TOOL WEAR AND INCLUSION BEHAWOUR DURING TURNING OF A CALCIUM-TREATED QUENCHED AND TEMPERED STEEL USING COATED CEMEmD CARBIDE TOOLS*
Tool wear, layer formation and deformation behaviour of difkrent inclusion types in the flow zone of the chip and in the chip-tool interface have been studied during turning of a calcium-treated quenched and tempered steel SS2541 (En24, SAE4337) using Tic-Al&-_-coated cemented carbide inserts. The analysis of inclusions in the flow zone was based on semiautomatic image anaIysis in a scanning electron microscope. A large adherence in deformation behaviour between different incfusion types was observed. The iniluence of Merent inclusions on tool wear and on the forma&m of protective layers on the tool surface is discussed along with the active wear mechanisms on tool surface. Dissolution of the coatings into adhered inclusion phases is indicated as an important wear mechanism, especially when machining calcium-treated steel with high amounts of calcium-rich sulphides.
1. Intmductiun It is well established that inclusions in steel have a strong influence on tool wear in Mach. The effects of sulphur addition and calcium treatment have been studied extensively during the last two decades f l-S]. These investigations have led to a qualitative understanding of how different types of inclusions participate in the wear process. The information has been gained through studies of the static properties of inclusions or of the corresponding substances, of their morphologies in the as-rolled steels and through the wear patterns on the tools. It can be expected that the effect of the inclusions during wear is controlled by their behaviour in the heavily sheared zones of the chip or the workpiece facing the tool and on the chip-tool contact surfaces. This has, however, not been studied previously to our knowledge. Instead, the deformation behaviour of the inchrsions just before tool contact *Paper presented at The fnsttute af Met& 1st International Conference on the Behaviour of Materials in Machining, Stratford-upon-Avon, U.K., November S-IO, 1988.
Elsevier Sequoia/Printedin The Netherlands
has been estimated from their defo~ab~~ during hot rolling. This has been evaluated through the shape factors of inclusions in the work material. This means that the large differences in forming conditions during hot working and chip formation were not considered. The action of the inclusions during the tool contact was studied through the wear pattern on the tools and through the possible presence of protective layers caused by the inclusions on the tool surfaces. No direct observations of inclusion behaviour in tool contact were thus made. This means that information is missing on the actual behaviour of inclusions inside the chip and on the contact surfaces between the chip, the workpiece and the tool. The purpose of the present paper is f%stly to describe in a qu~ti~tive way how different inclusion types in a calcinm-treated quenched and tempered steel deform in the heavily sheared zone of the chip close to the tool surface, the secondary deformation zone or the flow zone. This is accomplished by shape factor determination for inclusions in sections of chips. An image analyser based on the scanning electron microscope @EM) is utilized, Chips from turning tests using coated cemented carbide cutting inserts have been studied. The behaviour of different inclusions in actual contact with the tool surfaces, both rake and flank, is examined primarily on the side of the chip contacting the tool and on the machined workpiece. The influence on tool wear and layer formation is discussed for different inclusion types. Secondly, the tool surfaces are studied for traces of inclusions and wear marks. The wear of the coatings at different positions of the cutting insert is examined and the active tool wear mechanisms during machining are identified. This paper is devoted to wear mechanisms rather than to tool life data. 2. Work material The present study is performed on a calcium-treated tempered steel of the type SS2541 (En24, Sm4337). The the steel is given in Table I. The steel was chill cast and diameter of 85 mm and studied in quenched and tempered mechanical properties are given in Table 2.
quenched and composition of hot rolled to a con&tion. The
3. Inclusions The amount of different types of inclusions in the work material is shown in Fig. 1. Mixed sulphides (Mn,Ca)S make up the largest part of the incl~ion TABLE 1 Chemical composition of the cat&m-treated the machining tests Element AWWWG? (wt.%)
quenched and tempered steel SS2541 used in
211 TABLE 2 Mechanical properties of the work material in quenched and tempered condition Yield strength Ultimate tensile strength Fracture elongation (5d) Area reduction Hardness HB5/750
870 MPa 1000 MPa 19% 61% 300
0.10 Pfllr//lAOthers Duplex Caaluminate 7 r I5 %OPSz
0 2 = z
MnS Fig. 1. The amount of different inclusion types in the work material.
volume in the steel. The other main types are manganese sulphides and duplex calcium aluminates. The deformation behaviour during chip formation has been analysed for these three inclusion types. Mn3 is very common in both untreated and calcium-treated steels. MnS and CaS are completely soluble in each other at high temperature. In the same steel, mixed sulphides with very low calcium content can be found as well as almost pure CaS. Although the relative number of mixed sulphides with very high calcium content is often low, they are assumed to be very important for the wear of the coatings on the tool surface. Hardness and deformability of the mixed sulphides depend on the calcium content. Duplex inclusions are typical for calciumtreated steels. In most cases they are made up of an oxidic inclusion surrounded by (MnCa)S. The composition of the oxidic core may vary, depending on pre-deoxidation. The plastic deformation of an inclusion can be quantified using the length-to-thickness ratio (shape factor). The value of this shape factor depends on inclusion type, applied load, temperature and also particle size. For inclusions with high deformability, such as Mn3, the shape factor value increases with increasing particle size. In Table 3 examples are given of shape factor values for inclusions in longitudinal sections of rolled bar prior to machining. When the inclusions pass through the different deformation zones during chip formation, they may be further deformed.
212 TABLE 3 Examples of experimentally determined shape factor values (length f thickness) for different types of inclusions in rolled bar prior to machining Inclusia
(Mn,Ca)S Duplex calcium aluminate
Shape factor 6 3.5 3.4
TABLE 4 Machining test conditions and tool material data mg Speed v=208 m min-’ s=O.3 mm rev-’ Feed Depth of cut a=2 mm Tic -A1203 -TiN-coated cemented carbide (GC415 from Sandvik Coromant) Geometry IS0 TNMM 160408
4. Machining tests Chips from turning tests were used in the inclusion studies. Cutting conditions and tool material data are listed in Table 4. The cutting inserts were coated with three different types of coatings: next to the cemented carbide matrix a Tic coating 5 pm thick, followed by an AlsO coating 3 pm thick and finally by a very thin (0.1 pm) TiN coating. The identification of the different types of coatings on the worn cutting inserts was based on energy-dispersive X-ray (EDX) analysis in the SEM. 5. Image analysis of inclusions in the flow zone For the analysis of inclusion deformation behaviour in the flow zone, an SEM-based automatic image analyser has been used. The system combines automatic image analysis (geometrical classification) with a qualitative EDX analysis of individual inclusions. The system description, working principles and examples of practical applications in inclusion analysis are given in ref. 9. Software and a methodology for the analysis of inclusions in the flow zone have been developed. The analysis is based on the study of chips from the machining tests. Therefore no special types of experiments (e.g. quick stops) have to be performed and large amounts of test samples are easily available. The chips were normally coated with nickel to prevent edge rounding during grinding and polishing. Specimen preparation included mounting,
surface of chip
\ structur;l\ line
SHAPE FACTOR = XLlYL
:AXIAY= cot Q
Fig. 2. Schematic view of inclusion appearance in the image analyser during the analysis of inclusion deformation behaviour in the flow zone.
grinding and diamond polishing according to conventional methods. Longitudinal, tmetched sections of the chips were studied. The appearance of the inclusions in the image analyser is shown schematically in Fig. 2. The chip surface facing the tool is included in the image frame in order to calculate the shear strain and the distance from the tool rake face to the inclusion. The high resolution of the SEM makes it possible to detect very small and heavily deformed particles. The evaluation of geometrical parameters (inclusion size, shape, position and shear) and the X-ray analysis of each inclusion are performed fully automatically. 6. Deformation
types in the flow
In this section the deformation behaviour in the flow zone will be discussed for the different inclusion types in the work material. For each inclusion type the shape factor values are given as a function of the shear in the flow zone (or distance from the tool) and of the inclusion size. Shape
factor values in the workpiece before machiniug are indicated on the lefthand side of the diagrams (W). Of vital importance for tool wear and layer formation is the behaviour of the inclusions when they are coming in contact with the tool surfaces. Inclusions which have been sliding against the rake and flank faces of the inserts have been studied on the chip surface and on the workpiece surface respectively. For the chips investigated the thickness of the flow zone was approximately 30% of the chip thickness.
MnS inclusions close to the tool surface are heavily deformed because of the high shear strain in the flow zone. Especially for relatively large inclusions, extremely high shape factors (greater than 20) were recorded close to the rake face (Fig. 3(a)). When the particle size decreases, so does
Fig. 3. (a) Shape factor values for MnS as a function of the shear in the flow zone. (b) Heavily deformed manganese sulphides in the now zone.
shape factor value. Close to the tool the shape factors for small MnS are relatively low. This is due to fragmentation of long, heavily deformed sulphides into smaller particles with low shape factor. The plastic deformation of the sulphides is much higher in the flow zone compared to the primary shear zone. This can be concluded by comparing the recorded values with the shape factors for inclusions in the workpiece prior to machining (W). MnS inclusions in the flow zone are shown in Fig. 3(b). Close to the tool surface the sulphides are so elongated that they are hard to resolve, even when using SEM. The high deformability of the sulphides makes them very hard to observe on the chip surface. Occasionally, traces of thin sulphide hlms can be detected. MnS probably causes no mechanical, abrasive wear of the tool surface. The hardness of the inclusions is too low. At low cutting speed (V < 150 m nun-‘) MnS layers can be detected on carbide tools. These layers may protect the tool and decrease the diffusion type of wear. However, at high cutting speed the strength of the sulphide is too low to withstand the contact load during machining. Adhered sulphide phase will therefore be tom away and no layer build-up will occur.
6.2. Mixed s&phi&, (Mn,Ca)S The shape factor values for (Mn,Ca)S are generally lower than for MnS when data for the same inclusion size are compared. Large (Mn,Ca)S inclusions can reach high values (X,/u,= 15), but in most cases the values lie in the internal X,/Y-- = 4-6 (Fig. 4(a)). Probably (Mn,Ca)S inclusions with different calcium content are included in the analysis, which explains the unexpected variations in shape factor value for small inclusions.
Fig. 4. (a) Shape factor values for (Mn,Ca)S as a function of the shear in the flow zone. (b) Comparison of the deformability in the flow zone between an (Mn,Ca)S and sn Mns. (c) An (Mn,Ca)S and a calcium-containing oxide on the chip surface.
The difference in deformability between MnS and (Mn,Ca)S is illustrated in Fig. 4(b). The size of the particles and the distance from the tool surface are approximately the same for both inclusions. The MnS is much more elongated. Due to their low deformability in the flow zone, (Mn,Ca)S can be observed on the chip surface. Figure 4(c) shows an (Mn,Ca)S lying close to a calcium ahnninate. Both particles appear to be consumed in a slicingoff process, leaving the inclusion surface at the same level as the chip surface. No cavities can be seen in the interface between inclusion and steel matrix. It is assumed that in order to form layers on the tool surface, it is necessary for the inclusions to be sliced off and smeared out in the chip-tool interface. The mixed sulphides often play an spout role in the fo~ation of layers on carbide tools, both as free inclusions and as part of duplex inclusions. Their higher hardness compared to the MnS gives the (Mn,Ca)S a suitable deformation behaviour in the flow zone and in the chip-tool interface. Similar to MnS, the abrasive wear due to (Mn,Ca)S is thought to be negligible. However, in some cases tool wear due to dissolution may be increased because of an tmfavourable combination of tool material and composition of the (Mn,Ca)S inclusions. 6.3. Duplex calcium aluminate Duplex calcium-containing inclusions are common in calcium-treated steels. In most cases the oxidic part of the inclusions is either a calcium aluminate or a CaAl silicate. In this section the behaviour of duplex calcium aluminates wiII be discussed. In the investigated materials their normal appearance is a practically spherical, calcium and ahnninium-containing, oxidic core surrounded by a mixed sulphide, (Mn,Ca)S. When the shear in the flow zone is low, high shape factors can be registered (Fig. 5(a)). Increasing shear close to the tool face results in decreasing shape factor. No influence of inclusion size can be detected. This somewhat confusing result is contrary to those for MnS and ~,~a)S, but can be explained by studying Fig. 5(b). A high shape factor at low shear levels corresponds to plastic deformation of the sulphide phase (Fig. [email protected]
)). When the shear is increasing, the sulphide is separated from the calcium aluminate, leaving the oxide with only remnants of strIphide phase. This process results in a decrease in shape factor for the duplex particle. In Fig. [email protected]
) a calcium aluminate with only a thin sulphide tail is shown close to the tool. Even though the shear and the temperature are very high in this part of the flow zone, the spherical shape of the oxide is untiected. In Fig. 5(c) the remaining sulphide phase can be seen on the chip surface as a dark rim around the calcium aluminate. The duplex inclusion appears to be sliced off without any cavity formation in the inclusion-matrix interface, similar to the (Mn,Ca)S inclusions. Both the calcium ahuninate and the surrounding (Mn,Ca)S phase are thought to contribute to the layer formation on carbide tools. This conclusion is supported by the favourable deformation behaviour in the chip -tool interface. The absence of plastic deformation of the calcium aIuminates in
Fig. 5. (a) Shape factor values for duplex calcium aluminates as a function of the shear in the flow zone. (b) Duplex calcium aluminates in the flow zone: (hl) 30 q, (b2) 5 w from the tool surface. (c) A duplex calcium aluminate on the chip surface.
the flow zone indicates that the hardness of the oxides is substantially higher
than for the mixed sulphides. The calcium-containing oxides are thought to contribute to the abrasive wear of the tool. However, the abrasive wear is much less than for A1203 inclusions in untreated steels. An important effect of the calcium treatment is to eliminate the harmful, abrasive ahunina inclusions and to replace them with more favourabIe ~~ci~-~ou~g oxides. 6.4. InclM behaviow on the tool Jkmk The deformation behaviour of inclusions sliding against the flank face of the tool can be studied on the workpiece surface. For calcium-containing inclusions the deformation behaviour is simiiar to the behaviour on the chip surface. lhe oxides are sliced off and ~~ci~~on~ sulphides are heavily deformed and smeared out in the same way as in the flow zone. 7. Tool wear 7-f. Wear pattern A worn cutting insert is shown in Fig. 6 after 10 mm of machining. Several different zones with different wear of the coatings can be distinguished along the cutting edge. The differences in wear rate and composition of the tool surface between different parts of the tool are considerable. On the very cutting edge (zone A) the A&O:, coating and the thin TiN coating are still present. In zones B and E on the rake and flank surfaces respectively the TiN and A120, coatings are completeIy consumed and only part of the Tic coating is still left. In the nose area in zone B the Tic coating is worn through and work material fills a crater in the cemented carbide. Adhered
Fig. 6. View of a worn tool after 10 min of turning with thin CaS inclusion layers on the tool surface, especially in zones A, B and C; cutting speed 208 m min-‘. A, A1203+TiN; B, Tic; C, Alz03; D, (Mn,Ca)S inclusion layer; E, TIC; F, A1203.
inclusion layers were observed on the tool surface. Considerable differences in layer composition and layer thiclmess between different areas were detected. In zones A and B the tool surface is covered with a thin CaS layer. In zone D, 0.35475 mm from the cutting edge, the tool surface is covered with a relatively thick (Mn,Ca)S inclusion layer. The composition of the inclusion layers was analysed using an Auger technique. Using depth profiling it was shown that no significant amounts of elements other than manganese, calcium and sulphur were present in the layers. The oxygen content was negligible. The different zones with different wear and surface composition are clearly visible in the backscattered image in Fig. 7(a). In Fig. 7(b) the remaining A1203 coating on the cutting edge is visible and also the narrow section through the Al,Oa coating (zone C) marking the transition from the worn ‘DC coating in zone B to the (Mn,Ca)S layer in zone D. Figure 8 shows a part of the (Mn,Ca)S layer on the rake face (zone D). The thickness of the layer is approximately 10 pm. The cracks in the layer are probably due to thermal stresses during the machining tests. Because of the cracks, small pieces of the layer have been lost. The exposed tool surface under the inclusion layer displays no wear marks and corresponds to the appearance of an unworn tool surface. The chip is apparently not sliding against the surface of the coatings but on the relatively thick inclusion layer. The wear of the tool is probably negligible. Even more striking is the appearance of the flank surface (Fig. 9). Zone E, where the AlzOa coating is completely consumed, constitutes almost the whole contact area. In the same way as on the rake face, a thin, dark AlzOa zone (F) indicates the transition from zone E to the unaffected tool surface outside the contact area. Below zone F, where the contact between workpiece and flank surface ends, small particles similar to sol.idified drops can be seen along the whole cutting edge. Using EDX analysis, several Merent elements can be detected in these particles: manganese, calcium and sulphur (from
Al,O, + TiN
A'203 (Mn,Ca)S inclusion layer
Fig. 7. (a) Overdl view of the rake face. (b) Wear of the coatings close to the cutting edge.
the inclusions in the steel); aluminium and titanium (from the coatings on
the tool); chromium, silicon and (iron) (from the steel matrix). F’igure 9 indicates that a molten inclusion phase wets the work material-tool interface during machining. The characteristic wear pattern with a completely consumed A1203 coating close to the cutting edge and relatively thick inclusion layers on the rake face appears after only approximately 2 min of machining. The
Fig. 9. The flank face in the middle of the cutting edge.
AlaO coating is thus rapidly worn through. When the Tic coating next to the cemented carbide is exposed, the wear rate of the coating decreases.
Based on extensive machining tests and tool wear studies for several QT steels with different inclusion content, a number of different wear mechanisms have been identified [lo]. The most important wear mechanisms are thought to be the following. (A) Abrasion due to hard oxide inclusions. This mechanism is important when machining untreated steels containing abrasive A120s inclusions. (B) Dissolution of the coatings into adhered inclusion phases. (C) DSusion-controlled wear. For the particular steel investigated in the present work, mechanism B is thought to be most important for the wear of the coatings. The abrasive wear is thought to be of minor importance because of the low amount of oxide inchrsions. The calcium ahuninates in the steel are relatively favourable compared to AlsO3 inclusions from the abrasive wear point of view. Furthermore, the extensive formation of relatively thick inclusion layers on the rake face does probably reduce the importance of diSusion-controlled wear. Therefore, during turning of the caM.nn-treated steel, the wear of the coatings close to the cutting edge is assumed to occur principally in the following way. (1) Adhesion of calcium-rich strIphide inclusions on the tool surface. (2) The mehing temperature of the adhered sulphide decreases to the temperat~e in the interface between tool and work material owing to dissolution of other elements from the coatings into the inclusion layer.
(3) Melting of the inclusion layer and wetting of the contact surfaces, resulting in rapid dissolution of the coatings on the cemented carbide. The wear mechanism is illustrated schematically in Fig. 10. This type of wear is assumed to be active on both rake and flank surfaces of the tool. The mixed sulphides with high calcium content are assumed to be very important for the wear of the coatings even though they are relatively few. The strength of the sulphides and the resistance to plastic deformation increase when the calcium content in the inclusions increases. The abilie of the adhered inclusions on the tool surface to withstand the contact load during machining is also increased and therefore the inclusion layer is not tom away from the tool surface. F’ure MnS and (~n,Ca)S with low calcium content are heavily deformed on the contact surfaces and are easily removed from the tool surface by the action of the work material. The melting temperature for pure CaS is approximately 2400 “C [ 111. Addition of manganese to the sulphide decreases the melting temperature to a minimum at approximately 1500 “C for a composition of 72 mol.% MnS. A similar influence of other elements present in the work material tool interface (aluminium, titanium, iron and others) can be expected. It seems reasonable that the melting temperature of adhered CaS inclusion layers can come down to the temperature on the too1 surface.
F’ig_ 10. Schematic view of the wear of the coatings close to the cutting edge. A, unworn tool. B, adhesion of calcium rich sulphide inclusions. C, decrease in melting temperature of the adhered sulphide due to ~olu~on of elements from the coatings into the inclusion layer. D, melt&g of the inclusion layer, wetting of the contact surfaces and rapid ~olu~on of the CO&IlgS.
The decrease in melting temperature of the adhered inclusions is not necessarily a diffusion-controlled process. Inclusion layer, coatings and steel are in very intimate contact in the cutting zone. The different elements can therefore be considered as already mixed on the contact surfaces owing to the high contact loads and high temperatures. After initial wetting of the contact surfaces, the dissolution of the coatings does probably proceed rapidly. The mixture consisting of inclusions from the steel, dissolved coating and work material is pressed out of the contact zone and follows the chip or workpiece, or can be found on the rake or flank surfaces of the tool. Adhered CaS has been found on all three types of coatings - Tic, AlaOs and TiN. The proposed wear mechanism is assumed to be important for all three coatings. However, the difference in wear rate is considerable between different coatings. The wear rates have been estimated to be approximately 3 pm mix? in the AlaO coating and less than 1 pm min-’ in the Tic coating for the machining conditions used. The tool wear studies indicate that turning of a calcium-treated steel containing calcium-rich mixed sulphides using an AlaO,-coated tool is unfavourable from the tool wear point of view.
Tool wear, layer formation and inclusion behaviour have been studied during turning of a calcium-treated quenched and tempered steel using coated cemented carbide inserts. The deformation behaviour of different inclusion types in the flow zone of the chip has been analysed using semiautomatic image analysis. The following conclusions can be drawn. (1) The deformation behavior-u in the flow zone is different for different inclusion types. Manganese sulphides, MnS, and mixed sulphides, (Mn,Ca)S, are plastically deformed. This is opposite to the behaviour of oxidic inclusions. Calcium aluminates do not deform in the flow zone even though the temperature and shear strain are very high. (2) The ability to deform plastically in the flow zone is not a prerequisite for an inclusion type to contribute to the layer formation on the tool. Of vital importance for tool wear and layer formation is the behaviour when the inclusions are sliding against the tool surfaces. (3) Manganese sulphides become extremely deformed in the flow zone and in the chip -tool interface. No layers of MnS are formed on the tool surface at high cutting speed (V = 208 m ruin-‘). The strength of the sulphide is probably too low to withstand the contact load and the adhered sulphide layers are tom away. The soft MnS inclusions do not cause any mechanical, abrasive wear of the tools. (4) Harder, calcium-containing inclusions such as (Mn,Ca)S and calcium aluminate are continuously sliced off and smeared out during inclusion contact with the tool. This behaviour is assumed to be favourable from the abrasive wear point of view and to be a prerequisite for an inclusion type contributing to layer formation.
(5) Turning of the calcium-treated steel resulted in inclusion layers on the tool surface. The differences in layer thickness and layer composition between different areas were considerable. Thin CaS layers were adhered to the coatings close to the cutting edge. Relatively thick (Mn,Ca)S layers were formed further out on the rake surface. (6) For the investigated calcium-treated steel the wear of coatings close to the cutting edge is assumed to be mainly due to dissolution of the coatings into adhered CaS inclusion layers. A relatively low amount of calcium-rich sulphides in the steel is assumed to be important for the wear of the coatings. (7) The results show that the inclusion layers on the tool surface do not act as passive ditfusion barriers between steel and tool. Instead, the inclusion layers interact with the coatings and the combination of inclusion layer and coating composition are of importance for the wear rate. In particular, the machining of a calcium-treated steel containing calcium-rich sulphides using an AlsOa-coated tool is an unfavourable case.
This research work has been carried out with financial support from Sandvik Coromant AB, Kloster Speedsteel AB, Ovako Steel AB, Bofors AB and the National Swedish Board for Technical Development, which is gratefully acknowledged. The authors would also like to thank the members of the research committee, namely Rainer Lepp5nen (Ovako Steel AB), Bjijrn Olsson (Sandvik Coromant AR), Lennart Sibeck (Bofors AB) and Henry Wisell (Kloster Speedsteel AR), for valuable discussions.
References 1 R. A. Joseph and V. A. Tipnis, in Iqfluewe of MetaUurgy on Machinability, American Society for Metals, Metals Park, OH, 1975, pp. 55-72. 2 V. A. Tipnis, in Sulphide Inclusions in Steel, American Society for Metals, Met& Park, OH, pp. 480-495. 3 D. C. Hilty and J. W. Fsrrel, Iron and SteelmaJcer, 2 (1975) 20-27. 4 T. Tshenouchi and K. Suzuki, Pram. Iron Steel Inst. Jpn, 18 (1978) 344-351. 5 G. M. Faulring and S. Ramskngsm, MetaU. Pram. A, 10 (1979) 1781-1788. 6 T. Ksto et al., in R. W. Thompson (ed.), Proc. Carlf. on Iqjkence of MetaUurgy on the
Machinabiltiy ofEngine&ngMaterids, 7
8 9 10 11
September1982,13 - 15,American
Society for Metals, Metals Park, OH, 1983, pp. 323-337. Z. Palmsi, Met. Technol, 11 (1984) 34-37. H. Ksnksanpti et al., Mater. Sci. Technd, 3 (1987) 155-158. T. Werlefors and S. Ekelund, Scan&. J. MetalL, 7 (1978) 60 - 70. A. Nordgren and A. Melander, Mater. Sci. Technd, 5 (1989) 940-951. C.-H. Leung and L. H. van Vlack, J. Am. Ceram Sot., 62 (1979) 613-616.