Internal Crystallisation Method to Produce Oxide Fibres and Heat Resistant Composites

Most efforts in developing advance materials to be used in machines for transportation and power generation have been always directed towards enhancing the energy efficiency of the machines, which means decreasing fuel consumption per a unit of useful work. Researchers all over the world have done a lot to reach the goal including developing structural materials with higher specific characteristics to enhance the ratio of a payload to weight of the structure, improving aerodynamics, applying appropriate coatings on hot details, cooling them, etc. However, enhancing the use temperature of structural materials for hot parts of machines has been and will be always a most efficient way to improve the energy efficiency of transportation and power generation.


Introduction
Most efforts in developing advance materials to be used in machines for transportation and power generation have been always directed towards enhancing the energy efficiency of the machines, which means decreasing fuel consumption per a unit of useful work. Researchers all over the world have done a lot to reach the goal including developing structural materials with higher specific characteristics to enhance the ratio of a payload to weight of the structure, improving aerodynamics, applying appropriate coatings on hot details, cooling them, etc. However, enhancing the use temperature of structural materials for hot parts of machines has been and will be always a most efficient way to improve the energy efficiency of transportation and power generation.
Most heat resistant materials for heavily loaded structural elements, those being Ni-based superalloys are approaching their physical limit set by their microstructural stability and melting points. Perhaps, this limit will be around 1100°C and a dependence of the maximum use temperature on year of the alloy development is asymptotically going to this limit.
The importance of developing heat resistant material beyond nickel superalloys can be clearly demonstrated by comparing the history of enhancing the use temperature of superalloys and the history of developing aviation jet engines expressed in fuel efficiency of them. Both histories are presented in Figure 1.

Refractory metal alloys.
Active development of CMCs started with works by R. Naslain [3] and has now reached a stage of the active implementation [4]. The main efforts in this direction have been focused on silicon-carbide-fibres/silicon-carbide-matrix composites certainly due to a higher creep resistance of SiC fibre as compared to polycrystalline oxide fibres. Still, oxide/oxide composites have been developing to occupy an appropriate niche [5,6,7].
Metal matrix composites occur now to be in shadow of CMCs; however, it will be shown in this chapter that they have quite prospective future provided a number of the problems are solved.
The mainstream in the development of refractory metal alloys is alloying the matrix by silicon and some refractory metals to form complex silicides particles in the matrix. The main problem in the development of such alloys is to tailor their microstructure to reach an appropriate balance between creep resistance, fracture toughness and resistance to gas corrosion of the  [1] and decreasing the specific fuel consumption of aviation engines [2] with the time of their developing. The use temperature is determined by rapture stress 150 MPa at 1000 h. material to satisfy strict requirements set by the loading and heating conditions of a gas turbine blade surrounded by combustion products containing oxygen. Modern niobium alloys [8,9], unlike those developed earlier [10] are characterized by high fracture toughness and oxidation resistance, which certainly allows using them at temperatures up to about 1200°C.
A further enhancement of the use temperature of heat resistant materials will be, perhaps, possible by appropriate tailoring molybdenum-based alloys. The present mainstream in this direction is developing alloys in the Mo-Si-B system [11]. Despite an essential enhancement of gas corrosion resistance of such alloys as compared to well-known molybdenum alloys [12], the balance between creep resistance and fracture toughness does not seem to have been reached yet [13].
A fundamental difference between metal alloys and metal matrix composites is a correlation between strength and fracture toughness of these two types of structural materials. Strengthening metal alloys, as a rule, limits dislocation movements and, hence, decreases crack resistance; reinforcing metal matrix with high strength brittle fibres yields obviously an increase in the strength and can yield an increase in the fracture toughness at the same time [1]. Therefore, development of MMCs promises better results than that of metal alloys.
To go on with the development of heat resistant MMCs, now some scientific and technical problems should be solved. Fabrication technology of appropriate fibres and composites with sufficiently high oxidation resistance are among the most important problems. The present chapter is to focus mainly on these items. We start with oxide fibres of single crystalline and eutectic structures and then proceed with discussing MMCs containing such fibres. Finally prospects of the internal crystallysation method to obtain new types of fibres and composites will be briefly discussed.
no higher than 0.5 -1.0 mm/s, which makes productivity rate of the process low, so that the cost of the fibres is too high to use them in structural applications. The macrostructure and strength of sapphire fibres depend strongly on the crystallization rate. Developers of μ-PD method, Japanese researchers [17], did actually turn up down a classical scheme of EFG, which simplified slightly growth procedures. They claim that the usage of the μ-PD yields crystals with smaller thermal strains compared to other growth methods and using this method makes possible to grow crystals even from incongruent melts. Still, the method does not allow an essential decrease in the fibre cost as compared to EFG.
LHPG was used, perhaps for first time, to grow single-crystal Cr-doped Al 2 O 3 fibres from a small source rods melted locally on its end by a CO 2 laser [18]. Then it has been used rather intensively to grow mainly optical fibres [19], although attempts to grow structural fibres are also known [20]. There are some advantages of the method: the absence of a crucible allows growing sufficiently pure crystals; a small volume of the melts yields an increase in both thermal efficiency despite of a low efficiency of the laser heating and and mass exchange around the process zone that would be useful in growing such materials as mullite, which is characterized by a complicated phase diagram.
The methods just described have been used to produce continuous oxide fibres of high quality, which are effectively used in various non-structural applications (laser optics, sensors, scintillations, etc.). At the very beginning of the composite era, in 70s of the last century, there was a hope to use sapphire fibres produced by EFG in structural applications so their high temperatures strength was a subject of an intensive study [21,22,23,24]. However, high cost of the fibres produced by crystallisation of the melt has eventually led to losing interest to such fibres as reinforcement in heat-resistant composites technology.
Internal crystallisation method invented by Mileiko and Kazmin was described for the first time about 30 years ago [25,26]. Since then a main scheme of the method as well as some variation of it has been published in a number of the papers (see references [27,28,29,30,31,32,33]). All these papers are concerned with oxide fibres. However, recently it was shown a possibility to produce a larger family of the fibres. I particular, in Section 7.2 prospects of the internal crystallisation method composites with silicide fibres will be described.
A schematic of the method is presented in Figure 3. A molybdenum carcass with continuous channels in it, which is prepared by diffusion bonding of an assemblage of the wire and foil (step 1 in Figure 3), is infiltrated with an oxide melt (steps 2 and 3) by the capillary force. The melt is then crystallised in the channels to form fibres in an oxide/molybdenum block (step 4). This is a main scheme of the ICM, which can be varied to attain a particular goal. For example, to obtain sapphire fibres of a homogeneaous crystallographic orientation, a seed oriented in an appropiate manner is used as shown in Figure 4.  Finally, molybdenum is dissolved in a mixture of acids to get a bundle of the fibres ( Figure 5). At the present time, the length of the fibre bundle can be up to 250 mm. Fibre volume fraction in a molybdenum block is normally about 0.4. The fibre has a cross-sectional shape close to that shown shematically in Figure 3, a characteristic size of it is usually 50 to 300 μm. A number of the blocks, which can be crystallised in one process as shown in Figure 4 is limited only by the size of the thermal zone of the crystallising machine. A set shown in Figure 4 contains after the crystallisation process neary 40,000 fibres with characteristic cross-sectional size of 100 μm.

Structure and mechanical properties of fibres produced by ICM
The microstructure and mechanical properties of fibres produced by ICM are determined mainly by the crystallization process of the fibre in the channels of the auxiliary molybdenum matrix.

Microstructure of ICM-fibres
A qualitative model of the process is suggested in references [26]. Suppose we have a furnace with two temperature zones "I" and "II" ( Figure 6) and there is no direct thermal exchange between them. After the infiltration, the specimen is located within Zone I and its temperature is T I , which is higher than melting point T m of the fibre material. Now if the specimen is moved instantaneously in the direction of the arrow by distance ΔL, the temperature profile along the specimen starts changing as shown by lines marked with t = 0, t 1 , t 2 ,... When some overcooling, say ΔT, is reached at time t 3 , spontaneous crystallization within length L * occurs. As a result of the process a solid-liquid boundary in each channel is emerged. Then the events can go on in a number of the ways depending on a concrete material.

Sapphire
It is well known that sapphire has no preferable crystallization direction. Hence, if a channel effective diameter is sufficiently small then each crystal on the solid side of the solid-liquid boundary can serve as a seed for further crystallization and the fibres obtained have random crystallographic orientations [26]. It is important to note that the probability of having a single crystal occupying the whole channel transverse section increases with the cross-sectional sizes of the channels decreases. A layer of the fibres in a tensile specimen obtained by using ICM, which illustrates the scheme described is shown in Figure 7.

Yttrium aluminium garnet (YAG)
Fabrication technology of bulk crystal of YAG (Y 3 Al 5 O 12 ) normally includes high temperature annealing of crystals to restore YAG since cooling an overheated melt of the Y 3 Al 5 O 12 composition yields decomposition of YAG into alumina and perovskite YAlO 3 . This stage of the process while obtaining single crystal YAG fibre by using ICM should be avoided because it can cause additional stresses in fibres as recombination of YAG is accompanied by a volume change. Hence, while crystallising garnets by using ICM an overheating of the melts is to be excluded [31,34].

Mullite
Mullite has been known as most creep resistant oxide [35,36]. That is why, obtaining single crystal mullite fibres looks as a challenge for the composite community. However, some reasons including a complicated phase diagram of the alumina-silica system [37] set serious problems in crystallising the fibres from the melt while using either EFG or LHPG crystallisation methods.
Using ICM allows crystallising single crystal mullite fibres [34,35,38]. Crystallisation of a raw melt composed of a mixture of alumina and silica in the channels of a molybdenum block is a more complicated process than that of sapphire. The process was studied in details in references [34,35]. The fibre microstructure is determined by the crystallization process.
To study specific features of crystallisation of the melt containing a mixture of Al 2 O 3 and SiO 2 , a number of the raw compositions, n Al 2 O 3 :SiO 2 , were used with the value of n = 1.5, 1.8 and 2.05. As in the case of sapphire, the fibre microstructure is changing in the upper portion of the block where crystallisation starts and then remains fairly unchanged. The steady state growth yields the fibres of nearly single-crystalline structure. Typical cross-sections of the fibres corresponding to lowest and highest values of n are presented in Figure 8.
Consider first a fibre obtained from the raw mixtures corresponding to n = 2.05 ( Figure 8). The X-ray microanalysis shows that the black phase in SEM-pictures has atomic ratio Al:Si slightly higher than that in the precursor (n = 2.05). Despite of an excess of alumina in a fibre on the average, in the sharp corners of some fibres a "white" phase can be seen. With the value of n decreasing, an excess of silica in the raw material yields larger areas occupied by the "white" phase ( Figure 8a). The "white" phase occurs to be enriched with silica and it is certainly a glassy phase. Again, a composition of the "black" phase corresponds to mullite. Impurities contents in the two phases of the fibres indicate that a growing mullite crystal pushes the impurities, which exist in the melt, to a silica enriched melt that solidifies at a lower temperature to form inclusions of the glassy-phase.
The microphotographs presented and the data of the The X-ray microanalysis [34] reveal some important features of the fibre microstructures: 1. The composition of mullite in the fibres is close to 2Al 2 O 3 :SiO 2 independent of the initial composition of the precursor. A rather broad field of the existence of mullite in various phase diagrams published explains deviations from the exact 2:1 molar ratio.

2.
Mullite is optically transparent, which is characteristic to single crystal. The content of impurities in the mullite phase does not reach the sensitivity of the experimental method.

3.
Inclusions of the glassy phase are located mainly at the fibre periphery.
Optical microscopy [32] shows that the fibers consist of a mosaic of single crystal areas up to about 5 mm in length. The c-axis of the single crystal individuals are misalligned up to ±3°w ith respect to the fiber axis.
The events occurred in the volume of spontaneous crystallisation in the case of mullite fibres are similar in some details to those while crystallising sapphire fibres and different from it in other details. Observing changes of the microstructure of mullite fibres crystallised from the melts with n < 2 along the length of an oxide/molybdenum block provides an insight into the growth mechanisms of mullite that yield a microstructure of the fibre on the steady state growth portion. The microstructure of fibres at the top of an oxide/molybdenum block is presented in Figure 9. One can see a number of the mullite crystals divided by silica-based phase. For example, in an area marked by a circle, three crystals are seen. The direction of all crystals along the fibre axis coincides certainly with c-axis of the mullite crystal. However, their angular orientations can vary.
Crystallisation proceeds with growing some crystals of various angular orientations. The process is accompanied with pushing a liquid, which is to form the glassy phase, towards the oxide/molybdenum interface. Obviously, the crystallisation process depends on many parameters such as pulling rate, temperature gradients, melt composition, etc.

Oxide eutectics
Special features of crystallization process in the channels of a molybdenum carcass determine particularities of the eutectic microstructure of the fibres. Perhaps, most important feature is a non-plane solid-liquid boundary in the channels due to heat sinks into molybdenum carcass. Still, characteristic microstructure of the eutectics is observed in the fibres obtained by using ICM, Figure 10, 11 and Figure 12 [39]. A non-homogeneity of the microstructures is a result of the features of crystallization process mentioned above. The dependence of the characteristic microstructure size of the Al 2 O 3 -Y 3 Al 5 O 12 eutectic fibres on the pulling rate related to crystallysation rate ( Figure 11) is certainly close to a usual one.

Room temperature strength of fibres produced by ICM
First of all, it should be noted that the strength of fibres produced by using ICM as well as that of all structural fibres depends on their surroundings: the results of testing separate fibres can drastically differ from the effective strength of the fibre in a matrix [40].
Another point to be mentioned concerns methods of measuring strength This is important since the fibres have an unusual shape of the cross-section.
The strength of separate fibres is measured by looping a fibre around a series of the rigid cylinders of successively smaller diameters and calculating the bending strength [41]. The maximum fibre stress corresponding to the rigid cylinder of radius R is calculated according to where E is the Young's modulus of the fibre material and d is the distance between the plane surfaces of a fibre. To a first approximation, σ is the fibre bending strength at a length equal to the average distance between the fibre breaks.
Results of measuring sapphire fibre bending strength are presented in Figure 13. It should be pointed out that coating the fibre with a thin layer of silicon oxi-carbide by uing CVD process yields an essential increase in the fibre strength. This effect is a result of healing fibre surface defects by the coating material.
An effect of stress concentration in the vicinity of a defect in the fibre can be enhanced as a result of temperature gradients. Hence, annealing fibres in a furnace either without or with small temperature gradients leads to an increase in the fibre strength as shown in Figure 14 and Figure 15 for sapphire and single crystalline YAG fibre, respectively. 11 and 12. Hence, the strength of eutectics is expected to increase. However, the experiments yield a strength/crystallisation-rate dependence with a maximum at a some value of crystallization rate (Figures 16 and 17). This phenomenon remains to be studied.

High temperature strength of fibres produced by ICM
The measurements of high temperature strength of ICM-fibres [33] confirm early finding that single crystalline oxides retain their nearly room temperature strength up to high temperatures. Obviously, it is difficult to perform high temperature tests of ICM-fibres in the same manner as it is done at room temperature. However, the method of producing the fibres provides a way to test fibres that is to test oxide/molybdenum specimens. This way has been used from the very beginning of the developing of the internal crystallisation method.
Early experiments with oxide-fibre/molybdenum-matrix composite specimens described in [26,27] were conducted in tension. Then most tests have been performed under bending; such tests can be done with simple specimens. To calculate the fibre strength in molybdenum carcass with volume fraction of the channels of about 40% one needs to know mechanical properties of molybdenum used for the carcass in fully recrystallised state. The necessary data have been obtained in tensile tests, see Figure 18, and the strength/temperature dependence is approximated by a polynom of the 3 rd order. The strength data obtained for some fibres are presented in Figure 19 and Figure 20. One can see that (i) the effective strength of ICM-sapphire fibre in the molybdenum matrix does not significantly differ from that for the fibres obtained by EFG method, (i) the fibre strength obtained in bending tests is slightly higher than that obtained in tensile tests, which should be expected.

Creep characteristics of fibres produced by ICM
Creep characteristics of ICM-fibres are measured by testing oxide-fibre/Mo-matrix composites obtained at step 4 in Figure 3. It should be noted that oxide melts wet molybdenum excellently and the fibres have been crystallised in the channels of the molybdenum matrix, so the fibre/ matrix interface is considered to be an ideal; hence, the fibre strength is not affected by possible surface flaws. Therefore, testing fibres in their mother matrix yields their virgin properties. Any further processing of the fibres changes their properties. Note that the fibre characteristics are easily obtainable as a contribution of the matrix, which is fully recrystallised molybdenum, to the creep resistance of oxide/molybdenum composites even at the lowest temperature, 1000°C, is negligible, less than 10 MPa [27].
An analysis of the bending tests performed under a varying load yields approximate creep characteristics of the composite in a particular specimen [42]. This allows excluding a number of the factors, which determine a usual scatter of the creep data of a batch of the material. Performing such tests yields a characterisation of each specimen by its own set of the creep characteristics. The analysis is based on the following assumptions:     Figure 19. Temperature dependence of the tensile strength of sapphire fibres obtained by both the ICM and EFG-method. Experimental data are after Mileiko and Kazmin [27], Hurley [22] and Shahinian [21]. The present data are obtained by testing sapphire/molybdenum specimens in 3-point bending. Figure 20. High temperature strength of some oxide fibres produced by ICM. The data were obtained by testing oxide/ molybdenum specimens [33].

A composite is characterised by identical creep behaviour under tension and compression
where η n , σ n , and n are constants. The value of η n can be chosen arbitrary. In what follows, η n =10 -4 h -1 , which means that σ n is the stress to cause 1% creep strain for 100 h. We call this value as creep resistance of a material on 100-hours time base.
A solution of a creep problem for a beam under bending yields a dependence of the deflection rate, f˙ , of the beam at its centre on applied load Q [43]. For a beam of rectangular cross-section of height 2h and width b we have: for 3-point bending. The solution was obtained neglecting a contribution of shear deformations to the displacement.
Therefore, measuring f˙(Q) yields the value of exponent n directly and then applying Eq. (3) gives the value of σ n .
The results of testing some fibres are presented in Figure 21, Figure 22 and Figure 23. In these Figures corresponding data for bulk single crystals are also plotted.  Comparing creep resistance of ICM-fibres and bulk crystals for yttrium-aluminium garnet shows an excellent correspondence of creep resistance of bulk crystals and corresponding fibres grown from melt. The data presented provide also an additional confirmation of the validity of the method of evaluation of creep properties of ICM-fibres by testing in bending oxide/molybdenum specimens. Creep resistance of Nextel 720 fibre (α-Al 2 O 3 +mullite) evaluated from experimental data presented in [50] is also shown in Figure 24. A number of important conclusions can be drawn, in particular:

1.
In temperature interval from 1100 to 1600°C, values of the creep resistance of single crystalline YAG and mullite as well as that of alumina-YAG-eutectic fibres obtained by using ICM are nearly the same. YAG fibre looks slightly better than the others. Still, their creep resistance can be certainly enhanced by crystallising them in the <111> direction.

2.
Single crystalline mullite fibre produced by ICM does not seem to be superior to, say, YAG fibres. Its creep resistance differs essentially from the experimental point by Dokko et al.

3.
Polycrystalline oxide fibres obviously lose their creep resistance below a temperature of 1200°C certainly due to an intrinsic behaviour of grain boundaries (see Figure 24).

Preliminary notes
First of all, it should be noted that there are three most important characteristics of heat resistant composites, those being creep resistance, fracture toughness and oxidation resistance. In the case of nickel alloys as matrices for oxide fibres composites sufficiently high oxidation resistance and fracture toughness are to be expected. That is why we will focus here on creep behaviour of nickel-based composites reinforced with ICM-fibres.
On a fundamental study stage of developing composites the main bulk of creep tests are conducted by bending simple composite specimens and then calculating tensile creep characteristics of the composites according to the method developed in [42]. The dependence of displacement rate f˙ of the specimen on load Q for a specimen with a circular cross-section of radius R is similar to Eq. (3): Such an approach to the creep tests simplifies and speeding up the procedure. The experiments can be even more speeded up if the microstructural creep model developed in reference [51] is used in the analysis of experimental data [48].
The model mentioned yields a dependence of composite stress σ on steady state creep rate ε where fibre characteristics are determined by the Weibull based strength/fibre length depend- β ; matrix characteristics are given by the power low of matrix creep, σ = σ m ( ε η m ) 1 m ; n = m + β + mβ, d is a characteristic fibre diameter; λ is a function of fibre/matrix interface strength α, 0 < α ≤ 1, α = 1 correspond to the ideal interface, which means that its strength is equal to the matrix strength.
Let us approximate Eq. (5) by a power function normally accepted while interpreting experimental data on creep of any materials. An important point is a possibility to calculate the value of exponent q in the approximation Rewriting Eq. (5) as and searching for values of q and C to provide the best approximation of Eq. (7) by Eq. (6), which means to find a minimum to the integral ( ) in which the integration limits are an interval of creep rates of interest. In the present context they are normally 10 -4 and 10 -1 h -1 .
A comparison of exponent q calculated and the exponent values obtained in the experiments with 2-steps loading of specimens is presented in Figure 25. One can see, that the calculations yield an acceptable result.
Note that a developing of oxide-fiber/nickel-based matrix composites has been meeting a number of the problems. The first one that is finding appropriate oxide fibres has been solved by inventing ICM.
The authors affiliated to NASA described what seemed to be the second problem. They found that sapphire fibres degraded during a liquid infiltration fabrication process. This was observed in testing fibres extracted from the matrix [49]. They concluded that the sapphire fibre is not an appropriate reinforcement for nickel alloys. The experiment was correct, but the conclusion was not completely correct. It will be shown below that the matrix melt interacts with the fibre and causes the fibre degradation but after melt crystallised the matrix heals the defect resulting from the interaction provided there is a good adhesion between the fibre and matrix. Moreover, preexisting defects can also be healed [42,50]. In fact, the conclusion mentioned can be a kind of the delusion. The real problem is to find ways to organize a strong interface between an oxide fibre and nickel-based matrix.

Fabrication of nickel-based composites reinforced with ICM-fibres
Three commercially available Russian superalloys have being used as matrix materials. The alloy marked as VKNA-4U contains Al, Cr, W, Ti, Co, Mo and C. That marked as VKNA-25 differs from VKNA-4U by the presence of 3.5% rhenium. The GS-32 alloy contains Al, Cr, Mo, W, Ta, Co, Nb, Re, C, B.
Since Ni-based alloys do not normally wet oxides, pressure casting is used to produce the composites. To fabricate specimens for mechanical testing, a special simple mould was designed that is composed of external quartz tube with one end closed, internal alumina tube with open ends, porous alumina plug at the top of the fibre bundle, and alumina bed at the bottom of the quartz tube [51]. Temperature and pressure of argon gas in the infiltration process were 1550°C and 5 atm, respectively. Crystallisation of the matrix was performed in the axial temperature gradient to make either a single crystalline or columnar microstructure of the matrix. The diameter and length of the specimens were ~5 and ~50 mm, respectively.

Mechanical properties of nickel-based composites reinforced with ICM-fibres
Creep experiments were carried in 3-point bending by either step-wise loading and in such case the value of the exponent in a creep power law is determined directly or by a single load, in which case the exponent is calculated as shown above. Then tensile creep characteristics of the composites are calculated according to a procedure also described above. Some examples of dependencies of the creep resistance (accepted as a stress to cause 1% creep strain for 100 h) on fibre volume fraction are presented in Figure 26.  [44 ]. The test temperature is 1150°C.
It can be seen that for a particular composite creep resistance can reach a value of 150 MPa at a temperature of 1150°C. It is important to note that the density of the composites under consideration is between 6.5 and 6.8 g/cm 3 ; the density of modern superalloys are approaching 9 g/cm 3 .
A compilation of the data presented in Figure 26 is plotted in Figure 27 as a correlation of the maxima of creep resistance of the composites reinforced with various oxide fibres on fibre volume fraction. Such presentation that shows by arrows the possible developments demonstrates clearly an effect of particular fibre/matrix combinations on the creep properties of the oxide-fibre/nickel-based-matrix composites. MMCs that is the existence of a maximum on a dependence of the strength on fibre volume fraction. This remark is important for a discussion of strength/temperature dependence for an oxide-fibre/high-entropy-matrix composite below, in Section 5.

Composites with a high entropy metal matrix
An invention of high entropy alloys (HEAs), which are normally composed of five or more elements each with concentration between 5-35 atomic percent [53,54] stimulated the opening of a new chapter in the high-temperature metal matrix composites because (i) among such alloys there are some with both high melting points and sufficiently high oxidation resistance [55] and (ii) equiatomic compositions of HEAs promises an easier solution of the wetting problem while dealing with oxide reinforcements. Hence, the author's research group together with a Ukrainian group performed the first experiments aiming at the understanding of prospects of oxide-fibre/HEA-matrix composites [56].
Fibres of Al 2 O 3 -Al 5 Y 3 O 12 and Al 2 O 3 -Al 5 Y 3 O 12 -ZrO 2 eutectics were used as reinforcements. A high entropy alloy containing Fe, Co, Ni, Cr and W with a melting point of about 1450°C was chosen as the matrix. The alloy is composed of two phases, those being a base BCC phase and inclusions of a μ-phase of the Fe 7 W 6 -structure containing all the elements of the alloy, but with their ratios differing from that of the base phase [57].
Specimens were again produced by pressure liquid infiltration of a fibre bundle located in quartz casting mould with the matrix melt. In these experiments, a molybdenum foil 25 microns thick covered the inside surface of the mould to prevent the interaction between the HEA melt and silica. The necessary gas pressure in this case, 1 -1.5 atm., is much lower than that in the case of infiltrating oxide fibres with nickel superalloys, which is no less than 5 atm. This is indicative of good wetting in this system, which determines a sufficiently high strength of the fibre/matrix interface.
The microstructure of the composite is illustrated in Figure 29. The results of The X-ray microanalysis of the matrix shows that the base phase (grey phase in the Figure) contains all the elements in nearly equi-atomic ratios, tungsten being an exception. The composition of an inclusion (white phase) is close to Fe 1-(x+y+z) Cr x Co y Ni z W 6 . In these experiments, all specimens have relatively large fibre volume fractions, between 40 and 50%.
Composite specimens were tested in the 3-point bending to measure the strength at 20 -1300°C. Temperature dependence of the strength is presented in Figure 30. The experiments yields two interesting observations.  First, the room temperature strength values are relatively low and this is to be the case since the specimens have large fibre volume fractions. The dependence of strength of a brittle-fibre/ ductile-matrix composite on fibre volume fraction V f has a maximum. An inherent reason for decreasing composite strength at V f ≥ V f (max) is arising a crack in the composite over a fibre cluster as a result of a single fibre break at a rather low stress followed by fracturing a number of adjacent fibres. The crack size arising because of fracture of a fibre cluster can be too large for the matrix to dissipate the elastic energy and arrest the crack. The probability of fibre cluster formation increases with the fibre volume fraction increases. A corresponding model of the composites failure behaviour was developed by Mileiko [1]. Recent experimental results revealing such behaviour of oxide-fibre/nickel-based matrix composites are published in [56]. A particular value of V f (max) depends on fracture toughness of the matrix, fibre strength characteristics and a degree of non-homogeneity of fibre packing. Usually, for composites produced by infiltration of a fibre bundle with a matrix melt, V f (max) < 0.3 for room temperature behaviour because of a rather poor homogeneity of the fibre distribution in a specimen crosssection. Hence, the low strength of specimens at room temperature is characteristic of composites with larger fibre volume fraction and the same composites with V f < 0.3 would definitely be characterised by a much higher strength.
Secondly, the strength of composites with V f = 0.4 -0.5 at room temperature and at 1200°C is approximately the same. Two factors contribute to this effect. First, this can be a result of high fibre/matrix interface strength, which results in loading the fibre to high stresses. The second factor is a shift of critical value V f (max) of fibre volume fraction to higher values as a result of an increase in the matrix fracture toughness with temperature.
Composites of such a type are to be developed for high temperature use. Hence, creep properties of the composites are of importance. At a prelimary stage of the work described in [60] no creep tests have been performed yet, the results of strength tests together with a previous experience of the present author in studying creep of fibrous composites [51,52,58] allow formulating some expectations. The theory of creep behaviour of MMCs [51] and a large bulk of the experimental data [56,59,60] show that the most important structural parameter of composites with oxide fibres produced via liquid infiltration route is the fibre/matrix interface strength determined by a degree of wetting in the system. If a matrix melt wets a fibre then the matrix heals the surface defects that yields an increase in the effective fibre strength by a factor as high as 5 in the case of a fibre obtained by the internal crystallisation method (see above Figure 13). The effective fibre strength determines creep resistance of composites at temperatures, at which fibre creep is negligible. The oxide fibres used in these experiments do not creep at a temperature of 1200°C (see Section 3). Also it is important to note that good wetting and, consequently, a strong interface make the basis of high creep resistance of the composites with large fibre volume fractions. Therefore, the composites described in short communication [56] can be expected to have a high creep resistance at temperatures to 1200°C.

Molybdenum-based composites reinforced with ICM-fibres
As was mentioned in Section 1 the development of refractory metal alloys for high temperature usage meets a problem in reaching an appropriate balance of three main characteristics of such alloys those being creep resistance, fracture toughness (damage tolerance), and oxidation resistance. Hence, the developments of fibrous composites with refractory alloy matrices can obviously be considered as a way to overcome a problem of reaching the balance between creep resistance and fracture toughness. Recent results show a possibility to obtain such composites with an enhanced oxidation resistance [61,62]. Measurements of oxidation resistance and creep properties were performed on specimens composed of unalloyed molybdenum matrix (containing 0.04% of impurities) and fibers of sapphire, single crystalline YAG and a number of alumina-garnets eutectics (Al 2 O 3 -R 3 Al 5 O 12 where R is Y, Yb or Tb). Three sets of the experiments were performed to evaluate oxidation resistance of the composites.
First, specimens of molybdenum and molybdenum-matrix composites reinforced with alumina-YAG eutectic and sapphire fibers of the same sizes were kept in air at a temperature of 1000°C for up to 1 h. A decrease in the mass of the specimens with time is plotted in Figure  31. A significant difference in the oxidation kinetics of molybdenum reinforced with the fibre containing yttrium compared to both pure molybdenum and molybdenum reinforced with sapphire fibre is obvious. Then the specimen reinforced with the Al 2 O 3 -Y 3 Al 5 O 12 eutectic fibre was exposed to a temperature of 1300°C for nearly 2 h. For comparison, a "fresh" pure molybdenum specimen was also tested. An essential difference in the oxidation of the composite and unreinforced pure molybdenum can be clearly seen. A hypothesis of formation of a sufficiently stable layer of yttrium molybdate on the composite surface attacked by oxygen can explain the result obtained. The second series of the experiments was carried out to prove the hypothesis just formulated by a direct observation of the surface of a composite with molybdenum matrix and single crystalline YAG fibre heated in air at a temperature of 1000°C for 1 h. In Figure 32, one can see a fibre of a nearly tri-angular shape being originally single crystalline YAG. After heating, a layer in the fibre in a vicinity of the fibre/matrix interface occurs, which looks different from the centre of the fibre. The X-ray microanalysis was performed in three points marked on the micrograph. It was shown that (i) a chemical compound close to yttrium molybdate, Y 2 Mo 3 O 12 arises on the surface of both the molybdenum matrix and oxide fibre; (ii) the fibre has lost yttrium oxide and the composition of a layer arising near the interface is close to alumina. Examination of a specimen surface gives exactly the same result ( Figure 33): The The X-ray microanalysis performed over areas 60X60 microns around points 4 -6 shows that the surface is coated with a layer of the yttrium molybdate and depletion of yttria in the fibres is also observed. The only Y 2 O 2 -MoO 3 phase diagram published in 1982 and cited in [63] shows that the melting point of Y 2 Mo 3 O 12 is 1310°C and this sets a limitation for the usage of the composite. Note that the specimens in this experiment had no coating, so a large quantity of yttria was withdrawn from the fibre exposed to a free surface. Obviously, a special coating will prevent molybdenum from oxidation and formation of the molybdate to heal possible defects in the coating will occur only in a vicinity of the damaged area of the coating.
The third set of the experiments was a study of oxidation kinetics of Mo-matrix composites with alumina-based eutectic fibres containing Y 3 Al 5 O 12 , Tb 3 Al 5 O 12 and Yb 3 Al 5 O 12 . Exposure temperature for these experiments was lower than in the first two sets, 700°C, and exposure time larger, up to 10 h. The oxidation kinetics is presented in Figure 34 together with that for pure molybdenum. The data for molybdenum is taken from reference [64]. It should be noted that those data have been obtained by exposing specimens in atmosphere with at a partial oxygen pressure of 76 mm Hg, which is much lower than that in normal air atmosphere. Still, the oxidation rates of the composites are much smaller.
A layer of a compound containing both molybdenum and yttrium is found on the surface of the specimen with Al 2 O 3 -Y 3 Al 5 O 12 eutectic fibres with atomic ratio Mo:Y ≈ 2:1, which corresponds to Y 2 Mo 4 O 15 . Comparing this observation with the data obtained in the second set of the experiments leads to a hypothesis on the formation of a sequence of the molybdates with enhancing the exposure temperature. Nearly the same is found on the surface of a composite with eutectic fibers containing lanthanide elements. Again the Mo:Tb ratio is approximately 2. In the absence of published phase diagrams of a corresponding oxide system, it remains to expect an occurrence of Tb 2 Mo 4 O 15 on the specimen surface.
Tensile creep tests of the composites with Al 2 O 3 -Y 3 Al 5 O 12 -eutectic fibre as model materials were conducted in vacuum at temperatures 1000 -1300°C [27]. The dependence of rupture stress on time for temperature 1300°C is presented in Figure 35 together with the data acquired from experimental results published by Jain and Kumar [11], which describe creep rupture behaviour of molybdenum alloy of sufficiently high oxidation resistance. Creep rupture properties of the composite with a low density (~ 7.5 g/cm 3 ) on 100 h time base are better than those of the alloy. At lager rupture time the difference between values of the rupture stress for the composite and alloy will be even larger. This is because creep resistance of metal matrix composites with fibres of creep resistance much higher that of the matrix is determined by the fibre strength (see Eq. (5)) that yields a slop of the stress/time dependence in log-log coordinates for the composite much smaller than that for an alloy.

Prospects of the internal crystallisation method
The internal crystallisation method can be further developed to produce fibres and composites differ from those described in the previous sections. We limit ourselves just with two examples to illustrate the work in progress.

Composite oxide fibres
Fibrous composites containing inherently brittle matrix and brittle fibre can behave in a nonbrittle manner provided a fibre/matrix interphase allows either fibre/matrix delamination or   Figure  24), so the usage of single crystalline or eutectic fibres of the ICM type is wanted. Coating of ICM fibres is possible [68] but not a convenient procedure.
Therefore, a new design of the fibrous microstructure able to arrest cracks should be looking for. A possible solution is a composite oxide fibre containing a main ingredient, say sapphire, and a "weak" inclusion, say a kind of hexaaluminate, obtained by using the internal crystallization method [69]. To crystallise such fibres, raw mixtures of oxides containing 2 -3 -4.5 -6 mass per cent of calcia in alumina were prepared to be used in the ICM process. Pulling rate of the oxide/molybdenum blocks varied from less than 1 mm/min to 800 mm/min. In the first experiment, just to show a possible microstructure of hexaaluminate, pure CA6, without alumina was crystallised, the result is presented in Figure 36. It can be seen that the flat surface of the fibre is coincided with the basis crystallographic plane of CA6. Also it is obvious that the fibre is not single crystalline and the fibres of such type cannot be sufficiently strong.
Composites with Al 2 O 3 -CaAl 12 O 19 fibre and brittle molybdenum matrix (molybdenum in a block shown in Figure 3 is brittle being fully recrystallised) demonstrate quasi-plastic behaviour as can be seen in Figure 37a. Fracture surface (Figure 37b) of a composite specimen of the same type reveals fibre pull out.  Hence, there is a hope to develop composite oxide fibres to be an effective reinforcement for brittle matrix.

Silicide-fibre/molybdenum-matrix composites
As shown in Section 1 a mainstream in the development of heat resistant materials beyond nickel superalloys is searching for molybdenum alloy with a necessary balance of three main characteristics, those being creep resistance, damage tolerance and oxidation resistance. A goal to develop an alloy with the use temperature as high as 1300°C is expected to be reached by introducing into molybdenum matrix brittle particles of complex silicides. However, the alloys with sufficiently high creep and oxidation properties remain to be optimized to satisfy damage tolerance requirement. At the same time, it was mentioned in Section 1 that reinforcing a ductile metal matrix with brittle fibre can lead to a composite of an enhanced fracture toughness properties. Hence, it is interesting to look at prospects of the internal crystallization method to make silicide-fibre/refractory-metal-matrix composites. Preliminary experiments with silicide-fibre/molybdenum-matrix composites were conducted recently [70]. A most difficult couple of the ingredients to deal with is a molybdenum carcass and molybdenum disilicide, MoSi 2 , with the melting temperature 2020°C and this couple was chosen for the first experiments. The MoSi 2 melt was expected to interact with molybdenum to form silicides with a larger content of Mo, Mo 5 Si 3 or even Mo 3 Si. In fact, the MoSi 2 -melt/molybdenum interaction yields a drastic change in the fibre shape and its composition. The fibre lost its plane surfaces, Figure 38, they show traces of a complicated process of the chemical interactions. The Mo:Si atomic ratio measured in the points marked in the area of the fibre is about 3.4, which is close to silicide Mo 3 Si. The The X-ray microanalysis of the matrix reveals a presence of silicon, certainly in the form of inclusions of Mo 3 Si of small size since the The Xray spectrum obtained from a cross-section of a specimen ( Figure 39) shows only molybdenum and Mo 3 Si. Figure 39. The X-ray spectrum obtained from a cross-section section of the molybdenum-silicide-fibre/molybdenummatrix specimen.
Creep tests of specimens were conducted in bending at a temperature of 1410°C and the results of the experiments were interpreted on the basis of the method outlined above. One specimen was tested under a step-wise changing of the load, and the value of n occurred to be around 5. All the data were obtained by assuming n = 5; they are presented in Figure 40.
It is important to note that value of n = 5 was evaluated for a composite with fibre volume fraction V f = 0.2. Because the value of n increases with V f increasing (see Figure 25, in which the exponent was denoted as q), the value of creep resistance for composites with V f different from 0.2, which were calculated assuming n = 5, are to be corrected as shown in Figure 40 by the arrows. This would make the dependence of the creep resistance for fibre volume fraction close to linear, which is characteristic for composites with a strong fibre/matrix interface.
Therefore the preliminary data just described show that the ICM is a prospective way to obtain silicide fibrous composites.

Conclusions
1. The internal crystallisation method (ICM) can be a base for developing effective technologies to produce a large family of oxide fibres with mechanical properties similar to those of the fibres produced by using traditional methods.

2.
ICM-fibres can be used to obtain heat resistant composites with a number of the matrices including nickel superalloys and recently invented high entropy alloys.
3. An appropriate choice of the chemical composition of oxide reinforcements for refractory matrices can be a way to reach a necessary balance of the main properties of heat resistant materials, those being creep resistance, fracture toughness and oxidation resistance.

4.
Composite ICM-fibres containing highly anisotropic oxide inclusions can be an effective reinforcements for brittle matrices to produce non-brittle composites.

5.
Silicide fibrous reinforcements can also be produced by using ICM.