Search Documents

By Lawrence A. Roe

The magnetic reflux classifier, which utilizes the combined effects of magnetic fields and a hindered settling classifier, is a new tool for determining the quantity and quality of middlings in fine-sized magnetite concentrates. Results are given for processing a typical taconite ore, and a sketch of the apparatus is included. IN examining magnetite ores and beneficiated products it often becomes necessary to make critical studies of the amount of grinding necessary to produce the desired degree of magnetite liberation. In the past this has been accomplished by laboratory heavy-liquid tests, which provide a method for selectively removing middling particles and free magnetite from various-sized fractions. Examination of the various products under the microscope results in fairly accurate determination of the degree of liberation. The method is quite efficient on sizes coarser than 325 mesh. Thus the heavy-liquid method of middling separation was satisfactory until the advent of present day magnetic taconite studies. When magnetite concentrates ranging from 70 to 100 pct —325 mesh are studied it becomes apparent that older methods of determining liberation size are not satisfactory and that there is need for a new method. For example, some of the low-grade magnetite ores of the Wisconsin and Michigan iron ranges require grinding to 100 pct —325 mesh to produce a magnetic concentrate containing less than 12 pct silica. Examination of concentrates from such ores often reveals that many of the middling particles consist of only very minor proportions of iron mineral. Thus it becomes important to be able to determine the degree of grinding necessary not only for complete liberation, but also for liberation of only 80, 85, or 90 pct of the total iron mineral content. Actually, complete liberation is never attained, but is often used to designate that degree of liberation necessary for production of high-grade concentrates. A rougher concentrate, produced after elimination of a coarse-sized tailing, can usually be subjected to a second grinding stage and concentrated into a higher grade product than could be produced from the same crude ore with one stage of grinding resulting in the same overall size reduction. This fact adds to the importance of being able to determine partial degrees of liberation on any magnetite ore. Standard laboratory methods such as heavy-liquid separation, microscopic grain counts, Davis tube magnetic separation, magnetic flocculation, classification, flotation, and others often are not applicable, or are prohibitive because of time requirements when large numbers of fine-sized magnetite samples are investigated. The Davis tube magnetic separator is an efficient tool to use in rejecting the non-magnetic mineral particles from an ore sample. The middlings discarded by the tube separator usually are so low in iron content that they can be considered relatively unimportant in liberation studies. This condition is caused by the extremely high flux density used in the Davis tube. This flux density ranges from four to eight times the flux density produced by most of the powerful commercial machines in use today. Thus the problem resolves itself into a search for a method of selectively removing middlings from Davis tube magnetic concentrates which will be both rapid and efficient. Those methods showing most promise in the development of a process for isolating middlings from extremely fine-sized magnetic concentrates were flotation and magnetic flocculation. The use of flotation to remove middlings from magnetic concentrates is reported in the literature.'.' The flotation process is effective in removing middlings from a magnetite concentrate, but physical entrapment of fine-sized free magnetite in the silica-bearing froth is an undesirable feature. The flotation method of removing middlings requires time, effort, and precise control of many variables, and does not meet the required degree of middling isolation. Magnetic Flocculation Magnetic flocculation has long been resorted to"-" in efforts to upgrade magnetite concentrates. One of the new magnetic taconite plants now under construction on the Mesabi Range includes magnetic flocculation in the flowsheet' as an accessory process to remove high-silica middlings and free silica which has been mechanically entrapped in magnetite flocs. The use of magnetic flocculation as a laboratory method of making precise separation of middlings was further investigated, since it offered a rapid, simple method of accomplishing the desired result. Magnetic flocculation involves the subjection of a magnetic concentrate to a strong magnetic field, passing the concentrate in a highly flocculated condition to a hydroseparator or other classifiers of various types, and removing free silica and middlings as overflow products. In an attempt to utilize simple

Jan 1, 1954

By E. F. Poncelet

Forty years ago A. A. Griffith developed a theory explaining why brittle materials displayed such low tensile strengths.' He based his views on two points. First, he found himself compelled to assume that all brittle materials are replete with flaws, cracks, and other defects that act, although quite invisible, as large stress raisers. Second, he applied the "theorem of minimum potential energy," which says that the total potential energy of a system must pass from the unbroken to the broken condition by a process involving a continuous decrease in potential energy. By this means he satisfactorily accounted for the noted low strength of such solids and also for the wide spread obtained in experimental measurement of these strengths. So successful has the theory been that it is favored by some to this day. Unfortunately this theory is of limited use beyond the explanation of these two noted phenomena and it is keenly felt that a better theoretical insight into the physics of the fracturing process is needed as the volume of experimental evidence accumulates. The author proposes in the following to build on the fundamentally sound concepts of Griffith and, with the help of increased theoretical knowledge over that available to Griffith, develop a mechanism for frac-ture which will provide far greater understanding of the experimental evidence accumulated to date than is possible from the original Griffith idea. THE GRlFFlTH THEORY Very little progress indeed can be made without accepting the first postulate of Griffith which supposes all brittle solids to be full of microcracks. It would be difficult indeed to find a better mechanism for the small strength of such brittle materials, in conjunction with the fact that the energy that must be expended for comminution is by no means small. The postulate of the existence of the microcracks permits the breakup of the various bonds a few at a time by concentrating the stress at the tip of the progressing crack, while the total energy expended is the same as if they all had been ruptured simultaneously. The only flaw in the argument is that no reasonable explanation has been proposed to account for the genesis of such cracks. Indeed their very presence is in violation of the Griffith second postulate, the potential energy theorem. This theorem is straightforward for isothermal processes, and, in spite of Griffith, there is some doubt that treating the problem isothermally is legitimate. The surface energy of bodies is a free energy, not a potential energy as stated by Griffith, and the production of new surface free energies is not necessarily an isothermal process. There is ample evidence to the contrary. Generally speaking, if heating a body increases its surface area, then, by virtue of Le Chatelier's principle, any increase of that area by other means will tend to lower its surface temperature. Lord Kelvin calculated the actual cooling that resulted in drawing out a film of liquid.2 R. A. Houston calculated the surface cooling that resulted in stretching a metal wire.3 These calculations were made by applying the Carnot cycle to the process and evaluating the thermodynamics thereof. IRREVERSIBILITY OF THE FRACTURING PROCESS While Griffith was very careful not to say so, the impression gained from studying his papers is that he considered the fracturing process as reversible, that is, a succession of quasi-equilibrium states. There is ample evidence that it is not. The indication that the new surfaces produced by the propagation of a crack are cooler than the original body points to an irreversible heat flow from the interior to the new surfaces to equalize the temperatures. If the process be reversible, any crack accidentally formed should immediately close up as, in the absence of any strain energy, the potential energy would thereby be lowered. The fact that they do not, constitutes a paradox. Such paradoxes are nothing new where certain phenomena that propagate from minuscule nuclei are assumed to be reversible. Such is, for instance, the condensation of a pure saturated vapor that is suddenly chilled by adiabatic expansion. At the beginning the tiny droplets that are formed should be only a few angstroms in size, but the vapor pressure at such droplets is so high that they should evaporate at once. A similar situation arises if a saturated pure solution becomes super-saturated upon cooling; the first tiny crystal nuclei should dissolve as fast as

Jan 1, 1964

By Hobart M. Kraner

THE mineralogy of blast furnace sinter is of interest because its mineral content is one of the important factors contributing to its character. There are so many other factors affecting the properties of the sinter, however, that it is well to mention them here. The proportion and character of the raw materials, that is, raw ores, concentrates, returns, and fuel, as well as the mixing and the water content, all have a marked effect on the physical properties of the product and the degree to which sintering action can be carried on. The process of sintering is a relatively fast operation. In as much as appreciable time is required to carry on processes of fusion in such masses of low thermal conductivity, large lumps of hematite ore frequently remain unfused and partly unchanged in state of oxidation in the sintering process. The kind, the grain size, and the amount of fuel used affect both the completeness of the fluxing reaction and the prevailing atmosphere. The rate of reduction in laboratory tests is not only dependent upon the state of oxidation of the sinter but also upon the sizing and porosity. Atmosphere and temperature affect the state of oxidation of the iron oxide, and the atmosphere alone may determine the ferrous minerals that finally develop. The rate and extent of cooling, the type of coolant, the subsequent handling, and screening all have serious effects upon the type of sinter that eventually enters blast furnace bins. The degree to which actual fusion or fluxing takes place in the sintering operation has a marked effect upon density. A sinter which has been extensively fused by high content of fuel in the batch will no doubt have a higher weight on the bulk basis than one which had a lower fuel content. As high temperatures are required to do this job, the iron oxide under these conditions will be largely magnetite. Sintering at low temperatures to produce larger proportions of hematite means a decrease in the amount of liquid formed and a much more sensitive bonding process. In this case the liquid must be distributed more uniformly and thereby used more efficiently than would be the case where higher temperatures were permitted to prevail more or less indiscriminately. Where coarse ore particles are used in a sinter mix it is not expected that any particles coarser than 1/4-in. can be fused and incorporated in the system to such an extent that the gangue contained within these lumps will have been converted or fused by the sintering process. It is for this reason that coarse ore, returns, or both, in a sinter usually result in a sinter which breaks easily and at the same time may contain some of the original minerals of the lump, such as quartz and hematite. In examination of sinters at Bethlehem Steel Co. minerals such as quartz and corundum have been found, none of which are considered normal associ- ates of wustite or magnetite. Some degree of heterogeneity or lack of equilibrium is not unusual in the sintering process. The differences in specific gravity between hematite and magnetite might be ample reason for poor strength in a not very well sintered mass containing coarse particles of. ore or returns. The shrinkage taking place in a lump of hematite in its conversion to magnetite by temperature and/or atmosphere is appreciable. Sintering of ores as it is carried out is crude chemistry, for the grain size is relatively coarse, the application of heat is certainly not uniform, and the time factor is inadequate for other than partial completion of reactions. Coarse lumps of coke or coal cause local heating around these centers, and fuel which is too fine may result in such slow burning that sufficiently high temperatures are not always obtained. High temperatures are essential to the work required. The Swedish practice of sintering is established on the basis of producing an easily reducible product high in hematite. This is achieved through uniformity of grain size in the sinter mix and close control of the temperature through careful regulation of fuel and sintering rates. This produces a sinter which is very tough in character and which has a high degree of porosity. Although the hematite content is not produced upon cooling by drawing air through the mass, there would be greater possibility of accomplishing this reaction with this type of sinter than is the case in American practice. In the latter, the temperatures are so high that temperature alone converts most of the mass to magnetite. The grains are so coarse in the final product that together with the fluxed condition it would be difficult to reoxidize them to hematite upon cooling. An examination of the iron-oxygen diagram' shows that hematite does not exist above 2651°F. It also shows that there is no liquid in the pure magnetite-hematite system until 2881°F is reached. On the other hand, in the system magnetite-wustite liquids exist at considerably lower temperatures than this. It will be seen, therefore, considering only the iron oxides, that the bonding action obtained in America in sinters comes about through considerable temperature and/or reducing conditions that produce compositions containing even less oxygen than is contained in magnetite or than results from the fusion of silicates. The bonding obtained from the iron oxides is encouraged by the reducing conditions that prevail in the vicinity of the fuel particles in a mass of this sort, where temperatures are above 2600°F. As magnetite and wustite are opaque, they do not lend themselves to petrographic study by transmitted polarized light. The silicates found in sinter and the glass that has not crystallized transmit light and can be studied by these methods in which indices of refraction and other optical properties of anisotropic crystals lead to their definite identification. The index of refraction is the only property that can be measured in glass under the microscope, and this is a clue to its probable approximate composition.

Jan 1, 1954

By R. M. Hays

Prior to use of this model siphon finisher, the only means of evaluating metallurgical performance of the siphon finisher was by pilot plant studies. These are costly and require large quantities of material. A model siphon finisher and a test procedure which permits the evaluation of siphon finisher metallurgical performance has been developed. Metallurgical results between the model siphon finisher and pilot plant or commercial units show fairly good correlation. The upgrading of magnetic taconite concentrates by hydraulic concentration has been practiced for several years. In the 1940's, Braund1 used a multistage upflow classifier to upgrade magnetized magnetic concentrates. In present taconite plants, the SiphonSizer, marketed by Dorr-Oliver, Inc., has been particularly suitable for upgrading magnetic concentrates. When the hydraulic SiphonSizer is used to upgrade the concentrates in the finishing portion of the overall flowscheme, the unit will be called a siphon finisher to distinguish this application from classification. Prior to this work the only means of evaluating the metallurgical performance of the siphon finisher was by pilot plant studies which are costly and require large quantities of material. Therefore, it was desirable to develop a model siphon finisher and a test procedure which would simulate siphon finisher performance and require only small samples of material. This model would permit the evaluation of small samples such as drill core samples and be useful in evaluation of commercial siphon finisher design and performance. THE SIPHON FINISHER The siphon finisher, as shown in Fig. 1, is a tank equipped with a system of water jets which distribute the hydraulic water uniformly across the bottom of the vessel with an energy release for agitation in the teeter bed. The feed to the siphon finisher is intre duced through a feed well in the center of the unit. The silica and middling particles overflow a peripheral weir into the overflow launder. The magnetite fraction is discharged by a siphon or a modulated discharge valve. The siphon or the discharge valve is usually operated automatically by a float assembly which is responsive to the depth of the teeter bed. Various interface sensing devices have also been

Jan 1, 1968

By E. J. Roberts

The theory is developed that the tons ground through a given mesh per day in a wet ball mill is proportional to the percent plus that mesh in contact with the balls and the net power applied to the balls at this point. A grindability test is described. DURING the course of a study of the fundamentals of classification in 1937, the need for a more basic understanding of the action of a ball mill became acute. Unless one knows how classification affects grinding, one cannot hope to effectively improve on classification. The methods of evaluating grinding efficiency that depend on surface developed were studied but soon discarded for two reasons: 1. There was no apparent method which could be generally used to give a reliable figure for the actual new surface developed as a result of grinding. Subsequent papers have not changed this conclusion. 2. The practical evaluation of grinding in the main ore dressing applications was in terms of the percentage retained on a screen which passes 90 to 99 pct of the material and not in terms of surface area. The Probability Theory With the background of our experience in the field of closed-circuit grinding, together with the papers of Lennox,1 Gow,2 Gaudin,8 Fahrenwald,4 Coghill, and others, the approach of the theoretical physicist was then tried. The thought was somewhat as follows: When one grinds in a ball mill, a given expenditure of power leads either to a certain number of point to point blows per hp-hr or to a certain distance of line contact per hp-hr, depending on whether the action of the balls is considered to be cascading or rolling. It is also assumed that the balls actually come together on each blow or during the roll. Then a volume of slurry will be covered per minute which is some function of the size of the particle being considered (see fig. 1). All particles coarser than this size will be reduced through this size. This volume of slurry contains a certain weight of ore, depending on the percent solids and the density of the solids. If we fix the percent solids and the density of the solids and let w be this certain weight of ore in the volume covered, then, in mathematical terms, what we have just postulated is, w —— 8 hp (a) dt If W is the total weight of ore present in the mill, then we can write. W w/8 hp (b) W dt and if C is the cumulative percent plus the size chosen at the start of the time interval dt, w w c/dt W 8 hp x c (c) wc But wc/100 is the weight plus the size chosen which at 100 wc the close of time dt is finer than that size, and W is the decrease in the percent plus of the whole mass of ore or —dC. Then, —W dC/dt 8 hp x C. (d) In other words, the mesh tons ground through a given size per unit of time is proportional to the hp and the percent plus the mesh. A crude analogy would be to picture a 1-ft-wide steam roller going down the road at 1 ft per sec. If we place one egg on the road per square foot, one egg will be smashed per second. If we place a dozen eggs per square foot, a dozen eggs will be crushed per second. Similarly, if all the particles in w are plus the mesh, i.e., C=100, we should have a maximum rate of reduction. If only 10 pct of them are plus the mesh (C=10), we would have only one tenth the maximum rate; if only 1 pct are plus the mesh, the balls have a hard time finding anything to work on. This is where the term "probability theory" comes from. The chances of the balls crushing a particle through a given mesh depends directly on the concentration of particles coarser than this mesh in the general pulp in the mill. Giving W the units of tons and dividing equation (d) through by W, we obtain -dC hp ----- = k---— C [1] dt ton where k is a constant for any one size of particle, density of solid and moisture content of pulp. Eq 1 is the rate equation for a first order reaction and says that the rate of decrease of the percent plus a given mesh with time is directly proportional to the hp per ton applied to the body of ore and to the percent plus the mesh in the ore mass as a whole. Since it is a differential equation, it only

Jan 1, 1951

By E. J. Roberts

The theory is developed that the tons ground through a given mesh per day in a wet ball mill is proportional to the percent plus that mesh in contact with the balls and the net power applied to the balls at this point. A grindability test is described. DURING the course of a study of the fundamentals of classification in 1937, the need for a more basic understanding of the action of a ball mill became acute. Unless one knows how classification affects grinding, one cannot hope to effectively improve on classification. The methods of evaluating grinding efficiency that depend on surface developed were studied but soon discarded for two reasons: 1. There was no apparent method which could be generally used to give a reliable figure for the actual new surface developed as a result of grinding. Subsequent papers have not changed this conclusion. 2. The practical evaluation of grinding in the main ore dressing applications was in terms of the percentage retained on a screen which passes 90 to 99 pct of the material and not in terms of surface area. The Probability Theory With the background of our experience in the field of closed-circuit grinding, together with the papers of Lennox,1 Gow,2 Gaudin,8 Fahrenwald,4 Coghill, and others, the approach of the theoretical physicist was then tried. The thought was somewhat as follows: When one grinds in a ball mill, a given expenditure of power leads either to a certain number of point to point blows per hp-hr or to a certain distance of line contact per hp-hr, depending on whether the action of the balls is considered to be cascading or rolling. It is also assumed that the balls actually come together on each blow or during the roll. Then a volume of slurry will be covered per minute which is some function of the size of the particle being considered (see fig. 1). All particles coarser than this size will be reduced through this size. This volume of slurry contains a certain weight of ore, depending on the percent solids and the density of the solids. If we fix the percent solids and the density of the solids and let w be this certain weight of ore in the volume covered, then, in mathematical terms, what we have just postulated is, w —— 8 hp (a) dt If W is the total weight of ore present in the mill, then we can write. W w/8 hp (b) W dt and if C is the cumulative percent plus the size chosen at the start of the time interval dt, w w c/dt W 8 hp x c (c) wc But wc/100 is the weight plus the size chosen which at 100 wc the close of time dt is finer than that size, and W is the decrease in the percent plus of the whole mass of ore or —dC. Then, —W dC/dt 8 hp x C. (d) In other words, the mesh tons ground through a given size per unit of time is proportional to the hp and the percent plus the mesh. A crude analogy would be to picture a 1-ft-wide steam roller going down the road at 1 ft per sec. If we place one egg on the road per square foot, one egg will be smashed per second. If we place a dozen eggs per square foot, a dozen eggs will be crushed per second. Similarly, if all the particles in w are plus the mesh, i.e., C=100, we should have a maximum rate of reduction. If only 10 pct of them are plus the mesh (C=10), we would have only one tenth the maximum rate; if only 1 pct are plus the mesh, the balls have a hard time finding anything to work on. This is where the term "probability theory" comes from. The chances of the balls crushing a particle through a given mesh depends directly on the concentration of particles coarser than this mesh in the general pulp in the mill. Giving W the units of tons and dividing equation (d) through by W, we obtain -dC hp ----- = k---— C [1] dt ton where k is a constant for any one size of particle, density of solid and moisture content of pulp. Eq 1 is the rate equation for a first order reaction and says that the rate of decrease of the percent plus a given mesh with time is directly proportional to the hp per ton applied to the body of ore and to the percent plus the mesh in the ore mass as a whole. Since it is a differential equation, it only

Jan 1, 1951

By P. Somasundaran

Flotation of minerals has usually been discussed in terms of solid-liquid interfacial phenomena. This paper discusses the relative importance of phenomena such as collector adsorption at other interfaces formed between the bubble, liquid and solid. Adsorption of collector ions of various chain lengths at liquid-gas and solid-gas interfaces has been obtained. The data have been analyzed under conditions corresponding to those of flotation to show the significant role of collector adsorbed on the bubble surface in determining the bubble-mineral attachment required for flotation, and consequently to suggest consideration of this role in the development of better flotation conditions. Vacuum flotation of quartz with alkyl ammonium acetate reported in the foregoing paper1 showed that the collector chain length has an effect on flotation even at concentrations below those necessary for hemi-micelle association of the hydrocarbon chains at a solid-liquid interface. This chain length effect might be explained by the following arguments: Hydrocarbon chains find it energetically favorable to be in a medium of low dielectric constant. Considering the possibility of an iceberg structure at the solid-liquid interface, such a structure would produce a region of lower dielectric constant at the solid-liquid interface than in the liquid phase, because the dielectric constant of ice is 2 to 3,2 while that of water is 80. Depending on chain length, this difference will certainly cause even individual ions to adsorb in higher quantities at the solid-liquid interface. Another mechanism to consider is the possible formation of dimers, trimers, etc and chain length would influence this formation. But, above all, the chain length effect on flotation at concentrations of low adsorption at the solid-liquid interface could be due to the significant but often-neglected role of the collector adsorption at the bub- ble surface. Collector ions or molecules are adsorbed both at the solid-liquid interface and the liquid-gas interface. Below the concentration for hemi-micelle association at the Stern plane adjacent to the surface, the bulk of the collector ions adsorbed at the solid-liquid interface are in the diffuse part of the double layer. If such systems, where there is a gas bubble coming in contact with the solid surface, are to achieve some degree of electroneutrality, those collector ions would have to migrate from the diffuse part of the double layer to the region directly adjacent to the surface during the induction period when the bubble is making contact with the solid. The time necessary for this migration would contribute to the induction period required for creation of the solid-gas interface. It seems reasonable that the induction period could be reduced sufficiently for flotation to occur if these collector ions were carried to the surface by the bubble. In an attempt to ascertain whether or not the bubble could possibly have sufficient collector ions at its surface to meet the requirement mentioned above, the adsorption density of the collector ions at the liquid-air interface was determined. For this purpose, the surface tension of alkyl ammonium acetate solutions was determined as a function of concentration. Adsorption densities of the collector at the liquid-gas interface were calculated using the Gibbs adsorption equation: where y is the interfacial tension and ri and pi are the adsorption density and the chemical potential, respectively, of the chemical species, i. Using the convention that the adsorption density of water at the interface is zero, the following relationship is obtained for the liquid-gas interface where lg refers to liquid-gas interface and RA+ refers to the alkyl ammonium ions. For dilute solutions, rf^-wri^&rfwW [3] where C is the concentration of the collector ions in solution. l?:: can now be evaluated using Eq. 3 and surface tension measurements.

Jan 1, 1969

By H. Eklund, R. T. Hukki

This article presents a simple arithmetic derivation for the relationship between sharpness of classification and circulating load, as well as world-wide industrial data which are compatible with the derived equation. It is shown that the sharpness of classification decreases with increasing circulating load. Based on the evidence available a comparison is presented between the performance of the present-day hydraulic classifiers. In a continuously operated closed circuit grinding system the tonnage and the fineness of the components may be indicated as follows: Eq. 8 expresses the relationship between sharpness of classification and circulating load in closed grinding circuits. If CL = 0, E = 100% according to the definition used. This, however, is a purely theoretical case because any apparatus to be called a classifier must produce the conventional two products. Furthermore, if CL = m, E =0%. In order to visualize the meaning of Eq. 8, the effect of factors d (fineness or percentage of under-size of classifier overflow at size x) and b (fineness or percentage of undersize of classifier sands at size x) should be studied. By the most rigid standard, factor d should be given the value of 100%. By so doing the size of classification is defined as that screen size x through which 100% of the classifier

Jan 1, 1965

By L. Weyher, H. L. Lovell

This discussion is restricted to a very specific application of the cyclone - its use as a hydro-cyclone in the cleaning of fine coal. It is hoped that the development of the present data will assist in the further elucidation of existing qualitative models, and also in an enhanced application of the hydrocyclone and provide a base for further research. Discussed are the mechanisms of gravity separation in hydro-cyclones, test procedures, experimental observations, and industrial applications from these studies. Within the general area of unit operations, a considerable effort has been expended in theoretical and empirical research associated with cyclone operations. The motivation involves the industrial requirement for devices and techniques to manipulate small particles at low operating costs, a situation required in coal preparation. There are good reasons why the cyclone helps to satisfy this demand, because it combines operational simplicity (no moving parts) with versatility and high throughput capacity per unit area. The cyclone is one of the most widely applied tools in the process technologies. There are areas for potential cyclone application which have been only meagerly explored or not considered at all. This discussion is restricted to a very specific application of the cyclone - its use as a hydrocyclone in the cleaning of fine coal. It represents an initial presentation of a more comprehensive report. Coal preparation can appropriately claim the hydrocyclone since most of its predominant industrial uses have originated in this area and have spread into other applications, including the broad field of ore dressing, the petroleum industry, chemical processing, nuclear engineering and food technology. There is one operational principle common in all cyclones: A fluid is tangentially introduced into a cylindrical or conical container under pressure. A predominantly irrotational flow field1 is established within the cyclone (Fig. 1). The uniqueness of this flow is indicated by the variation in tangential velocity across the cyclone radius, which reaches a maximum a short distance from the center and decays toward the cyclone wall. In contrast, the tangential velocity in a centrifuge increases uniformly toward the wall. Usually two products are discharged from the cyclone. They differ in that one discharge consists predominantly of that portion of the feed fluid which moves along the wall axially away from the feed inlet, while the other consists predominantly of fluid spiralling along the central regions of the cyclone, in an axial direction opposite to that of the wall current. Fig. 2 indicates these discharge points schematically. The latter product may be termed the vortex-product, since it is discharged with the central vortex-current. In the case of conical cyclones, the former product may be referred to as the apex-product. Although the names "over- and underflow" are widely used, many cyclones are operated from extreme angular positions thus the proposed terms are less ambiguous. The cyclone feed fluid can be laden with solid particles or nonmiscible fluid drops. Such a feed component is discharged from the cyclone, after some residence time, into the one or the other product. Yet, the orifice through which a particle will leave the cyclone is dependent upon a number of factors. These factors, in a statistical sense, determine the path of the particle within the cyclone. Thus, different separation phenomena are possible and different process results attainable by the controllable variation of the particle path. It is then, to attain a fuller understanding of this particle path, to which this study is ultimately directed. It is anticipated that the detailed study of the data being developed will allow further elucidation of existing qualitative models. However, the development of a quantitative description of cyclone operation in terms of a mechanism is not currently feasible. The flow conditions in a cyclone are not fully understood even for a pure fluid, without the added complexity of the inclusion of multiple size-density particles. It is hoped that the development of the present data will also assist in an enhanced application of the hydro-cyclone and provide a base for further research. Some practical applications drawn from the data are suggested. Cyclone operations have been classified according

Jan 1, 1967

By D. A. Elgillani, M. C. Fuerstenau, M. C. Kuhn

The species responsible for flotation when xanthate is added as collector for pyrite is dixanthogen. Measured oxidation potentials of these systems show that xanthate should be oxidized to dixanthogen, and the presence of dixanthogen on the pyrite surface is confirmed by infrared absorption. Zeta potential measurements are presented and related to collector adsorption. In spite of the normal presence of pyrite in most sulfide ore deposits, only a modest amount of study has been devoted to the xanthate-pyrite flotation system in the absence of activators and depressant~.'-~ As a result, little is known about the mechanism of xanthate adsorption on pyrite. From the few studies that have been made, however, a number of significant observations have been forthcoming. For example, the dependence of flotation response on the state of oxidation of the surface was noted in two studies — one conducted by Plante and Sutherland4 and a later one by plaksin.~ In another study, Gaudin and wilkinson3 found that ferric xanthate was not present on a pyrite surface. In a later investigation, however, Gaudin' and his colleagues did extract a small amount of ferric xanthate into benzene when pyrite was ground in an evacuated mill with water, benzene, and xanthate. This quantity was apparently

Jan 1, 1969

By C. C. Martin, M. C. Fuerstenau, R. B. Bhappu

Experiments revealed that quartz could not be floated in conductivity water at any pH with a long-chained sulfonate as collector. Various cations, Fe+++, Al+++, Pb++, Mn++, Mg++, Ca++, are shown to function as activators when the pH is such that the cation hydrolyzes. As the solubility product of the various cation-sulfonates was exceeded at the concentrations involved, precipitated cation-hydroxy-sulfonates, e.g. Fe (RSO3)2OH must be functioning as the collector in these systems. A number of interesting and important phenomena were revealed during a study of the response of beryl to sulfonate flotation.' Perhaps the most important observation is the effect of various cations on flotation response. Experiments showed that in the vicinity of pH 3.0, about three ppm Fe+++ increased recovery of leached beryl from about 60 pct to 100 pct, whereas about four ppm Ca+ + reduced the recovery from 60 pct to 2 pct. With the exception of ferric iron, all of the other cations investigated reduced flotation recoveries drastically at this pH. As postulated in the paper Sulfonate Flotation of Beryl,' even with low concentrations of polyvalent cations and sulfonate (on the order of one ppm), cation-sulfonates are precipitated immediately in solution. That is, blueish-white clouds were noted to form after given additions of salts to known solutions of sulfonate. Furthermore, it was also shown that one-hundred-fold more ferric iron than necessary to form the precipitated cloud had to be added to effect complete flotation. In other words, some form of solid ferric sulfonate was found to function as the collector in these systems. The form of the compound was suggested to be Fe(RCO3)2OH and that bonding to the surface occurs through the hydrogen of the hy-droxyl. As the form of this compound was shown to be dependent on the extent of hydrolysis of ferric iron, it is to be expected that ferric iron will be unique in some pH region. Further, if the presence of hydroxyl in the neutral precipitated compound is of primary importance, then other cations should function similarly but at different values of pH, since they will hydrolyze at different concentrations of hydroxyl ion. To determine whether the premise of hydrolysis is correct, the study was extended to quartz systems in the presence of various cations at various values of pH. EXPERIMENTAL MATERIALS AND METHOD Sodium alkyl aryl sulfonate was chosen as the collector for the experimental work. The reagent, sup plied as a courtesy sample by the Shell Chemical Co., has the following physical properties: Physical form Solid — Finely ground Sulfonate content, pct wt 95 to 97 Molecular Weight 450 to 470 Number of carbon atoms in hydrocarbon chain 25 to 30 Melting Point, OF 250 to 260 Solubility Soluble in water with gel formation at concentrations higher than 25 pct by weight sulfonate This particular reagent was chosen because of its high active sulfonate content, long-chain length, and solid form. With the exception of the frother and sulfonate, all other chemicals were of reagent grade quality. Conductivity water was used in the experimental work. This water, made by passing distilled water through an ion exchange column, had an average measured conductivity of one micromho. Experiments were conducted at room temperature in a small glass flotation cell with the following procedure: 1) A redetermined amount of water and salt solution (e.g. CaC12 . 2H2O) were combined and the pH adjusted to a given value. 2) A given amount of sulfonate was added so that final solution volume was 130 cc. 3) One drop of Dow 250 frother was added. 4) Five g of quartz were added and the system conditioned for five minutes. 5) The pH of the system was measured (termed flotation pH).

Jan 1, 1963

By L. D. Keller

The application of hydraulic concentration to large scale operations has been effectively accomplished by the use of the SiphonSizer. Operation has been simple, automatic and essentially trouble free. The use of the primary desliming SiphonSizer on magnetic taconites has shown an improvement in grade of concentrate and an increase in recovery over conventional hydroseparators. Magnetic concentrates have been further deslimed and upgraded in SiphonSizers by floating a coarse middling product away from the heavier concentrate. The application of the Siphon-sizer to closed circuit grinding has met with only partial success to date. Hydraulic concentration of minerals has been practiced for a good many years. Until recently, however, the practice has been limited to equipment using a multiplicity of small pockets and relatively low tonnages and flows. The adaptation of the SiphonSizer to the concentration and classification of magnetite has changed this concept. Some fifty commercial units varying in size from a 16-ft diam circular unit to a 25x100-ft rectangular unit are currently in operation on magnetic taconites. Additional units up to 38 ft diam are under construction. Depending upon the application in the flowsheet, upgrading varies from 0.75% to 14% Fe in a single stage. To those not familiar with taconite processing, the importance of the struggle to remove economically as much silica as possible may not be fully appreciated. Perhaps some figures pertaining to transportation costs will underscore this important process. If the silica content of the concentrate from a 10,000,000 ton per year plant could be reduced by I%, a savings in boat freight, at $2.00 per ton, would amount to $200,000 annually! THE SlPHONSlZER The SiphonSizer, as illustrated in Figs. lA, 1B and 2, is a hydraulic classifier in which the hydraulic water is distributed uniformly across the bottom of the vessel by means of a system of headers (1 and 2), jet pipes (3), and nozzles (4). The depth of the teeter bed is automatically maintained by the siphon-float assembly. The feed is introduced through a feed well (5) and separated into two fractions by the effects of rising water and the density of the teetered bed. The magnetite fraction is discharged by the siphon (12) while the fine slime and light middling overflow a peripheral weir (9) to the overflow launder (6). The bed depth is controlled automatically by float valves (7), which are responsive to changes in the height of clear water supported in the superelevation tubes (8) by the density of the teetered bed. On top of the float (7) is a needle valve (10) which permits

Jan 1, 1963

By R. R. Kimpel, L. G. Austin

An equation is derived describing fragment size for single fracture of brittle solids. It contains the equations of earlier workers as special or approximate cases and, in addition, is capable of describing coarse distributions in which a few large particles are produced on fracture. The equation can be fitted well to single fracture experiments where over-breakage is avoided. In 1956 Broadbent and callcottl described a method of analyzing the unit operation of grinding, applicable to homogeneous breakage of brittle materials. In essence, the grinding process was split into two parts: 1) the fractional amount of a given size which is selected for grinding per unit time, 2) the size-weight distribution of the fragments produced by breakage of this certain size (primary breakage distribution function). The theory was formulated using matrix algebra. They attempted an experimental verification of the theory using an assumed primary breakage distribution but the results were not very satisfactory. Austin and Gardner2 showed that this assumed distribution was widely different from experimental values and that it led to the poor agreement of predicted size-weight distributions with experiment found by Broadbent and Callcott. In addition, Austin and Gardner2,3,4 reformulated the original approach in terms of differential equations of grinding and demonstrated that the approach led to a very satisfactory description of the grinding process, providing that the descriptive parameters were measured experimentally. Recent work5 has substantiated these findings and the distribution functions of several materials ground under different conditions in different mills have been determined experimentally. The fraction underweight versus size used by

Jan 1, 1965

By G. L. Tufford, R. L. Bleifuss

Low-grade manganiferous iron ores of Minnesota's Cuyuna Range, in general, do not respond to conventional mineral dressing techniques because of their fine-grained texture. Reducing these ores to magnetite with subsequent magnetic separation and flotation have not given satisfactory results because, in many instances, concentrate grade and manganese recovery are low. Metallization of iron and concurrent reduction of higher manganese oxides to MnO, as in the R-N process, followed by recovery of iron by magnetic separation and manganese by leaching, seems to be a technically feasible process. High-temperature x-ray diffraction studies show that solid-state reactions between iron and manganese, with the formation of intermediate iron-manganese oxides, occur during reduction and affect recovery of both iron and manganese from the final product. Solid state reactions also occur between FeO, MnO, and SiO, at temperatures as low as 900°C. The kinetics of these reactions are such that they become an important factor in the treatment of ores by the R-N process. These investigations suggest a temperature and gas composition cycle that would maximize both iron and manganese recovery. During the past ten years, iron ore shipments from the Cuyuna Range of Minnesota have decreased sharp1y. Because of adverse grade, structure, and cost, many Cuyuna ores are no longer competitive with high-grade foreign ores and domestic pellets. The low-grade manganiferous iron ores from the Cuyuna Range, in general, do not respond to conventional mineral dressing techniques because of their fine-grained texture. Reducing these ores to magnetite for subsequent magnetic separation and flotation has not given satisfactory results because concentrate grade and manganese recovery are often low. However, a technically feasible process appears to be metallization of the iron and concurrent reduction of manganese oxides to MnO with recovery of metallic iron by magnetic separation and extraction of manganese from the tailings by leaching. In 1962 the R-N Corp., with a grant from the Area Redevelopment Administration, conducted an amenability study of the metallization of manganiferous iron ore from the Cuyuna Range. The results of this preliminary investigation were sufficient to interest the Economic Development Administration in sponsoring further studies of the process with emphasis on secondary recovery of manganese from the nonmagnetic tailings. Inasmuch as the metallization process has considerable influence on the final form of the manganese - its texture and distribution between the magnetic concentrate and nonmagnetic tailings-the investigation included a study of metallization of the ore as well. Therefore, the investigation conducted at the Mines Experiment Station covered three interrelated phases: (1) laboratory R-N roasting studies and magnetic separation tests, (2) laboratory chemical extraction experiments of the nonmagnetic tailings, and (3) a mineralogical study of the crude ores and their various roast products. High-temperature x-ray diffraction investigation of the Fe-Mn-Si0,-0, system is the main topic of this paper and an extension of the mineralogical work. GENERAL GEOLOGY The Cuyuna Range is known for its complicated geologic structure, complex orebodies, and generally indifferent response to attempts at conventional beneficiation. Several geologists have made intensive studies of the Cuyuna Range, including Leithl (1907), ~darns' (1910), Zapffe3 (1925), Grout and Wolff4 (1955), and schmidt5 (1963), whose study was the most comprehensive of all. The Range is covered with glacial deposits - till, out-wash, and lake sediments -up to 250 ft thick. Bedrock stratigraphic units of the Cuyuna Range consist of three tightly folded Precambrian sedimentary formations. The lower unit is the Mahnomen formation, composed of 2000 or more feet of argillite, slate, siltstone, and quartzite. The middle unit is the Trommald formation, which varies from 45 to 500 ft in thickness and comprises the major iron-bearing formation. The upper unit is the Rabbit Lake formation, consisting of over 2000 it of argillite, slate, and interbedded lenses of lean iron formation. In the central part of the Cuyuna Range, the units have been folded into a series of

Jan 1, 1969

By T. C. King

AT the new Graham-Central Mill of Eagle-Picher, near Galena, Ill., material is simultaneously dewatered and introduced into the ball-mill scoop boxes by the use of variable-speed, 24-in. spiral classifiers. This arrangement has proved satisfactory in that the moisture content of the wet feed is reduced, thus allowing maintenance of density within the mills, and a constant feed at any desired rate to the mills is provided. In this mill the primary feed is wet crushed to minus 1/2 in. and preconcentrated by jigs to produce a bulk lead-zinc-iron concentrate and a waste tailing. The jig concentrate is then dewatered by a conventional dewatering classifier and stored in bins ahead of the ball mills. Even though this material is well dewatered, the discharge at the bin bottom is sloppy because of the gradual drainage of the water from the surface of the particles. Conventional belt feeders with vertical cutoff gates are unsatisfactory because of the flushing characteristic of the wet sand. The new installation employs 24 in. x 10 ft, 9-in. Simplex double-pitch spiral classifiers for feeders from the bin bottoms to the ball mill scoop boxes (fig. 1). These units are arranged with straight tanks and are driven by 5 hp motors through a U. S. Vari-Drive to provide a 4 to 1 speed ratio. Each bin discharges through a chute into the rising side of the spiral. The flushing characteristic of the material from the bottom of the bin is utilized to maintain a steady feed. A small amount of water tapped into the bottom of the bin insures uniform feed to the classifier. Excess clear water drains over the weirs. Feed rate to the mills is controlled by the variable-speed drives. The feed delivered by the spiral at a given speed setting is relatively dry, uniform, and steady, and as such is easily controlled by the operator. The classifiers are set on a pitch of 5 in. per ft and have a sand-raking capacity of from 8 to 32 tons per hr. It is believed that the solution of this problem offers possibilities for the storage of deslimed rod-mill sands and other similar products of the industry.

Jan 1, 1951

By T. C. King

AT the new Graham-Central Mill of Eagle-Picher, near Galena, Ill., material is simultaneously dewatered and introduced into the ball-mill scoop boxes by the use of variable-speed, 24-in. spiral classifiers. This arrangement has proved satisfactory in that the moisture content of the wet feed is reduced, thus allowing maintenance of density within the mills, and a constant feed at any desired rate to the mills is provided. In this mill the primary feed is wet crushed to minus 1/2 in. and preconcentrated by jigs to produce a bulk lead-zinc-iron concentrate and a waste tailing. The jig concentrate is then dewatered by a conventional dewatering classifier and stored in bins ahead of the ball mills. Even though this material is well dewatered, the discharge at the bin bottom is sloppy because of the gradual drainage of the water from the surface of the particles. Conventional belt feeders with vertical cutoff gates are unsatisfactory because of the flushing characteristic of the wet sand. The new installation employs 24 in. x 10 ft, 9-in. Simplex double-pitch spiral classifiers for feeders from the bin bottoms to the ball mill scoop boxes (fig. 1). These units are arranged with straight tanks and are driven by 5 hp motors through a U. S. Vari-Drive to provide a 4 to 1 speed ratio. Each bin discharges through a chute into the rising side of the spiral. The flushing characteristic of the material from the bottom of the bin is utilized to maintain a steady feed. A small amount of water tapped into the bottom of the bin insures uniform feed to the classifier. Excess clear water drains over the weirs. Feed rate to the mills is controlled by the variable-speed drives. The feed delivered by the spiral at a given speed setting is relatively dry, uniform, and steady, and as such is easily controlled by the operator. The classifiers are set on a pitch of 5 in. per ft and have a sand-raking capacity of from 8 to 32 tons per hr. It is believed that the solution of this problem offers possibilities for the storage of deslimed rod-mill sands and other similar products of the industry.

Jan 1, 1951

By Douglas W. Fuerstenau, Kalanadh V. S. Sastry

A mathematical model is developed for flotation in a countercur-rent column where continuously generated air bubbles rise through a downward flowing pulp. The model is based on the assumption of axially dispressed plug flow of both the liquid and air-bubble phases. An appropriate flotation-rate expression, which takes into account the changing concentration of solid particles on the bubble surface, is formulated and applied to the column flotation process. Although a complete solution of the resulting equations is not possible for the general case, solutions for specific limiting cases have been obtained. From these, it is possible to predict theoretically the dependence of the column performance on certain variables. Specifically, concentration profiles as a function of height in the column have been calculated for various values of the characteristic dimensionless parameters. The effect of axial dispersion on the perforance of the equipment is pointed out. In recent years, there has been considerable progress in delineating the fundamental physicochemical principles underlying the unit operation of flotation and in advancing methods for kinetic analysis and process modeling of flotation.' In general, nearly all such process research has been concerned with the analysis of conventional flotation equipment and circuits. From these investigations, the possibility of developing and designing flotation apparatus for more efficient separation appears limited with conventional equipment. How- ever, recently there has been an attempt to apply coun-tercurrent flow techniques to flotation in the context of a flotation column.- some of the results from pilot-plant studies2 indicate possibilities for success of this process at least in obtaining higher-grade concentrates though not higher recoveries. A flotation column operates on countercurrent principle in which bubbles continuously generated at the bottom of the column rise through the downward flowing slurry. A schematic diagram of the flotation column is shown in Fig. 1. The physicochemical principles of bubble-mineral attachment, using this flotation column, are essentially the same as those which prevail in conventional equipment, the only differences being in the flow patterns of the various phases. The salient operating feature of the column as compared to those of a conventional cell is that neither mechanical nor turbulent pneumatic agitation is employed. The column is

Jan 1, 1971

By F. W. Bowdish

A theoretical and experimental laboratory study has shown that the rate of breaking of oversized particles in a ball mill is proportional to 1) the concentration of such particles, and 2) the area of the balls, provided that the balls are larger than a certain size relative to the particles. If all undersized material were removed as formed, the rate of breaking would be almost independent of the quantity in the mill. Rate equations are given which are similar to those for zero, first, and second order reactions. Deviations of small balls from the equation for a second order phenomena are quantified by a factor called "nipping efficiency." Some data are correlated by a dimension-less equation, which is used to estimate nipping efficiencies from data not in the correlation. The investigation described here was undertaken after a literature survey revealed that little was known concerning the mechanism of breaking of particles within a ball mill and the effect of this mechanism upon grinding rates. One paper, by Roberts1 , offered a theory concerning the mechanism of breaking, and a rate equation for the breaking of particles larger than some size to particles smaller than that size. That paper provided the starting point for a theoretical and experimental investigation of grinding under carefully controlled conditions. THEORETICAL CONSIDERATIONS In his paper on the probability theory of ball mill grinding, Roberts theorized that "... a given expenditure of power leads either to a certain number of point to point blows per bp-hr or to a certain distance of line contact per hp-hr ..." He assumed that the balls touch each other during each blow or during moving contact. Then he said that a volume of pulp, dependent upon the size being considered, would be covered per minute within which all oversized particles would be crushed to particles smaller than that size. Naturally, the higher the concentration of oversized particles in the pulp, the higher the rate of which they would be broken. From this reasoning Roberts obtained the differential equation -dC/dt = k hp/ton C (1) where k is a constant for any size of particle, density of particle, and moisture content of pulp. C is the percentage of oversize in the material to be ground, t is the time, and hp/ton is the power input per ton of material in the mill. He then integrated the equation to log Co/Ct = k/2.3 hp/ton t , where Co and Ct are the percentage of oversize at time zero and t respectively. By always using the same quantity of material in a mill and assuming that it required constant power to run the mill, he was able to combine the horsepower per ton factor with the other constants to give an equation for a first order rate.

Jan 1, 1961

By G. A. Parks, C. Kovacs

The xanthate flotation of chrysocolla can be improved by preheating the ore to 500° to 600°C. Addition of water vapor reduces the preheating temperature required. The temperature range of flotation improvement correlates well with the range over which dehydroxylation occurs. This observation is consistent with changes in the structure and properties of the mineral as detected by measurement of solubility, D.T.A., T.G.A., X-ray diffraction, and infrared absorption and with recently proposed explanations of chrysocolla flotation. A brief experiment by one of the authors several years ago showed that exposure of chrysocolla to high temperature water vapor resulted in increased recovery by subsequent xanthate flotation. Recently renewed interest in flotation and thermal properties of chrysocolla1-5 prompted further study of this effect and more detailed reporting.6 This paper explores the response of pure chrysocolla recovery in Hallimond tube flotation tests to the time and temperature of exposure and to changes in the water vapor pressure. Brief attention is paid to the effects of such treatment on accessory minerals. A single experiment with flotation of chrysocolla from an ore is reported. MATERIALS Chysocolla from Inspiration, Gila County, Ariz. was crushed in a porcelain mortar, cleaned of visible and magnetic impurities, and screened to obtain -48 +65 mesh (Tyler) material for flotation test work. The clean material contained 28.5% Cu and was identified as chrysocolla by comparison of X-ray diffraction, D.T.A., and infrared absorption data with that reported for similar materials by others.4,5,7 Samples of malachite and chalcopyrite, prepared in a similar way, were also identified by X-ray diffraction. The sources of these materials are unknown. A sample of chysocolla-bearing oxidized copper ore from Pima, Ariz. was crushed and dry ground for thermal activation and subsequent flotation. PROCEDURES Activation: The apparatus used for exposure of materials to high temperature water vapor is illustrated in Fig. 1. The furnace temperature was controlled to ±5°C Water vapor pressure was controlled by pre-saturating a nitrogen carrier stream with water in an adjustable thermostatted saturator (multiple bubbler). A water manometer and capillary orifice flow meter were used to measure nitrogen flow rates for activation and for subsequent Hallimond tube tests. Gas flow rate and temperature were adjusted prior to introduction of samples. Exposure time was measured from the moment of introduction of the sample to the furnace, hence includes an initial drop below the control temperature. After heating, the samples were removed from the furnace and stored at room temperature in air. Representative samples of activated materials were examined by X-ray diffraction and infrared absorption.

Jan 1, 1967

By S. C. Lindsay, D. J. I. Evans

With each year that passes hydrometallurgical processes are being more widely used to recover base metals from ores and concentrates. Generally these processes involve liquid-solid separation of metal-bearing liquors from barren residues. This may be done by countercurrent decantation, thickening and filtration, or filtration or centrifuging only. Choice of method is governed by chemical and physical characteristics of the pulps; cost of labor, power, materials, and waste disposal; and even by the amount of capital available when building plans are under way. Since all these controlling factors change with time and location and with technological advances, the separation method must be re-evaluated when new plants or expansion programs are contemplated. At the Sherritt Gordon refinery in Fort Saskatchewan, Alberta, an ammonia pressure leaching process recovers nickel, copper, and cobalt from the nickel concentrate produced at Lynn Lake, Manitoba.' Pregnant liquors are separated from the barren leach residues in a thickener-filter liquid-solids circuit. The decision to use this type of separation was based on a summation of economic factors during the pilot plant and design stages. Although countercurrent decantation was not incorporated in the refinery flowsheet, the problems solved in thickening the leach residues and the experience gained in operating the commercial plant are pertinent to countercurrent decantation, particularly where it is applied to slow-settling leach slurries or slurries having a relatively high vapor pressure of a valuable, obnoxious, or poisonous vapor. The nickel sulfide concentrate is leached under pressure with air and ammonia to dissolve the nickel, copper, and cobalt and most of the sulfur, leaving a residue of iron and siliceous material to be discarded.2,3 During a subsequent distillation stage to recirculate part of the ammonia used for leaching, copper is recovered from the solution as a high-grade sulfide. After further purification to remove traces of copper and undesirable sulfur compounds, the nickel is precipitated by hydrogen under pressure at an elevated temperature. The cobalt, together with unreduced nickel, is precipitated as a mixed metal sulfide by hydrogen sulfide for eventual cobalt recovery. The end solution is then evaporated to produce a crystalline ammonium sulfate byproduct. As shown in Fig. 1, the nickel concentrate is leached in a continuous, two-stage, countercurrent circuit. In the first stage, or adjustment leach, the fresh concentrate is partially leached (60 pct) in the liquor produced in the second stage, or final leach. This system is necessary to produce a solution suitable for treatment in the copper sulfide precipitation step. Both stages are carried out at 170" to 180°F under a total pressure, with air, of 100 to 120 psig in large, horizontal, water-cooled autoclaves. The autoclaves are divided by overflow weirs into four compartments to give a certain amount of staging during leaching. In each compartment a relatively slow-speed (100 to 150 rpm) agitator maintains the nickel concentrate in suspension. Adjustment leaching, carried out in two autoclaves in parallel, is followed by a thickener-filter liquid-solids separation. Final leaching is done in two parallel trains of three autoclaves using aqueous ammonia to extract metal values from the adjustment leach residues. This too is followed by a thickener-filter circuit. Here the residues are repulped once with aqueous ammonia and twice with process water and refiltered after each repulping stage, to insure a maximum recovery of soluble metal, before being discarded to the tailings pond. The adjustment leach is followed by a seven-disk Dorr-Oliver American filter, with an agitator in the filter boot to keep the residues in suspension. Agitators have also been installed beneath the four ten-disk filters following the final leach thickener. Slurry Characteristics: A typical analysis of the Lynn Lake nickel concentrate, produced by differential flotation,' is given in Table I. The concentrate comprises 35 to 49 pct pentlandite, (Fe, Ni)S; 3 to 6 pct chalcopyrite, CuFeS,; 1 to 4 pct pyrite, FeS2; and 24 to 42 pct pyrrhotite, Fe1-2S (where x varies between 0 and 0.2) together with various siliceous minerals such as actinolite, hornblende, and talc. During leaching, the sulfur in the pentlandite, chalcopyrite, and part of the pyrrhotite is successively oxidized through the thiosulfate and poly-thionate radicals to sulfate and sulfamate." Simultaneously the nickel, copper, and cobalt are converted to soluble ammine complexes and the iron in the pentlandite, chalcopyrite, and part of the pyrrhotite is oxidized to a spongy hydrated iron oxide. The hydrated iron oxide is readily broken down by the agitation needed to sustain the concentrate in suspension and to maintain a satisfactory mass transfer of the oxygen from the air to the slurry. Table I shows typical chemical, screen, and

Jan 1, 1961