The following article is a component of the July 1999 (vol. 51, no. 7) JOM and is presented as JOM-e. Such articles appear exclusively on the web and do not have print equivalents.

Observing and Measuring Solidification Phenomena at High Temperatures

TABLE OF CONTENTS
INTRODUCTION THE DOUBLE HOT THERMOCOUPLE TECHNIQUE CONFOCAL SCANNING LASER MICROSCOPY PHENOMENON OBSERVED WITH CSLM Carbon-Saturated Iron Low-Carbon Aluminum-Killed Iron DROPLET SOLIDIFICATION STUDIES ACKNOWLEDGEMENTS References

INTRODUCTION

Both announcements were historically significant. NSC, with partner Mitsubishi Heavy Industries (MHI), announced that their highly secret strip-casting project at the Hikari works was producing significant quantities (20,000 tonnes per month) of commercial-quality stainless steels. BHP and their partner IHI announced completion of their even more secret development of a similar scale project for low -carbon steels (Figure 1). Thus, by year end, it was clear that it was technologically possible to direct-cast 60 tonne heats of liquid steel and form coiled steel strip that is 1–3 mm thick, up to 1.345 m wide, and cast at speeds in excess of 60 m/min. In addition, it was shown that it is possible to develop a continuous-casting process for steel that produces a strip that is geometrically similar to a hot-rolled steel strip product. Both developers project that a commercial plant would be capable of producing between 300,000 tonnes and 500,000 tonnes per year.

Both casting technologies—the oscillating and the travelling mold, are based upon two completely separate technologies that result in the formation of the surface of the casting. The solidifying shell is formed by heat transfer through a molten oxide layer that separates the solidifying shell from the mold surface in the oscillating mold processes. In the twin-roll caster there is direct contact between the shell and the mold surface. Thus, shell formation and growth in oscillating casters is controlled by the physical characteristics of the oxide flux, while in the twin-roll caster it is controlled by contact resistance between the shell and the mold surface. Clearly, mold-heat fluxes are significantly higher in twin-roll casters (10–20 MW/m²) than in oscillating mold casters (1–5 MW/m²), and this results in significant differences in cast structure between the two processes.

Figure 1. Examples of twin-roll, cast, low-carbon steel coils cast by BHP/IHI, at Project M, Port Kembla, New South Wales. (Figure courtesy of BHP.)	Figure 2. A double- and single-hot thermocouple.

Although these two technologies are at the forefront of steel production, it is very difficult to improve or to determine the fundamental limitations of these processes as a detailed knowledge of the phenomena that control heat transfer, nucleation, and growth of the steel shell are not fully understood. Although the science of solidification is well developed,^1–3 the application of theoretical developments to plant operation is still an area of great challenge due to the complexity of the casting environment. One of the major issues in understanding solidification is to clearly determine the physical phenomena that occur during the various casting processes. For example, in the oscillating-mold processes, heat transfer through a liquid oxide film that is under a severe thermal gradient controls shell growth. In this case, the crystallization of the liquid oxide as it cools must be understood to determine the heat transfer and lubricational characteristics of the oxide film. In traveling-mold processes, the heat-transfer rate as the liquid contacts the mold surface controls surface quality and shell growth. Thus, the details of contact between a mold surface and a liquid control initial solidification phenomena.

The parameters controlling solidification in casting processes are specific to the details of the process. This has led to the development of a number of experimental techniques that are aimed at simulating actual casting environments. Three of these techniques are

• The double hot thermocouple technique (DHTT) that can be used to study crystallization phenomena in transparent oxides as a function of undercooling and under conditions of variable thermal gradient.
• The use of the confocal laser microscope to study the melting and solidification behavior of metals and oxides.
• The combination of video and mold instrumentation to allow the observation of undercooling and the calculation of initial heat-transfer rates upon contact between a metal droplet and a water-cooled copper mold surface.

THE DOUBLE HOT THERMOCOUPLE TECHNIQUE

a	b
Figure 3. (a) A schematic of the DHTT and (b) the experimental setup.

The mold slags are liquid oxides and easy glass formers under high cooling rates. Thus, it is necessary to describe the conditions under which glass formation is possible, the conditions for the initiation of solidification, the crystal morphology, chemistry and growth rate, and the time evolution of the fraction of solid. It is well known that the onset of crystallization in liquid slags must be a function of cooling rate and that to determine the solidification behavior of a liquid slag one must construct either isothermal time temperature transformation diagrams (TTT curves) or continuous cooling transformation diagrams (CCT curves). In addition, the growth rate, morphology, and solidified fraction of the slag under varying cooling rates are important in the determination of the effect of crystallization of the slag on heat transfer and rheology. Thus, a technique whereby the thermal field can be determined as the solidification process is observed was developed. This technique, the double hot thermocouple technique, combines the hot thermocouple technique with video observation and image analysis, allowing crystal-growth rates, morphologies, and solidified fractions to be determined under defined thermal conditions.

The concept of the single hot thermocouple technique dates back to the end of 19th century, and the hot thermocouple method itself was successfully developed by Ordway⁴ and Welch et al.⁵ in the 1950s. The progress of electronic development has made the hot thermocouple method easier and more reliable and has led to renewed interest in the technique.⁶ Ohta et al.⁷ applied the hot thermocouple method to the measurement of liquidus temperatures, the clarification of the existence of two phase regions in slags, and to the understanding of slag reactions. More recently (1993), Asayama et al.⁸ discussed glass formation in silicate slags using the hot thermocouple method.

Figure 4. A video clip of crystal formation in an undercooled oxide. (approx. 17Mb)	Figure 5. A video example of crystal instability due to fluid flow. (approx. 14Mb)

Ishii and Kashiwaya first developed the double hot thermocouple method and applied it to a microgravity experiment to determine the microstructural change of a superconducting oxide in 1992.^9–11 The technique was also used by Murayama et al.¹⁰ to study Marangoni flow in silicate slags under microgravity and by Kuranaga et al.¹¹ to measure the ultimate length of a silicate slag film to clarify the mechanism of the separation of a bubble from a liquid slag surface. The DHTT (Figure 2) was recently improved by Kashiwaya et al.^12,13 to allow the study of the thermal conditions that may be found in the mold of a continuous caster. The DHTT was developed from the hot-stage microscope. The hot thermocouple technique allows the sample to be mounted on a single hot thermocouple (SHTT) or between two thermocouples (DHTT). In Figure 2, the difference in set up between DHTT and SHTT can be seen. The DHTT makes it possible to set a desired temperature gradient between the two thermocouples, which can simulate the various thermal conditions a mold slag encounters in an operational continuous-casting mold.

Due to the low mass of the system (sample and thermocouple), high heating and cooling rates can be easily obtained. This is very useful for determination of TTT and CCT diagrams where fast cooling rates are required.

The experimental apparatus of the DHTT consists of two systems: an observation system and a thermocouple system. The observation system includes a microscope with a three charge couple device color video camera connected to a videocassette recorder. The thermocouple system includes the two thermocouples located in a reaction tube below the microscope, and the sample is placed between the thermocouples. Each thermocouple is connected to a separate thermocouple controller, which is a unique system to enable simultaneous measurement of temperature while heating a thermocouple. A computer controls the two thermocouple controllers making it possible to individually control heating and cooling conditions of the thermocouples. Figure 3 shows a schematic of the experimental apparatus and a photograph of the setup.

a	b	c
d	e	f
Figure 6. The fragmentation process at (a) t = 161 s, (b) t = 167 s, (c) t = 173 s, (d) t = 179 s, (e) t = 189 s, and (f) 199 s.

The effect of fluid flow also has a striking effect on the crystal stability. For example, it is often seen that growing crystals are not stable and break apart under the stresses imparted by the flow under conditions where a high rate of internal recirculation is observed. An example of this phenomenon can be seen in the video clip of Figure 5 and the photographs in Figure 6, where a growing equiaxed dendrite of calcium aluminate can be seen to fragment under the action of a shear flow that was generated by either Marangoni or natural convection. It is often noted that this mechanism leads to an increase in the number of crystals that form in the equiaxed zone. Thus, if one crystal nucleates and begins to grow, it spontaneously fragments under the action of fluid flow, and the fragments then also start to grow. The fragmentation thus leads to one nucleation event spawning numerous growing crystals. This effect is also seen when one has strong fluid flow across a dendritic array. For example, during melting crystal fragmentation leads to mass transfer from the solid phase to the hottest temperature in the system and enhanced melting rates. During growth, this phenomena is observed to lead to the formation of an equiaxed zone at temperatures just below the liquidus, where viscosities are low and fluid flow rates are high.

Figure 7. A CCT curve compared to TTT for an industrial mold flux.	Figure 8. A video clip of the effect of water vapor on high-temperature crystallization phenomena. (approx. 10Mb)	Figure 9. The CSLM setup.

The double hot thermocouple can also be used to study environmental effects on solidification. For example, in Figure 8, the precipitation of a high-temperature phase due to the introduction of water vapor to the sample is shown.

CONFOCAL SCANNING LASER MICROSCOPY

Table I. Microscope Specifications
Characteristic	Specification
Resolution	0.30 µm (1LM21W), 0.25 µm (1LM21H), field lens:at 100 times (NA=0.95)
Scanning Speed	Horizontal:15.73 kHz, Vertical: 60Hz, 2:1 Interlace
Light Source	Laser RED:He-Ne 632.8 µm (light source output 1.5 mW at maximum 0.1 mW when assembled) Normal lighting: Halogen lamp
Output Signal	(VIDEO SYNC) Composite
Frame Memory	910*485*8 bi*2 pieces (for images and surface profiling)
Zoom	Electronic zoom *2, *4
Slow Scanning	2,4,8,32,64,128,256,512 (reduces scanning speed at magnifications shown left when normal is set at 1.)
Critical Dimension	Measuring range: 0.30 µm or above (1LM21W), 0.25 µm or above (1LM21H)
Measurement	Repeatability (3): 0.03 µm
Surface Profiling	Measuring range: 0.1 µm to 5.7 µm (6.5 mm optionally available) Repeatability (3): 0.03 µm
Instrument Size	Camera head: 105*350*140 (W*H*D) mm 4.7 kg Control unit: 420*105*490 (W*H*D) mm 13.0 kg Monitor: 320*322*356 (W*H*D) mm 9.0 kg YM mirror base: Approx. 56.0 kg (according to specifications)
Power Source	AC115V 50/60Hz: A maximum of 8 A power supply is required when mounting full range of options
Power Consumption	Camera head: supplied by the control unit Control unit: 150 W at maximum Monitor: Approximately 37 W

Figure 10. A video illustration of the melting behavior of an Fe-C alloy. (approx. 10Mb)

Figure 11. A video example of the solidification behavior of an Fe-C alloy. (approx. 13Mb)

Figure 12. A video of the melting behavior of a low-carbon steel. (approx. 21Mb)

Figure 13. A video example of a high-temperature phase transformation. (approx. 12Mb)

Figure 14. A video of high-temperature growth of a carbide in a low-carbon steel. (approx. 26Mb)

Although confocal microscopy has been used quite extensively in biological research,¹⁴ it is only recently that confocal optics combined with a laser has been utilized for metallurgical studies. To the best of the authors' knowledge, there are currently only two systems in the world that are equipped to study phenomenon under steel-making conditions; these are located in Tohoku University (Sendai, Japan) and the University of Wollongong (Australia). Work related to clean-steel production and other metallurgical applications has been pioneered and developed by T. Emi, formerly at Tohoku University (Japan) and currently SIPA professor at the Royal Institute of Technology, and co-workers.

Real-time investigation of crystal growth in the Fe-C alloys were carried out in CSLM by observing the melt surface under argon with a temperature gradient of a few K · mm^–1 across the surface.¹⁵ The investigators were able to observe and measure, among other things, that the limits in the growth rate when -solidification resulted in transitions from planar to cellular and then dendritic growth; planar to cellular transition occurs through sequential perturbations at 30–65 µm intervals in the advancing front; and the cell-tip stability observed in Fe-83C was in good agreement with the theory proposed in literature, and that nonmetallic inclusions could be seen to sometimes get engulfed into the solidifying front.

Direct observations of collision, agglomeration, and cluster formation of Al₂O₃ and CaO-Al₂O₃ inclusions on the surface of molten-steel samples where carried out under argon atmosphere^16,17 at 1,673 K. It was concluded in the case of Al₂O₃-Al₂O₃ particle interaction that clustering occurred sequentially through the formation of intermediate aggregates, loose structured dendrites, and finally densification to more compact units through sintering. A strong long-range attraction (10^–16 N) due to a capillary attraction force between the solid particles and the gas/melt interface was responsible for the agglomeration. The magnitude and reach of the force increased with increasing Al₂O₃ particle size, but was not affected by the bulk chemistry of the steel melt. In the case of CaO-Al₂O₃ and CaO-Al₂O₃-SiO₂ inclusions, it was found that for solid particles, the collision was caused by capillary attraction (as in the case of Al₂O₃- Al₂O₃ particles); when a liquid inclusion touched a solid cluster, it would spread and cover the solid cluster; and liquid particles did not attract each other due to capillary attraction and merged together only by being pushed by the solidifying front. In all cases thermocapillary flow was found not to play an important role in agglomeration or cluster formation.

The solubility of manganese and sulfur in Fe-Ni melts were studied by observing the formation of MnS on the surface of Fe-Ni alloys with CSLM.¹⁸ MnS was found to precipitate heterogeneously on Al₂O₃ inclusions and grow to either pyramidal- or rod-like shapes. Since the cooling rate did not have a strong influence on the nucleation temperature and since the temperatures for nucleation and dissolution (during heating) were close, the reaction was considered close to equilibrium. The equilibrium constant calculated from the CSLM observations were, however, somewhat larger than what had been reported in literature.

CSLM was also used for studying the interaction of nonmetallic inclusion particles with the advancing melt/solid steel interface.¹⁹ Beyond a critical-growth velocity, the solidifying front engulfed the particle by forming a bump perturbing out of the interface. The critical velocity above which engulfment occurred and below which particle pushing occurred for a particle with radius R was found to be v = 60/R for solid particles and 23/R for liquid particles.

Recently, Dippenar et al.²⁰ reported an in-situ study of the solidification of steel. The relation between cell-tip radius and the growth rate required to form dendrites during solidification of carbon steel that was found through CSLM was in agreement with theoretical predictions. Morphological instability and finger-like growth was observed during the solid state to transformation. Furthermore, the peritectic reaction rate in carbon steel was measured, and the rate-limiting step was attributed to carbon diffusion.

PHENOMENON OBSERVED WITH CSLM

Carbon-Saturated Iron

Figure 11's video clip illustrates the solidification of a carbon-saturated iron sample observed from the top. The crystal density varied with undercooling, with the expected outcome that a higher undercooling resulted in higher grain density. The film, however, shows the case for a relatively low undercooling, thus allowing the grain evolution more time before impingement. Nonmetallic inclusions are pushed ahead of the evolving grains.

Low-Carbon Aluminum-Killed Iron

a	b
Figure 15. (a) A schematic and (b) photograph of the experimental setup for interfacial heat transfer behavior.

After solidification, the sample contained drops of a liquid phase on the solid-metal surface. The liquid drops could be either a low melting slag phase or highly segregated molten liquid. The fact that they were semitransparent, however, suggests that it may be a slag phase. Figure 13 shows the surface after solidification, when most of the liquid drops have disappeared. A solid-state transformation front proceeds, starting from top left.

The sample is shown in Figure 14 after solidification, with various nonmetallic inclusions clearly visible on the surface. An almost perfectly spherical particle can be seen on the upper lefthand corner. A plate- or rod-like phase is seen to occur with a certain periodicity, originating from a grain boundary. It is possible that this is a carbide phase. This film shows the growth of one of these phases. Thus, CSLM can be used to track the formation of precipitates in liquids as well as solids in real time and real conditions.

DROPLET SOLIDIFICATION STUDIES

a

b

Figure 16. (a) Measured mold and chill temperatures vs. time and (b) heat flux vs. time at 90° superheat for droplet solidification studies.

Various investigators have measured rapid heat transfer between solidifying metals and copper molds by monitoring the surface temperature of the casting with a pyrometer^21–23 or a photo diode sensor.²⁴ L. Strezov et al.²⁵ investigated the initial heat-transfer behavior using a substrate embedded in a inclined moving paddle similar to melt/roll contacting geometry of the meniscus region of a twin-roll caster. Measurements in a prototype twin-roll caster and laboratory strip casters have also been conducted.^26–28 While considerable uncertainties remain in properly understanding and predicting the rapid heat transfer, it has been shown that:

• A fast data processing system is key for any experimental study on rapid heat transfer.^24,27
• There is a peak in the heat-transfer coefficient.^24,25
• This peak decreases with increasing superheat, but increases with higher mold roughness.²⁵
• The maximum profile depth (Pt) and mean height are better parameters than Ra to evaluate the effect of superheat on cast surface profile.^29–31
• Phenomena that occur within the first 20 ms of contact between a metal and a mold can control subsequent heat-transfer behavior.²⁷

Figure 17. A video of the direct observation of droplet solidification. (approx. 6Mb)

In order to study the phenomenon under controlled conditions an apparatus was designed at Carnegie Mellon University for millisecond resolution of interfacial heat-transfer behavior. The basic approach was to let a sample of molten metal fall on a copper mold. As shown in Figure 15, the experimental setup consists of a copper mold, quartz cylinder, induction furnace, gas purification system, and data acquisition system. The important features in the apparatus are a rapid data acquisition system that handles millisecond resolution; an elaborate thermocouple design inside the mold to capture the surface temperature; and a gas purification system to control the atmosphere. An example of the measured results is shown in Figure 16.

In order to conduct the measurement properly, it is critical to control the fall of the droplet. The process had to be monitored continuously in real time and conditions to ensure that the fall was occurring as expected; a video recording system was designed to record the process as it happened. An example of the process can be seen in the film on Figure 17. It is noteworthy that the shell solidification front and thermal front can be studied with this technique. This novel approach of simultaneous in-situ observation and measurement of rapid heat transfer enables a coupling between the interfacial heat transfer and droplet solidification.

ACKNOWLEDGEMENTS

References
1. W.A. Tiller, The Science of Crystallization: Microscopic Interfacial Phenomena (New York: Cambridge University Press, 1991).
2. W.A. Tiller, The Science of Crystallization: Macroscopic Phenomena and Defect Generation (New York: Cambridge University Press, 1991).
3. W. Kurtz and D.J. Fisher, Fundamentals of Solidification (Clausthal-Zellerfeld, Germany: Trans Tech Publications, 1986).
4. F. Ordway, J. Res. Natl. Bur. Stand., 48 (1952), p. 512.
5. J.H. Welch, J. Sci. Instrm., 31 (1954), p. 458.
6. T. Yanagase et al., Proc. of Inter. Conf. Sci. and Tech. Iron and Steel (Tokyo: ISIJ, 1971), p. 516.
7. Y. Ohta, K. Morinaga, and T. Yanagase, Bulletin Japan Inst. Metal., 19 (1980), p. 139.
8. E. Asayama, H. Takebe, and K. Morinaga, ISIJ International, 33, (1993), p. 233.
9. Y. Kashiwaya, D. Jukawa, and K. Ishii, JASMAC-8 (1992).
10. Y. Murayama, Y. Kashiwaya, and K. Ishii, CAMP-ISIJ, 7(1994), p. 891.
11. T. Kuranaga, Y. Kashiwaya, and K. Ishii, Int'l. Symp. on Adv. Mat. and Tech. for 21st Cent. (Sendai, Japan: JIM, 1995).
12. Y. Kashiwaya, C.E. Cicutti, and A.W. Cramb, ISIJ International, 38 (1998) (4), p. 357.
13. Y. Kashiwaya et al., ISIJ International, 38 (4) (1998), pp. 348–356.
14. C.J.R. Sheppard, Confocal Laser Scanning Microscopy (Heidelburg, Germany: Springer-Verlag, 1997).
15. H. Chikama et al., Materials Transactions, JIM, 37 (4) (1996), pp. 620–626.
16. H. Yin et al., ISIJ International, 37 (10) (1997), pp. 936–945.
17. H. Yin et al., ISIJ International, 37 (10) (1997), pp. 946–955.
18. N. Yuki, H. Shibata, and T. Emi, ISIJ International, 38 (4) (1998), pp. 317–323.
19. H. Shibata et al., ISIJ International, 38 (2) (1998), pp. 149–156.
20. R.J. Dippenaar, H. Shibata, and T. Emi, Materials 98, Vol. 2, Engineering Materials (Wollongong, Australia: Wollongong University Press, 1998), pp. 445–450.
21. H. Mizukami, T.Suzuki, and T. Umeda, Tetsu-to-Hagane, 77, No.10 (1991), pp. 1672–1679.
22. W. Liu, G.X. Wang, and E.F. Matthys, Int. J. Heat Mass Transfer, 38 (8) (1995), pp. 1387–1395.
23. G.X. Wang and E.F. Matthys, "On the Heat Transfer at the Interface Between a Solidifying Metal and a Solid Substrate," Melt Spinning, Strip Casting and Slab Casting, ed. E.F. Matthys and W.G. Truckner (Warrendale, PA: TMS, 1996), pp. 205–237.
24. H. Todoroki et al., Proc. of Electric Furnace Conf. (Warrendale, PA: ISS, 1996), pp. 585–590.
25. L. Strezov and J.Herbertson, ISIJ International, 38 (9) (1998), pp. 959–966.
26. K. Miyazawa et al., Int. Conf. On New Smelting Reduction and Near Net Shape Casting Technologies for Steel, vol. 2 (Pohang, Japan: ISIJ, 1990), pp. 745–750.
27. J.C. Grosjean et al., I&SM, (20 August 1993), pp. 27–32.
28. T. Mizoguchi and K.Miyazawa, Proc. of the First European Conf. on Adv. Materials and Processes, vol. 1 (Aachen, Euromat, 1989), p. 93–99.
29. H. Todoroki et al., "Solidification of Steel Droplets Against a Water Cooled Copper Mold," Proc. of the McLean Symposium (Warrendale, PA: ISS, 1998), pp. 155–175.
30. H. Todoroki, Y. Suzuki, and A.W. Cramb, "Initial Solidification of Iron, Nickel and Steel Droplets," I&SM, 26 (4) (1999), pp. 57–71.
31. H. Todoroki et al., "Initial Solidification of Iron, Nickel and Steel Droplets" (Paper presented at the 1998 TMS Annual Meeting, San Antonio, TX, 15150;19 February 1998).

C. Orrling, Y. Fang, N. Phinichka, S. Sridhar, and A.W. Cramb are with the Department of Materials Science and Engineering at Carnegie Mellon University.

For more information, contact A.W. Cramb, Carnegie Mellon University, Department of Materials Science and Engineering, Pittsburgh, Pennsylvania 15217; (412) 268-2964; e-mail cramb@andrew.cmu.edu.