Behavior of Nonmetallic Inclusions in Front of the Solid-Liquid Interface in Low-Carbon Steels SEI KIMURA, Y. NABESHIMA, K. NAKAJIMA and S. MIZOGUCHI The present study is concerned with the interaction phenomena of nonmetallic inclusions in front of a moving solid-liquid interface. The in situ observation was done in a high-temperature experiment by using a laser microscope. Alumina inclusions in an aluminum-killed steel with low oxygen content exhibited the well-known clustering behavior. The velocity of the advancing interface first increased while approaching the particle, but became stagnant during engulfment and increased again after that. Alumina-magnesia complex inclusions in a magnesium-added steel with high oxygen content were very finely dispersed in the molten pool. These inclusions escaped from the advancing interface during solidification, but gathered again at the retreating interface during remelting. The tiny inclusions were thought to behave just as tracer particles of a local flow. The velocity of particles was measured on a video image, and the significant acceleration or deceleration was found near the interface. It was concluded that the flow was induced by the Marangoni effect due to the local difference in temperature and oxygen content in front of the interface, particularly in the case of a higher oxygen content. However, the flow was weak in the case of a low oxygen content.

I. INTRODUCTION

AMONG various types of nonmetallic inclusions, oxide and sulphide inclusions have been thought harmful for common steels. However, there are some positive roles in these inclusions. Oxide inclusions act as the trapping sites of hydrogen atoms in enameled steel and prevent the coating from stripping off. Sulfide inclusions also improve the machinability of free-cutting steels. Furthermore, it has been well known in welding that the tiny oxide inclusions act as the nucleation sites for very fine acicular ferrite crystals in the bond metal, providing very good ductility.[1–5] In the rolled steel products too, many fine intragranular ferrite crystals nucleate at some oxide inclusions inside of austenite grains. This technique is based on the concept of “Oxide metallurgy,”[6] in which an important role is given to oxide inclusions as inoculants for the heterogeneous nucleation of the phase transformation and the precipitation. Therefore, the purpose of the present study is to understand the phenomena of oxide particles at the solid-liquid interface in order to disperse them during solidification. There have been many studies performed to date, but the majority of these dealt with aluminum alloys or other materials in low-temperature experiments.[7–15] The only exception is the work done for steel by Emi and his coworkers.[16,17] In the present article, a novel result of in situ observation carried out on steel by using a laser microscope will be reported in detail. II. EXPERIMENTS A confocal scanning laser microscope was used to carry out the experiment of an in situ observation of nonmetallic SEI KIMURA, Research Fellow dispatched from Kobe Steel Company, Kakogawa, Japan, Y. NABESHIMA, Postgraduate Student, K. NAKAJIMA, Associate Professor, and S. MIZOGUCHI, Professor, are with the Institute for Advanced Materials Processing, Tohoku University, Sendai 980-8577, Japan. This article is based on a presentation made in the “Geoffrey Belton Memorial Symposium,” held in January 2000, in Sydney, Australia, under the joint sponsorship of ISS and TMS. METALLURGICAL AND MATERIALS TRANSACTIONS B

inclusions in front of the solid-liquid interface. The principle and the method of operation of the laser microscope have been described in detail elsewhere.[16,17] Two samples were used. Sample A is a normal low-carbon, aluminum-killed steel cut out from the continuously cast slab. Sample B is a low-carbon steel of high oxygen content specially made in the laboratory. The 100 g sample A was remelted under Ar gas flow in an alumina crucible in a 5 kW electric resistance furnace. Subsequently, the 5 g Fe-10 mass pct Mg pressed powder cake was added to the molten sample A, and the melt was cooled by switching off the power. When Mg metal was added to the melt, strong boiling took place, and the melt absorbed oxygen from ambient air entering from the top port. Sample B, of high oxygen content, was, thus, obtained. Chemical analysis was made twice for sample B, before and after the in situ observation experiment, in order to confirm the oxygen content. Table I shows the results of the chemical analysis. A small piece of each sample was machined into a disc (4.3 mm in diameter and 2 mm in height), mirror polished, and melted in an alumina crucible (5.5 mm o.d. and 4.5 mm i.d.) under ultra-high-purity Ar gas. The temperature was measured at the bottom of the holder of a crucible. The power was controlled to always keep the temperature near the liquidus. Special attention was paid to not melt the sample completely, but to leave a thin solid shell of steel at the periphery of the crucible. This solid shell prevented nonmetallic inclusions existing in the melt from being absorbed into the crucible. In order to observe the behavior of nonmetallic inclusions at the solid-liquid interface, the temperature of the melt was slightly decreased by decreasing the power carefully. Then, the solidification started and the solid-liquid interface advanced. In the next experiment, the temperature of the melt was slightly increased by increasing the power carefully. This time the remelting took place and the interface retreated. The movement of inclusions in front of the moving interface was, thus, observed at a magnification of up to 2100 times with resolution of 0.5 mm. The images were viewed on a computer monitor and recorded on videotape VOLUME 31B, OCTOBER 2000—1013

Table I. Chemical Composition of Specimens (Mass Percent) A (before) B (before) B (after)

0.04 0.04 0.04

,0.01 ,0.01 ,0.01

0.19 0.02 0.02

Fig. 1—Typical example of pure alumina inclusions showing clustering (sample A).

0.006 0.006 0.006

0.006 0.006 0.006

0.032 0.002 0.002

,0.0001 ,0.0001 ,0.0001

0.0021 0.1314 0.0581

Fig. 2—Typical example of alumina-magnesia inclusions showing no interaction (sample B).

at 1/30 a-second intervals. The position of each small particle and the interface was traced on the video pictures at the 1/ 30-second interval. Thus, the velocity of the particles could be measured as a function of the time from the start and the distance from the moving interface. After the observation, the sample was gas quenched. The chemical composition of the inclusions on a quenched specimen was analyzed by scanning electron microscopy and electron probe microanalysis (EPMA). III. RESULTS A. Comparison of the Behavior of Alumina Inclusions and Alumina-Magnesia Complex Inclusions (Samples A and B) Figure 1 shows a typical example of alumina inclusions clustering each other. Three particles can be seen at 0 seconds. The largest particle suddenly attracts first the larger particle, A, at 0.033 seconds, and then the smallest particle, B, at 0.233 seconds. However, alumina-magnesia inclusions behave in an entirely different manner. Figure 2 shows the tiny particles uniformly dispersed on the molten pool, with no coagulating or clustering tendency at all. Figure 3 is the magnification of the particles on the surface of the sample after quenching. The composition of these particles was analyzed by EPMA and found to be 93 pct alumina-7 pct magnesia in mass. B. Engulfment of Pure Alumina Inclusions at the SolidLiquid Interface (Sample A) The sequence of “engulfment” of a single alumina inclusion at the interface is shown in Figure 4. When a particle 1014—VOLUME 31B, OCTOBER 2000

Fig. 3—Particles of alumina-magnesia inclusions after quenching (sample B).

arrives at a place near the interface, the flat shape of the interface is modified to make a bump in the shadow of the particle. The bump top is then made concave to form a narrow channel between the particle and the interface. The position of the advancing front was measured inside and outside the shadow of a particle and plotted with time, as shown in Figure 5. The slope of the line is the velocity. Note that the broken line shown in the figure is the apparent position of the front line above the particle after 30 seconds. There is a big gap in the position at this time when the engulfment is completed, because the front line jumps up suddenly from the bottom to the head of the particle. Therefore, this gap of about 30 mm in position corresponds to the diameter of the particle. Finally, the other part of the front line caches up with the apparent position of the front line over the particle. In order to avoid confusion and to show the actual velocity of the front line, the “effective” position METALLURGICAL AND MATERIALS TRANSACTIONS B

constant velocity of 1.6 to 2.3 mm/s. However, three regions are seen for the position inside the shadow. In the first region, the particle arrives at the front, and the advancing front velocity significantly increases due to the cooling effect. The mean velocity is 3.8 to 4.5 mm/s. In the second region, the interface is too close to advance further, because of microsegregation in the channel. The front velocity is as low as 0.1 to 0.3 mm/s during this period of engulfment. In the third region, the particle is completely engulfed and the front line passes over the particle. The front velocity recovers but is about 0.9 mm/s, smaller than the value outside the shadow. C. Repulsion of Alumina-Magnesia Inclusions from the Advancing Interface during Solidification (Sample B)

Fig. 4—Sequence of the engulfment of a single pure alumina inclusion at the advancing interface (sample A).

Figure 6 shows the sequence of the repulsion of aluminamagnesia inclusions from the advancing interface. The particles locating at the front line of the interface suddenly escape from the interface when it starts to move. Note the movement of the typical examples of four particles, A through D, within 0.7 seconds. In Figure 7, the distance from the interface of each particle of different size is plotted as a function of the time after the start of escaping. It looks as if the curves are parabolic. Figure 8 shows the velocity of the particles moving away from the interface as functions of the distance. Each plot is widely scattered and shows a zigzag motion. But, generally speaking, the particles are continuously accelerated near the interface. The velocity becomes very high, about 100 times greater than the velocity of the advancing interface of 1.7 mm/s. The value appears to be different for each particle, but it is not systematically dependent on the size. D. Attraction of Alumina-Magnesia Inclusions to the Retreating Interface during Remelting (Sample B) When the solid-liquid interface retreats in remelting, many inclusions suddenly gather and pile up at the interface, as shown in Figure 9. The typical examples of particles, A, B, and C are traced on the frame. Note that the time shown is negative, showing the past sequence of particle A reaching the interface at 0 seconds. The distance between the interface and the inclusions is plotted in Figure 10 as a function of the time before reaching the retreating interface. The curves, again, look roughly parabolic. Figure 11 shows the velocity of each particle plotted as a function of the distance. The data are widely scattered, but the deceleration is clear near the interface. The gathering speed is, again, roughly 100 times greater than that of the retreating interface (3.3 mm/s). IV. DISCUSSION

Fig. 5—Position of the front line of the advancing interface inside and outside the shadow of the particle (sample A).

was plotted by subtracting the diameter of the particle from the apparent position above the particle after the engulfment. It can be seen in Figure 5 that the interface always advances smoothly outside the shadow of a particle, with a METALLURGICAL AND MATERIALS TRANSACTIONS B

A. The Difference in the Behavior of Alumina Inclusions and Alumina-Magnesia Inclusions Yin et al.[16] clearly showed that the strong attraction between the solid alumina inclusions within the critical distance was due to the capillary force. The origin of the capillary force is the deformation of a meniscus surrounding a particle.[18] Therefore, the wettability between the particle and liquid is the most influential factor. According to the equilibrium phase diagram, the liquidus temperature of 93 VOLUME 31B, OCTOBER 2000—1015

Fig. 6—Sequence of the repulsion of alumina-magnesia inclusions from the advancing interface (sample B).

Fig. 7—Distance between the interface and the alumina-magnesia inclusions as the function of time after the start of escaping in solidification (sample B).

pct alumina-7 pct magnesia inclusions is as high as 1900 8C and, thus, the inclusions are solid at the melting temperature of steel. The angular shape of the quenched particles is the direct evidence. The contact angle of pure alumina with molten iron is about 135 deg, while that of pure magnesia is 1016—VOLUME 31B, OCTOBER 2000

Fig. 8—Velocity of the escaping particles as the function of the distance from the interface in solidification (sample B).

about 95 deg.[19] The datum has not been available for 93 pct alumina-7 pct magnesia, but it is certainly lower than that of pure alumina. The value may be even further lower, considering the effect of the high oxygen content in sample B. This reduction in contact angle will give rise to the better wettability, effectively preventing particles from clustering. METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 9—Sequence of the gathering of alumina-magnesia inclusions to the retreating interface (sample B).

Fig. 10—Distance between the interface and the inclusions as the function of time before reaching the retreating interface in remelting (sample B).

B. Marangoni Flow near the Moving Solid-Liquid Interface The modification of inclusions by itself does not account for the rapid movement of the particles in front of the advancing or the retreating interface. The movement is very rapid and quite localized near the interface. The effect of particle METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 11—Velocity of the gathering particles as the function of the distance from the interface in remelting (sample B).

size is not clear. The size (the mass) should affect the deformation of the meniscus.[10] However, this effect does not account for the present results. Apart from the conventional analyses, it is now considered that the particles are just the floating tracer, and the movement of particles is the result of the liquid flow carrying them. The most-possible explanation for such a local flow VOLUME 31B, OCTOBER 2000—1017

Table II. Properties of Steel Melt Melting point, Tm (K) Density, r (kg m23) Viscosity, m (kg m21 s21) Heat capacity, Cp (J K g21 K21) Thermal conductivity, l (W m21 K21) Thermal expansion coefficient, b (K21) Thermal diffusivity, a 5 l/r Cp (m2 s21) Diffusivity of oxygen, D (m2 s21) Prandtl number, Pr 5 n / a Schmidt number, Sc 5 n /D Equilibrium distribution coefficient of oxygen, k

­C 50 ­z Fig. 12—Area of the mathematical analysis.

1. Numerical analysis model When a gradient of temperature or concentration exists on the surface of a melt, the shear stress is induced due to the gradient of surface tension. The flow can be estimated by using the basic equations of (1) continuity, (2) momentum, (3) energy, and (4) mass transport. The following assumptions are also necessary. (1) No slip exists between the solid and liquid. (2) The heat flux is directed from the center of a melt pool to the periphery (Figure 12). (3) The temperature at the solid-liquid interface is kept constant at the melting point. (4) Instead of mass transfer on the surface, equilibrium partitioning of oxygen at the solid-liquid interface is considered. The interface is considered to be at equilibrium, and the third assumption is adopted since the rate of solidification or fusion is very small (only 1.7 or 3.3 mm/s) in the present experiment. The latent heat of solidification or fusion must be taken into account in order to estimate accurately the temperature profile in front of the interface. However, the overall heat flux from the hot spot to the periphery in the microscope is dominant in the present experiment, and the latent heat is just neglected for the first approximation. The numerical calculation was done for the liquid cylinder (1.25 mm in radius and 1.0 mm in height) with the boundary conditions given in the following equations. Boundary conditions at the liquid surface at the solidliquid interface: ­ur ­g ­g ­T 5 5 , uz 5 0 ­z ­r ­T ­r 2l

­T 5q ­z

1018—VOLUME 31B, OCTOBER 2000

[3]

At the solid-liquid interface:

will be the Marangoni flow that is induced by the local difference in the surface tension. It is due to the change in the temperature and concentration profile in front of the interface. Hence, a simple numerical calculation has been done in the area of analysis, as shown in Figure 12 by using the highly simplified marker and cell (HSMAC) method.

m

1808 7.03 3 103 6.93 3 1023 7.92 3 102 32.1 1.25 3 1024 5.77 3 1026 1.5 3 1028 0.171 65.7 0.022

[1] [2]

2D

ur 5 uz 5 0

[4]

T 5 Tm

[5]

­C ­C 5 2D 5 (1 2 k) Vi Ci ­r ­z

[6]

The physico-chemical properties and the dimensionless parameters are listed in Table II. The following special parameters are also introduced to make the numerical estimation: Rayleigh number Ra 5

gb DTa3 an

[7]

Thermal Marangoni number MaT 5

­g/­T DT a ma

[8]

Concentration Marangoni number MaC 5

­g/­C DC a mD

[9]

where ur is the velocity of melt flow in the r direction, uz is the velocity of melt flow in the z direction, g is the surface tension, l is the thermal conductivity, q is the heat flux, k is the equilibrium distribution coefficient, a is the radius of the cylinder, g is the acceleration of gravity, a is the thermal diffusivity, b is the thermal-expansion coefficient, m is the viscosity, n is the kinematic viscosity, Vi is the advancing velocity of the solid-liquid interface, C is the solute concentration, Ci is the solute concentration at the solidliquid interface, D is the diffusivity of solute, T is the temperature, Tm is the melting point, and DT and DC are deviations of T and C, respectively. 2. The result of numerical analysis At first, the effect of the gravitational convection can be ignored in this case, because the Rayleigh number is on the order of 1022 and is negligibly small compared to the thermal Marangoni number, which is on the order of 10. Second, Mac is in the order of 102 to 103 and is much greater than MaT. This means that the concentration difference–driven METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 13—(a) through (d ) Schematic illustration of Marangoni flow classified in the four cases.

Marangoni flow is dominant. However, there is an uncertainty in the very small value of oxygen diffusivity in the measurement. Since MaC is inversely proportional to the diffusivity, it should drastically change according to the small value of oxygen diffusivity. The profiles of temperature, concentration, and velocity due to thermal Marangoni are calculated in various cases. However, the concentration Marangoni number is not calculated in the same way in the present work. Based on the results, a schematic map is made, as illustrated in Figure 13. There are four cases. The first and the second cases are solidification and the third and the fourth cases are remelting. The upper half is the case of a higher oxygen content, and the lower half is the case of a lower oxygen content. Note the profiles of the surface tension, temperature, and oxygen concentration. Takiuchi et al.[20] have made it clear that the thermal gradient of surface tension changes from negative to positive according to the oxygen content in the molten iron. In case (a), where the oxygen content is high, ­g/­T . 0. The gradient of g due to the thermal and concentration profiles is negative in both cases. Hence, the added effect causes Marangoni flow to be strongly directed from the interface to the center. In the second case, (b), where the oxygen content is low, ­g/­T , 0. The thermal gradient of g is opposite the concentration gradient of g. Hence, the two effects are cancelled out, so that the overall flow may be weak or stagnant. In the third case, (c), where the oxygen is high, the effect of the concentration Marangoni flow becames dominant due to the dilution effect by remelting. Hence, the overall flow is strongly directed from the center to the interface. In the fourth case, (d), where the oxygen is low, both Marangoni flows are weak but the direction is METALLURGICAL AND MATERIALS TRANSACTIONS B

the same. The overall effect will be the weak flow from the center to the interface. Figure 14(a) shows the result of calculations of temperature and surface-velocity profiles in the case of 100 ppm oxygen. This applies for both solidification and remelting. The profile of velocity first sharply increases at the interface, reaches the maximum, and then decreases to zero toward the center. Note that the sign is negative and the flow direction is from the interface to the center. The range of the measurement is in the region very close to the interface on this diagram, while the range of calculation is much wider than the measurement. Comparing the measurement in Figure 8 to the calculation, it can be seen that both are in fairly good agreement as a tendency of acceleration. However a complete agreement is not obtained in the peak value and its peak distance. A further tuning of calculation and observation is still necessary. Figure 14(b) is the comparison of normalized concentration profiles for solidification and remelting. They are obviously opposite in sign. Since MaC À MaT , the concentration Marangoni flow will upset the thermal one in remelting. This is the reason why the particles rapidly gather at the interface during remelting, as shown in Figure 11. Similarly, Figure 15 shows the result of calculation for a very low oxygen content (10 ppm). The direction of thermal Marangoni flow is opposite that in Figure 14(a), because of the negative temperature dependency of g in a low oxygen concentration. The flow direction is from the center to the interface. The concentration profile is very small, but opposite in sign in solidification and remelting. Thus, the sum of both thermal and concentration Marangoni flow will be roughly stagnant in solidification and weak toward the interface in remelting. This is the reason why VOLUME 31B, OCTOBER 2000—1019

(a)

(a)

(b) Fig. 14—(a) and (b) Velocity, temperature, and concentration profile for the thermal Marangoni flow for high oxygen steel (bulk concentration, Cbulk 5 100 ppm).

the interaction between alumina particles and the interface in a low-oxygen sample was not at all as strong as that of alumina-magnesia inclusions in a high-oxygen sample. In other words, the Marangoni flow is weak, but the attraction of particles to each other is very strong and becomes dominant in aluminum-killed steel. V. CONCLUSIONS The behavior of alumina and 93 pct alumina-7 pct magnesia inclusions at the solidification front of steel was observed in the in situ experiment. The conclusions are as follows. 1. Alumina inclusions in low-oxygen steel easily coagulate 1020—VOLUME 31B, OCTOBER 2000

(b) Fig. 15—(a) and (b) Velocity, temperature, and concentration profile for the thermal Marangoni flow for low oxygen steel (bulk concentration, Cbulk 5 10 ppm).

2. 3. 4. 5.

with each other, but the movement at the interface was weak. The velocity of the advancing interface changes in due course due to “engulfment.” The 93 pct alumina-7 pct magnesia inclusions do not coagulate as alumina inclusions do. The inclusions escape very rapidly from the advancing interface, but gather again at the retreating interface. The rapid movement of particles is due to the Marangoni flow induced by the changes in both the thermal and the concentration profiles in front of the moving interface. REFERENCES

1. O. Grong and D.K. Matrock: Int. Met. Rev., 1986, vol. 31, pp. 27-48. METALLURGICAL AND MATERIALS TRANSACTIONS B

2. D.J. Abson: Weld. World, 1989, vol. 27, pp. 77-101. 3. J.M. Dowling, J.M. Corbett, and H.W. Kerr: Metall. Trans. A, 1986, vol. 17A, pp. 1611-23. 4. Z. Zhang and R.A. Farrar: Mater. Sci. Technol., 1996, vol. 12, pp. 237-60. 5. J.M. Gregg and H.K.D.H. Bhadeshia: Acta Mater., 1997, vol. 45, pp. 739-48. 6. J. Takamura and S. Mizoguchi: Proc. 6th Int. Iron and Steel Congr., IISC, Nagoya, Japan, 1990, vol. 1, pp. 591-604. 7. D.R. Uhlman and B. Chalmers: J. Appl. Phys., 1964, vol. 35, pp. 2986-93. 8. J. Poeschke and V. Rogge: J. Cryst. Growth, 1989, vol. 94, pp. 726-38. 9. C. Koerber, G. Rau, M.D. Cosman, and E.G. Cravalho: J. Cryst. Growth, 1985, vol. 72, pp. 649-62. 10. S. Mohanty, F.H. Samuel, and J.E. Gruzlesky: Metall. Trans. A, 1993, vol. 24A, pp. 1845-56. 11. D.M. Stefanescu, A. Moitra, A.S. Kacar, and B.K. Dhindaw: Metall. Trans. A, 1990, vol. 21A, pp. 231-39.

METALLURGICAL AND MATERIALS TRANSACTIONS B

12. D. Shangguan, S. Ahuja, and D.M. Stefanescu: Metall. Trans. A, 1992, vol. 23A, pp. 669-80. 13. F.R. Juretzko, B.K. Dhindaw, D.M. Stefanescu, S. Sen, and P.A. Curreli: Metall. Mater. Trans. A, 1998, vol. 29A, pp. 1691-96. 14. Q. Han and J.D. Hunt: J. Cryst. Growth, 1994, vol. 140, pp. 406-13. 15. K. Mukai and W. Lin: Tetsu-to-Hagane´, 1994, vol. 80, pp. 527-32. 16. H. Yin, H. Shibata, T. Emi, and M. Suzuki: Iron Steel Inst. Jpn. Int., 1997, vol. 37, pp. 936-45. 17. H. Shibata H. Yin, S. Yoshinaga, T. Emi, and M. Suzuki: Iron Steel Inst. Jpn. Int., 1998, vol. 38, pp. 149-56. 18. V.N. Paunov and P.A. Kralchevsky: J. Coll. Sci., 1993, vol. 157, pp. 100-12. 19. K. Ogino, A. Adachi, and K. Nogi: Tetsu-to-Hagane´, 1973, vol. 59, pp. 1237-44. 20. N. Takiuchi, T. Taniguchi, Y. Tanaka, N. Shinozaki, and K. Mukai: J. Jpn. Inst. Met., 1991, vol. 55, pp. 180-85.

VOLUME 31B, OCTOBER 2000—1021