The primary purpose of the research covered by this report was the development of a method for induction melting titanium in a water-cooled copper crucible. The results reported represent findings obtained during the melting of a minimum number of ingots; therefore, only tentative conclusions can be drawn regarding the effects on metal quality of many of the operating procedures.

The most important result of the research was the development of a method for melting titanium by induction heating in a water-cooled copper crucible. The method is an improvement of the method of Schippereit, Leatherman, and Evers; a similar crucible design was used, but for this investigation a cover of slag was used to insulate the ingot from the crucible walls. The addition of this slag cover eliminated arcing between the crucible and the ingot. One run was conducted during the course of this research with no slag cover in the crucible. During this run, arcing between the ingot and the crucible segments was noted, and damage to the crucible occurred. These experimental results confirmed, the value of the slag as an electrical insulator between the ingot and the crucible.

The chief advantage of this melting technique over vacuum-arc melting and electroslag melting is the capability of melting loose titanium sponge and scrap. This capability eliminates the need for fabrication of the consumable electrode required for both vacuum-arc melting and electroslag melting. Titanium scrap can also be easily consolidated into a usable form by this technique.

One possible application of this process would be as a first-melting operation to convert titanium sponge into an electrode for vacuum-arc remelting. This would eliminate the compaction and welding steps needed for preparing the first-melt electrode and would also constitute the first-melting step in the present vacuum-arc-melting technique. Ingots prepared by this modified melting technique had slightly higher levels of impurities than ingots prepared by conventional double vacuum-arc melting. The inability of this process to remove hydrogen from sponge with high hydrogen and the residual fluorine content resulting from melting with calcium fluoride slag are considered possible serious problems to this process, but improvements in equipment and procedures should reduce other impurities to an acceptable level.

A complete study has not been made of the effect of furnace pressure on the level of impurities in titanium melted by this technique. Attempts have been made to melt at low furnace pressure with only moderate success, and most melting was done at one-third of an atmosphere of helium. Vacuum-distilled sponge melted at low pressure had lower levels of impurities than similar metal melted at a pressure of one-third of an atmosphere, but a study of the effect of intermediate pressures was not undertaken. One attempt to melt leached-and-dried sponge at low pressure was unsuccessful because of excessive outgassing of the molten pool; however, melting at intermediate pressures between zero and one-third of an atmosphere might be effective.

Results obtained with this equipment indicate that chances of satisfactory scaleup are very good; however, failure of the large-scale tests made by the Air Force based on Schippereit’s work raises a question of possible problems associated with scaleup. The authors believe that Schippereit’s small-scale tests were successful because a combination of high power input and small ingot size caused distortion of the molten metal away from the crucible walls similar to the distortion of the pool noted in this work. In the Air Force’s large-scale work; the bulk of metal melted was too large to be similarly held away from the wall, and good electrical contact between the ingot and crucible occurred. Tests conducted with and without slag in the 3-½- inch-diameter crucible confirmed both the effectiveness of the slag as an insulator and the need for the slag even on runs of this size, Barring unforeseen difficulties, the slag should be equally effective as an insulator between the ingot and the crucible in large-scale runs.

The fact that this technique makes possible the melting of loose titanium sponge and scrap is encouraging, but it is not known whether gains made by eliminating electrode fabrication would result in lower melting costs.

Further tests with this equipment are now in process; the results will be the subject of subsequent reports.

The Bureau of Mines has been active in research related to the vacuum-arc melting of reactive and refractory metals for many years and recently has actively pursued a program of research on alternate melting techniques, including techniques involving induction heating. This report describes an induction-melting technique by which titanium can be melted in a split, water-cooled copper crucible. The term “inductoslag” melting has been applied to this technique to correspond to the term “electroslag” used to describe the Hopkins process now used extensively for melting specialty steels and superalloys. As in electroslag melting, the ingot was melted under a cover of molten slag, but melting was accomplished by induction heating.

Titanium, because of its reactive nature, is melted almost exclusively by consumable electrode vacuum-arc melting. This melting process provides a relatively economical means of melting in vacuum and in water-cooled copper crucibles, but has the disadvantages of requiring the fabrication of a consumable electrode and of lack of control of the melting rate. Studies are being conducted by the Bureau to evaluate the Hopkins “electroslag” process for titanium melting.

Numerous attempts have been made to utilize induction melting for melting reactive metals. Notable among recent attempts was the work of Schippereit, Leatherman, and Evers of Battelle Memorial Institute. In their work, a segmented, water-cooled copper crucible was used to contain a charge of titanium which was heated inductively and melted. While this technique worked satisfactorily for small-scale ingots, work conducted by the U.S. Air Force on a 14-inch-diameter crucible of the same design failed because molten metal shorted across the slits in the crucible and caused crucible failure.

Earlier Bureau research resulted in the development of an induction-melting technique which did not require the use of a split crucible. Results of this work have been reported previously for melting titanium ingots up to 2 inches in diameter, and studies are still in process on melting larger titanium ingots.

The work of Schippereit probably came nearest to achieving the desired results, even though the system he proposed did not function on large-scale equipment. His work with a segmented crucible was closely related to work conducted as long ago as 1926 by Siemens, who proposed the use of a split, water-cooled crucible of this type.

The large-scale work conducted by the U.S. Air Force on Schippereit’s design also involved the unsuccessful use of an oxide coating on the interior crucible surface to prevent shorting across the crucible slits. A thin coating of beryllium oxide was placed on the inner surface of the crucible, but, when a titanium ingot was melted in the crucible, this coating failed, and the crucible was damaged by arcing between the ingot and the crucible walls. Work with this equipment was terminated because of consistent failure of the crucible whenever melting of the ingot occurred. It was concluded that arcing to the crucible walls occurred with the 14-inch-diameter crucible but not with the small-scale crucible (2-½-inch-diameter) because of the difference in frequencies used. In the small-scale studies, the frequency was 2,000 to 3,000 cycles per second (cps) as compared with only 60 cps for the large-scale work.

This report describes studies conducted in crucibles similar to those of Schippereit, however, in these studies, calcium fluoride was used as a slag cover for the ingot. Titanium ingots 3-½ inches in diameter were successfully prepared by induction melting under a cover of molten slag. A thin layer of this slag solidified against the cold wall of the crucible. This layer of solid slag provided good electrical insulation between the ingot and the crucible and eliminated all tendency for arcing between the ingot and the crucible. Invention report MIN-1296, dated November 20, 1967, describes this process; the patent application is pending.

Discussion

Description of Equipment

Based on preliminary experiments, a copper crucible with four slits was constructed, and provisions were made to enclose the unit in a vacuum chamber. This crucible, shown in figure 1, has a 3-½-inch ID, and the segmented section was 8-½ inches long. The segments of the crucible were bolted to a lower section, 2 inches long with a 3-½-inch ID. The lower section was not segmented but served as a base on which to bolt the crucible segments and to provide an additional length of cooled crucible through which the solidified ingot was withdrawn. The crucible sections were insulated from each other by an aluminum oxide thermocouple insulator, and the assembled sections were insulated from the lower furnace section by a micarta ring. The lower crucible section was bolted directly to the furnace bottom.

Figure 2 shows the crucible, complete with work coil, installed in the furnace chamber. The work coil was seven turns of ½-inch, heavy-walled

copper tubing and was insulated with Teflon tubing. With the work coil located near the top of the crucible, heating was most intense approximately 2-¼ inches below the top of the crucible. In this way, the ingot melted in the upper part of the crucible and was cooled and solidified in the lower part. Power for melting was supplied by a 10,000-cps motor-generator rated at 75 kilowatts.

The furnace body shown in figure 2 provided a controlled atmosphere for melting. A 260-cfm, single-stage blower backed by a 21-cfm mechanical pump was used to evacuate the chamber with melting normally conducted in approximately one-third of an atmosphere of helium. The furnace was also equipped with sidefeeding units for adding both calcium fluoride and titanium sponge. Ingots formed were withdrawn by the shaft through the bottom of the furnace. The top sidefeeder could be replaced with a similar shaft through the furnace top if the metal to be melted was in rod form.

Figure 3 is a longitudinal section of the crucible showing the relative position of the work coil, ingot, and slag cover during melting. As shown in the figure, melting took place at the top of the ingot. Heating of the ingot maintained a pool of molten slag at the top of the ingot although a layer of this slag solidified against the water-cooled copper wall of the crucible.

The starting stub was a 5-inch length of titanium approximately 3-¼-inch OD. This diameter, slightly less than the inside diameter of the crucible, prevented the crucible segments from being shorted out and also left space for a slag cover to form when melting started. A shoulder at the bottom of the starting stub prevented powdered or molten slag from running out of the bottom of the crucible. The starting stub was long enough for this shoulder to be below the segmented portion of the crucible.

Figure 4 is a cross section of the segmented portion of the crucible and shows the detail of the slit construction. The aluminum oxide (Al2O3) thermocouple beads shown in figure 4 were used as a convenient means of maintaining the spacing between the segments of the crucible during assembly. They also provided a seal to prevent molten slag from leaking out of the crucible. Once the molten slag ran into the slots and against these insulators, the slag solidified and formed a layer of solid calcium fluoride between the metal and the aluminum oxide.

The shape of the ingot cross section in figure 4 is representative of the portion of the ingot below the melting zone. The solidified slag around the ingot was normally 1/16 inch thick or less. During melting, the molten pool was distorted into the shape shown in figure 3. The cover of slag around the ingot was thicker near the top of the ingot, and most of the slag was molten. A thin layer of solid slag was observed at the top of the slag pool, however, and it was assumed that a continuous sleeve of solid slag existed at all times.

The levitating forces acting on the molten metal caused a part of the metal to be lifted above the surface of the molten slag. Levitation of the metal was probably exaggerated because of the buoyant forces exerted by the molten slag surrounding the metal. This levitation of the metal probably also aided in preventing metal from shorting across the slits in the crucible. The field strength was more intense at the crucible slits than in the area between two slits; consequently, metal adjacent to the slits was forced inward. When viewed from above, the metal surface protruding from the surface of the slag

had the shape of a four-pointed star with the points located midway between the slits.

The question arises as to whether the slag, once molten, was heated by induced current flow in the slag itself. The absence of any visible levitation of the slag indicated that the slag was heated only by contact with the metal.

Melting Procedure and Results

Melting experiments were conducted with a variety of forms of titanium including previously melted swaged rods, vacuum-distilled sponge, leached-and-dried sponge, and scrap. Calcium fluoride slag, prepared by fusing reagent-grade calcium fluoride, was used for all runs. A complete discussion of slag treatment is given by Ausmus and Beall.

While melting procedures varied slightly with different starting materials, the general procedure for all runs was to establish a pool of molten metal and slag in the crucible and then to add metal to this pool. Metal was added by drip melting from the lower end of a consumable feed rod or by side-feeding loose sponge metal or scrap into the top of the crucible. As material was added to the pool, the resulting ingot was withdrawn through the bottom of the crucible.

Success in melting depended on first establishing a full pool of molten metal and slag in the crucible, particularly during runs when loose sponge was added to the pool by sidefeeding. The procedure developed consisted of placing an initial charge of the metal to be melted in the crucible with the required amount of slag. The initial charge of metal was a 2-½- to 2-¾-inch-diameter cylinder approximately 3 inches long. Pressed compacts of titanium sponge were used when melting sponge, and either pressed sponge compacts or machined ingot sections were used when drip melting feed rods of previously melted titanium.

This initial charge of metal was placed on top of the 3-¼-inch-diameter starting stub previously described, and the annular space between the metal and the crucible wall was filled with granular calcium fluoride slag. The starting stub was attached to a water-cooled control rod through the bottom of the furnace, and this rod was used to adjust the level of the ingot in the crucible and to withdraw the ingot as it formed.

After the starting stub and slag were added to the crucible, the metal to be melted was loaded into the furnace either as a feed rod attached to a motor-driven control rod or as loose, granular sponge metal in a manually controlled sidefeeder. Solid feed rods were either swaged rods of previously melted material or compacted sponge bars. The furnace was evacuated and then backfilled with helium to approximately one-third of an atmosphere.

At the start of melting, the level of the initial charge of metal in the crucible was placed at the midpoint of the work coil. When this initial charge was heated inductively, the slag surrounding it melted and formed an insulating layer between the metal and the crucible wall. With sufficient heating, the metal melted and welded itself to the top surface of the starting stub. Power supplied to the work coil was limited to approximately 30 kilowatts during the first 2 or 3 minutes until the metal and slag in the crucible melted. The power level was then increased to 45 to 50 kilowatts, and, as soon as a full pool of molten metal had formed, the metal to be melted was added to the pool.

When melting from a solid rod, the rod was fed downward until its lower end was submerged in the slag in the crucible. This placed the lower end of the feed rod in the hot zone, and metal at the end melted and transferred to the pool of molten metal in the crucible. As the feed rod melted, it was fed downward, and the resulting ingot was withdrawn through the bottom of the crucible. Melting continued until the feed rod was consumed. The remaining stub of the feed rod was withdrawn from the crucible, power to the work coil was terminated, and the ingot was allowed to cool.

When the ingot was sufficiently cool, it was removed from the furnace, and the thin layer of slag which had formed on the ingot as melting progressed was removed. Figure 5 is a photograph of an ingot produced from previously melted, vacuum-distilled sponge. The feed rod for this ingot was a 2-inch- diameter swaged rod. The lower 5 inches of this ingot was the starting stub, and the upper 4-¾ inches consisted of metal drip-melted from the feed rod plus a 2-½-inch-diameter machined ingot section 2-½ inches long. Melting results for this ingot are summarized in table 1. The low melting rate and relatively high energy utilization for this ingot reflect difficulties experienced with control of the feed rod; however, better results during melting of solid feed rods could probably be obtained with additional melting experience.

When loose sponge was melted, sponge was dropped into the pool from the outlet of the sidefeeder located 11 inches above the top of the crucible. This was done as soon as a full pool of molten metal and slag had formed in the crucible. The rate of addition was limited by the rate at which the solid sponge was assimilated by the molten pool. Makeup slag was added as the slag cover was depleted by solidification of a slag layer on the outer surface of the ingot. As metal was added, the ingot was withdrawn to maintain a constant level of the ingot in the crucible, and melting was continued until the supply of sponge metal in the sidefeeder was exhausted. The ingot was allowed to cool and was then removed from the furnace.

Figure 6 shows an ingot prepared from vacuum-distilled sponge, and figure 7 shows an ingot prepared from leached-and-dried titanium sponge. The ingot melted from leached-and-dried titanium sponge was expected to be more difficult to melt than vacuum-distilled sponge because of the higher content of volatile impurities, but no serious difficulties were encountered when melting this material, although the melting rate was somewhat lower, and the ingot sidewalls were rougher. A somewhat smaller initial charge of sponge in the crucible or an initial charge of previously melted material was used to reduce excessive outgassing when melting started. Melting results for these two ingots are also summarized in table 1.

Ingots were also induction melted from titanium scrap to determine the utility of this process for reclaiming scrap. The scrap used consisted of broken tensile specimens, chopped sheet, sections of ingots, and other forms of solid titanium. No attempt was made to utilize machine chips. Figure 8 shows an ingot prepared from scrap similar to

that shown beside the ingot. Because of the difficulty of sidefeeding this scrap, the bulk of the metal melted was simply placed in the crucible along with a supply of calcium fluoride slag. The ingot shown was prepared by filling the crucible with scrap and slag three times and melting the three charges one on top of the other. No difficulties were experienced in melting this material, and the resulting ingot was excellent.

During the melting of this ingot, a sufficient number of pieces of scrap were side-fed into the molten pool to determine whether chunks of metal could be added in this manner. Ten pieces, each weighing 10 to 15 grams, were dropped into the pool from a tube approximately 11 inches above the top of the crucible. All pieces dropped quietly into the pool with virtually no splashing of metal. This was enough to show that the scrap could be added in this way if provisions were made to drop them from a shorter distance above the crucible.

Surface quality of the ingots produced by this melting technique was directly related to the quality of the starting materials, smoother surfaces being obtained from materials low in volatile impurities. The best surfaces were obtained from previously melted material, either in the form of swaged rods (fig. 5) or scrap (fig. 8). Ingots prepared from titanium sponge had slightly rougher surfaces than those prepared from melted stock, and, of the two sponge varieties studied, vacuum-distilled sponge (fig. 6) yielded ingots with better surfaces than those from leached-and-dried sponge (fig. 7).

All of the ingots were relatively sound internally except for shrink holes near the upper surface. No attempt was made to reduce the size of these shrink holes, although it is believed that hot-topping practices similar to those used for vacuum-arc melting and electroslag melting of titanium could reduce the size of these holes or possibly eliminate them entirely. Figure 9 shows the etched internal surface of an ingot produced from vacuum-distilled sponge. There was no subsurface porosity such as is characteristic of vacuum-arc-melted ingots. It would appear, from the size of the shrink hole and the large area of equiaxed grains, that the molten pool was quite large and perhaps represented almost half of the ingot shown. Further studies are needed, however, to establish accurately the boundary of the molten pool.

Sufficient data have not been obtained to determine the optimum ratio of slag to metal. This ratio varied with the type of material melted; somewhat more slag was needed for melting leached-and-dried sponge than for vacuum- distilled sponge or previously melted metal. For runs conducted in the 3-½- inch diameter crucible described, the weight of slag used was approximately 10 percent of the weight of metal melted. The ratio of slag to metal would decrease with increased ingot length since a large percentage of the slag used was tied up in the pool of slag at the top of the ingot, and the dimensions of this pool of slag would not change with ingot length. The ratio of slag to metal might also be expected to decrease with increased ingot diameter; however, no runs were conducted in which the ingot diameter was varied.

The greatest advantage of this melting technique is the capability of melting titanium sponge and scrap without fabricating an electrode. Ingots can be melted more rapidly and with fewer kilowatt-hours per pound by vacuum-arc melting and by electroslag melting but only when the titanium is in electrode form. Fabrication of consumable electrodes from titanium sponge is an expensive step in the conversion of sponge to ingot, and consolidation of titanium scrap into consumable electrodes for remelting is likewise expensive and difficult. Loose titanium sponge and scrap can be melted by this induction- melting technique with ease; in fact, it was easier to melt loose materials than solid rods.

Ingots prepared by inductoslag melting titanium sponge can be used as consumable electrodes for vacuum-arc remelting into a final ingot. Induction melting would eliminate the pressing and welding operations required to produce a first-melt electrode for vacuum-arc melting and would also constitute the initial melting step. One purpose of this work was to develop such a melting scheme and to compare the resulting ingots with ingots prepared by conventional double vacuum-arc melting.

Induction-melted ingots prepared from all the varieties of starting materials discussed previously were suitable for electrode stock for vacuum-arc remelting. No machining of the outer surface of the ingot was necessary. Ingot length was limited by the capacity of the experimental induction furnace, and it was necessary to weld several ingots together to form an electrode for vacuum-arc remelting. A typical electrode produced by welding four induction- melted ingots together is shown in figure 10. This electrode, which was approximately 3-3/8 inches in diameter and 24 inches long, was vacuum-arc remelted into the 5-inch-diameter ingot shown in figure 11. Analyses of the final ingot and of the four induction-melted ingots used for the electrode are included in table 2. Also included in table 2 are analyses of ingots prepared from the same sponge lot by conventional double vacuum-arc melting and by electroslag melting. Samples of the four induction-melted ingots were taken

from the upper portion of each ingot, and analyses of these samples are less representative of the overall ingot composition than if samples had been obtained from several locations in the ingot. Consequently, a fifth ingot, induction melt 111, was prepared from the same sponge lot by induction melting, and this ingot was more thoroughly sampled. Results of analyses of this ingot are also shown in table 2.

Analyses of a similar set of ingots prepared from leached-and-dried titanium sponge by induction melting and by other techniques are given in table 3. The titanium sponge from which these ingots were prepared was typical, magnesium-reduced domestic sponge. This sponge contains an appreciably greater quantity of hydrogen and other volatile impurities than vacuum-distilled sponge. These impurities can be reduced to acceptable levels by double or triple melting in consumable-electrode, vacuum-arc furnaces. However, ingots prepared from this sponge by electroslag melting contained up to 200 ppm hydrogen, and, for this reason, electroslag melting was not considered a satisfactory melting technique for domestic leached-and-dried sponge. Hydrogen content of the ingots produced by induction melting was only slightly lower than for ingots produced by electroslag melting.

The ingot shown in figure 12 was prepared by vacuum-arc remelting the induction-melted ingots 103 through 106. Analyses of this ingot, SA 26,512, are included in table 3. Hydrogen content of the arc-melted ingot was still above the acceptable level. The aluminum content of this ingot was appreciably higher than the corresponding ingot, SA 26,496, prepared from vacuum-distilled sponge. However, the aluminum content of the ingot (SA 26,155) prepared from this sponge by double vacuum-arc melting was also high.

Table 3 also includes analyses of a fifth induction-melted ingot, number 112, which was subjected to a more thorough analysis than the ingots used for remelting.

Most of the ingots prepared by induction melting were melted at a furnace pressure of approximately one-third of an atmosphere of helium. This furnace pressure was arbitrarily chosen to correspond with that used during studies of the electroslag melting of titanium. A partial backfill, which had been chosen for electroslag melting to reduce losses of calcium fluoride, was for the same reason chosen for induction melting. However, a number of runs were also attempted in the induction furnace at lower furnace pressures. Induction-melted ingot 108, included in table 2, was prepared from vacuum-distilled sponge at a reduced furnace pressure. For this ingot, melting was initiated at the usual one-third atmosphere pressure, but, as soon as a molten pool had been established, the vacuum valve was opened, and the furnace pressure was reduced to less than 1,000 millitorr. At this low pressure, difficulty with electrical discharges within the furnace was experienced, and the run was terminated.

A similar run was attempted with leached-and-dried titanium sponge, but outgassing of the molten pool was so violent at low pressure that no ingot was obtained. The improvement in the impurity content of ingot 108 and the obvious outgassing of the molten pool during the one run conducted at low pressure with leached-and-dried sponge indicate the need for further study of the effect of furnace pressure on metal quality. Operation at reduced furnace pressure should be the best method of lowering the hydrogen content of ingots prepared from leached-and-dried sponge. A detailed study of the effect of furnace pressure between zero and one- third atmosphere has not been undertaken at this stage of development of the melting process.

One run was also attempted in this equipment without slag in order to prove the necessity of using slag to prevent arcing between the ingot and the crucible. For this run, the usual initial charge of metal was placed in the crucible, and provisions were made to sidefeed vacuum-distilled sponge into the crucible as melting progressed. The furnace was backfilled to one-third of an atmosphere of helium, and heating of the metal in the crucible was initiated. As soon as a pool of molten metal formed and molten metal ran against the sides of the crucible, heating of the metal decreased, and it was impossible to maintain a full pool of molten metal. Sponge added to the crucible by side-feeding did not melt completely.

When the metal was removed from the crucible, evidence of arcing between adjacent segments of the crucible was noted. The slits in the crucible were pitted at the level of most intense heating, and metal which had been melted had run into the slits in the crucible. No further attempts were made to operate without slag because of the negative results obtained during this run.