calletl by Fowler and Guggenheim2 the "quasi-chemical" method which has been shown to lead in certain cases to results identical with those obtained by the method of Bethe.3 The quasi-chemical method has been extended by Takagi4 to give general formulas for the free energy of binary solid solutions based on several lattices at all compositions and temperatures. Specifically, he treats the four systems typified by: I. ß—brass; 2. a— brass; 3. AuCus; 4. AuCu. When the temperature is so high that no long-range order exists, these formulas reduce to the same form for all lattice types, namely: F - FO = (E - EO) - T(S - S°) = ½N(c — x)(2vAB — vAa - Vbb) + M1/2KTzN[x log x + 2(c - x) log (c - X) + (I - 26 + X) log (1 - 2c + x)] -kT(z - I)N[c log c+(I-c) log (I -c)] Eq 1 Here the symbols have the following meanings: F—Helmholtz free energy E—nergy 5—ntropy k—the gas constant per molecule z—coordination number, (e.g. I2 in a-brass). N—total number of atoms c—atomic fraction of A Vaa, vbb, vab—nergy per bond of the corresponding bonds log means log, throughout The superscript ° denotes the thermodynamic fuliction in the standard state, which has been chosen to be pure A or B with the same crystal structure as the solid solution. The quantity x is the ratio of the number of A-A bonds to the total number and for minimum F satisfies the equilibrium condition Vaa + Vbb — 2Vab _ (c — x)2 kT l0g x(i - 2C + x The free energy may also be written in terms of /La and B, the chemical potentials or partial free energies per alom:

Jan 1, 1949

By Lester Guttman

The connection between short-range order and thermodynamic activities in binary solid solutions has been pointed out by Birchenall1 who calculated approximate values of the energy of interaction of Zn and Cu and of the degree of short-range order in alpha-brass at one composition and two temperatures. The purpose of this note is to call attention to a somewhat more accurate theory than that used by Birchenall and to extend the calculations to a wider range of temperatures and compositions as a means of testing its applicability. This is an example of the well-known general method of comparing measured thermodynamic functions of a system with those calculated from a microscopic model by means of statistical mechanics. In this case, mathematical approximations must be used which make difficult any decision about the validity of the model. Birchenall's statistical method resembles closely that calletl by Fowler and Guggenheim2 the "quasi-chemical" method which has been shown to lead in certain cases to results identical with those obtained by the method of Bethe.3 The quasi-chemical method has been extended by Takagi4 to give general formulas for the free energy of binary solid solutions based on several lattices at all compositions and temperatures. Specifically, he treats the four systems typified by: I. ß—brass; 2. a— brass; 3. AuCus; 4. AuCu. When the temperature is so high that no long-range order exists, these formulas reduce to the same form for all lattice types, namely: F - FO = (E - EO) - T(S - S°) = ½N(c — x)(2vAB — vAa - Vbb) + M1/2KTzN[x log x + 2(c - x) log (c - X) + (I - 26 + X) log (1 - 2c + x)] -kT(z - I)N[c log c+(I-c) log (I -c)] Eq 1 Here the symbols have the following meanings: F—Helmholtz free energy E—nergy 5—ntropy k—the gas constant per molecule z—coordination number, (e.g. I2 in a-brass). N—total number of atoms c—atomic fraction of A Vaa, vbb, vab—nergy per bond of the corresponding bonds log means log, throughout The superscript ° denotes the thermodynamic fuliction in the standard state, which has been chosen to be pure A or B with the same crystal structure as the solid solution. The quantity x is the ratio of the number of A-A bonds to the total number and for minimum F satisfies the equilibrium condition Vaa + Vbb — 2Vab _ (c — x)2 kT l0g x(i - 2C + x The free energy may also be written in terms of /La and B, the chemical potentials or partial free energies per alom:

Jan 1, 1949

By L. S. Darken

It has been known for sometime that in an ionic lattice, such as that of Ag,S or FeO, the migration velocity of the anion may differ markedly from that of the cation, the cation being usually the more mobile. It has been shown, for example, that during the formation of silver sulphide tarnish on silver, the tarnish layer grows principally by the motion of silver cations through an essentially stationary silver sulphide lattice. Similarly, the growth of wüstite scale on iron proceeds principally by motion of iron ions, this being accomplished with the aid of vacant iron lattice sites into which the iron may readily move. Likewise investigations of solid solutions of various salts show that there is usually considerable difference in the mobility of the different cations. Measurements of the mobility of various metals in liquid mercury indicate that the various metals move at different rates relative to the mercury. In view of the foregoing it seems reasonable to suppose that when any single phase solution is acted upon by a field, the atoms of different elements respond in different ways; in particular, that the force arising from a particular concentration gradient in a binary alloy would cause the atoms of one component to move with a drift velocity which is in general different from that of the atoms of the other component. The recent experiments of Smigelskas and Kirkendall' support this viewpoint, since they found, in essence, that during diffusion in alpha brass zinc moves appreciably faster than copper. This was evidenced by the relative motion of inert molybdenum wires. Such a phenomenon is incompatible with the place interchange theory of diffdsion proposed by several authors, most recently by Bir-chenall and Mehl.2 This theory has, however, also been questioned3v4 for other reasons, principally because of the high energy required for direct atomic interchange. The rather large difference between self-diffusivity, as determined by tracer technique, and mutual diffusivity, as determined with the aid of another element at low concentration, might also be interpreted as favoring the above idea; however. Birchenall and Mehl have pointed out that there is another possible interpretation of this difference; this is discussed later. A rather interesting and pertinent observation was made by Schwartz, who noticed that tungsten lamp filaments, heated by direct current, commonly fail by necking down in the region at the temperature gradient near the positive end, with a concomitant swelling at the corresponding position near the negative end. He introduces considerable auxiliary evidence to show that this phenomenon is occasioned by the motion of tungsten cations which migrate in the high temperature region from the positive end toward the negative end. The mechanism of this migration is not clearly understood. If, however, this experimental evidence be accepted, we are forced to conclude that alloying elements in the tungsten may migrate by the same or a Similar mechanism, whatever it may be, at a greater Or lesser rate, thus giving rise

Jan 1, 1949

By L. S. Darken

It has been known for sometime that in an ionic lattice, such as that of Ag,S or FeO, the migration velocity of the anion may differ markedly from that of the cation, the cation being usually the more mobile. It has been shown, for example, that during the formation of silver sulphide tarnish on silver, the tarnish layer grows principally by the motion of silver cations through an essentially stationary silver sulphide lattice. Similarly, the growth of wüstite scale on iron proceeds principally by motion of iron ions, this being accomplished with the aid of vacant iron lattice sites into which the iron may readily move. Likewise investigations of solid solutions of various salts show that there is usually considerable difference in the mobility of the different cations. Measurements of the mobility of various metals in liquid mercury indicate that the various metals move at different rates relative to the mercury. In view of the foregoing it seems reasonable to suppose that when any single phase solution is acted upon by a field, the atoms of different elements respond in different ways; in particular, that the force arising from a particular concentration gradient in a binary alloy would cause the atoms of one component to move with a drift velocity which is in general different from that of the atoms of the other component. The recent experiments of Smigelskas and Kirkendall' support this viewpoint, since they found, in essence, that during diffusion in alpha brass zinc moves appreciably faster than copper. This was evidenced by the relative motion of inert molybdenum wires. Such a phenomenon is incompatible with the place interchange theory of diffdsion proposed by several authors, most recently by Bir-chenall and Mehl.2 This theory has, however, also been questioned3v4 for other reasons, principally because of the high energy required for direct atomic interchange. The rather large difference between self-diffusivity, as determined by tracer technique, and mutual diffusivity, as determined with the aid of another element at low concentration, might also be interpreted as favoring the above idea; however. Birchenall and Mehl have pointed out that there is another possible interpretation of this difference; this is discussed later. A rather interesting and pertinent observation was made by Schwartz, who noticed that tungsten lamp filaments, heated by direct current, commonly fail by necking down in the region at the temperature gradient near the positive end, with a concomitant swelling at the corresponding position near the negative end. He introduces considerable auxiliary evidence to show that this phenomenon is occasioned by the motion of tungsten cations which migrate in the high temperature region from the positive end toward the negative end. The mechanism of this migration is not clearly understood. If, however, this experimental evidence be accepted, we are forced to conclude that alloying elements in the tungsten may migrate by the same or a Similar mechanism, whatever it may be, at a greater Or lesser rate, thus giving rise

Jan 1, 1949

By Paul A. Beck, John Towers, W. D. Manly

Recent work on grain growth in high purity aluminum and in a solid solution type alloy of aluminum and magnesium1 showed that the isothermal increase of the average grain diameter D with time follows the equation D = K(tg + A)2 Eq 1 Here tg is that portion of the total annealing time 1, which is available for grain growth, that is the annealing time less the time necessary for complete recrystallization. The parameters A and n depend only on the temperature. A has the dimension of time, and is usually of the same order of magnitude as the time for recrystallization: R. In some instances R may be substituted for A in Eq I. This is the case for example at long periods of annealing or at high temperatures, where both A and R become small in comparison with t. In such instances Eq I is simplified to D = K, tn Eq 2 For high purity aluminum the exponent n increases in the ratio of about I to 5 in. the temperature range of 350' to 6w°C. 70-30 brass has been long known to exhibit grain growth of the continuous type, such as occurs in pure metals. For this reason, it was of interest to determine whether grain growth in brass also follows Eq I and 2 as it does in aluminum. Isotherma1 grain growth data for 70-30 brass were published in recent years by R. S. French2 and by H. L. Walker.' Fig I shows a log D vs. log t plot of French's data, together with the 500, 600 and 700°C data by Walker for 42.8 pct reduction by rolling. In such a log D vs. log t plot Eq 2 corresponds to a straight line with a slope of n. As seen in Fig I, the data of French and the 600°C data of Walker can be represented by straight lines. However, the 500 and 7c0°C data of Walker give curved lines and the average slope of his 500°C line is very different from the slope of the line representing French's data for the same temperature. Furthermore the curvature of the 700°C line suggests a tendency for grain growth to stop after 8 hr. If confirmed, this would constitute a definite deviation from Eq 2, not explainable on the basis of the specimen thickness effect found in high purity aluminum.' Walker's specimens were 0.125 in., or 3.18 mm thick, while the deviation occurs at a grain size of approximately 0.4 mm. A further point of divergence arose recently, when J. E. Burke suggested4 on the basis of Walker's data, that, instead of Eq 2, grain growth in 70-30 brass follows the equation D - D, = K . tgn Eq 3 Here D, is the grain size as recrystallized at the temperature in question. Although it was possible to show5 that, for pure aluminum, for which isothermal grain growth has been carefully investigated, Eq 3 is incompatible with the experimental facts, the situation remained somewhat obscure with respect to 70-30 brass. In view of the discrepancies between the published data on brass from different

Jan 1, 1949

By John Towers, Paul A. Beck, W. D. Manly

Recent work on grain growth in high purity aluminum and in a solid solution type alloy of aluminum and magnesium1 showed that the isothermal increase of the average grain diameter D with time follows the equation D = K(tg + A)2 Eq 1 Here tg is that portion of the total annealing time 1, which is available for grain growth, that is the annealing time less the time necessary for complete recrystallization. The parameters A and n depend only on the temperature. A has the dimension of time, and is usually of the same order of magnitude as the time for recrystallization: R. In some instances R may be substituted for A in Eq I. This is the case for example at long periods of annealing or at high temperatures, where both A and R become small in comparison with t. In such instances Eq I is simplified to D = K, tn Eq 2 For high purity aluminum the exponent n increases in the ratio of about I to 5 in. the temperature range of 350' to 6w°C. 70-30 brass has been long known to exhibit grain growth of the continuous type, such as occurs in pure metals. For this reason, it was of interest to determine whether grain growth in brass also follows Eq I and 2 as it does in aluminum. Isotherma1 grain growth data for 70-30 brass were published in recent years by R. S. French2 and by H. L. Walker.' Fig I shows a log D vs. log t plot of French's data, together with the 500, 600 and 700°C data by Walker for 42.8 pct reduction by rolling. In such a log D vs. log t plot Eq 2 corresponds to a straight line with a slope of n. As seen in Fig I, the data of French and the 600°C data of Walker can be represented by straight lines. However, the 500 and 7c0°C data of Walker give curved lines and the average slope of his 500°C line is very different from the slope of the line representing French's data for the same temperature. Furthermore the curvature of the 700°C line suggests a tendency for grain growth to stop after 8 hr. If confirmed, this would constitute a definite deviation from Eq 2, not explainable on the basis of the specimen thickness effect found in high purity aluminum.' Walker's specimens were 0.125 in., or 3.18 mm thick, while the deviation occurs at a grain size of approximately 0.4 mm. A further point of divergence arose recently, when J. E. Burke suggested4 on the basis of Walker's data, that, instead of Eq 2, grain growth in 70-30 brass follows the equation D - D, = K . tgn Eq 3 Here D, is the grain size as recrystallized at the temperature in question. Although it was possible to show5 that, for pure aluminum, for which isothermal grain growth has been carefully investigated, Eq 3 is incompatible with the experimental facts, the situation remained somewhat obscure with respect to 70-30 brass. In view of the discrepancies between the published data on brass from different

Jan 1, 1949

By T. R. Graham, R. G. Feustel, J. R. Long, R. S. Dean

The comprehensive study of the copper-manganese-zinc alloy system in the Bureau of Mines Laboratories has so far been principally concerned with alloys that lie within the alpha solid solution field of the system. Two regions, one centering about 65 per cent copper, 10 manganese and 25 zinc, and the other centering about 70 per cent copper, 20 manganese, 10 zinc, have been carefully examined. These have indicated that when the alloys are made with electrolytic manganese considerable amounts may be readily introduced and that the strength and hardness are increased without excessive losses in ductility. In the course of the work of establishing the limits of the alpha solid solution range,3 enough of certain of the alloys was prepared to permit determination of their- physical properties at a future date. The 60 per cent copper alloys concerned here are part of that series. They cross through the alpha field and into the adjacent alpha plus beta and alpha plus X fields, and are of interest because they offer some information on the effects of manganese on alloys lying in these fields. Composition and Fabrication of Alloys The chemical composition of the alloys involved is given in Table I. They were intended to contain 60 per cent copper with manganese increasing from 5 to 25 per cent in 5 per cent steps and zinc decreasing from 35 to 15 per cent. The analyses are so close to the intended composition that the alloys mill be referred to hereafter by their norninal compositions. They were finished into sheet by cold-rolling and intermediate annealing procedurzs calculated to produce sheet $i6 in. thick, cold-reduced by 20, qo, 60, and 80 per cent in thickness. Physical properties were determined on standard flat tensile specimens with 4-in. gauge length, although the test gauge section used was 2 in. The properties were obtained on cold-worked material and on material annealed at temperatures ranging from 700' to r500°F. after 60 per cent cold reduction. Since an annealing temperature of 1200°F.

Jan 1, 1946

By J. R. Long, T. R. Graham, R. G. Feustel, R. S. Dean

The comprehensive study of the copper-manganese-zinc alloy system in the Bureau of Mines Laboratories has so far been principally concerned with alloys that lie within the alpha solid solution field of the system. Two regions, one centering about 65 per cent copper, 10 manganese and 25 zinc, and the other centering about 70 per cent copper, 20 manganese, 10 zinc, have been carefully examined. These have indicated that when the alloys are made with electrolytic manganese considerable amounts may be readily introduced and that the strength and hardness are increased without excessive losses in ductility. In the course of the work of establishing the limits of the alpha solid solution range,3 enough of certain of the alloys was prepared to permit determination of their- physical properties at a future date. The 60 per cent copper alloys concerned here are part of that series. They cross through the alpha field and into the adjacent alpha plus beta and alpha plus X fields, and are of interest because they offer some information on the effects of manganese on alloys lying in these fields. Composition and Fabrication of Alloys The chemical composition of the alloys involved is given in Table I. They were intended to contain 60 per cent copper with manganese increasing from 5 to 25 per cent in 5 per cent steps and zinc decreasing from 35 to 15 per cent. The analyses are so close to the intended composition that the alloys mill be referred to hereafter by their norninal compositions. They were finished into sheet by cold-rolling and intermediate annealing procedurzs calculated to produce sheet $i6 in. thick, cold-reduced by 20, qo, 60, and 80 per cent in thickness. Physical properties were determined on standard flat tensile specimens with 4-in. gauge length, although the test gauge section used was 2 in. The properties were obtained on cold-worked material and on material annealed at temperatures ranging from 700' to r500°F. after 60 per cent cold reduction. Since an annealing temperature of 1200°F.

Jan 1, 1946