Microstructural evolution on continuous casting of

0 downloads 0 Views 2MB Size Report
In the present work, the evolution of microstructure in the nickel based superalloy Inconel 713C was investigated for the vertical continuous casting of small ...
Microstructural evolution on continuous casting of nickel based superalloy Inconel* 71 3C F. ZupanicÏ, T. BoncÏina, A. KrizÏ man, and F. D. Tichelaar In the present work, the evolution of microstructure in the nickel based superalloy Inconel 713C was investigated for the vertical continuous casting of small cross-section rods (10 mm dia.), using several microstructural characterisation techniques. Microstructural evolution was greatly affected by the mould design, average casting speed, and distance from the rod surface, but the strongest in¯ uence was from the alternating drawing mode consisting of the drawing stroke, the resting period, and the reverse stroke, which caused periodic orientation changes of columnar g crystal grains. Combined with other casting parameters this determined the local solidi® cation conditions, in¯ uencing the formation and growth of g dendritic grains and primary and eutectic MC carbide, as well as the stress distribution in the solidi® ed shell, which caused shell deformation and the appearance of the inverse segregation and, occasionally, hot tears. A physical model is presented to explain the in¯ uence of casting parameters on microstructural evolution on the continuous casting of this alloy. MST/5043 Dr ZupanicÏ ([email protected]),Mr BoncÏina, and Dr KrizÏman are in the Faculty of Mechanical Engineering, University of Maribor, Smetanova 17, SI ± 2000 Maribor, Slovenia and Dr Tichelaar is in the Laboratory of Materials Science, Delft University of Technology, Rotterdamseweg 137, 2628 AL Delft, The Netherlands. Manuscript received 28 March 2001; accepted 1 August 2001. # 2002 IoM Communications Ltd.

Introduction According to one de® nition,1 the term continuous casting means all processes in which the castings are longer than the casting mould. However, the advantages of continuous casting apply only if the casting is much longer than the mould, and this is the case with most industrial continuous casting processes. The main bene® ts of continuous casting are increased productivity and quality of the products, as well as reduced production costs. Therefore, it is not surprising that the majority of semi® nished metal products are being cast continuously. For a long time, many alloys have been continuously cast in large quantities (e.g. aluminium, copper alloys), but some alloys have not yet been continuously cast commercially. The latter group includes the vacuum melted, cast nickel based superalloys, which are mainly used in high temperature applications in the automotive and aerospace industries. However, in recent years, activities have been initiated to apply continuous casting also to this alloy group,2 because of the high production costs and other drawbacks of the current technique, outlined below. The usual way of manufacturing these alloys is by vacuum induction melting (VIM) and casting the melt into steel moulds, under a vacuum or a protective atmosphere, in the form of remelted ingots or electrodes.3 In the existing conventional pouring process, considerable expense is involved in terms of mould equipment, mould assembly, and the steel tubes used for casting the alloy bar. The intricate and expensive refractory pieces that must be used, and the labour costs involved in preparing the moulds and breaking them down for stripping after casting, represent signi® cant additional costs. The conventional process results in much scrap material, because of the numerous cutoffs, bar ends, or runner system scrap. In addition, conventional cast bars will always have some porosity in the centre. As a consequence, during bar cutting, special precautions must be taken to prevent ingress of potentially harmful cutting debris into the central cavities. Conventionally cast bar routinely goes through a *Inconel is a tradename of the Inco group of companies.

DOI 10.1179/026708302225003640

® nal spot grinding operation to remove any small visible imperfections on the surface. Owing to these drawbacks of conventional casting, it is not surprising that alternatives to this process have been sought. One possibilitly would be continuous casting. The most important reason why the vacuum melted, cast nickel based superalloys have not been continuously cast until now is that no technology on the world market enabled both melting and continuous casting under a vacuum and/or a protective atmosphere at temperatures up to 1650 °C. The greatest problem was ensuring reliable attachment of the continuous casting device to the holding furnace. This has been solved recently by a modular lining of the holding furnace, consisting of high precision, injection moulded alumina modules. The second obstacle was that a low annual production of many nickel based superalloys did not justify investment in the continuous casting technology. In addition, there is no information in the open literature regarding whether, and if so, which cast nickel based superalloys are capable of being continuously cast. Therefore, it was decided to study the behaviour of some nickel based superalloys on continuous casting using laboratory scale equipment. The alloy Inconel 713C was chosen as the initial study alloy for two reasons. First, its annual production is relatively high, in comparison with other nickel based superalloys. Second, it possesses excellent high temperature strength and ductility,4 which are important factors in reliable continuous casting. In the conventional casting of remelted ingots, the as cast structure is of limited interest because it affects neither the casting procedure nor the remelting process.5 On the other hand, in continuous casting, microstructural control allows optimisation of the casting parameters, and increases the safety and reliability of the process. As a consequence, an understanding of the microstructural evolution and knowledge of the physical and mechanical properties at temperatures close to the alloy solidus temperature represent essential prerequisites for successful continuous casting. Therefore, the present study was conducted to investigate (a) the microstructural evolution during solidi® cation and (b) the in¯ uence of the casting parameters. Materials Science and Technology

July 2002 Vol. 18

811

812 ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C

Experimental procedure CONTINUOUS CASTING TRIALS AND CASTING CONDITIONS The chemical composition of the Inconel 713C used in the present work was (wt-%) Ni ± 13.15Cr ± 6 .04Al ± 0 .78Ti ± 2 .11Nb ± 4 .19Mo ± 0.15C ± 1 .42Fe ± 0 .13Si ± 0.012B ± 0 .25W. The casting conditions are given in Table 1. Continuous casting experiments were carried out using a pilot scale setup consisting of a Leybold Heraeus vacuum induction melting furnace and a Technica Guss vertical continuous caster (Fig. 1a). The standard practice of melting and continuous casting of nickel base superalloys adapted after initial trials was as follows. Before casting, a dummy bar was introduced through the cooler up to the zirconia nozzle, which had a slightly smaller diameter (9.7 mm) than that of the mould (10 mm). Then the alloy ( ~ 14 kg) was melted in an alumina crucible under a vacuum of approximately 10 ­ 3 ± 10 ­ 2 mbar and heated to the casting temperature, which was normally 100 K above the alloy liquidus temperature. As soon as the casting temperature was attained the vacuum chamber was ® lled with protective argon gas, and the withdrawal device began to pull the dummy bar out of the cooler in the casting direction. From the three different withdrawal modes available, uniform, pulsating, and alternating, only the alternating drawing mode (Fig. 1b) allowed reliable continuous casting: with uniform and pulsating modes, breakouts occurred after a few cycles. The alternating mode is often applied in similar continuous casting machines in commercial use.6 A casting cycle consisted of three sequential stages: the drawing stroke, the resting period, and the reverse stroke. The cycle length was de® ned as the difference between the length of the drawing stroke and the length of the reverse stroke. In addition to being affected by the drawing mode, the solidi® cation conditions can also be in¯ uenced by the melt superheat, the average casting speed, the use of a secondary cooler, etc.7 a continuous casting setup; b alternating drawing mode: LDS length of drawing stroke, LRS length of reverse stroke, CL cycle length; 1: drawing stroke; 2: resting period; 3: reverse stroke

CHARACTERISATION TECHNIQUES Specimen preparation was carried out using standard mechanical grinding and polishing techniques. Specimens for optical and scanning electron microscopy (SEM) were lightly etched by immersing into a solution consisting of 37 mL HCl and 6.3 g K 2 Cr2 O7 in 12.5 mL distilled water. Scanning electron microscope studies were carried out using a Jeol 840A instrument ® tted with a Link analytical energy dispersive X-ray (EDS) microanalyser. Thin foils for TEM investigation were made by electrolytic thinning using a solution of 10% perchloric acid and 20% glycerol in methanol. Table 1 Conditions of continuous casting experiments Casting condition

Length of draw, mm Duration of drawing stroke, s Resting time, s Length of reverse stroke, mm Duration of reverse stroke, s Casting speed, mm s ­ 1 Melt temperature, °C Water ¯ owrate, L min ­ 1 Inlet water temperature, °C Outlet water temperature, °C

1

2

3

4

5 0.353

7 0.443

8 0.470

8 0.470

0.15 0.30

0.15 0.30

0.15 0.30

0.10 0.30

0.063

0.063

0.063

0.063

8.33

10.17

11.17

12.33

1450 20

1420 20

1400 30

1400 30

30

38

38

37

30

43

43

43

Materials Science and Technology

July 2002 Vol. 18

1

Schematic representation of continuous casting setup and characteristics of alternating drawing mode

Electron transparent thin regions were scarce; therefore, specimens were additionally ion beam thinned. Transmission electron microscope studies were carried out using Philips CM30T and CM30UT-FEG (® eld emission gun) instruments. The Inconel 713C alloy usually solidi® ed with a dendritic morphology, with eutectic pools in the interdendritic spaces. . It was more convenient to estimate the cooling rate T using the secondary dendrite arm spacing k 2 rather than the primary dendrite arm spacing or size of the eutectic pools. Bhambri’s empirical relationship 8 between secondary dendrite arm spacing k 2 and local solidi® cation time tf was applied to Inconel 713C according to k

n 2 ~Atf ~A

DT TÇ

n

:

:

:

:

:

:

:

:

:

:

: (1)

with constant A~6 .79 610 ­ 6 m s ­ n and n~0 .43 taken from Ref. 8, and the solidi® cation range DT of 65 K determined using differential thermal analysis.9 The secondary arm spacing k 2 was measured, and local cooling rates calculated using equation (1).

Results and discussion SURFACE APPEARANCE The surface of the continuously cast rod was bright in appearance, with a characteristic metallic lustre. This

ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C 813

a casting condition 2; b casting condition 4

3 a photograph; b schematic representation: PWM primary witness mark, SWM secondary witness mark, CL cycle length, GW gap width, RD rod diameter, RDW rod diameter at primary witness mark

2 Surface appearance of continuously cast Inconel 713C

indicates that solidi® cation in the copper mould and cooling in the protective argon atmosphere effectively prevented formation of a scale on the surface. Therefore, subsequent grinding or shot blasting of the rod, which are typical steps in the current production of remelt ingots, would be unnecessary. Careful observation of the rod surface is important because it may help in both interpretation of the macrostructure and understanding of the processes taking place during solidi® cation, and their dependence on the casting conditions. Two types of surface marks were observed on the rod surface with the naked eye (Fig. 2); these are often termed `witness marks’.6 The primary witness marks were located periodically along the rod. The distance between them corresponded very closely to the cycle length. At the preset cycle lengths of 4 .7, 6 .7, and 7 .7 mm, the actual lengths were 4 .60, 6 .65, and 7 .53 mm, respectively. The difference between the preset and actual cycle lengths can be attributed to solidi® cation shrinkage and thermal contraction of the alloy. It was observed that, at the primary witness marks, the rod diameter was smaller (between 9 .3 and 9 .5 mm) than at other parts of the rod (9 .7 mm), implying that a primary witness mark was in fact a gap (Fig. 2). The bottom edge of the gap was sharp (except for the fastest cooling speed) and rotationally symmetrical around the rod axis. On the other hand, the top edge was rounded, and its distance from the bottom edge varied along the circumference. Therefore, the gap width varied from 0 .5 to 1 .2 mm, increasing with increasing casting speed. The secondary witness marks were observed between the primary marks as thin bands of a slightly different colour from other parts of the rod. These bands formed closed loops (Fig. 2). Their distance from the top edge of the primary witness mark varied around the axis, sometimes coinciding with the top edge (casting condition 1, 0 ± 2 mm; casting condition 4, 0 ± 3 mm). The rod diameter was only slightly smaller at a secondary witness mark than elsewhere.

MACROSTRUCTURE AND ORIENTATION OF c GRAINS Longitudinal cross sections of the rods clearly showed a periodic macrostructure (Fig. 3) caused by the alternating drawing mode. The wavelength of the periodic pattern corresponded closely to the cycle length. Figure 3 reveals the

Macrostructure of continuously cast Inconel 713C: r resting period + reverse stroke, d drawing stroke, p primary witness mark, s secondary witness mark

size and orientation of the c grains, as well as the size and location of the primary and secondary witness marks. At the rod surface a shallow chill zone was present, composed of randomly oriented crystal grains. The thickness of the chill zone increased slightly with increasing casting speed (also with the decreasing melt temperature). Otherwise, columnar grains prevailed. They extended from the chill zone to the rod centre, except for the highest casting speed (casting condition 4), in which case some coarse equiaxed grains were present at the rod centre. Within a casting cycle the orientation of c grains changed from orientations almost parallel to the radial direction (perpendicular to the rod surface) towards orientations more inclined to the casting direction. The latter grains in fact grew in the opposite direction to the casting direction. It is very likely that radially oriented grains formed and grew during the drawing stroke, whereas c grains inclined to the casting direction grew during the resting period. These two different regions are clearly shown in Fig. 3. The orientation of c grains is expressed by the angle h , de® ned as the angle between the casting direction and the growth direction of the dendrite trunks in a grain. According to the de® nition, for grains inclined towards the casting direction h lies between 0 and 90 ° (0 °5h 590 °), for those growing perpendicular to the surface h ~90 °, and for those inclined opposite to the casting direction h is between 90 and 180 ° (90 °5h 5180 °). The orientation of columnar grains that formed and grew during the drawing stroke depended on the casting conditions. At the smallest casting speed of 8 .33 mm s ­ 1 (casting condition 1), the orientation of c grains was between 90 and 105 °; they were inclined in the direction opposite to the casting direction. For casting conditions 2 (casting speed 10.17 mm s ­ 1 ) the grains were oriented at h º 60 ° at the beginning of the drawing stroke, but during most of the drawing stroke they were almost perpendicular to the mould wall (h ~90°). At the highest casting speeds of 11.17 and 12.33 mm s ­ 1 (casting conditions 3 and 4), the angle h was always smaller than 90°, usually between 75 and 85°. The casting conditions also in¯ uenced the orientation of grains formed during the resting period and the reverse stroke. For casting conditions 1 and 2 the whole range of angles from h ~90 ° up to h ~180 ° was observed. At casting conditions 3 and 4 the range of orientations of columnar grains was narrower (from ~ 90° to 130 ± 140 °). In contrast with grains formed at the slowest casting speed, columnar grains that formed during casting conditions 3 and 4 usually did not originate from the rod surface, but were separated Materials Science and Technology

July 2002 Vol. 18

814 ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C

from it by a fairly thick layer of equiaxed grains. It is shown below, that these grains had formed at or close to the surface, but were later remelted during the next drawing stroke. The primary witness marks were positioned in the region of crystal grains inclined opposite to the casting direction, indicating that they must have formed during the resting period and the reverse stroke. The primary witness marks, especially their bottom edges were approximately at the same height on the both sides of the rod, but the gaps differed in shape and width. The secondary witness marks were located in the ® rst half of the drawing stroke, usually at its very beginning. It is evident that they were accompanied by a small depression on the rod surface.

MICROSTRUCTURE OF CONTINUOUSLY CAST INCONEL 713C At the surface was a 0 .2 ± 0. 5 mm thickness chill zone consisting of randomly oriented crystal grains. Figure 4 compares microstructures formed at various distances from the rod surface during most of the drawing stroke. Close to the surface, c grains with a cellular or cellular ± dendritic morphology were occasionally observed, implying a very high temperature gradient in the liquid phase during their growth. In these grains, MC carbide (where M indicates Nb, Mo, Ti, or Cr) was present in the form of elongated discrete particles, mainly aligned parallel to the dendrite trunks (Fig. 4a). The minimum secondary dendrite arm spacing in dendritic grains was 1 .4 mm, which suggests that the cooling rate was ~ 2350 K s ­ 1 as calculated by applying Bhambri’s relationship (equation (1)). In the chill zone, dendrites with a much larger secondary dendrite arm spacing ( ~ 10 mm) embedded in MC ± c eutectic were occasionally observed. At these sites a small deformation of the rod surface could be seen (gap deformation 10 ± 15 mm, length ~ 150 mm, depth 75 mm). These sites were mainly present in the ® rst part of the drawing stroke and their frequency increased with increasing casting speed. They were designated `tertiary witness marks’ because they had a similar structure to the secondary witness marks, but were much smaller. In the columnar grains extending from the chill zone up to the rod centre, the secondary arm spacing increased proportionally with distance from the rod surface. Table 2 gives the measured secondary dendrite arm spacings together with the calculated cooling rates. The results showed that the . local solidi® cation rate T at a distance of 0 .5 mm from the surface was 150 ± 250 K s ­ 1 and was ~ 20 K s ­ 1 at the centre of the rod. It is obvious that the distance from the surface had a dominant in¯ uence on the secondary dendrite arm spacing. Increasing the casting speed slightly decreased the secondary dendrite arm spacing, but its in¯ uence was not as pronounced. The MC carbides were usually present as a constituent of the MC ± c eutectic, but primary MC carbides were also observed in the interdendritic space. Some of them had been entrapped by the c dendrites during growth. As expected from results of previous studies,8 the size of the eutectic pools increased with increasing distance from the rod surface, i.e. with decreasing cooling rate (Fig. 4). The microstructure around the primary witness marks showed several peculiarities (Fig. 5). In addition, there were

a cellular ± dendritic morphology, rod surface; b, c c dendrites with c ’ precipitates, primary MC carbide (arrow), MC ± c eutectic pools (M), 2.5 mm and 5 mm from rod surface, respectively

4

Microstructure of continuously cast rod at given distances from rod surface (SEM)

some signi® cant differences in microstructure between the various casting conditions, the main transition occurring between casting conditions 2 and 3. For casting conditions 1 and 2, the columnar grains beneath the primary witness marks were oriented in the direction opposite to the casting

. Table 2 Secondary dendrite arm spacings l2 and estimated cooling rates T Casting condition

Grain orientation

0.5 mm from surface . T, K s ­ 1 k 2 , mm

2.5 mm from surface . T, K s ­ 1 k 2 , mm

5 mm from surface . T, K s ­ 1 k 2 , mm

2 2 4 4

Perpendicular Inclined Perpendicular Inclined

4.5 5.2 3.7 5.8

6.6 6.2 6.1 7.7

10.6 10.7 10.1 11.3

Materials Science and Technology

169 121 267 94

July 2002 Vol. 18

69 80 83 49

23 23 26 20

ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C 815

a arrow indicates sharp border between chill zone and columnar grains, casting condition 1; b arrow indicates partially remelted columnar grain, casting condition 2; c remelted regions above and below primary witness mark, casting condition 4; d remelted region above primary witness mark, casting condition 4: arrows in c and d indicate remains of remelted columnar grains

5 Microstructure close to primary witness mark (P)

direction; however, they had the same characteristics as columnar grains formed during the drawing stroke. The secondary dendrite arm spacing increased with increasing distance from the surface; however, at the same distance from the surface, k 2 was slightly larger than in the grains growing during the drawing stroke (Table 2). In columnar grains formed above the primary witness mark the secondary dendrite arm spacing did not follow the same dependence. Namely, the crystal grains next to the chill zone (Fig. 5b, arrow) had a much larger k 2 ( ~ 10 mm; . T ~ 25 K s ­ 1 ), as would be expected from their remoteness from the surface; they appeared to be remelted. At the gap root, equiaxed dendrites embedded in MC ± c eutectic were present. In the chill zone above the primary witness mark, crystal grains were equiaxed, with a composition apparently

equal to the nominal composition of the alloy, but the grain size was smaller than in the chill zone formed during most of the drawing stroke (Fig. 5). For casting conditions 3 and 4 there was no sharp border between the chill zone and columnar grains above the primary witness marks (Fig. 5c and d). The equiaxed zone was between 200 and 300 mm in thickness; and, within the zone, some equiaxed grains with k 2 º 10 mm were observed. The angle h of grains growing from beneath the primary witness marks did not exceed 130 ° (Fig. 5 d). Furthermore, there often appeared beneath the primary witness marks a 0 .5 ± 1 mm thickness layer of equiaxed grains (Fig. 5 c), with coarser dendrite arm spacing k 2 ( ~ 10 mm). The secondary witness marks appeared at different distances from the primary marks, but always in the ® rst Materials Science and Technology

July 2002 Vol. 18

816 ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C

a inverse segregation, arrow indicates MC + c eutectic pool; b arrow indicates hot tear

6 Microstructure at secondary witness mark (S)

part of a drawing stroke. All secondary witness marks had some common characteristics (Fig. 6). At these marks the solid shell was deformed, resulting in a small depression (intrusion) approximately 400 ± 600 mm in length and 50 ± 100 mm in depth. Under the intrusion there was usually a region consisting of mainly equiaxed grains with k 2 º 10 mm, embedded in the interdendritic space consisting of MC ± c eutectic (arrow in Fig. 6 a). The dimension of the region was usually 100 ± 200 mm, with a tail, which protruded several hundred micrometres towards the centre of the rod. The tail consisted predominantly of MC ± c eutectic, and the dendrite tips in contact with the eutectic were rather coarse. The columnar grains embracing this region were usually heavily inclined, below the secondary

witness mark in the direction opposite to the casting direction, and, above, in the casting direction. In more than 95% of cases there were no visible cracks at the secondary witness mark. One is presented in Fig. 6b (arrow). This crack has de® nitely formed during solidi® cation, because it follows the dendrite shape; it is obviously a hot tear. The observed hot tears were fairly short (approximately 40 ± 50 mm) and always started at the rod surface. In some cases a heavy deformation was connected with the hot tear, but in others there was no visible deformation. For casting condition 2, TEM was used to investigate the size and distribution of c ’ precipitates at the rod centre and approximately 1 .5 mm from the rod surface (Fig. 7). Specimens were taken in the region formed during the drawing

a drawing stroke, 1.5 mm from rod surface; b drawing stroke, rod centre: inset shows selected area diffraction pattern, small spots are superre¯ ections of c ’ precipitates, large spots are re¯ ections of c matrix + c ’ precipitates

7 Images (TEM) at given locations for casting condition 2

Materials Science and Technology

July 2002 Vol. 18

ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C 817

8 Division of casting cycle into three stages: stage 1, resting period and reverse stroke; stage 2, beginning of drawing stroke; stage 3, remaining part of drawing stroke

stroke, as well in the region formed during the resting period and the reverse stroke. Table 3 gives the average size of the coarse precipitate population; the ® ner precipitate population (2 ± 5 nm) was observed only in the very narrow regions between the coarse precipitates (arrow in Fig. 7b). Precipitates at the rod centre were in both cases larger than those close to the surface. In addition, precipitates formed in the part of the rod that solidi® ed during the resting period and the reverse stroke were ~ 30% smaller than precipitates in other parts. Usually, the foil thickness was larger than the precipitate size; therefore, the precipitates overlapped, making quantitative determination of the precipitate density dif® cult. Nevertheless, based on micrographs of thin foil regions without much overlap of precipitates within the TEM foil, and based upon foil thickness measurements, it was concluded that precipitate density was at least within + 30% of the same density in the different areas. This was further con® rmed by microhardness measurements along the centreline of the dendrite trunks. The microhardness was apparently insensitive to distance from the rod surface, location during the casting cycle, and average casting speed (inclined dendritic grains 338 + 30 HV0 .01, perpendicular dendritic grains 317 + 16 HV0 .01).

(a)

(b)

(c)

(d)

MODEL OF PROCESSES TAKING PLACE DURING SOLIDIFICATION An attempt is made in the following to explain the microstructural evolution on continuous casting by dividing the whole casting cycle into three stages (Fig. 8). The ® rst stage comprises the holding period and the reverse stroke, the second stage is the beginning of the drawing stroke, where many transition phenomena occur, and the third stage represents the remaining drawing stroke. The third stage is the simplest to understand and is therefore dealt with ® rst. In stage 3 the rod is pulled downwards in the casting direction, and the melt comes continuously into direct contact with the Cu ± Be mould (Fig. 9a), resulting in high cooling rates and melt undercooling below the alloy liquidus temperature. In the undercooled melt (on the surface of the mould), copious nucleation of randomly oriented crystal grains occurs. Solidi® cation parameters, the most important being the growth rate R and the temperature gradient in the liquid G,1 0 determine the growth morphology, which is predominantly dendritic. 8 Since most of the heat is transferred through the mould wall, the crystal grains grow towards the rod centre. In the grain selection process, only crystal grains Table 3

Size of g’ precipitates at various locations for casting condition 2

Location

Average size, nm

Centre: drawing stroke 1.5 mm from rod surface: drawing stroke Centre: resting period and reverse stroke 1.5 mm from rod surface: resting period and reverse stroke

53 48 38 34

a melt comes into direct contact with mould, chill zone forms, columnar grains with [001] direction best aligned with heat ¯ ow direction grow up to rod centre; b columnar grains grow into inlet nozzle, step is formed on rod surface because diameter of inlet nozzle is smaller than diameter of rod, new grains are nucleated close to inlet nozzle; c melt ¯ owing into gap between mould and rod causes remelting and disintegration of some columnar grains, grain remains ¯ oat in melt and coarsen, they serve as nuclei upon subsequent cooling; d melt ¯ ows into gap between mould and rod without causing remelting of solid shell: at c and d primary witness marks form

9

Schematic representation of microstructure evolution at given stages of casting cycle: v(t) casting speed

with [001] axis most closely aligned with the heat ¯ ux direction can overgrow other less favourably oriented grains and grow up to the rod centre. The preferred orientation of a crystal grain depends on the particular situation, especially at which location in the mould the grain selection takes place, and on the shape of the solid/liquid interface. If the solid/liquid interface was at the top of the mould, and the grain selection occurred there, then the crystal grains would be growing at an angle h 490°. If the grain selection took place, for example, in the middle of the mould, then c grains would be growing perpendicular to the mould wall (i.e. h º 90 ° as in Fig. 3 a). On the other hand, if there were a deep melt pool, then the crystal grains would be growing in the direction of the coldest melt, that is towards the bottom of the melt pool. In this case the angle h would be less than 90° (see Fig. 3b). How can casting conditions in¯ uence the orientation of crystal grains? With increasing casting speed the amount of heat extracted through the mould wall and consequently the growth rate R remain approximately constant. This results in both the solid/liquid interface Materials Science and Technology

July 2002 Vol. 18

818 ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C (a)

(b)

a tensile stresses in rod cause shell deformation; b dendritic grains separate causing melt ¯ ow into this region, inverse segregation occurs because melt is enriched in positively and depleted in negatively segregating elements: if melt had not ® lled space between grains, hot tear would have formed

10

Schematic representation of formation of secondary or tertiary witness mark

moving downwards and the melt pool growing deeper. As a consequence the angle h would decrease, which is exactly what was experimentally observed (Fig. 3). Just the opposite is expected by increasing the axial cooling (e.g. by applying a secondary cooler) or by increasing the melt superheat. In the present experiments, a simultaneous increase of the casting speed and decrease of the melt superheat resulted in the formation of a deep melt pool, which, at casting speed 4, even led to the formation of equiaxed grains at the rod centre (Fig. 3b). As a result of solidi® cation shrinkage the solid shell separates from the mould. This decreases the heat ¯ ux through the mould wall, and, combined with the additional heat resistance because of the thicker shell, increases the solidi® cation time. Therefore, the primary and secondary arm spacing, as well as the eutectic pool size, increase with increasing distance from the surface. It is expected that these conditions are not changed signi® cantly by increasing the casting speed, and this was con® rmed by experiments (Fig. 3 and Table 2). Owing to the relative motion of the rod against the mould caused by the rod pulling and thermal contractions, friction forces occur, which give rise to stresses in the shell. These stresses are likely to be ampli® ed by the forces arising from rod acceleration. When the stresses overcome a critical stress they may cause plastic deformation of the shell (Fig. 10). On deformation, crystal grains can separate along grain boundaries, and also the distance between the trunks in a grain can increase.1 1 If the grains are not yet completely solid, then the pressure of the melt in the intergranular and interdendritic spaces falls, and the melt from the melt pool ¯ ows into these regions. Resistance to the melt ¯ ow represents a low permeability of the interdendritic or intergranular spaces and high melt viscosity.1 1 If resistance to the melt ¯ ow is too high, then hot tears may form; this usually happens in the sensitive region where the fraction of liquid phase lies between 0 .90 and 0 .99. 1 2 Otherwise, only inverse segregation occurs, because the melt ¯ owing into the interdendritic (or intergranular) region is enriched with positively (partition coef® cient k51) and depleted in negatively (k41) segregating elements owing to microsegregation. The melt ¯ ow can also cause the partial remelting of crystal grains, their disintegration, and coarsening of dendrite arms. It is expected that, with higher casting speed, the acceleration stresses increase and the shell thickness decreases, resulting in more frequent occurrence of the so called tertiary witness marks; this is in line with the above observations. Materials Science and Technology

July 2002 Vol. 18

At the beginning of stage 1 the rod stops. Since, during the previous drawing stroke, at least a thin shell has covered the entire mould surface, the melt is no longer allowed to come into direct contact with the mould. During the resting period the axial component of the heat ¯ ux becomes more important, and with competitive crystal growth, the grains with h 490 ° can overgrow other grains, because their [001] directions are best aligned with the heat ¯ ux direction. In the undercooled melt between the columnar grains and the nozzle, new grains can form, whose orientation is even better aligned with the axial direction (Fig. 9 b). In this stage, the c grains grow into the inlet nozzle which has a smaller inner diameter than the mould. As a result, a step is formed on the rod surface. Casting parameters can in¯ uence the length of grains grown into the inlet nozzle. The length can be increased by decreasing the melt superheat and increasing the resting period. In the present experiments a decrease of the melt superheat had a stronger effect than a shortening of the resting period (compare Fig. 3 a and b). The reverse stroke compensates for solidi® cation shrinkage and thermal contraction, breaking possible contacts between the mould and the rod. If the length of the reverse stroke were to exceed the solidi® cation shrinkage and thermal contraction, then rod deformation could occur, which was not observed in the present experiments. At the initial stage of the next drawing stroke (stage 2), the metallostatic pressure tries to force the melt into the gap between the mould and the already solid shell. As the melt penetrates the gap, it is cooled, and heat is extracted through the mould wall, as well as through the solid shell. If the melt solidi® es before ® lling the gap completely, a primary witness mark forms (Fig. 9 d, and Fig. 5). It is very likely that the cooling rate in the gap is very high, resulting in the formation of many small equiaxed grains (Fig. 5a). Heat being dissipated through the solid shell causes heating of the shell. The temperature increase depends at least on the melt superheat and the melt agitation. If only a small amount of heat is transferred from the melt in the gap to the solid shell, only a slight increase in temperature is expected. In this case, the grains next to the gap can be heated in the three phase liquid + MC + c region. Prolonged holding of c dendrites in this region causes coarsening of the secondary arms. In the present experiments, this happened with casting conditions 1 and 2 (Fig. 5 b). On the other hand, if the amount of heat transferred to the shell is high, the shell can be remelted, or at least the crystal grains that have grown close to the nozzle surface during the previous stage 1, can disintegrate (Fig. 9c). The parts of the grains that have not remelted completely ¯ oat in the melt and coarsen. When the temperature decreases again, they serve as growth centres (indicated by arrows in Fig. 5 c and d). The remelting happened frequently at higher casting speeds. At the beginning of the drawing stroke the forces resulting from rod acceleration are much stronger, and they cause more severe shell deformation than during stage 3. The process is expected to be qualitatively the same as explained for formation of the tertiary witness marks (Fig. 10). However, the deformed volume is much larger (Fig. 6 a), so it was possible in the present experiments to observe it at smaller magni® cations, or even with the naked eye (the secondary witness mark). As a result, hot tears were occasionally observed (Fig. 6 b). Nevertheless, no breakouts occurred, mainly only inverse segregation. This indicates that the alloy Inconel 713C is reasonably suitable for being continuously cast. This simple solidi® cation model shows that equalising the diameter of the inlet nozzle and the mould, or appropriately adapting the mould design, could eliminate the primary witness marks. Furthermore, the secondary witness marks are the result of shell deformation, which may in turn result in the appearance of inverse segregation and/or hot tears. When continuous casting the more sensitive alloys (more

ZupanicÏ et al. Microstructural evolution on casting of Inconel 713C 819

prone to hot tearing, for example with a broader solidi® cation range), it would be necessary to use mould material with a low coef® cient of friction to reduce stresses induced in the solid shell during drawing, as well as to decrease the maximum acceleration at the drawing stroke. With these changes it would also be possible to cast continuously using the uniform and pulsating drawing modes. In this case the macrostructure would be more uniform, similar to the structure formed during stage 3. With a decrease in casting speed and/or by changing the cooling system to obtain a higher temperature gradient in the axial direction, it would be possible to obtain mainly uniaxial growth of crystal grains during all stages of the continuous casting cycle and, therefore, increase the resistance of the alloy to hot tearing and breakouts.

Conclusion s In the present work, the microstructural evolution of the nickel based superalloy Inconel 713C was investigated for the vertical continuous casting of small cross-section rods (10 mm dia.), using several microstructural characterisation techniques. The results allow the following conclusions to be drawn. 1. Microstructural evolution was greatly affected by mould design, average casting speed, and distance from the rod surface, but the strongest in¯ uence was from the alternating drawing mode consisting of the drawing stroke, the resting period, and the reverse stroke. Combined with other casting parameters this determined the local solidi® cation conditions, in¯ uencing formation and growth of c dendritic grains and primary and eutectic MC carbide, and induced stresses in the solidifying shell. 2. Primary witness marks formed because of the unequal inner diameters of the inlet nozzle and the Cu ± Be mould, and could therefore be eliminated by proper mould design and/or by adjusting casting parameters. 3. The secondary witness marks arose from shell deformation during the drawing stroke, which caused inverse segregation and/or formation of hot tears. The latter could endanger the reliability of continuous casting. 4. The casting parameters used had only a minor in¯ uence on the size and distribution of c ’ precipitates.

Future work will be mainly directed towards optimisation of both mould design and casting parameters to (a) minimise stresses acting on the rod during casting, thus avoiding the appearance of secondary witness marks and making continuous casting more reliable, and (b) control conditions in the mould more tightly, hence achieving the required microstructure and properties.

Acknowledgements The authors would like to thank Derek Hendley from Ross & Catherall, Shef® eld, UK for providing the alloy and giving useful advice regarding the melting procedure.

References 1. e. herrmann and d. hoffman (eds.): `Handbook on continuous casting’ , V ± VI; 1980, DuÈ sseldorf, Aluminium-Verlag. 2. : Foundry Trade J., 1999, 173, 52 ± 55. 3. j. r. davies (ed.): `ASM specialty handbook, heat resistant materials’ , `Metallurgy, processing and properties of superalloys’ , 221 ± 254; 1997, Materials Park, OH, ASM International. 4. `Alloy digest, data on world wide metals and alloys, Inconel 713C’ ; February 1959, Upper Montclair, NJ, Engineering Alloy Digest. 5. m. durand-charre: `The microstructure of superalloys’ , 53; 1997, New York, Gordon & Breach Science Publishers. 6. j. zalner and s. e. taylor: Iron Steel Eng., February 1985, 37 ± 44. 7. i. anzel, a. c. kneissl, and a. krizman: Metallurgia, 1997, 51, 620 ± 624. 8. a. k. bhambri, t. z. kattamis, and j. e. morral: Metall. Trans. B, 1975, 6B, 523 ± 537. 9. f. zupanic, t. boncina, a. krizman, and f. d. tichelaar: J. Alloy. Compd., 2001, 329, 290 ± 297. 10. w. kurz and d. j. fisher: `Fundamentals of solidi® cation’ , 63 ± 92; 1989, Aedermannsdorf, Switzerland, Trans Tech Publications. 11. m. rappaz, j.-m. drezet, and m. gramaud: Metall. Mater. Trans. A, 1999, 30A, 449 ± 455. 12. i. farup and a. mo: Metall. Mater. Trans. A, 2000, 31A, 1461 ± 1472.

Materials Science and Technology

July 2002 Vol. 18