About one stable and three metastable eutectic microconstituents in the Fe–W–C system

Abstract The Fe–W–C system has been the subject of numerous investigations and the findings have led to quite different interpretations in the iron-rich corner, justifying the need for new experimental studies. In the present study, some 12 compositions are prepared using both the powder metallurgy technique and the casting of melted alloys. DTA experiments, long time annealing supplemented by the phase separation technique and controlled solidification experiments are carried out. The micrographs of the corresponding specimens reveal significant sedimentation of WC carbides, an observation which must be carefully taken into account in interpreting the micrographs. The liquid/solid equilibria between 1423 and 1543 K are investigated and the corresponding partial isothermal sections of the Fe–W–C system are compared to the calculated phase equilibria based on thermochemical models. Four ternary eutectic microconstituents γ/M 6 C/cementite, γ/WC/cementite, γ/M 6 C/graphite and γ/WC/graphite, are observed by experiments. The microstructures are characterised and the four eutectic temperatures are established both experimentally and by calculations. The experimental eutectic temperatures lie within a range of 10 K but well below the temperatures predicted by thermodynamic modelling and the calculated temperatures within a range of 20 K. This explains why metastable microconstituents can so readily be formed.


Introduction
The Fe-W-C system may be considered as the fundamental system for interpreting the structure of many commercial alloys such as high-speed and tool steels. The system has therefore been the subject of extensive investigations. Solid state equilibria have been studied by numerous authors [1][2][3] but little experimental works has been done on liquid/solid equilibria. In 1931, Takeda [4] published the first systematic experimental studies on the system. In 1968, Jellinghaus [5] experimentally determined the basic entire phase diagram. In 1978, Shevchuk et al. [6] provided information on phase equilibria involving the liquid phase. In 1985, Gabriel et al. [7] used a phase separation technique to obtain new data concerning equilibria involving the liquid phase. In 1985, a compilation by Rivlin [8] underlined the differences between the findings of different authors. In 1997, Shtansky et al. [9] described the mechanism and kinetics of carbide transformation in Fe-W-C steels at 973 K. More recent studies have focused on the decomposition of austenite in Fe-W-C steels [10][11][12].
The first diagrams based on calculations were proposed by Uhrenius [13] in 1980, based on some new experimental data from Å kesson [14]. This was followed in 1987, by a new thermodynamic evaluation of the system by Gustafson [15], again using the above-mentioned experimental results.
Nevertheless the liquid-solid equilibria in the iron-rich corner remain a point of controversy. The studies reported to date yield phase diagrams with major differences in the extent of the primary phase fields, and in particular, the primary WC carbide phase field. The main points of controversy concern the eutectic temperatures, gross compositions and the phases constituting the various ternary eutectic microstructures in the iron-rich field. Surprisingly, up to four eutectic microconstituents are reported, all of them involving austenite (c phase) and two other solid phases: graphite, WC, cementite (Fe 3 C) and g-carbide (M 6 C). The ternary eutectic temperatures are reported to lie within a range of 58 K (values are reported in Table  1). The ternary eutectic point liquid McWC + graphite is considered to represent the stable state.
The WC, MC 1Àx ,M 6 C, M 3 Ca n dM 2 C carbides have been reported to be microconstituents of liquid/solid equilibria. The Fe-W-C system shows similarities with the Fe-Mo-C system. The n phase (Fe 2 (Mo, W)C) identified and studied by numerous authors [2,10,16,17] is involved in liquid/solid equilibria in the Fe-Mo-C system and only in solid/solid equilibria [2] in the Fe-W-C system. Structural similarity with cementite has been recognized and the n phase is therefore often referred to as molybdenum cementite. In 1992, Shen et al. [18] characterised the n carbide in an Fe-13.4W-3.1C (mass percent) alloy maintained for 2 h at 1333 K. The morphology was not significantly changed in relation to the microstructure of the as-cast sample with primary WC blocky carbides observed alongside ternary eutectic M 6 C carbides (Fe-33W-15C in at.%) and M 3 C. However, M 3 C (Fe-0.4-3W-23.5C in at.%) and n carbides (Fe-16.5W-20C in at.%) were identified in a small proportion.
The experimental procedure used in preparing the alloys plays an important role. The microstructures are quite different when obtained on as-cast specimens following solidification of a melt, or after long time annealing of sintered specimens. Shen et al. [18] characterised a ternary eutectic with c, cementite and M 6 C and proposed an alternative metastable liquidus surface that was rather different from those found in the literature. The crystallisation paths were determined by quench interrupted solidification during thermal analysis [19] (QDTA). It was shown that the eutectic microconstituent is transformed under 1273 K, M 6 C disappears and WC carbides grow with a plate-like morphology.
This ternary eutectic structure with c, cementite and M 6 C was observed in commercial alloys containing W, such as tool steels or cast irons. However, these alloys contain many additional elements that may modify the eutectic structure.
The aim of the present study is to clarify these discrepancies. It focuses mainly on phase equilibria involving the liquid phase and in particular on reactions leading to the various ternary eutectics.

The electromagnetic phase separation technique
Equilibria involving a liquid phase were investigated using the technique of electromagnetic phase separation as described by Allibert et al. [20,21]. The starting materials are placed in an alumina crucible encased in a graphite container and heated in a medium frequency induction furnace (10 kHz), under helium atmosphere. The temperature is measured by an alumina-sheathed Pt-30% Rh/Pt-6% Rh thermocouple immersed in the melt. The thermocouple is calibrated against the Ni melting point. Induction currents heat the sample and create inside electromagnetic forces related to the electrical power supplied in the coil. The specimen temperature increases up to the melting range. Electromagnetic forces result in a stirring movement of the liquid metal. Solid particles present in the stirred liquid bath are dragged by the movement of the liquid and then trapped near the walls of the crucible. The temperature is then stabilized at the chosen value. After isothermal treatment lasting several hours, an equilibrium state is reached. When the electric power is cut off in the coil, the quick cooling of the specimen results in the high-temperature configuration freezed in. The cooling rate is estimated at about 900 K/min above 1273 K. The chemical analysis of spatially separated phases makes it possible to arrive at the composition of the ex-liquid in the upper portion of the specimen and of one or two solid phases in equilibrium at the studied temperature.
Nine alloy compositions were prepared by melting compacted powders of Fe, WC and C and a proportion of electrolytic iron to facilitate induction melting. 150 g ingots were produced for each composition (compositions of the raw materials and the alloys are presented respectively in Tables 2 and 3). The following procedure was carried out: • Heating at around 873 K for 2 h. • Heating in He atmosphere up to 1800 K for about 15 min to ensure the rapid dissolution of the most refractory elements into the molten iron.  Polished vertical sections of the specimens were then investigated. In each of the eight experiments, the solid phases separate efficiently from the liquid. The distribution of the phases in areas including the solid/liquid boundary is shown in Figs. 1a and 1b. The alloys are rapidly solidified. Consequently, the area corresponding to the ex-liquid is characterized by a fine microstructure. The arrows in Fig. 1b indicate a boundary limiting a liquid area. The coarse solid phases are located below this boundary. One solid phase is detected in samples 3.1 and 3.2 (namely c) and in sample 5 (WC). Two solid phases are detected in samples 2, 7 and 8 which are WC + c in samples 2 and 7 and M 6 C+c in sample 8. Both W-rich coarse carbides appear with a bright contrast. The microanalysis of the treated samples showed a good homogeneity of the solid phases, indicating that thermodynamic equilibrium had been reached. Composition of the solid particles was analysed by electron microprobe analysis. The overall composition of the ex-liquid phase separated by the electromagnetic process was determined by chemical analysis of a specimen sampled in the quenched liquid area. Tungsten and iron were analysed by the Inductively Coupled Plasma (ICP) technique and carbon by the combustion method.

DTA experiments
The investigations performed by Shen et al. [18] were supplemented by the examination of three more composi-tions similar to the eutectic composition: alloy A (Fe-19.8W-3.1C in mass percent), alloy B (Fe-16W-3.98C) and alloy C (Fe-12.6W-3.96C). In all, seven compositions were investigated (compositions of the alloys are presented in Table 4). Raw materials used were Fe, W and C powders (compositions of the initial materials presented in Table 2). Ingots of about 100 g of presintered pellets were melted in a cold crucible inductive furnace under argon atmosphere and cast in a water cooled mold. Settling of heavy carbides is avoided by this process and specimens display a fine, homogeneous distribution of phases, suitable for DTA experiments.
Specimens were cut into two parts. One was used for the chemical analysis of carbon by combustion methods, while the other was machined into small cylinders (8 mm long, 5 mm in diameter) for DTA experiments. In a DTA furnace, heating was carried out under argon at a rate of 300 K/h up to a temperature of 1643 K for alloy A, 1583 K for alloy B and 1523 K for alloy C. Melting was ensured by maintaining the alloys at these temperatures for 3 min before slow cooling at a rate of 120 or 300 K/h.
In the solidification range of temperatures, DTA thermograms exhibit several peaks. The high temperature peak is poorly defined and cannot be attributed to a well-defined reaction since the primary phases probably form with severe supercooling. However at the end of the solidification period a well-defined peak is detected, enabling the invariant reactions to be determined.
The samples produced by both procedures were then observed by optical micrography and scanning electron microscopy (SEM), and characterised by X-ray diffraction Alloy 2 was duplicated with alloy 7 in order to explore two annealing temperatures.

Homogeneous liquid
Coarse solid area

Results and discussion
Thermo-Calc software [22] (r-version) and the TCFE5 database [23] were used to calculate the phase diagrams. This database is based on the thermodynamic description of the Fe-W-C system by Gustafson [15] and on a new description of the cubic WC phase by Jonsson [24].

Determination of solid/liquid equilibria
The composition of the phases in equilibria at the three investigated temperatures is reported in Table 5. The corresponding partial isothermal sections at 1423, 1473 and 1543 K are plotted in Figs. 2-4. The corresponding tie-lines are calculated and compared to the experimental ones.

Section at 1423 K
The results shown in Fig. 2 are consistent with the calculated isothermal section. However, the W content of the liquid in the liq./WC/c equilibrium is found to be higher than in the calculations (point m1). The experimental level of C in the liq./c equilibrium is also higher in the liquid phase.

Section at 1473 K
The five equilibria determined at this temperature are reported in Fig. 3 on the calculated isothermal section. Concerning the three-phase field liq./WC/c the experimental point m2 is shifted to a higher W content than the cal-

Section at 1543 K
Only one composition (8) was investigated at this temperature (Fig. 4). The liq./M 6 C/c tie-triangle was determined experimentally. For this composition, calculations predict only a two-phase equilibrium liq./WC. However, the compositions analysed for the coexisting phases are in good agreement with the experimental results reported by Å kesson [25] for the liq./M 6 C/c equilibria at 1593 K and both sets of results confirm the extension of the above-mentioned three-phase field to a lower range of tungsten compositions.

Positions of the univariant lines
Each  The values were established by analysis of the three elements Fe, W and C. Fe content, not indicated here, is the complement to 100%. a According to the calculations, WC carbide is also present in these alloys. b According to the calculations, this alloy is in the two-phase field liq./WC. c To be noticed that a small amount of iron-between 1% and 3%-is detected in WC (see Section 3.2).  Table 5. Symbols 1 and 2(+): gross compositions of alloys 1 and 2. Symbol m1: experimental composition of the liquid phase at 1423 K of alloy 2 in Table 5. It should be noted that the shift between the nominal composition WC and the analysed values induced a slight tilt of the tie-lines.

Solidification paths
Solidification paths are usually established by DTA and QDTA experiments but in the case of the Fe-W-C system, this is not easy because for the seven compositions investigated (Table 4) the crystallisation paths were dependent on the solidification rate. Also the link between the DTA-determined temperatures and the micrographs is not direct. Observation of all the specimens reveals a change in microstructure between the top and bottom of the vertical section of the sample. A high proportion of settled precipitated WC carbides is observed at the bottom of the specimen. The settling effect is enhanced at low solidification rates. In addition the microconstituents detected along the section are different and seem to be related to overall variations in W concentration. Fig. 6 shows a selection of typical micrographs.

Alloy A (Fe-19.8W-3.1C)
At least four phases are detected: c, cementite, M 6 C and WC. These latter carbides formed first and then sink to the bottom of the ingot on account of their higher density. Their facetted shapes, either pyramidal or cubic, point to free growth in the liquid. The next steps in crystallisation process are more difficult to determine since several phases formed with a primary phase morphology: c austenite dendrites and cementite needles. These phases probably grow in the melt since their development is never stopped. The occurrence of several primary structures may be explained by a difficult nucleation of the phases. It should be noted that a facetted morphology, as in the case of the WC particles, is usually related to high supercooling. Thus substantial chemical supersaturation is necessary for WC nucleation in the liquid, otherwise metastable phases will tend to take place. The WC blocky carbides with angular shapes illustrate the morphology of primary WC carbides without there being any physical connection with the ternary eutectic. Two eutectic microconstituents can be distinguished above the settled WC carbides: c/M 6 C/cementite in a substantial proportion and c/WC/cementite.

Alloy D (Fe-30.8W-3.4C)
This alloy, previously investigated by Shen et al. [18], presents a similar microstructure and also contains primary WC carbides as a sediment at the bottom of the ingot. In the upper part, both primary austenite dendrites and primary cementite needles are observed. The matrix is formed by a very fine eutectic microconstituent, the three phases being identified as c,M 6 C and cementite.  Table  5. Symbol 8(+): gross composition of alloy 8. Symbol m3: experimental composition of the liquid phase at 1543 K of alloy 8 in Table 5.

Alloys B (Fe-16W-3.98C) and C (Fe-12.6W-3.96C)
Again, coarse faceted tungsten WC carbides settled to the bottom of the ingot and two primary phases appear above, c austenite dendrites and large graphite flakes. In the interdendritic regions, three phases are observed c, fine graphite flakes and M 6 C carbides, under the morphology of two eutectic microconstituents, one binary c/M 6 C, and the other ternary c/M 6 C/graphite. Although cementite might be expected for such compositions [2,13], it was not found.
Carbide analyses are presented in Table 6. Cementite and M 6 C carbides are stoichiometric with respect to C. In cementite, the tungsten content is about 2-3 at.% W. In M 6 C carbides, the tungsten content is limited to 35 at.% W. This proportion (Fe 3.5 W 2.5 C) corresponds exactly to the experimental values proposed by Pollock [1] even if a larger extent is proposed by Holleck [26] up to Fe 3Àx W 3+x C in solid state equilibria. The WC carbides are stoichiometric with respect of C. They are also usually expected to be perfectly stoichiometric with respect to W  [18].A l l compositions are given in mass percent. The areas labelled I-IV correspond to the micrographs presented in Fig. 8; In part D, an insert shows the bottom of the ingot in which the carbides are observed as a sediment like in part A, B and C. but our analysis found a low iron solubility of about 0.2-3 at.% Fe. This fact has already been observed by Uhrenius [13]. Usually iron is considered to have negligible solubility in WC [11]. In an earlier investigation concerning an alloy containing 4% Cr (Fig. 7), Shen et al. [18] discriminated two classes of tungsten carbides, one with nearly no iron in WC (Fe 0.01 W 0.99 C 0.99 Cr 0.01 named WC) and the other with a low iron content (Fe 0.17 W 0.86 C 0.93 Cr 0.04 named WC Fe ). The authors pointed out a relationship between the composition and morphology of carbides. As presented in Fig. 7, the WC carbides of about 40 lm exhibit a geometric facetted shape with an even surface. A compact zone is observed on the border of the carbides. The WC Fe carbides, though facetted, appear with a mottled surface as if containing tiny inclusions.
An interpretation is proposed assuming that WC Fe carbides are formed during a solid state transformation. It is assumed that the carbides which form first from the liquid are MC 1Àx due possibly to an easier nucleation than WC carbides. On subsequent cooling, this metastable carbide undergoes an internal transformation into WC. Then tiny iron islands are embedded in WC Fe carbides during cooling. This could explain why iron analysis lies in a range between 1% and 3%.

The ternary eutectic.
Four eutectic microconstituents were identified in our present study, depending on the solidification rate and the initial compositions which can be classified into four groups (see Table 7). Their microstructure is presented in Fig. 8. Each micrograph is representative of the characteristic morphology for this ternary eutectic. Correlations can be made with the calculated respective proportions of the phases.
High W low C: c/WC/cementite (II Fig. 8). It may be observed that WC carbides, though facetted, are not as massive as the primary carbides observed at the bottom of the ingot.
Low W low C: c/M 6 C/cementite (IV Fig. 8). The three phases have a closely interlinked morphology, clearly indicating coupled growth. A similar morphology was found for the ternary eutectic in the FeMoC system [16]. The structure is fine enough to allow the global composition to be checked with the microprobe analyser. The corresponding invariant point is located at 4.2 C (±0.2%) and 18.6 W (±1.3%). DTA measurements exhibit a thermal arrest at 1358 K which can be ascribed to this eutectic reaction. The c/M 6 C/cementite microconstituent is found to systematically occur in the as-solidified alloys. The measurements of the corresponding eutectic temperature of 1358 K are perfectly reproducible.
High W high C: c/WC/graphite (I Fig. 8). In DTA specimen B, two different eutectic structure were observed close to one another with both austenite and graphite. In the area located on the left of the micrograph I (dark band) the third phase is M 6 C carbide with a fishbone-like morphology. In the middle of the micrograph, the third phase is WC carbide with a broken needles shape.  Low W high C: c/M 6 C/graphite (III Fig. 8). The ternary eutectic microconstituent can be found at the interdendritic junction (arrows on the micrograph III) which is the last area to be solidified. It seems difficult to identify since the proportion of M 6 C is very low, which is confirmed by the calculation. In general, it seems that in some of the earlier investigations referred to in Table 1, these types of very fine microconstituents were not detected or taken into account. In fact they can be observed only under specific conditions of enlargement and contrast. Furthermore, the eutectic carbides are not always stable at temperatures of a few tens K under their eutectic temperature. This may explain some of the discrepancies noted between the various investigators (Table 1). We calculated the liquidus projection. The main problem to make a correlation between the equilibrium phase diagram and our experimental findings is that most of the phases which form first are not the ones expected. The stable phases do not nucleate readily and are in this case replaced by less stable ones. To take into account these assumptions, different liquidus surfaces were drawn, suspending WC, or graphite, or WC and cementite. The results are presented in Fig. 9 in a similar way to those in Fig. 8. The ternary eutectic obtained with calculations involving all the phases liable to be found is supposed to be the stable one. This is c/WC/graphite (Fig. 9b), which is only be observed in alloy B. However, calculations can be carried out assuming that a stable phase does not nucleate and is replaced by a less stable phase. If WC and cementite are suspended in the calculation (Fig. 9d), the ternary eutectic c/M 6 C/graphite is found. If WC alone is suspended (Fig. 9c), the ternary eutectic c/M 6 C/cementite is then found and its composition is close to our experimental value (Point E). If graphite is not considered in the calculations (Fig. 9a), the ternary eutectic is c/WC/ cementite.
It should be noted that if we consider the temperature of each ternary eutectic, the classification for these values is almost the same as that for experimental measurements and calculations (Table 7). However there is an overall shift of about 45 K between experimental and calculated temperatures. All these eutectic temperatures lie within a very narrow range of values. This may explain why so-called metastable structures are so readily obtained even in the case of slow-cooling conditions.

Conclusion
Our investigation was based on a careful interpretation of the microstructure obtained with different solidification conditions, taking into account the settling effect of WC carbides. Four different ternary eutectics were identified within in a narrow range of temperatures: c/WC/cementite, c/M 6 C/cementite, c/WC/graphite and c/M 6 C/graphite. Their occurrence was found to depend on the solidification rate and the composition. Equilibrium states involving the liquid phase were determined experimentally. At 1423 and 1473 K, liq./c tie-lines and the liq./WC/c tie-triangle were obtained. At 1423 K, the liq./WC, liq./graphite, liq./c tie-lines and the liq./WC/c tie-triangle were determined, defining the liquid domain. At 1523 K, the liq./M 6 C/c tie-triangle was obtained. The position of the liq./WC + c and liq./WC + M 6 C monovariant lines was specified.
Concerning the formation of the four ternary eutectics, metastable calculations supplied a good basis. These experimental results suggest that this system should be reassessed by taking into account a larger M 6 C and WC 1Àx liquidus surface. Bratberg [27] recently proposed some new data concerning the WC carbides. The integra-tion of all these new data in a new thermodynamic evaluation should allow a better description of the Fe-W-C system.