Development of the Molten Slag Reduction Process -1 Characteristics of Closed Type DC arc Furnace for Molten Slag Reduction

Transcription

1 ISIJ International, Vol. 58 (2018), ISIJ International, No. 10 Vol. 58 (2018), No. 10, pp Development of the Molten Slag Reduction Process -1 Characteristics of Closed Type DC arc Furnace for Molten Slag Reduction Toshiya HARADA, 1) * Hiroshi HIRATA, 1) Takashi ARAI, 2) Takehiko TOH 3) and Takahiro YAMADA 4) 1) Process Research Laboratories Nippon Steel & Sumitomo Metal Corp., 16-1 Sunayama, Kamisu, Japan. 2) Plant Eng. & Facility Management Center Nippon Steel & Sumitomo Metal Corp., 20-1 Shintomi, Futtsu, Japan. 3) Advanced Technology Research Laboratories Nippon Steel & Sumitomo Metal Corp. Now at Nippon Steel & Sumikin Technology Co. Ltd, 20-1 Shintomi, Futtsu, Japan. 4) NS Plant Designing Corp., Nakabaru Tobata-ku Kitakyushu-City, Japan. (Received on March 23, 2018; accepted on June 29, 2018; J-STAGE Advance published date: August 24, 2018) To develop a slag reduction process (IBX process: Iron Bath for X), three kinds of slag reduction tests were carried out using different pilot scale and commercial scale DC arc furnaces. The goals are to recover the valuable elements such as Fe or P from the steelmaking slag and to modify the slag composition. The main target is to use molten slag directly in the reduction process and to establish a stable operation for improving the energy consumption, productivity, and processing cost. In the present paper, we describe the results obtained in a pilot scale closed-type 2 MW DC arc furnace (Test 1), commercial scale open-type 30-MW DC arc furnace (Test 2), and newly designed closed-type 4-MW DC arc furnace (Test 3). Reduction efficiency and the metallurgical properties in DC arc furnaces, the mechanism of slag foaming, and the influence of the atmosphere in the furnace were discussed based on the experimental results. Finally, model calculations of the flow patterns in the furnace were carried out for further comprehension of the metallurgical phenomena. From these results, it was confirmed that the closed-type DC arc furnace can achieve efficient and stable reduction of steelmaking slag with hot charging. KEY WORDS: BOF slag; reduction; arc furnace; recovery; slag foaming. 1. Introduction In Japan, eight million tons of BOF slag (Basic oxygen furnace slag) are produced annually, and it is mainly used for civil construction. However its uses are restricted due to properties such as high ph leaching or hydroscopic expansion. On the other hand, BOF slag contains valuable elements, such as iron and phosphorus. The concept of full utilization of BOF slag was proposed by Kubodera et al. in ) According to this concept, molten BOF slag is charged to the electro-melting furnace and reduced. The reduced slag is used as cement, the hot iron with high phosphorus content is recycled after dephosphorization, and the dephosphorized slag is used as phosphate fertilizer. Kubodera et al. insisted that the direct use of hot slag is indispensable from the economical aspect. Though the concept is not realized yet, the need for such a process is increasing. The profit of iron recovery corresponds to more than 2% of the iron yield in the steelmaking process. Phosphorus is an important element for agriculture and various kinds of industrial products, though its supply is dependent * Corresponding author: harada.pq4.toshiya@jp.nssmc.com DOI: on import in Japan. 2) If the recovered phosphorus from the slag could be converted to the effective phosphate product, it would be a great contribution as a new phosphorus resource. Reduction and modification of BOF slag would extend the range of usage of BOF slag. Waste materials such as fly ash or used refractory, which contain silica and alumina, can be used as slag modifier. Moreover, due to the deterioration in quality of iron ore, the production amount of BOF slag will increase in the near future, which raises the steel production cost. However, the increase in slag amount and phosphorus content is rather profitable in this process. That is, this process converts BOF slag and related waste materials to various kinds of valuable products without generating any waste material. The required energy for producing hot iron is comparable to that of the blast furnace process by using the heat of the hot slag. For the smelting reduction process, there are several options, such as the converter-type, 3) shaft-type 4) and electric arc furnace-type. 5) From these options, the EAF process was adopted, considering the suppression of greenhouse gas and the possibility of using molten slag. In case of shaft-type furnaces molten slag is not available. Converter-type furnaces might accept the molten slag, but when the molten slag is charged into the furnace, the total iron content as 2018 ISIJ 1934

2 oxides in the slag ((T,Fe)) is raised temporarily. In such a case it would be difficult to suppress slag foaming in the converter due to the vigorous stirring. On the other hand, the characteristics of smelting reduction in arc furnaces are not well known. Generally speaking, the stirring effects in arc furnaces are poorer than that in converter-type furnaces. To investigate the metallurgical properties, factors causing slag foaming, the necessity of preventing air sucking, or the possibility of stable operation for the direct use of molten slag, test operations in three types of arc furnaces were conducted and the results were compared. In Test 1 cold slag and carbon were added into the hot spot (i.e. the high-temperature area of the bath, on which the arc jet impinges) through the hollow electrode under a reducing atmosphere. In Test 2 only carbon was added into the molten slag bath prepared in the furnace, where air was always introduced from the slag door. In Test 3, hot slag was charged into the furnace and carbon was added through the feed tube under a reducing atmosphere. The purpose of the present work is to not only raise the reduction efficiency, but to simultaneously suppress the slag foaming caused by hot slag charging and to achieve stable operation. It is antinomy in a sense. Therefore, generally disadvantageous conditions for reduction, such as no gas stirring or low metal carbon content, were also investigated, and the possibility of compatibility was pursued. 2. Experimental 2.1. Closed type-2 MW DC arc Plant Test (Test 1) The experiments were carried out using a 2 MW closed-type DC arc pilot plant. Here, closed-type means that there is no opening for sucking air into the furnace. The purpose was to see the features of the reduction of cold slag in the arc furnace. Figure 1 shows the schematic diagram of the pilot plant. The materials used were two types of pulverized BOF slag, anthracite, silica sand, and bauxite. The composition of the slag is shown in Table 1. These slags were both produced at steelmaking shops. The granular materials were added into the furnace through the hollow electrode. The tapping hole is equipped at the height of 200 mm from the furnace bed. About three ton of molten iron was always present in the furnace as a hot heel, which was prepared by melting cold iron and FeSi before starting the test operation. BOF slag was usually added at 1 t/h with anthracite at the corresponding feed rate. The standard heat time was 2 hours. The required energy for melting and reduction was supplied by the arc jet. After the material and energy feed almost two ton of reduced slag and generated iron were discharged through the same tap hole. As there is no opening in the furnace and the electrode was perfectly sealed, no air was sucked into the furnace and the inner atmosphere was kept as a reducing one Open-type 30 MW DC arc Plant Test (Test 2) As the second step, a commercial scale DC arc furnace was used. 6,7) Figure 2 shows the schematic diagram of the arc furnace used for the experiment. The furnace capacity was 120 ton/heat and the feed power during operation was around 30 MW. The purpose was to see the characteristics of reduction of oxidized slag by adding carbon onto the molten slag bath in the large-scale DC arc furnace. At first, 85 ton of steel scrap was charged with 2 to 2.5 ton of coke for raising the carbon content. Quasi-BOF slag was prepared in the furnace by charging 40 ton of BF (Blast Furnace) hot metal after melting half of the steel scrap, and by oxidizing hot metal with oxygen and adding required fluxes, such as burned lime and calcium aluminate. The target slag composition was (CaO)/(SiO 2 ) = 1.2, (T.Fe) > 15%, and (Al 2 O 3 ) = 12%. where (CaO), (SiO 2 ), (T.Fe), and (Al 2 O 3 ) denote the contents of CaO, SiO 2, total iron, and Al 2 O 3, respectively, in mass%. The target metal and slag temperatures were and K. The formed slag volume was 6 7 t/heat, though it depended on the silicon content Fig. 1. Schematic diagram of the furnace in Test 1. (Online version in color.) Table 1. Composition of the material slags. [mass%] T.Fe CaO SiO 2 MnO MgO P 2O 5 Al 2O 3 Slag A Slag B Fig. 2. Schematic diagram of the furnace in Test 2. (Online version in color.) ISIJ

3 in the hot metal. Once slag was formed, granular coke was added into the slag bath with 50 kg/min for 20 minutes mainly through the injection tube inserted from the slag door. During carbon addition, slag and metal sampling was made every five minutes to observe the behavior of each composition. The Slag door was opened so that the foamed slag overflowed out of the slag door. Air was always sucked through the slag door due to the negative pressure in the furnace, the inside was an almost entirely air atmosphere. Such a furnace is called open-type here Closed-type-4 MW DC arc Plant Test (Test 3) The third step test was carried out by using newly installed closed-type DC arc furnace with a capacity of 4 MW. 8,9) The schematic diagram of the pilot plant is shown in Fig. 3. The furnace was equipped with tilting-type slag container for charging molten slag onto the hot metal bath. The purpose was to confirm the possibility of a stable reduction treatment by charging molten slag directly into the furnace. To prevent slag foaming caused by rapid reactions between carbon in metal and molten slag, the feed rate was controlled by the tilting angle. The openings around the electrode or the spout of the slag container were sealed completely, so that the air was not sucked into the furnace. The furnace scale had an inner diameter of 3.1 m and an inner height of 2.5 m. Granular materials such as coke (1 mm), silica (20 μm), or alumina (< 3 mm) were added through the feed pipe equipped near the electrode on the furnace cover with a small amount of carrier gas, mostly under the influence of gravity. There are two tapping holes. One is for metal tapping and is located at a height of 200 mm from the bottom hearth. The other one is for slag tapping and is located at a height of 400 mm from the bottom hearth. The amount of hot heel metal was a minimum of 10 ton, and 3 to 6 ton of reduced slag was always present in the metal bath. A part of the molten slag tapped from commercial scale LD converter in the steelmaking shop was transferred to the pilot plant by the ladle car. Concerning the composition of the used slag, (T.Fe) was 20%, (P 2 O 5 ) was 3.5% on average. About 4 ton of converter slag was charged to the slag container for one heat and discharged to the arc furnace continually by tilting the container. Granular coke was added continuously through the material feed tube at the furnace cover. The power supply was 2.5 to 4 MW. After feed the required electric power, coke, and fluxes, sampling and temperature measurement were performed using an automatic device. Then, the slag tap hole was opened and the reduced slag was discharged. The amount of hot metal in the furnace was increased gradually by reduction of iron oxide in the slag. When the metal surface level approached the level of the slag tapping hole, metal was discharged. The duration of one test heat was 1 to 2 hours, and the cycle time was about 3 hours. One campaign test operation lasted 5 days, of which the first two days were spent preparing the metal bath. In total, 7campaign tests were performed. 3. Results and Discussion 3.1. Reduction of Fe x O and P 2 O 5 The results of the slag reduction obtained in each test are shown in Figs. 4 and 5. Here, the reduction degree, R is defined as the ratio of the decrease in (T.Fe) or (P 2 O 5 ) during reduction to the content in the initial slag before charging as shown in (1) and (2). R ( TFe. ) ( TFe. ) / ( TFe. ) 100 [%]... (1) Fe i f i RP ( P2O5) i ( P2O5) f /( P2O 5) i 100 [%]... (2) i: composition of the charged slag into the furnace f: composition of the slag in the furnace after reduction In Test 1 and 3, that is, in the closed-type DC arc furnaces, (T.Fe) after reduction was lowered to less than 2%, and (P 2 O 5 ) was lowered to almost less than 0.3%. Both reduction degrees of (T.Fe) and (P 2 O 5 ) reached almost 90% Fig. 4. (T.Fe) in the slag before and after reduction in each test. (Online version in color.) Fig. 3. Schematic diagram of the furnace in Test 3. (Online version in color.) Fig. 5. (P 2O 5) in the slag before and after reduction in each test. (Online version in color.) 2018 ISIJ 1936

4 or more, in spite of not using gas bubbling or stirring. On the other hand, in Test 2, that is, in the open-type furnace, the reduction degrees of (T.Fe) and (P 2 O 5 ) are much lower than those of Test 1 and 3. The reason is discussed in a later section. Figure 6 shows the slag and metal temperature at each heat in Test 1 and 3. They were measured after reduction by immersing the disposable probe with built-in thermocouple automatically to an adequate level. The measurement position was at about 1/2 radius of the furnace and 300 and 100 mm above the furnace bed for the slag and metal temperatures respectively, where metal bath depth was 200 to 250 mm in both Tests. The target metal temperature was K. There was a large temperature gradient in the metal and slag bath. The slag temperature was always higher than the metal temperature, because there was no mixing between slag and metal and the heat transfer to the metal phase was poor. The difference of temperature depends on the excess of the power supplied over the required heat. This feature is considered desirable for a reduction process, if the reaction occurs mainly in the high temperature slag layer. Figure 7 shows the relationship between the phosphorus distribution ratio and (T.Fe) of each heat after reduction. As a reference, two lines obtained by Healy s Eq. (3) 10) are shown in the diagram for the conditions of and K, and an average value of 37.5% as (CaO). According to Fig. 6, most of the slag temperatures of Test 1 and Test 3 were between and K. Fig. 6. Relationship between slag and metal temperature. (Online version in color.) P log %CaO P T... (3) 25. log % TFe. There is a correlation between the phosphorus distribution ratio and (T.Fe), even in such a low (T.Fe) region, though the deviation is not small. It appears that (P)/[P] tends to be somewhat lower than the value estimated by Healy s equation. This deviation arises from the gap of temperature and oxygen potential in the slag and metal, in addition to the variation of (CaO). However, existence of a temperature gap raises the equilibrium line of (P)/[P], when a high slag temperature is used for calculation, because the temperature that affects the phosphorus equilibrium at the interface should be lower than the slag temperature. In the same way, the oxygen potential gap lowers the equilibrium line of (P)/[P], if (T.Fe) in slag is used for calculation. Because the oxygen potential, which affects the equilibrium, is lower than that of the slag. However the gap is considered to be small in this case, because the (T.Fe) level is low enough. Another factor which lowers the actual (P)/[P] value is the reaction site. If the main reduction occurs in the high temperature region below the arc jet in slag, the reducing condition is much better than that at the bulk interface between the slag and metal. In such a case, it is possible that the reduction proceeds beyond the apparent equilibrium level. Therefore, in the case of DC arc furnaces it can be said that a high reduction efficiency can be obtained without vigorous stirring, or in other words, the stirring effects in DC arc furnaces is sufficient for the reduction of iron oxide or phosphate in slag Accumulation of Phosphorus in Metal Figure 8 shows the behavior of the phosphorus content during each heat after reduction. P 2 O 5 in slag is easily reduced and phosphorus content in the metal is raised linearly. Slag A was used from Heat 1 to 9, and 21 to 28, and Slag B was used from Heat 11 to 18. At Heat 10, 19 and 20, iron ore fines were used for a special purpose. The P 2 O 5 content of Slag A is 0.49% and that of Slag B is 1.42%. Therefore, the slope between Heat 1 to 9 is gentle, and the slope between Heat 11 to 18 is steep. It is notable that the phosphorus content reaches 0.9% and is almost saturated Fig. 7. Relationship between phosphorus distribution ratio and (T.Fe). (Online version in color.) Fig. 8. Behavior of each metal composition at heat end during the campaign test in Test 1. (Online version in color.) ISIJ

5 from Heat 21 to 28. The ratio of phosphorus and iron content in Slag A is 0.21 : 18.7, corresponding to 1.1% of the phosphorus content in iron. This is quite reasonable, taking the reduction efficiency into account. That is, when the phosphorus content in the metal is low enough, it increases almost linearly by slag reduction. However, once it reaches to a certain level of phosphorus content determined by the composition of material slag, the increase ceases and the phosphorus content becomes saturated Property of Slag Foaming It is necessary to prevent slag foaming and overflow in the reduction process, both during and after molten slag charging, because (T.Fe) increases rapidly when the molten slag is charged. What causes the slag foaming was investigated in Test 1. The behavior of carbon and silicon content at each heat after reduction and the timing of slag foaming are shown in Fig. 8. Whenever boiling occurs, [C] in metal drops, that is, decarburization occurs. In other words, the operation is stable as long as reduction takes place in the slag phase and there is little interaction between slag and metal. Figure 9 shows the carbon addition ratio to the stoichiometric amount of carbon for slag reduction in each heat. It is found that slag foaming occurs, only when the amount of added carbon does not exceed a value 1.2-times greater than the stoichiometric amount. Of course, the threshold is not universal and it depends upon the carbon content in the metal, for example. During the first four heats slag foaming did not occur, though the amount of added carbon did not reach a value 1.2-times greater than the stoichiometric amount. In this period, silicon content was at a fairly high level, as shown in Fig. 8. Therefore we consider the reason that the silicon in metal reduced the slag at the interface without CO generation. Here, the increase in silicon content would be caused by delay of melt down of the bulk metal in the bottom area. Before start up, 70 kg of FeSi was put on the cold iron as deoxidizer. It was melted down in the Fig. 9. Carbon addition rate at each heat in Test 1. (Online version in color.) Fig. 10. Behavior of (T.Fe) and [C] during separate feeding of iron ore and carbon in Test 1. (Online version in color.) 2018 ISIJ 1938

6 early stage, but it took long time for the bulk metal to melt down to the bottom layer. Thus the silicon content in the molten metal region was very high for a few heats, which was caused by poor stirring of the metal layer in the DC arc furnace. From these results, it can be concluded that the key factor for preventing slag foaming is to suppress the metal decarburization. Therefore, the interaction at the interface between the slag and metal should be suppressed and (T.Fe) in the slag should be kept at a low level by adding the appropriate amount of excess carbon. To observe the reduction of the molten slag bath with high (T.Fe) by carbon addition in Test 1, iron ore (pellet feed) and carbon (anthracite) were added separately. The result is shown in Fig. 10. First 200 kg of iron ore was added three times with the rate of 400, 600 and 800 kg/h. After, 40 kg of carbon was added three times at a rate of 100 kg/h. During iron ore addition, decarburization did not occur unexpectedly, but (T.Fe) in the slag was increased. Only when (T.Fe) reached about 5%, was the carbon content decreased, accompanied by foaming. As soon as the feed material was switched from iron ore to carbon, foaming stopped and (T.Fe) decreased very quickly. Then, the carbon content in the metal began to rise, where, 120 kg of coke corresponds to 3.6% of [C] in metal. However, the actual increase of carbon content in the metal was 0.25%, which corresponds to only less than 7% of the added carbon. These results suggest that the feed material does not react with metal directly. Added iron oxide only increases the (T.Fe) in the slag, and the added carbon remains in slag layer and reacts with oxides mainly. These results also suggest that decarburization of metal at the interface promotes the slag foaming. According to Ogawa et al., 11) the CO bubbles formed at the interface by decarburization are very fine and they tend to accumulate in the slag phase due to their low rising velocity. This can well explain the observed phenomena Influence of the Atmosphere in the Furnace In Test 2, the influence of the air atmosphere on the reduction was investigated. Figure 11 shows the behavior of (T.Fe) and [C] during and after the period of coke addition in Test 2. Here, coke was added through the top lance with the height of 500 mm. Though (T.Fe) decreased and [C] increased while the coke was added, once coke addition ceased, (T.Fe) began to increase and [C] began to decrease obviously. This means that reduction and oxidation proceed in parallel during carbon addition under the air atmosphere in the open-type furnace. Once carbon addition ceases, the oxidation rate surpasses the reduction rate. As a result, (T.Fe) is increased for a period of time and decarburization occurred in the metal. The mechanism of oxidation by air above the slag is considered to be as follows. Oxygen above the slag oxidizes from Fe 2+ to Fe 3+, and the formed Fe 3+ is transferred to the interface between the slag and metal, where Fe 3+ is reduced by carbon or iron in the metal. As a result, (T.Fe) increases and [C] decreases, as if the oxygen penetrated through the slag phase to the metal layer. Therefore, it is important to prevent air sucking into the furnace and to keep the oxygen partial pressure above the slag surface as low as possible, to achieve efficient slag reduction. Figure 12 shows the relationship between (T.Fe) and [C] after reduction treatment in Test 1, Test 2, and Test 3. From the diagram, three features should be highlighted. The first is that (T.Fe) can be lowered even when [C] is as low as 2%. The second is that the upper limit of (T.Fe) increases as carbon content is decreased. The third is that the lower limit of (T.Fe) obtained in Test 2 is about 3% and higher than that obtained in Test 1 and 3. The first feature suggests that the main reduction occurs in the slag layer and carbon content does not affect the reduction degree. The variation of (T.Fe) is caused by the test conditions, such as unit carbon consumption or sampling timing. The second feature suggests that higher (T.Fe) slag can exist stably in a lower [C] region. This property is desirable for the reduction process with hot slag charging, because after hot slag charging, (T.Fe) in slag rises temporarily. If [C] in the metal is kept at a low level, metal decarburization and slag foaming are suppressed and FeO can be reduced to a satisfactory level, even under low [C] conditions. The third feature corresponds to the result of Fig. 4. The reason is thought to be the difference of oxygen partial pressure in the furnace as described above. There are two other different conditions between Test 2 and Test 1 and 3. One is the difference in scale. The furnace in Test 2 was of commercial size. The inner diameter of the furnace was 3.8 and 1.9 times larger than that of Test 1 and 3, respectively. However, the decrease in (T.Fe) was nearly complete Fig. 11. Behavior of (T.Fe) and [C] during and after coke addition in Test 2. (Online version in color.) Fig. 12. Relationship between (T.Fe) and [C] in open and closed type furnaces. (Online version in color.) ISIJ

7 after 20 minutes, though the amount of added carbon was sufficiently high. Therefore this is not likely. Second is the difference in carbon content. In Test 2 the content was 1.5 to 2%, whereas it was 2 to 4.5% in Test 1 and 3. However there is a remarkable difference in (T.Fe) at around 2% carbon content in Test 2 and Test 1 and 3. Therefore, the most important factor is the atmosphere in the furnace. Figure 13 also shows the phenomena that illustrate the effect of introduced air into the furnace. Each time the temperature is measured, the probe hole at the furnace cover is opened, and air is sucked into the furnace from the probe hole. The diagram shows the behavior of the CO 2 concentration in off-gas and slag surface level measured by a micro wave sensor. At this time, no material was fed to the furnace and a certain level of power was supplied to maintain the bath temperature. Each time the temperature was measured, the CO 2 concentration increased and showed a peak. The height of the peak decreased gradually with time. It is thought that the carbon particles floating on the slag surface were oxidized by the introduced air and CO gas was generated for a short time. The decrease of the peak height indicates that the carbon drifting in the slag layer is consumed. Concerning the behavior of the slag surface level, ten to thirty minutes after the CO 2 peak the slag level began to rise and maintained a certain level. This means that oxygen from the introduced air penetrated the slag layer and decarburization occurred at the interface. As a result, the generated foamy slag raised the slag surface level. The foamy slag settled down quickly, when the next measurement was about to start and the electric power was cut off. Thus it is envisaged that the main reaction zone is a hot spot beneath the arc frame, and when it disappears by extinguishment of the arc frame, the reaction also ceases quickly. 4. Simulation of the Flow Pattern 4.1. Flow Pattern in the Furnace To investigate the flow pattern of slag and metal in the Fig. 13. Influence of the air suction into the furnace in Test 3. (Online version in color.) Fig. 14. Boundary conditions for the MHD (Magneto-hydro-dynamics) simulation. (Online version in color.) 2018 ISIJ 1940

8 furnace, a Magneto-hydrodynamics (MHD) simulation was performed using the commercial code FLUENT by using user defined functions (UDF). The model was based on the configuration of the DC arc furnace used in Test 3, which is shown with boundary conditions in Fig. 14. The calculation was made not using a half cylindrical model but a 3-dimensional full model. The inner diameter of the furnace was mm and the thickness of the slag and metal layer were both 200 mm. The input power was 2.5 MW. Figures 15 and 16 show the calculation results of the flow pattern in the slag and metal layer. A downward stream is formed at the center in both the slag and metal layers due to Lorentz forces. Another stream in the radial direction occurs on the slag surface, which is caused by the gas drag force. There are two main circulating flows in the slag layer and it is reinforced by DC arc jet rather than Ac arc jet, because both Lorentz and gas drag forces are larger in the case of DC arc jet. The interface between slag and metal is flat. Though there is a counter-flow at the interface near the center, the flow rate is not very high. Therefore the interaction between the slag and metal seems to be very small Characteristics of DC arc Furnace According to the flow patterns and the metallurgical results, the characteristics of the DC arc furnace can be considered as follows. Some of the added carbon particles go directly through the hot zone at the center, where most of the carbon particles react with slag quickly and are exhausted. (T.Fe) in slag decreases rapidly and the remaining unreacted carbon particles continue to drift in the slag layer. The drifting carbon suppresses the increase of (T.Fe), and weak interaction between the slag and metal suppresses decarburization of the metal. They both take a role in preventing slag foaming. Of course, decarburization occurs at the interface simultaneously with the dissolution of carbon particle. However, once (TFe) is lowered to some extent, carbon dissolution prevails over decarburization and carbon content begins to increase. If air exists above the slag surface, carbon particles floating on the slag surface would be burnt. Otherwise carbon particles continue to drift in the slag phase for a long time and finally dissolve to the metal or reduce the slag. In the metal phase, the stirring effect is very poor, which is detrimental to melting solid material. Fig. 15. Flow pattern in the slag and metal layer. Fig. 16. Flow pattern in slag layer ISIJ

9 However, it does not disturb the slag reduction, because the main reduction occurs not at the interface but in the slag phase. Reduced iron and phosphorus are transferred to metal phase without the aid of stirring. That is, metal phase is just a reservoir of reduced metal in the arc furnace. Certainly, there is a temperature gap between the slag and metal, but the high temperature in the slag raises the reduction efficiency and the gap is diminished by removal of heat from the slag layer due to the slag reduction. The DC arc furnace provides a hot spot in the slag phase as an ideal reduction site and provides slag circulation that transports carbon particles to the hot spot and homogenizes the slag phase. Slag reduction occurs mainly between drifting carbon particles and ambient slag at the hot spot region in the slag phase. Therefore, there is no need to increase the area of the interface by stirring or to raise the carbon content in the metal to the saturation level for improving the reduction efficiency. They are both effective at suppressing the reaction at the interface and preventing slag foaming. These characteristics of DC arc furnaces are suitable for the slag reduction process, especially with hot slag charging, in which slag foaming should always be controlled. 5. Conclusion To reduce the converter slag and recover valuable elements such as iron and phosphorus by direct use of molten slag, pilot and commercial scale test operations were performed using closed- and open-type DC arc furnaces. In parallel, a simulation of the flow patterns in the furnace was also carried out. Based on both results, the following conclusions were obtained. -More than 90% reduction of Fe x O and P 2 O 5 can be attained in DC arc furnace without gas stirring. -Slag foaming tends to occur, when slag reacts with carbon in the metal at the interface. It is effective in keeping (T.Fe) and [C] at low levels for suppression of slag foaming. -In the DC arc furnace, reduction takes place mainly between drifting carbon particles and ambient slag at the hot region in the slag phase. Therefore, intensive stirring at the interface or saturation of carbon in the metal is not necessary, which is effective for suppressing slag foaming. -Prevention of air sucking into the furnace is important, because air in the furnace deteriorates the slag reduction efficiency by reoxidation. -Closed-type DC arc furnaces are suitable for the reduction process with hot slag charging, because of their high reduction efficiency and ability to control slag foaming. Acknowledgements This study was partly based on results obtained from a project subsidized by the New Energy and Industrial Technology Development Organization(NEDO). We thank Dr. Guozhu Ye and Dr. Michael Lindvall in Swerea MEFOS, Dr. Rodney Jones, Mr. Glen Denton, Dr. Quinn Reynolds and other researchers in MINTEK in South Africa, project members of test operation at Muroran Works, engineers of Mitsubishi Steel MFG and all of our colleagues who made great efforts to construct the pilot plant and to carry out the test operations. Their contributions and comments are gratefully acknowledged. REFERENCES 1) S. Kubodera, T. Koyama, R. Ando and R. Kondo: Trans. Iron Steel Inst. Jpn., 19 (1979), ) H. Ohtake and K. Matsubae: Rin no jiten, Asakura Publishing, Tokyo, (2017), ) M. Tschudin, K. Brotzmann and C. Günther: Proc. Recycling and Waste Treatment in Mineral and Metal Processing: Technical and Economic Aspects, Vol. 2, TMS, Warrendale, PA, (2002), ) T. Yamamoto and M. Nakamoto: Proc. 1st Int. Conf. on Energy and Material Efficiency and CO 2 Reduction in the Steel Industry, ISIJ, Tokyo, (2017), ) G. Ye, E. Burström, M. Kuhn and J. Piret: Scand. J. Metall., 32 (2003), 7. 6) T. Harada, H. Hirata, T. Arai, H. Fukumura, T. Toh, G. Ye and M. Lindvall: Proc. 6th Int. Cong. on the Science and Technology of Steelmaking, The Chinese Society for Metals(CSM)/China Machine Press, Beijing, (2015), ) H. Hirata, T. Harada, T. Arai, H. Fukumura, M. Aono and S. Nishino: CAMP-ISIJ, 168 (2014), 779, CD-ROM. 8) T. Harada, H. Hirata, T. Yamazaki and T. Arai: CAMP-ISIJ, 174 (2017), 756, CD-ROM. 9) T. Harada, H. Hirata, T. Arai and T. Yamazaki: Proc. 1st Int. Conf. on Energy and Material Efficiency and CO 2 Reduction in the Steel Industry, ISIJ, Tokyo, (2017), ) G. Healy: J. Iron Steel Inst., 208 (1970), ) Y. Ogawa and N. Tokumitsu: Tetsu-to-Hagané, 87 (2001), ISIJ 1942