Chemical and Structural Characterization of Slag Compounds Formed in the Melting Processes to Produce Spheroidal Graphite Cast Irons

The aim of this research is to investigate the composition and phases present in the slags formed during the production of spheroidal graphite (SG) cast irons. This paper contains the results of the first part of such investigation which is focused on those slags generated in the induction furnace, i.e., solid slags formed on the melt surface and slags adhered to quartzite refractory lining. A group of slag samples of each type were obtained from melts prepared using different metallic charges. These samples were then characterized in order to determine their chemical and structural composition and to evaluate the influence of the raw materials used during the melting process on the amount of slag formed in each case. Three different techniques were used for analyzing the slag samples: X-ray fluorescence, X-ray diffraction and scanning electron microscopy with energy dispersive spectroscopy (EDS) microanalysis. Important differences have been detected among samples studied in this work that have revealed the detrimental role of aluminum on refractory linings. The knowledge has been successfully used to minimize the problems caused by adhesion of slags to refractory linings.


31
techniques were used for analyzing the slag samples: X-ray 32 fluorescence, X-ray diffraction and scanning electron 33 microscopy with EDS microanalysis. Important differences 34 have been detected among samples studied in this work 35 that have revealed the detrimental role of aluminum on 36 refractory linings. The obtained knowledge has been suc-37 cessfully used to minimize the problems caused by adhesion 38 of slags to refractory linings.

40
Keywords: spheroidal graphite cast irons, slag compounds, 41 induction furnace, refractory lining, X-ray diffraction, 42 X-ray fluorescence, scanning electron microscopy 43 44 45 Introduction 46 One of the main problems of spheroidal graphite (SG) cast 47 iron production is the formation of slag compounds that 48 can be formed in any of the various stages of the manu-49 facturing process. The consequence of the formation of 50 these slag compounds depends on the stage in which they 51 are formed. Slag inclusions can be found as an inclusion in 52 the manufactured parts, which is one of the most common 53 defects, 1,2 but they can also be found, as adhered products 54 in the refractory linings of melting furnaces. In this second 55 case, important reductions on the internal diameter of 56 refractory crucibles are detected which decrease the 57 effective capacity of melting furnaces. In addition to this 58 fact, the formation of such slag accumulations causes 59 cracks and erosion in the silica refractory lining and it 60 promotes failures on the inductor isolation systems. 3,4 The 61 other important source of active slag is the treatment of 62 base melts with magnesium ferroalloys. 5 These slag 63 compounds have to be properly removed from ladles in 64 order to minimize subsequent contamination problems on 65 pouring devices used in foundry plants. Otherwise such 66 slag products will be rapidly deposited on the refractory 67 lining, and a high risk of degradation will be present in the 68 pouring tools. In general, filters and/or proper filling sys-69 tems are commonly used in molds for avoiding the 70 appearance of slag inclusions on castings.

71
In order to avoid the problems related to the formation of 72 slag, it is important to know its chemical composition and 73 those phases present in the slag compound formed at dif-74 ferent stages of the production process of SG cast irons. 75 This information becomes useful to determine the affecting 76 chemical elements and which of them are present in the 77 different types of slag commonly found in SG cast iron 78 manufacture. Previous studies 6 on slag formed in spher-79 oidal and lamellar cast irons have shown that it mostly 80 consists of several oxides as FeO, MnO, SiO 2 , Al 2 O 3 and 123 Katz 8 showed the harmful effects of Fe-O-bearing slag 124 which promote the oxidation of valuable elements such as 125 C and Mn. On the other hand, this slag also sticks on the 126 silica refractory which is the most commonly used in 127 electric furnaces. Additions of SiC in the melting furnace 128 are recommended in this work to avoid this problem. It also 129 reported that such slag sources were sand residues present 130 in the foundry returns and oxidized compounds normally 131 found on raw materials surfaces in case of melting pro-132 cesses made with electric furnaces.
133 Considering the results of previous studies, two different 134 aims have been approached in this work: the study of the 135 chemical composition changes and of the different com-136 pounds found in slag samples depending on the raw 137 materials used for preparing melts and the minimization of 138 detrimental effects caused by slag which is stuck to the 139 refractory linings.

140
Experimental Work

141
In a first step, six different base melts were prepared in a 6-t 142 capacity medium frequency induction furnace (250 Hz, 143 4250 kW) using three different metallic charge composi-144 tions and two maximum temperatures (see Table 1). In all 145 cases, these raw materials were introduced in the furnace 146 crucible when a remaining amount of melt (around 4000 kg) 147 was present in it. The composition of each remaining melt 148 was the one that corresponds to a standard metallic charge 149 previously melted in the furnace (see Table 2). Note that the 150 FeSi alloy was only used with pig iron and steel crap-based 151 charges, while graphite and SiC were exclusively used for 152 preparing the melts with steel scrap. After melting and just 153 after achieving the maximum temperature, an alloy sample 154 and a slag sample were simultaneously extracted from each 155 base melt surface. Then, a second sampling was made after 156 skimming the surface of melts and then remaining them in 157 contact with open air for 45 min at each of the two selected 158 maximum temperatures. The first and the second groups of 159 samples will be identified as initial (I) samples and as final 160 (F) samples, respectively. These samples have been also 161 identified according to the metallic charge composition (PI 162 for pig iron, FR for foundry returns and SC for steel scrap) 163 and to the maximum temperature achieved during melting 164 process (00 for 1500°C (2732°F) and 45 for 1545°C 165 (2813°F), see Table 1). For instance, the PI00I code is used 166 to identify the samples that have been initially picked up 167 from the base melt prepared with pig iron at 1500°C 168 (2732°F).

169
In a second step, three slag samples adhered to different 170 refractory linings of the induction furnace have been 171 studied. These samples were removed from the refractory 172 surface at the end of the life span of the furnace linings 173 which duration was not systematically the expected one 174 due to failures on the inductor isolation system. In these 175 cases, the metallic charge composition was commonly used 176 in the manufacturing process of the foundry plant (Table 2) 177 and the melting procedure was similar to the one detailed 178 above. These samples will be identified in the text as 179 UC11, UC31 and UC32 where UC notation refers to ''usual 180 charges,'' the first number is the furnace identification and 181 the second one is the sample number.

182
In a third step of the present work, a set of experiments 183 were made in order to obtain more information about the 184 formation of slags stuck to the refractory linings. Thus, two 185 different metallic charges, one based on foundry returns 186 and the other one based on steel scrap, were exclusively 215 tetraborate (1/40 dilution) and 5 mg of lithium iodide as 216 surfactant factor was finally added to the mixture. This 217 mixture was melted at 1100°C (2012°F) in an induction 218 furnace (Perle'X-3) to obtain the 30 mm (1.18 inches) in 219 diameter pearls for XRF analysis. The fluorescence inten-220 sity was measured with a AXIOS Advanced wavelength 221 dispersion X-ray sequential spectrophotometer equipped 222 with a semiquantitative software program for elements 223 with atomic number higher than 9 (F), using as excitation 224 source a tube with a Rh anode. The quantification of the 225 elements is done using a calibration line previously made 226 with international reference geological samples pearl (di-227 lution 1/40) to analyze their chemical composition by XRF.

228
XRD was used to characterize the constituent phases 229 formed on each slag sample by means of a PANalytical 230 X'Pert PRO MPD q/q Bragg-Brentano powder diffrac-231 tometer 240 mm (9.45 inches) in radius. The slag samples 232 were crushed in an agate mortar until micrometer size.
233 Then, the sample was placed in a rectangular standard 234 holder 20 mm (0.79 inches) in length, 15 mm (0.59 inches) 235 in width and 1 mm (0.04 inches) in height in order to 236 obtain a flat surface by manual pressing of the powder 237 material using a glass plate.

238
Scanning electron microscopy and EDS microanalysis 239 (SEM-EDS) were used to corroborate the results obtained 240 from the two other techniques and to check the slag sam-241 ples' microstructure. For this purpose, the raw samples 242 were broken in small pieces and then were embedded in Author Proof 243 epoxy resin at room temperature. After conditioning the 244 embedded samples for metallographic inspection, they 245 were sputtered with carbon and then analyzed using a 246 ESEM Quanta 200 FEI, XTE 325/D8395 with observa-247 tion conditions of AV = 20.00 kV, WD = 10 mm 248 (0.39 inches) and intensity probe of 4.5 nA. Secondary 249 electron mode (SE Image) and backscattered electron mode 250 (BSE Image) were also used for characterizing the slag 251 samples.

Slags Generated in Induction Furnaces
254 After melting the metallic charge and then stopping the 255 induction power in the furnace, the slag formed is normally 256 found as scabs which are floating in the melt surface. These 257 slags are formed in the melt surface areas that are close to 258 the refractory lining, but then they become aggregated in a 259 crust form found in the central area. Once extracted from 260 the melt and then cooled at room temperature, the slag 261 shows an apparent vitreous morphology and a dark gray 262 color. A second inspection of melts surfaces after skim-263 ming process shows that more slags were gradually formed 264 in the surface of melts. Color and morphology of these 265 recent slag compounds are similar to the ones initially 266 obtained. The chemical composition ranges of the prepared 267 melts are shown in Table 3. It is noted that the alloy 268 obtained when using the steel scrap-based charge shows 269 aluminum, manganese and zinc contents that are signifi-270 cantly higher than the two others. On the other hand, alu-271 minum has been also added by means of SiC and FeSi 272 products that were used for adjusting the melt composition 273 in these cases (Table 1). However, the last product does not 274 seem to be very relevant in this sense as aluminum contents 275 are higher for alloys prepared with charges based on 276 returns than for those prepared using pig iron as main raw 277 material (Table 3). On the other hand, the use of returns 278 seems to be the cause of the highest silicon contents 279 observed on the base alloys investigated. The highest sulfur 280 levels are found when using pig iron-based charges due to 281 the high content of this element commonly found in this 282 raw material.

283
The results obtained from the XRF analyses performed on 284 the twelve slag samples collected from base irons can be 285 seen in Table 4. In this table, only those values higher than 286 1.00 wt% are shown. The results indicate that silicon oxide 287 is the main constituent of all these samples. Thus, SiO 2 is 288 shown to be the most important oxidation product in melts 289 as silicon is the main alloying element and this element 290 exhibits a high tendency to be oxidized. As can be 291 expected, the four slag samples obtained from the melt 292 prepared using foundry returns show the highest SiO 2 293 contents (Table 4) due to the sand adhered to these raw 294 materials and owing to the high silicon content found on 295 the corresponding base melts (see Table 3). The SiO 2 296 content is also higher in the slag samples extracted after 297 45 min than in those collected just after melting. This 298 result could be related to the progressive oxidation of sil-299 icon while keeping melts in the furnace at a given tem-300 perature. However, a similar effect due to the use of high 301 temperatures has not been detected.

302
The aluminum oxide content in the slag samples obtained 303 from the two steel scrap-based melts is higher than the ones 304 extracted from the pig iron-based melts. Although the exclu-305 sive SiC addition made when preparing the steel scrap-based 306 melt can be related to this fact, another important available 307 source of aluminum can be the own steel scrap. The Al 2 O 3 308 content is also higher in the samples obtained from melts that 309 were remaining in the furnace than the ones obtained just after 310 melting the metallic charges. This is probably due to the 311 progressive oxidation of aluminum present in the liquid alloy 312 during the remaining time. The high CaO levels found in the 313 slag samples collected from the melts prepared using steel 314 scraps and pig iron can be explained by the addition of the FeSi 315 ferroalloy. Clear tendencies when comparing the chemical 316 compositions of the slag samples obtained at 1500°C 317 (2732°F) or at 1545°C (2813°F) are not observed.

318
The XRD analyses carried out on all the slag samples 319 studied in the present work support the results obtained by 320 XRF. Floating slags generated in the induction furnace are 321 mainly composed by amorphous phases. The two main 322 phases with crystalline structures that have been found on 323 these samples are quartz and cristobalite (SiO 2 ). Quartz is 324 the stable phase at temperatures lower than 867°C 332 The other oxides detected in the XRF analyses but not 333 found as crystalline phases by XRD are contained in the 334 mentioned amorphous part of the slag. Two micrographs 335 included in Figure 1 show how the SiO 2 crystals grow in 336 the amorphous matrix of the FR45I and FR45F slag sam-337 ples. The SEM observation of these crystalline phases 338 shows dendritic-type morphologies typically found on 339 phases that nucleate and grow from the liquid alloy.  Figure 2 shows the XRD spectrum 359 and indexation of the three crystalline phases found in the 360 FR00I slag sample, i.e., quartz, cristobalite and pigeonite 361 (Mg,Ca,Fe)SiO 3 . On the other hand, Figure 3 illustrates the 362 crystalline growth of a silicate-type compound (a) found in 363 the same sample and the results obtained from the SEM-364 EDS microanalysis (b) of such compound.

365
Besides, some nondissolved particles of SiC and/or FeSi were 366 detected by XRD on samples obtained from melts prepared 367 using these additives (see Table 1). Subsequent SEM analyses 368 made on such samples confirmed these results.

369
As it has been stated before, the zinc content both in the 370 prepared melts and in the slag samples extracted after 371 melting becomes high when the steel scrap-based metallic 372 charges are used as the galvanized steel scrap from the 373 automotive industry is the main constituent of these char-374 ges. This fact is confirmed by means of the XRD analysis 375 made on the slag samples. Figure 4 shows the XRD 376 diffractogram recorded on the SC45F slag sample where 377 the minor crystalline phases are very easily identifiable 378 even though a high content of amorphous phases is found 379 in this sample. Thus, peaks of Zn 2 SiO 4 (willemite) and 380 ZnO are detected in addition to the ones that belong to 381 ZnAl 2 O 4 (gahnite), the latter only appearing in this slag 382 sample. Although all these zinc-bearing phases have not 383 been detected by SEM analysis because they are present in 384 a very minor amount, the microanalysis made on the 385 amorphous phase of the SC45F slag sample shows a small 386 peak of zinc (this sample contains the highest ZnO and 387 Al 2 O 3 contents of all samples analyzed) (see Table 4).

U N C O R R E C T E D P R O O F
394 total height. This affected area always remains in contact 395 with the liquid alloy even after tapping the furnace 396 according to the usual procedure of the plant. The thickness 397 of the slags found in this affected area ranges from 20 mm 398 (0.79 inches) to 150 mm (5.91 inches), and they seem to be 399 heavier than those floating slags directly obtained from 400 melts. Figure 5 shows a general view of a discharged 401 refractory lining of a furnace. The zone marked as 1 in this 402 figure is the worn part of the refractory lining. Notice that 403 the expected thickness of the used refractory lining can be 404 found in the upper levels (marked in Figure 5 with arrows).
405 The adhered slags were found below zone 1, and they 406 affect the whole section of the lining at this lower level 407 (zone 2 in Figure 5).

408
Quartzite refractory areas in contact with slag showed a 409 darker region with around 1 cm in thickness which is 410 marked by arrows in Figure 6a. A general view of a slag   Table 2). Here it is worthy to emphasize the rel-425 evant intensity of the aluminum peak even though this 426 element does not seem to be the most expectable one  432 Table 5 shows the chemical composition of the three slag 433 samples that were collected from the discharged refractory 434 linings. In this table, only the oxides with content higher 435 than 1.00 wt% were included. Surprisingly, all three sam-436 ples are mainly composed by Al 2 O 3 and MgO, while SiO 2 437 becomes now a minority oxide when comparing to data 438 include in Table 2. MgO and Al 2 O 3 are two of the oxides 439 included in Table 5 with the highest melting point, i.e.,

440
[2000°C (3632°F), so they should be more prone to be 441 stuck to the refractory material than the rest of possible 442 oxides. On the other hand, quartzite, i.e., the refractory 443 material used in the present work, is essentially composed 444 by SiO 2 which is considered as an acid oxide. Thus, one 445 can expect that the basicity and the amphoteric character-446 istics of MgO and Al 2 O 3 , respectively, also become a rel-447 evant cause of the reaction between these slags and the 448 refractory material.

449
The available source of magnesium seems to be the foun-450 dry returns used as raw materials; however, the sources of 451 aluminum are numerous. In this second case, possible 452 supplies are the use of additives as ferrosilicon, silicon 453 carbide, the use of steel scraps and also of foundry returns 454 as raw materials. According to this fact and to the high 455 Al 2 O 3 content found in all the slag samples obtained from 456 the refractory linings, it could be considered that aluminum 457 plays a very relevant role on refractory degradations in 458 electric furnaces and consequently on the life span reduc-459 tion in these devices. Table 5 also shows important 460 amounts of cerium oxide and lanthanum oxide in these slag 461 samples in comparison with those samples obtained 462 directly from melts. The presence of these two elements 463 should be related to the massive use of foundry returns as 464 raw materials in the melting furnace. Notice that these 465 returns are manufactured with the use of FeSiMg and of 466 inoculants which both contain rare earth elements.
467 Figure 8 shows a SEM micrograph and the corresponding 468 EDS microanalysis spectra obtained from four different 469 constituents found in the UC11 sample. The microanalysis 470 of the massive phase identified as 1 in this figure led to 471 record peaks of aluminum, magnesium and oxygen in 472 accordance with the results shown in Table 5. Notice that 473 any peak of silicon was not detected in this compound.
474 Another main phase (marked as 2) is composed by a group 475 of elements (calcium, cerium, lanthanum, silicon, sulfur  481 The XRD diffractogram shown in Figure 9 was obtained 482 from the UC32 slag sample. It can be seen that it contains a 483 much smaller amount of amorphous phases than the slag 484 samples obtained from the melts surface (see Figure 4). 485 The UC11 and UC31 samples exhibit a similar behavior. 486 This high crystalline degree must be related to the observed 487 heavy aspect of these slags when comparing to the floating 488 ones.

489
The most important crystalline phase found in the UC32 490 sample is the MgAl 2 O 4 (spinel) which is formed by 491 reaction between the two main oxides MgO and Al 2 O 3 492 present in these slags (Table 5). This result confirms the 493 relevant role of aluminum previously predicted in the 494 SEM inspections carried out on the slag-affected regions 495 of refractory (Figure 7). On the other hand, this phase 496 has been also detected in the SEM analysis of the UC11 497 slag sample shown in Figure 8 (phase 1). The crystalline 498 phase MgO (periclase) is also detected in the XRD index-499 ation shown in Figure 9. This fact indicates that an excess 500 of MgO which has not reacted with the Al 2 O 3 to form the 501 spinel is present in the UC32 sample. In fact, this sample 502 showed the highest MgO content (Table 5). Other minor 503 phases identified in Figure 9 are CaMgSiO 4 (monticellite) 504 and SiO 2 (quartz). The former compound has been also 505 detected in Figure 8 for the UC11 sample (phase 3).

U N C O R R E C T E D P R O O F
506 Comparing to the results obtained for the UC32 sample, the 507 XRD characterization of UC11 and UC31 samples also 508 showed a high crystalline degree, MgAl 2 O 4 (spinel) was 509 detected as the main crystalline phase and CaMgSiO 4 510 (monticellite) and (Mg,Fe) 2 SiO 4 (forsterite) were identified 511 as minor phases. It is worth nothing that CaMgSiO 4 and 512 CaS compounds had been already detected as phase 3 and 513 phase 4, respectively, in the SEM-ESD analysis performed 514 on the UC11 sample ( Figure 8). The (Mg,Fe) 2 SiO 4 (for-515 sterite) compound which is formed by the reaction between 516 MgO and SiO 2 is only present in the UC11 slag sample as 517 is shown the highest SiO 2 content (Table 5).
518 Origin of Slags Adhered to Refractory Linings 519 Once identified the spinel phase MgAl 2 O 4 as the main con-520 stituent of slags adhered to refractory materials, it is now 521 worthy to investigate the origin of this phase and some of its 522 influencing factors. As it has been described in the experi-523 mental section, two different metallic charge compositions 524 (mainly composed by foundry returns or by steel scrap) were 525 separately used during the whole life span of each refractory 526 lining of the furnace following a similar melting procedure. 527 Malfunctions owing to the presence of slag stuck to the 528 refractory lining were detected after 214 melting batches 529 when steel scrap-based charges were only used in the melting 530 furnace. However, no failure occurred after 724 melting 531 batches when exclusively using the return-based charges.
532 Another important difference is the amount of slags 533 adhered to the refractory linings at the end of their life 534 span. The lining where only steel scrap-based charges were 535 used for melting shows massive slags stuck to the entire 536 refractory ring located in the usual region described above.
537 However, only specific zones of the lining were found to be 538 affected when return-based charges were exclusively used 539 following a similar melting procedure. Thus, it can be 540 concluded that slag formation was more ''aggressive'' in 541 the first case, based on the use of steel scrap-based charges.

542
XRF chemical compositions of the two slag samples 543 collected from the linings are shown in Table 6 where 544 only those contents higher than 1.00 wt% are included. It 545 can be seen that the MgO, CeO 2 and La 2 O 3 contents are 546 higher for the sample obtained from the return-based 547 charges than for the one coming from the steel scrap-548 based charges. On the contrary, the Al 2 O 3 content is 549 much lower in the FR21 sample than in the SC31 sample.
550 These results are expected as foundry returns become a 551 notorious source of the three elements previously men-552 tioned (they were manufactured using a FeSiMg alloy and 553 inoculant), while steel scrap and the adjusting products 554 (SiC and FeSi) contain significant amounts of aluminum. 555 These auxiliary products should also be considered as the 556 source of Ca and Zn in case of the slag sample formed in 557 steel scrap-based melts.

558
Regarding the phases identified by XRD for these two slag 559 samples, the spinel MgAl 2 O 4 is the most abundant crys-560 talline phase on the FR21 sample (see the diffractogram 561 shown in Figure 10 below). Additionally, an important 562 amount of MgO (periclase) has been also found in this 563 sample. The crystalline phases detected on the SC31   Figure 10. XRD diffractogram and indexation of: the SC31 sample (above) and the FR21 sample (below). Figure 11. Schema of the steps proposed for the adhered slags formation in the quartzite refractory linings.  (Figure 10 above). In this case, 565 many phases that contain aluminum and calcium have been 566 found, being the most relevant the MgAl 2 O 4 spinel, Al 2 O 3 567 (corundum), CaAl 2 Si 2 O 8 (anorthite) and an aluminum-568 calcium oxide known as hibonite. These XRD outcomes 569 are in good agreement with the differences shown in 570 Table 6, all confirming the negative effect of aluminum on 571 life span of refractory linings.

International Journal of Metalcasting
572 A scheme that illustrates the mechanism and the probable 573 reactions involved in the formation of slags adhered to the 574 refractory linings is shown in Figure 11. Part of the alu-575 minum dissolved in the melt would react with the quartzite 576 giving aluminum oxide and silicon as final products. 577 Similarly, the magnesium dissolved can react with the 578 refractory material to obtain magnesium oxide and silicon. 579 Thus, these two oxides would be present close to the lining, 580 so they can react to form the spinel (MgAl 2 O 4 ) previously 581 characterized as the mainly phase of these detrimental slag 582 compounds.