Search Documents

By W. A. Griffith

Fractional replication of factorial design, a general method for planning experimentation and for analysis of data obtained, is described as applied to a flotation investigation. This method leads to determination of optimum process conditions with minimum experimental effort. Its advantages over simple factorial design are demonstrated. A METHOD for planning experimentation and for analyzing the data secured will be demonstrated. This method, termed fractional replication of factorial design, employs a relatively small number of individual experiments to determine which of a large number of process variables are controlling, to determine which combination of levels of these variables is most likely to produce optimum results, and also to predict what results will be obtained with a particular combination of conditions not already tested. Although the general method is not new, having been developed by Finney in 1945,' the extent to which it can improve the effectiveness of applied research generally has not been recognized by metallurgists. The fractional replication procedure is particularly useful in flotation experimentation and an example from such an investigation will be used in the paper. However, it has equal value in any investigation in which similar experimental difficulties are encountered. In developing a flotation process for a particular mineral separation, the investigator is inevitably confronted with the following difficulties: 1—There are a large number of potentially important process variables. 2—Results of individual experiments are not highly reproducible, due in part to the difficulty in precisely controlling all the variables. 3—Considerable effort is expended in conducting individual experiments. 4—There are practical limits on the number of individual experiments which can be made. In situations of just such a type, modern statistical methods of planning experimentation and analyzing data have their greatest value. Applications of one such technique, called factorial design, to problems of this type have been described by Dorenfeld and others.'-' The simple factorial design is an efficient procedure when the investigator hopes to provide a comprehensive understanding of the effects and interrelationships of a small number of variables over a limited range. In applied research, this is seldom the main objective. Rather, the investigator usually wishes to determine which of the many potentially important variables are in fact controlling, which levels of the controlling variables will provide opti- mum metallurgical results, and what these results will be at optimum conditions. Interest in detailed trends is limited to the controlling variables and to levels in the region of optimum conditions. Simple factorial design has serious deficiencies for such objectives and is not the most efficient method of experimental design. Deficiencies of Factorial Design In a simple factorial design, an experiment must be made at every possible combination of each level of every variable, once these have been chosen and the levels of each to be included have been decided upon. As the number of variables or levels of each increases, the experimental program quickly reaches prohibitive size. For example, an investigation of only four variables, each at four levels, requires 256 individual experiments. Often upon completion of such an extensive program, it is found that one or more of the variables has metallurgically unimportant effects or that a poor estimate has been made as to the appropriate range of levels to be investigated. The result is that only a small proportion of the data obtained falls in the range of real metallurgical interest. Indeed, it frequently can be anticipated that certain combinations of levels of variables will not produce results of interest, but they still must be included so that the essential balance, or orthogonality, of the design will be retained. It may be true that factorial design will provide the greatest amount of information from a given number of experiments, but it does not necessarily follow that it will lead to the desired information with a minimum number of experiments. Much of the information provided may be of little value. Advantages of Fractional Replication The disadvantages of simple factorial design are overcome to a great extent by a modification known as fractional replication. This is a technique for sampling systematically the potential data of a full factorial experiment, that is, the data which would have been obtained if the complete factorially designed program had been completed. Only a fraction of the total array of experiments is made, but the experiments are chosen in such a way that the important advantages of factorial design and the accompanying analysis of variance are retained. The data obtained from the first group of experiments are used to determine which of several variables are controlling and which levels of these variables are most likely to produce the desired result. Unimportant variables and levels of variables then may be

Jan 1, 1956

By W. F. Stowasser

Robert E. Hagen (Oliver Iron Mining Div., U. S. Steel Corp.)—Mr. Stowasser and his associates are to be congratulated on their achievements in developing the design of a pilot agglomerating plant, based on information obtained from experiments made in a laboratory sinter pot, and subsequently constructing and successfully operating this plant. In his Transactions article, May 1955 (vol. 201) Mr. Stowasser stated that the output of his heat treating strand was about 2 Lt per sq ft of grate area per day and that the fuel consumption was 0.7 to 0.8 million Btu per Lt of product. If it is assumed that the burning zone extends from the back of the ignition furnace to the beginning of the recuperation windbox, only about 1/3 of the grate area over the windboxes is utilized for burning pellets, and the actual rate at which partially dried pellets are burned is about 6 Lt per sq ft of grate area per day. It may be interesting to compare the above operating data with that obtained at the Extaca plant, where taconite concentrate is treated on a conventional sin- tering machine using limestone or hydrated lime additives with no recuperation of heat. The entire windbox area is used for burning, and a production rate of 2.5 Lt per sq ft of windbox area per day has been attained at a fuel consumption of 1.8 to 2.0 million Btu per Lt of product, the latter being considerably in excess of that reported by Mr. Stowasser, partly because a circulating load of returns amounting to 28 to 30 pct of the machine feed is used and because no attempt is made to recuperate heat from the waste gases. With the high capital and labor costs prevailing today, it seems that future development of the downdraft grate pelletizing process should be directed toward taking advantage of the high burning rate of predried pellets and increasing capacity, perhaps at the expense of fuel economy, by decreasing the proportion of grate area utilized by ancillary functions. For example, eliminating the cooling zone and using this area for burning would increase the proportion of grate area available for burning from 33 to 50 pct of the total windbox area, and a proportionate increase in capacity would result.

Jan 1, 1956

By R. B. Rosene, G. L. Foster, A. R. Hendrickson

he properties of some of the common fracturing oils, such as sand-falling rate, fluid loss and viscasity, are a function of the three major constituents of the oil. These are: asphaltic material, parafin waxes and a base oil. The base oil can be either high or low viscosity. Asphaltic material increases viscosity by thickening the oil, decreasing the rate of fall of sand through the oil. at temperatures up to 200" F. When large arnorrnts of parafin wax are present, much of the wax i.s not dissolved, but is present in the form of rnicroscopic crystals. 7'his, of course is limited to temperatures below the melting points of waxe.s. As a result, parafin waxes clo not function primarily as thickener.7, but rother as bullting materials. The functiori of wax, without asphaltic material, in a low viscosity base oil is uniquely unlike that of wax with asphalt coexistent in the oil. The resulting difierence in wax crystal size governs the sand-falling rate to a large extent. An ideal type of fracturing oil is one containing neither asphalt nor waxes. In such cases, the viscosity and resulting low sand-falling rate are apparently due to soluble resin- ous hydrocarbons, which are compatible with formation oils. INTRODUCTION Since the introduction of hydraulic fracturing as a method of stimulating production from oil and gas wells, a variety of fluids have been employed to open the fractures and carry the propping agent. Such fluids include gels, emulsions, and oils. In the latter category, the use of refined oils has increased tremendously, and it is probably safe to assume that such oils represent more than one-half of the total fluid used in fracturing. The general acceptance of refined oils for fracturing can be accounted for on the basis of their relatively low cost and general applicability. To minimize transportation costs, such refined oils are generally obtained within the local areas of operation. As a result there are several kinds of oil used in different areas, with a wide variation in the basic composition. These types may vary from a fair grade of lubricating stock to a highly viscous cut-back asphalt. or road oil. Originally, little attention was paid to the composition of the oils used for fracturing, and any oil which would carry sand was considered satisfactory. Recently, however, service companies and well operators have become more concerned with the type of oils used for fracturing. With this in mind, the present work was undertaken to evaluate and compare many widely used fracturing oils to determine which constituents were most desirable for fracturing purposes. Some of the important properties of fracturing fluids are sand-falling rate, viscosity, and fluid loss. Other properties, that are actually of no less importance, include flash point, pour point, emulsifying tendencies with formation brines, and compatability with the formation oils. This discussion has been limited, however, to the first mentioned properties in order to give some idea of the general composition of the refined oils, how they function, and why. PROPERTIES OF FRACTURING OILS The properties of fracturing oils are a function of the three major constituents of the oil. These may be designated as: (1) asphaltic material, (2) paraffin waxes and (3) a base oil, which may be of either high or low viscosity. The asphaltic material clas-

Jan 1, 1956

By G. K. Manning, W. E. Few

WORK on the effect of vacuum heat treatment of 0.040 in. diam sintered and wrought molybdenum wires has been sponsored by the Office of Naval Research at Battelle for the past several years. An electrical resistance furnace' capable of maintaining a vacuum of less than 0.1 micron of mercury at temperatures as high as 4000 °F was used for heat treating the wires. Early in the work, it was found that the introduction of a few microns pressure of oxygen into the furnace resulted in a rapid increase in the oxygen content of the wire specimens and the appearance of a discrete oxide phase. Subsequent elimination of the oxygen atmosphere caused the disappearance of the oxide phase. This was used to determine the solid solubility of oxygen in molybdenum over the temperature range of 2000" to 3000°F.2 It was observed that, with the disappearance of the oxide phase, the wires regained some ductility. Single-phased wires possessed from 5 to as much as 40 pct elongation in 2 in. Later, attention became centered on determining the cause of this wide range in ductility. Wires containing less than 0.002 pct C, less than 0.0007 pct 0, less than 0.0005 pct N, and less than 0.00003 pct H exhibited the erratic ductility noted above. These percentages represent approximately the limit of accuracy of presently available methods of analysis and thus it became apparent that chemical analysis would be of little help in the study. A number of possible explanations for the wide range in ductility were suggested during the course of the work. It was first suspected that variations in the grain size of the recrystallized wire specimens might be responsible. Wire from several lots of material was available. One of these lots tended to develop very large grains when recrystallized, the others all developed an ASTM grain size of 3 to 4 when heated above 2700°F. No correlation between grain size and ductility could be found. It was thought that back diffusion of oil from the diffusion pump might cause contamination of the wire, principally with carbon, but when wires were treated without cooling the cold trap in the pumping system,

Jan 1, 1956

By G. A. Croes, N. Schwarz

This paper reports the results of a series of model displacement experiments carried out for measuring the efficiency of the water-drive process. This series forms a continuation and extension of that described by Engel-herts and Klinkenberg. Detailed information has been obtained on the influence of the oil/water viscosity ratio. The results are represented in the form of a diagram from which both the oil recovery and the water/oil ratio can be easily read off as a function of the total production (oil plus water) for all values of the viscosity ratio M between I and 500. These resu1ts are viewed in the light of the two existing theories on water flooding, viz. that of Buckley and Leverett and that of Dietz. It is shown that if the Buckley-Leverett theory applies, the relation between the relative permeability ratio and saturation, which plays a major role in this theory, (.an be calculated with a high degree of accuracy from the results of the model experiments. Dietz's theory is found to be in agreement with the experimental results in a certain range of circumstances. A discussion of the physical background of the Buckley- Leverett and Dietz theories makes it clear why the latter fails to agree with some of the experimental results. INTRODUCTION For an economical exploitation of oil fields both a rational control of natural (primary) recovery processes and a proper application of artificial (secondary) recovery methods are essential. These require detailed information upon the dependence of the relevant flow processes on the variables by which they are governed. Three methods are available for obtaining this information: (1) (statistical) analysis of actual field data; (2) appropriately scaled model experiments; and (3) mathematical analysis of the physical problem. If sufficient data were available, the first method would, in principle, be the most reliable one, although it has the drawback that the basic relationships between the variables involved are to a large extent obscured by the natural inhomogeneities of most reservoirs. The most profitable procedure seems to be: (a) determining the fundamental relations for idealized reservoirs by methods (2) and (3); (b) checking the consistency of the results obtained by these two methods; and (c) establishing the range of applicability of these methods to actual reservoirs by comparison with the data accumulated under (1). In this paper items (a) and (b) are dealt with for the case of homogeneous loose sands initially saturated

Jan 1, 1956

By P. L. De Bruyn, R. J. Charles

The transfer of kinetic energy of translation into other forms of energy by impact is a fundamental process in most crushing and grinding operations. During and after the impact process the original source energy may be accounted for in any of the following possible forms: 1) Kinetic energy of translation of both the impacted and impacting objects. 2) Kinetic energy of vibration of the components of the impact system. 3) Potential energy as strain energy of the components of the system or in the form of residual stresses. 4) Heat generated by internal friction during plastic deformation or during damping of elastic waves. 5) New surface energy of fractured materials. At any instant during the impact process only the strain energy of the components of the system can contribute directly to the brittle fracture process. If fracture is the desired result, as in comminution, it would seem advantageous to choose or arrange the conditions of impact so that a maximum amount of the original kinetic energy could be converted to strain energy at some moment during a single impact. The present work deals with determination of these desirable conditions for a simple case of impact and application of the principles involved to general cases of impact.

Jan 1, 1956

By A. M. Gaudin, D. W. Fuerstenau

By streaming potential techniques, the zeta potential of quartz has been measured as a function of the concentration of dodecylammonium salts at different pH values. These experiments indicate that adsorption of dodecylammonium ions takes place as individual aminium ions until a certain critical concentration is reached in the double layer, at which point the adsorbed collector ions begin to associate into patches of ions, called hemi-micelles. Because of its effect on the surface potential of quartz, pH controls the adsorption of dodecylammonium ions. The value of cooperating agents and nonionized constituents of flotation systems is discussed. A PREVIOUS investigation' of the role of pH in the flotation of quartz with anionic collectors is here extended to an inquiry of how the cationic collector, dodecylammonium acetate, is adsorbed at the quartz-solution interface. The equilibrium amount of this collector adsorbed by quartz has already been measured as a function of concentration and pH.2,3 The weight of prior evidence has suggested that two mechanisms are involved in the adsorption of this collector by quartz, but with existing adsorption measurements alone it is not possible to determine what the mechanisms are. Gaudin and Bloecher2 postulated that a polymolec-ular adsorption coating begins to form before the first monolayer has been completed. Gaudin and Morrow4 indicated that a similar situation might exist in the film of dodecylammonium ions on hematite. A study of electrokinetic potentials at the quartz-solution interface provides some detail toward a fuller understanding of the adsorption mechanism. Experimental Results The zeta potential at the quartz-solution interface was calculated from streaming potential data measured with the apparatus described in detail in a thesis by Fuerstenau5 and briefly in a paper by Gaudin and Fuerstenau.1 The quartz used in these experiments was a 48/65-mesh fraction prepared in the manner described previously.' Zeta potential of quartz was obtained for various amounts of the amine salt at pH 4, 7, 10, 11. In

Jan 1, 1956

By F. W. Luerssen, J. F. Elliott

F LUSHING the early slag from a stationary open Fhearth having a high percentage of hot metal in its charge is necessary in order to remove silica from the system. The flush slag is strongly oxidizing and is somewhat acidic. It has, however, considerable capacity to extract phosphorus from the bath and it also removes considerable manganese. It seems probable that factors which control the distribution of phosphorus and manganese between slag and metal in the refining period also should be dominant in the flush and postflush periods. Several studies, as summarized elsewhere,1,2 support the viewpoint that conditions closely approaching equilibrium for these elements are rather readily established during the refining period. Over the years these studies have repeatedly demonstrated that 1—high slag v01ume, 2—low bath and slag temperature, 3—basic slag, and 4—strongly oxidizing slag favor rapid elimination of phosphorus from the bath to the slag. They also show that the following conditions favor retention of manganese in the bath: 1—low slag volume, 2—high bath and slag temperature, 3— basic slag, and 4—minimum oxidizing power of slag. When it is considered that the flush slag often carries as high as 75 pct of the manganese charged and only 25 to 60 pct of the phosphorus charged, it is evident that in removing silica much manganese is sacrificed but phosphorus removal is far from conplete. Because of overriding circumstances, this is accepted in most operations and actually it is considered to be inevitable. This may account for the fact that little attention has been paid to conditions affecting the elimination of phosphorus and manganese in the flush slag. A recent study of the behavior of various charge oxides has developed considerable information on the flush and postflush periods. Because the data are felt to be of general interest, they have been brought together and Presented in this paper. The object is to show the various factors in the flush and postflush periods which influence elimination of phosphorus and manganese. Physical Conditions During and After Flushing Physical conditions existing during the flush vary from plant to plant, from shop to shop, from furnace to furnace, and even from heat to heat. They are strongly influenced by the physical and chemical character of the charge oxide which is ordinarily necessary to provide sufficient oxidizing power early in the heat. Invariably the period is characterized by a vigorous reaction between the principal re-actants: the hot metal being added and the charge oxide. During the flush, it is probable that the slag acts to some extent as an oxidizer; but, because of the critical influence of the behavior of the charge oxid'e on flushing action, it seems apparent that the oxide itself is the dominant oxidizer. Fig. 1 shows the course of two heats which were selected as being typical of the group studied. Heat A was charged with 55 pet hot metal, based on the total metallics charged, and heat B had 57 pct hot metal. As indicated in Table I and Fig. 1, the melt-down slag, which is not usually voluminous and which is principally FeO, expands greatly in volume and will show rather high levels of SiO2, MnO, and P2O5 very soon after the beginning of the hot metal addition. Simultaneously, large volumes of CO are liberated which cause violent mixing of slag and metal. It is of interest to note that the time required to bring carbon down to a low level is very much longer than that required for the removal of silicon, manganese, or phosphorus. At the end of flush, carbon in the bath is still approximately 2 pct. When strongly reducing hot metal is brought into contact with strongly oxidizing conditions within the furnace! it is probable that the rate of mass transfer to the slag (and atmosphere) of silicon, manganese, phosphorus, and carbon initially depends principally on the rates at which the two participating phases are brought into contact That is, it depends on the nature of the various reactions. Later in the flush period, when the scrap is virtually all dissolved and the action of the bath has settled down to a steady and somewhat gentle boil, it is likely that other factors, such as the transfer of oxygen across the slag-metal interface, become dominant. The temperature of the slag-metal system is far from uniform. Heat is being driven by the flame down through the slag. Bubbling and surging of the metal also frequently brings portions of the bath in contact with the flame. At areas of contact between the ore and liquid metal, or slag and liquid metal, the oxidizing reactions generate much heat. On the other hand, scrap is being melted which tends to absorb large quantities of heat. Because the liquid bath is high in carbon, the steel scrap is brought into solution rapidly. This can proceed at a rather low temperature; and until much of the scrap has been taken into solution, the bath temperature would not be expected to increase appreciably. Consideration of these factors leads to the conclusion that during the flush period the slag should be rather hot and the bath relatively cold. Both observation and temperature measurements bear this out. Experimental Data The extended program of charge oxide evaluation permitted study of the widely varying conditions existing during the flushing period. Slag and metal analyses and bath temperatures reported herein (Tables I and 11) were obtained toward the latter portion of the work. Four different types of charge oxide, sinter, two types of hydraulic cement-bonded soft ores, and a pyrobonded agglomerate were used in the study. Although the heats reported were from only one 205 ton furnace, they show variations in flush slag analyses all the way from 25 pct FeO, which is typical with the use of a hard natural charge ore, to 45 pct FeO which resulted when a very poorly agglomerated fine ore was used. The physical behavior of the flushes showed a correspondingly wide variation from well controlled reactions to violent surges following periods of inac-

Jan 1, 1956

By W. D. Forgeng, R. L. Folkman, D. C. Hilty

The solubility of oxygen in molten Fe-Cr alloys has been determined at 1550° , 1600°, and 1650°C for alloys containing up to alloyshasbeenabout 50 pct Cr and found to decrease as chromium increases to 6 pet and then to increase gradually. Phase relations in the Fe-Cr-0 system at steelmaking temperatures have been evaluated and two previously unreported oxides have been identified. OPERATIONS in the melting and refining of chromium steels are dependent to a major degree on the reactions of chromium and iron with oxygen and are primarily directed toward controlling and modifying these reactions to secure maximum advantage in the utilization of chromium. Although fairly effective practices have been developed empirically, understanding of the specific reactions is limited. Moreover, the ability of molten steel to dissolve oxygen that subsequently precipitates as oxide inclusions during solidification is well known, so that clarification of the effect of chromium on the nature and mechanism of formation of these inclusions is desirable. Study of the problem indicated that increased knowledge of the fundamental Fe-Cr-0 system is essential to further technical and economic improvement. Consequently, an investigation of the influence of chromium on the solubility of oxygen in molten iron and of phases and phase relations in the Fe-Cr-0 system was undertaken at the Metals Research Laboratories of the Electro Metallurgical Co. as part of a general program on the effective utilization of chromium in chromium steel production. Experimental Procedure All of the experimental runs made during this investigation were carried out in the rotating crucible induction furnace which has been described in detail in a previous publication.' In review, the principle underlying this furnace is that rotational forces in the crucible cause the molten metal to assume a concave shape in such a manner that the slag is contained in the molten metal cup, thereby minimizing slag/crucible contact and subsequent reaction. Practically all of the heats were melted in commercial-quality magnesia crucibles of the type usually furnished for laboratory induction furnaces. All of the runs were made under an argon atmosphere after the furnace chamber had been degassed. Bath temperatures were controlled by Pt—Pt-10 pct Rh thermocouples. The basic furnace charge consisted of a premelted 12 to 15 lb electrolytic iron slug which had been cast to a shape convenient for fitting into the rotating furnace crucible. A typical analysis of the electrolytic iron slugs is given in Table I. For certain runs, it was desirable to melt down charges containing high initial chromium contents. In these cases the chromium was added to the electrolytic iron during the premelting operations. Chromium was added to the rotating furnace bath in the form of electrolytic chromium of the analysis given in Table 11. In all of the constant temperature runs, the standard procedure was to melt down the electrolytic iron or alloy slug under vacuum, admit argon to slightly greater than atmospheric pressure, and adjust rotational speed to develop a well formed "cup" on the bath surface. The melt was then saturated with oxygen by an addition of ferric oxide and the temperature of the bath adjusted to within +5°C of the desired level. At this point the heat was permitted to come to equilibrium prior to sampling or further addition. This time interval was varied from 15 min to 2 hr but generally was about 20 min. After sampling, the desired chromium addition was made and the bath again equilibrated prior to subsequent sampling. This cycle of addition and sampling continued for the duration of the run which lasted from 4 to 8 hr. At the end of this period, power was shut off and the bath allowed to solidify in the furnace. Quenched metal samples were taken from the equilibrated melts by means of the "Taylor sampler" which has been described by Taylor and Chipman.' Before being submitted to chemical analysis, all samples were carefully examined for surface slag occlusions and cold shuts. Later in the study, each sample was radiographed and examined metal-lographically in cross-section to determine the presence of internal defects. On the basis of these examinations, sound and apparently clean specimens were selected from the Taylor samples for chemical analysis.

Jan 1, 1956

By E. S. Greiner

Germanium and silicon have been plastically deformed in torsion at elevated temperatures. Slip took place on the four {Ill} planes. Dislocations, revealed by etch pits on a (111) face, occurred in rows that were traces of operative slip planes. THE plastic deformation of germanium and silicon by bending and tension at elevated temperatures has been reported by Gallagher.&apos; Transla-tional slip during plastic flow in these materials, according to Gallagher&apos; and Treuting,&apos; occurs on (111) planes. Recent work by Treuting&apos; indicates the direction of slip to be <1l0>. The present paper describes the plastic flow characteristics of germanium and silicon deformed by torsion. Torsional deformation was accomplished by twisting 1/4x1/4x2 in. specimens held in molybdenum grips and contained in an atmosphere of helium in a transparent quartz tube and heated by an external resistance furnace. The angular displacements were measured on a graduated circle with a pointer attached to the movable grip. Temperatures were measured with a chromel-alumel couple; the hot junction was within the quartz tube and in close proximity to the specimen. Photographs of samples of germanium and silicon twisted at 850" and 1180°C, respectively, are shown in Fig. 1. The germanium was twisted 230" per in. and the silicon 80" per in. These deformations are larger than those used in experiments to be described later and are presented only to emphasize the extensive plasticity of these materials. Germanium A single crystal specimen of germanium (n-type, 7 ohm-cm resistivity) with a (111) plane normal within 3" of the torsion axis, Fig. 2, was chemically polished with CP-4 solution" and then twisted 9" during the twisting of these specimens of germanium, the slip lines from the (lli) plane should have appeared in the four lateral faces instead of in only two. Perhaps the absence of rotational slip is brought about by the square cross-section. A cylindrical specimen would have a more uniform stress distribution in torsion than one with a square cross-section, and may deform by rotational slip. A specimen of germanium (n-type, 5 ohm-cm resistivity) twisted less than 8" per in. at 550°C, was cut to expose the (111) face 19" 28&apos; from the torsion axis, Fig. 4a. This face, after metallographic

Jan 1, 1956

By W. Rostoker, S. Spachner

A N alloy of 50 pct Ni-50 pct Fe is currently manu- factured in tape thickness down to 1/8 mil. In the annealed form, this alloy produces an almost square hysteresis loop. This has generally been related to the fact that by proper rolling and annealing practice a sharp cube, (100) [OOl], texture can be developed. As part of a program directed toward an understanding of the relationships between magnetic properties and structure of tape core materials, analyses were made of preferred orientations in tapes of various thicknesses. Annealed tapes of this alloy in 2, 1, 1/2, ¼, and Y&apos; mil gages were acquired from commercial sources. Pole figures constructed for the 2 and 1 mil tapes showed the expected sharp cube texture. However, the ½, ¼, and ? mil annealed tapes showed no evidence of any preferred orientations. A comparison of two transmission patterns is shown in Fig. 1. The origin of the absence of texture in the tapes of less than 1 mil thickness was explored in some detail. The basic questions were whether the heat treatment for the thinner tapes was wrongly selected, whether the initial cold worked textures were different, or whether the actual thickness of the tape exercised some influence on its annealing characteristics. Quantities of the five gages of tape were acquired in the as-rolled condition just prior to the final anneal. Pole figures of the cold worked textures were constructed for each of the tapes. Eye estimates of in-

Jan 1, 1956

By Milton Glicken

THE airborne geophysicist is a busy man these days. In his plane he may have the airborne magnetometer, the airborne scintillation counter, and the airborne electromagnetic surveying system. Each of these is an independent tool, but all require additional auxiliary equipment for locating the aircraft in space: recording altimeters and Shoran or aerial cameras. Now there is still another piece of equipment, the airborne magnetic gradiometer, an accessory to the magnetometer. To understand its uses, consider the function of the magnetometer itself. Aside from detecting magnetic ore, the airborne magnetometer finds greatest use in spotting intrusions of igneous material. Where there is enough contrast in magnetic susceptibility of igneous rock and adjacent formations, it outlines the intrusion. Certain minerals also influence the magnetometer directly, but with the exception of magnetite and possibly one or two others, their effect is weak and can be detected only when there is sufficient ore and the magnetometer flight passes very close to it. An igneous intrusion of infinite depth with vertical sides is represented on a magnetometer record by an anomaly, as in Fig. 1. Amplitude of the high depends on susceptibility contrast of the igneous rock. Generally speaking, the edge of the intrusion lies below the point of inflection of the curve, and this point, where the curvature changes from positive to negative on the magnetometer profile, would be near A in Fig. 1, with a counterpart, of course, on the other side. Location of the contact is one of the principal objects of the survey, but finding the precise point is not always easy, as inspection of the curve near A will show. Mineralization is often found at the contact zones, as at B. Magnetic effects, if detected, may be small, as in B&apos;, and when superimposed on the anomaly due to the instrusion they are very difficult to discern and analyze. Furthermore, if these small fluctuations are to be perceived by the magnetometer the vertical scale should be large. This increases the slopes of the anomaly and makes detection of small deviations and inflection points even more difficult. The airborne magnetic gradiometer was designed to help overcome these difficulties. What it presents is the first derivative of the magnetometer record with respect to time, that is to say, the slope at any point. Fig. 2 represents an actual magnetometer record (solid line) with the corresponding gradiometer record (dashed line) superimposed. Both records read from right to left. Vertical lines on the original magnetometer record are automatic steps designed to keep the pen from going off scale. The slope of any curve is greatest at the point of inflection or point where the curvature changes sign, and this point is a maximum (or minimum) on the gradi- ometer. The chief advantage of the gradiometer is that maxima or minima are much easier to see and to locate precisely; hence an accurate location for the point of inflection can easily be found. Note that points C and D are more sharply defined than C and D&apos;. Similarly the small fluctuations of the original record, so important to the interpreter, are far more clearly shown at E, F, and G, than on the original record at E&apos;, F&apos;, and G&apos;. Though not necessarily highs and lows on the gradiometer, they do show up clearly what would take a painstaking analysis to detect on the original magnetometer record. Will the gradiometer have a particular configuration which indicates an orebody? Not necessarily. The total intensity curve, or original magnetometer record, can display an orebody in various ways, depending on dimensions, orientation, latitude, and composition, as well as on direction, flight height, and instrumental sensitivity of the traverse. Where the total intensity can take on so many different shapes the gradiometer must vary too. It is generally recognized that interpretation of total intensity magnetometer records requires an expert analysis; the gradiometer can be of considerable assistance to the expert but it does not replace him. Mechanism of the gradiometer is simple. A Leeds & Northrup recorder in the aircraft records the magnetic gradient simultaneously with the total intensity, which is on another recorder. Fiducial marks are put on both records simultaneously and the speed of the paper through the recorders is kept the same on both. This makes it possible to place one record over the other for direct comparison. In the laboratory the flights are positioned on a map. Maximum and minimum points on the gradiometer, which can then be posted on the map at their proper locations, may be expected to fall along a trend crossing the direction of flight. Trends should indicate the edge of an intrusion, or some other important features, and when superimposed on the total intensity contour map help greatly to locate the points of inflection, or line of zero curvature.

Jan 1, 1956

By L. D. Jaffe

IN 1945-1946, the author measured temper brittle-ness in ingots of varying composition prepared by remelting SAE 3135 bar stock under vacuum. Since other investigators1-" have been referring to this hitherto unpublished work, belated publication seems worthwhile. Steel in 7 lb batches&apos; was induction melted with nickel additions in a magnesia crucible under a pressure of 10 microns of Hg or less and cast into a 2 in. diam chill mold. The compositions of ingots A, B, and C prepared in this way are shown in Table I; the vacuum melting lowered carbon, manganese, sulphur, and nitrogen content. Ingot B was homogenized 24 hr at 1095°C; ingots A and C were not homogenized. Blanks, M in. square, were cut from each ingot, austenitized 1 hr at 870°C, water quenched, tempered 1 hr at 595"C, and water quenched. Half the blanks from each ingot were given an embrittling treatment of 50 hr at 455°C and water quenched. A few specimens, embrittled and unembrittled, were finally heated 1 hr at 580°C and water quenched. Charpy specimens machined from the blanks were V-notched on the side closest to ingot mid-radius and paired as to radial position in the ingot, one of each pair having received the embrittling treatment and the other not. Specimens were broken in a standard impact machine at various temperatures, the same temperature, in general, being used for both bars of a pair. Fig. 1 shows the results. It is evident that temper brittleness, as measured by the increase in temperature of transition from brittle to tough failure introduced by the 455°C treatment, was decreased but not eliminated by reducing carbon to 0.006 pct, manganese to < 0.004 pct, and nitrogen to 0.001 pct. In another study," temper brittleness was apparently eliminated by -reducing carbon to 0.003 pct, with 0.80 pct Mn and 0.0015 pct N (ingot D)." • Analyses given for carbon below 0.02 pct were made by Jensen method; for nitrogen, by Kjehldahl method. Values quoted else-where&apos;-&apos;." for carbon content of ingot C and for nitrogen content of ingot D were obtained by conventional combustlon and by vacuum fusion methods, respectively, and are considered much less accurate. The 580°C final treatment removed previous em-brittlement (ingot B particularly), indicating that this was "reversible embrittlement."" Metallographic examination of ingot C with 1 pct nital etch revealed only ferrite; ingot A had carbide spheroids uniformly distributed in a ferrite matrix. With ethereal picric acid plus zephiran chloride etchant.&apos; no differences between embrittled and un- embrittled specimens of ingots A, B, or C were noted. Fracture of C and D below the transition temperature was predominantly transcrystalline, with traces of intercrystalline, in both embrittled and unembrittled conditions. The corresponding fracture of unembrittled B was mixed transcrystalline and intercrystalline; of embrittled B, predominantly intercrystalline with some transcrystalline areas. . . . Acknowledgment The efforts of J. C. Leschen, formerly of National Research CO~P., in preparing the ingots are acknowledged with thanks.

Jan 1, 1956

By R. C. Rumble

To /neet the need for a subsurface flowmeter capable of measuring low rates of flow, a new type flowmeter has been built and tested in the field. This instrument is self-contairzed, can be run on a wire measuritzg line, and may be run through 2-in tubing. It consists of a rotameter flow-measuring assembly used in conjuncfion with a conventional subsurface instrument clock and chart. Integral with the tool is an inflatable packer which, by effecting a seal between casing and tool, diverts all fluid through the metering assembly. Packers of different size adapt the tool to different sizes of casing or liner. The flowmeter is capable of measuring rates of flow as low as 4 BID. The flowmeter has been run experimentally in seven wells. It has yielded satisfactory resrrlts in five of the wells, and in the other two it failed to operate because of loose sand in the hole. INTRODUCTION One of the oldest problems facing the producing branch of the petroleum industry is the location of depths of entry of fluid into a wellbore and the determination of the amount of fluid entering at each depth. This problem is often encountered in oil and gas wells completed in single producing horizons as well as in multiple pays; it is a common problem in oil wells which produce excessive volumes of gas. Several subsurface instruments designed to detect and meter flow in the producing interval of wells have been described in the literature.1,2,3,4 Most of these, however, were apparently intended for use in wells producing at relatively high rates of flow, and consequently lack sensitivity to low flow rates. Because of this and of other possible disadvantages inherent in these instruments, development of the subsurface flowmeter described in this report was undertaken. DESIGN OF FLOWME&apos;I&apos;ER Design Requirements In considering the requirements for a flowmeter that would not be subject to the limitations of devices existing at the time work was started, attention was first given to the size of the tool. Such an instrument should be of sufficiently small diameter that it could be lowered readily in 2-in tubing in a producing well. The flowmeter should contain a recording unit so that it might be run on a wire measuring line with conventional reeling equipment. It should not be too long to be accommodated by a portable lubricator of the type used for running conventional subsurface instruments. The metering mechanism of the flowmeter should not be affected appreciably by the viscosity or density of the flowing fluid. Such an instrument should record flow rates in the range from less than 10 B/D to over 100 B/D. To achieve this, a packer should he provided to divert through the instrument all fluid flow in the wellbore. This packer should be of such nature that in the event of excessive flow the fluid would bypass the instrument instead of forcing it up the wellhore. General Description A flowmeter incorporating most of the desired features outlined above was developed and built. The instrument is an elongated tubular device terminated on

Jan 1, 1956

Jan 1, 1956

By W. Hodge, A. F. Haskins, R. M. Evans

Refractory boron compounds are shown to resist corrosion by molten zinc. Coatings were made from ferroboron and manganese boron by several methods: welding, hard facing, and pack diffusion; and techniques of coatings are very important. Mechanical failure of the diffusion coatings can be partially eliminated by applying them to type 416 chromium steels rather than carbon steels. Welded coatings made with a tungsten arc are better than those made by other welding methods. Sintered compacts of mixtures of iron and chromium borides developed strengths of about 30,000 psi. They resisted corrosion of zinc at 600°C and oxidation at higher temperatures. MOLTEN zinc rapidly attacks common metals of construction. The rate of attack is greatly amplified at temperatures over 500°C. However, there are many applications where it is necessary, or at least advantageous, to use metal construction, such as for galvanizing pots and zinc smelter equipment. Condensers and agitator linings used in continuous zinc smelting processes are eroded and corroded by violently agitated molten zinc. Pumps, tubing, and other parts to permit transfer of molten zinc would be desirable in many zinc handling operations. The development of metallic materials which are resistant to molten zinc also might permit its use as a heat-transfer medium in modern high efficiency power plants. In reviewing previous work on zinc-resistant metals, it was evident that most of the available data refer to corrosion at temperatures just above the melting point. Tungsten and high Mo-Fe alloys containing more than 80 pct Mo resist corrosion by molten zinc.&apos;, &apos; Imhoff states that because chromium is not attacked by zinc, chromium plating improves galvanizing pot life." Nitrided cast iron has been used in die-casting machines&apos; where the molten alloy is 96 pct Zn and 4 pct Al. Other metals that have been suggested as of possible value for zinc resistance are columbium, tantalum, and titanium.V he corrosion of steel by zinc has been discussed by many authors,z, "2 None of these materials resisted corrosion by molten zinc at the temperatures used in this investigation, i.e., 600" to 700°C. I. Exploratory and Resistant-Coating Tests To make a study of zinc-resistant metals, exploratory tests were conducted to sort out roughly those materials which seemed most worthy of further development. This preliminary study included both solid metals and surface coatings. Exploratory Corrosion Tests in Molten Zinc—In initiating a study of this problem, numerous samples of metals and alloys were evaluated by dynamic corrosion tests made at 440 °C. Each metal was tested in its most readily available form, which, because of shape and size, was often unsuited for standard corrosion testing. The equipment used for dynamic corrosion testing of all materials, except in a few initial experiments, is shown in Fig. 1. Samples were tied to tungsten rods with tungsten wire, immersed in the zinc, and revolved at 12 rpm about the circumference of a 4 in. diam circle. Speed of travel was about 12.5 fpm. The molten zinc was protected from oxidation by an atmosphere of carbon monoxide, or, at a later period in the program, by a 9:1 mixture of nitrogen and hydrogen. When the effect of both molten metal and atmosphere was desired, the zinc level was lowered. Special high grade zinc was used in all of the corrosion tests. Corrosion losses for materials tested are shown in Tables I and 11. Whenever the materials were available in suitable form, the losses are reported in mils per year. It was recognized that weight losses on irregularly shaped specimens did not show the corrosion rate but the important consideration—poor resistance to corrosion—could be determined readily.

Jan 1, 1956

By W. J. West, T. D. Mueller, J. E. Warren

Two methods are used to obtain the relatiorlship between ratio of gas to oil permeability and gas saturation. One method is to calculate a curve from field measurements of reservoir behavior: the other is to measure the relationship on a core .sample in the laboratory. The curves resulting from these TWO methods arc often comparer1 and seldom found to agree. The most important reason for the nonagreement is often overlooked. This reason is that the performance k9/k0 curve is calculated with the assumption of uniform pressure and saturution throughout the reservoir. To show the differerence between performance type kg/k0 curves and laboratory type kg/k0 curves. two performance type kg/k0 curves for two different production rates were calculaled from the production perforrnance of a hypothetical reservoir for which a laboratory type curve was assumed. Production performance of the hypothetical reservoir bras calcu- lated by use of international Business Machine&apos;s 701 Computer. The perfortnance type kg/k0 curves differed from the laboratory curve for the hypothetical reservoir. The performance type crrrve for the lowest production rate most nearly coincided with the laboratory curve. INTRODUCTION Predictions of oil reservoir performance are often made with the aid of complex numerical calculations. An important quantity required in these calculations is the ratio of gas permeability to oil permeability (kg/k0 )at a particular gas saturation. Two methods arc used to obtain a relationship between this ratio and gas saturation. The curves resulting from these two methods are called performance kg/k0 curves and laboratory kg/k0 curves. The performance curve is obtained from field measurements of reservoir behavior. Particular kg/k0 values are calculated from the producing gas-oil ratio, the cumulative reservoir withdrawals, the original oil in place, the average reservoir pressure, and the PVT properties of the reservoir fluids. The particular kg/k0 values are then plotted against the average rcservoir gas saturations for the entire history period. In order to use this derived kg/k0 relationship for performance predictions, a curve through these points is extrapolated to higher gas saturations. The laboratory kg/k0 curve is obtained from measurements made on core samples. Saturation is maintained uniform throughout the core during these measurements. The kg/k0 curves obtained with these two independent methods are often compared and seldom are found to be the same. Various reasons have been offered for the lack of correlation. One is that insufficient core? were obtained to derive a curve representative of the reservoir rock. Another is that laboratory methods have failed to reproduce reservoir conditions. The most important reason for the noncorrelation has been often overlooked because of the difficulty in evaluating its effect. This reason is that the performance or field kg/k0 curve is calculated with the assumption of uniform pressure and saturation throughout the reservoir. In any

Jan 1, 1956

By E. W. Hart, W. R. Hibbard

The room temperature flow characteristics of a series of Cu-Cr alloys are found to be related to the amount and characteristics of the chromium-rich precipitate. The results are consistent with the theory of Fisher, Hart, and Pry, which considers the trapping to dislocation loops around precipitate particles. THE use of dispersed phases for hardening metals, particularly high temperature alloys, has been practiced for several years. Current theories describing the effects of dispersions on the basis of dislocation theory have been proposed by Mott and Nabarro.l-4 rowan," and Fisher, Hart, and Pry.8 These theories recently have been discussed by Hart,&apos; who pointed out that the experimental data available for comparison with theoretical calculations for overaged alloys are confined to studies by Gensamer, Pearsall, Pellini, and Low,8 Shaw, Sher-by, Starr, and Dorn,9 and some hitherto unpublished data on Cu-Cr alloys which will be described herein. Certain difficulties were associated with the interpretation of the first two studies. Gensamer, Pearsall, Pellini, and Low were concerned principally with interparticle spacing and did not document particle sizes. Shaw, Sherby, Starr, and Dorn included all the data in their report which was necessary for the theoretical evaluation, but their study is subject to some question1" due to difference in prior history of various specimens and due to systematic differences in particle characteristics" which appeared to be associated with differences in the magnification of their microscopic examination. The current study was undertaken to obtain critical data in a manner which avoids some of these limitations. Specimens of eight Cu-Cr alloys remaining from a previous investigation&apos;&apos; were available in the form of 100 mil sheet of the compositions shown in Table I. At 200 mil this material had been previously an- nealed ½ hr at 1000°C, water quenched, and cold rolled 50 pct. Samples were further cold rolled to a thickness of 20 mil for a total reduction of 90 pct. Sheet tensile specimens of standard geometry with a gage length 2½ in. long and 0.20 in. wide were machined from this sheet. These specimens were then annealed in a dried N2-H2 atmosphere for the times and temperatures indicated in Table I. This anneal served two purposes: 1—to obtain a fairly uniform and nearly constant grain size of 0.015 to 0.022 mm average diameter, and 2—to precipitate a fairly uniform, nearly spherical over-aged dispersion of essentially pure chromium. A typical microstructure illustrating the grain size and the uniformity of the dispersion is shown in Fig. 1. A typical dispersion illustrating the range of sizes and shapes of the particles is shown in Fig. 2. The tensile specimens were tested in the Instron equipment described in ref. 11 at room temperature at a strain rate of 0.04 in. per in. per min. Three tests were run on each alloy with sufficient repro-ducibility that a single true stress-true plastic strain curve could be drawn through the three sets of data for each alloy. Typical curves are shown in Fig. 3.

Jan 1, 1956

By L. M. Clarebrough, A. K. Head, G. W. West, M. E. Hargreaves

RECENTLY Welber and Webelerl reported that during the annealing of fatigued copper no energy was released, but that energy was absorbed in the range of temperature 250" to 400°C. In view of the fact that strain hardening occurs during the fatigue of copper, this result was surprising and it seemed desirable to check the measurements. This note describes measurements made by a different technique which indicates that energy is stored during the fatigue of copper. Two test bars with gage length ¾ in. in diameter and 6 in. long were machined from OFHC copper and annealed in vacuum for 1 hr at 600°C. These were fatigued to fracture in an Avery pulsator at a frequency of 2000 cycles per min. The first specimen fractured after 0.25x10 cycles under a push-pull

Jan 1, 1956

By H. W. Mead, C. E. Birchenall

SELF-DIFFUSION of cobalt has been investigated by Ruder and Birchenall,1, 2 Nix and Jaumot,3 and Gruzin.4 The results for a given tem- perature differ by a factor of about six in the extreme cases with no really close agreement between any two sets of data. The first two groups of investigators employed the decrease of surface activity method. Since Ruder and Birchenall have shown that the values obtained from this method are sensitive to the surface preparation, it seemed desirable to perform a few check experiments by a method not open to this objection.

Jan 1, 1956