As a container material for growing single crystals, platinum has to perform three functions:

Experience has shown that it is the third function (retaining the optimal shape to determine the mechanical strength of the container) which requires the maximum amount of platinum. The container wall thickness is not important for functions [a] and [b], especially for [a]. (1).

This paper describes how the mechanical strength of the containers can be increased – not by increasing the thickness of the platinum – but by using platinum-based composites as the structural material (2, 3). Ceramic-coated platinum is a potential high-temperature material for this use, as it allows crucibles to be manufactured with less platinum, but in similar sizes and with improved properties (4, 5).

The evaporation of ceramic onto metal by plasma methods may be considered an appropriate commercial technology due to its simplicity and resulting lower costs (6). The composite crucibles, shown in Figure 1, can be produced by a process that includes four steps (7):

Contrary to expectations, the cohesive strength of such a platinum-alumina join is very strong. The coating does not separate from the platinum even, for instance, after using a composite crucible for 7000 hours to grow PbWO₄ single crystals. Also, the appearance of cracks in the ceramic did not lead to separation of the coating from the substrate.

Using plasma technology therefore provides a join of high cohesive strength, but the reason for this high cohesive strength is not yet understood, and the nature of the cohesion mechanism between the Pt and Al₂O₃ is not clear. In this paper the morphology of the transition layer at the interface between the metallic matrix and the ceramic coating, and the effects of this layer on the mechanical properties of platinum will be considered.

Experimental Procedure

It is well known that the adhesion characteristics of plasma coatings depend strongly on the geometry of evaporation (or shape of the metallic substrate) (6). In addition, the stress state of a cylindrical surface is different to that of a layered parallelpiped of comparable width and thickness (8). Hence, data obtained from ceramic coated flat samples would not explain the behaviour of a composite cylindrical container. Therefore, such laboratory samples of Pt-Al₂O₃ composite were not made for testing. Instead, test samples were cut from the platinum substrate of composite containers (wall thickness 0.5 mm) taken during the four stages of manufacture: Pt, Pt_sanded, Pt+Cer and Pt+Cer+An. These samples had a standard “double spoon” shape with a working area of 20 × 3 mm². The ceramic layer (3 mm thick) was separated from the metallic matrix prior to cutting the samples, as it was impossible to cut samples from a ‘metal-ceramic’ crucible. Similar metallic samples were also taken from containers which failed in use. Both the back surfaces of the samples (“platinum” and “transition layer” sides) were studied in detail in their initial states by optical and scanning electron microscopes (SEM). Tensile tests were carried out in a regime of semi-creep when the traverse rate was very slow: 0.1–1 mm h⁻¹ (applied stresses of 50, 70 and 90 MPa) at 1300°C in air, while some of the samples were stretched at room temperature (at a traverse rate of 1 mm min⁻¹ ). After that, their back and fracture surfaces were again studied.

Clearly, the information obtained cannot directly characterise the mechanical behaviour of a Pt-Al₂O₃ composite container, but it can allow an understanding of the adhesive mechanism between the platinum and the alumina coating.

Morphology of the Transition Layer between Pt and Al₂O₃

As the platinum crucible has been pressed out of rolled sheet without intermediate recrystallisation annealing, the platinum substrate is in a deformed state, see Table I. Treating this platinum surface with alumina particles (∼ 300 µm in size) leads to the appearance of a relief consisting of approximately flat areas oriented to the substrate plane at arbitrary angles. The shapes and sizes of these regions are similar to those of the alumina particles used for sanding the platinum. The depth of the relief (∼ 30 µm) also correlates with the particle size, as long as the work-hardening of the platinum, which prevents the alumina particles from penetrating into the metallic matrix, is taken into account. There are two kinds of alumina inclusions in the sanded platinum: large and small. On SEM images, they appear as dark regions, see Figure 2(a). The dimensions of the large inclusions, ∼ 20–30 µm, correspond to the pyramidal ends of alumina particles used for sanding. Therefore, the large inclusions may be considered to be the sharp tips of particles which have been stuck into the metallic matrix.

The average size of the small inclusions is ∼ 1 µm. Large and small inclusions are uniformly distributed on the surface; their positions are not connected with the distribution of surface defects caused by preliminary processing of the surface. Despite the appearance of the relief, sanding does not influence the microhardness of the “platinum” side of the samples because of the considerable thickness of the crucible walls (Table I).

After the evaporation of alumina and the separation of ceramic from the substrate, particles of Al₂O³ are observed on the platinum surface. The average size of the large inclusions has increased to ∼ 50 µm, see Figure 2(b), although their quantity per unit area of surface remains the same. By contrast, the dimensions and concentration of the small inclusions have not changed. The microhardness of the “platinum” side in the Pt+Cer samples has considerably increased, see Table I.

The stabilising anneal of the ceramic has no influence on the size of both kinds of alumina inclusions or their distribution on the surface, but “jogs” on the surface becomes round or fused, see Figure 2(c). However, the microhardness of the “platinum” side has decreased in comparison with the Pt and Pt_sanded samples (Table I). A coarse grained structure, in which all the crystallites are in the deformed state, appears on the “platinum” side, see Figure 3, while features of recrystallisation, such as grain boundaries or a new kind of relief, do not appear on the rough “transition layer” side. After evaporation of the ceramic and its stabilising anneal, the character of the distribution of the alumina inclusions on the surface remains the same.

The results obtained are important for understanding the cohesion mechanism between platinum and alumina in composite crucibles manufactured by plasma technology. The main feature is the “point”, “fragmentary” or “discontinuous” contact of the Al₂O₃ with the transition layer (9). Because of such a join, both the environment and the melted scrap could percolate between the metal wall and the ceramic coating, see Figure 4. However, this circumstance does not affect the usefulness of the containers. Indeed, the scrap never chemically reacts with the walls of the crucible, and the level of mechanical stress due to scrap percolation between wall and coating is not sufficient to break the cohesion of the join. Moreover, experience has shown that separating the ceramic from the platinum in used crucibles requires as much effort as in new ones, and the number and size of the large alumina inclusions in the platinum is the same in both cases. It should be particularly noted that small inclusions of alumina are absent on the samples.

The alumina particles either do not chemically react with platinum, or, if reactions do occur then the region where it happens is so thin that it cannot be detected by conventional SEM or X-ray techniques. Therefore, the join between the platinum and alumina may be considered as “mechanical” when the sharp ends of the ceramic particles penetrate into and remain in the metallic matrix. The large inclusions serve as anchors for the ceramic coating which “glues” to them during plasma evaporation. No joining of alumina to platinum is revealed in other places, although the ceramic replicates in detail the surface of the transition layer. The increase in size of the large inclusions, the preservation of their numbers and character of distribution on the surface after evaporation of the ceramic agree with this supposition. On the other hand, small inclusions can play the role of second phase particles which increase the yield stress of the metallic matrix by blocking dislocation motion as it takes place in dispersion strengthened metals.

Mechanical Properties of Platinum with a Transition Layer on the Surface

Identical mechanical behaviour is inherent in all the materials: Pt, Pt_sanded, Pt+Cer, Pt+Cer+An at room temperature (10). Samples fail after ∼ 3–5% elongation, and no necking is observed on their back surfaces, see Figure 5. Localisation of plastic deformation takes place near fracture surfaces, but happens in a very narrow region, see Figure 6. This is very unusual behaviour for a pure f.c.c. metal, which should have high plasticity (11). On the other hand, analysis of fracture surfaces (where necking to a line occurs even in the most strengthened material, such as Pt+Cer, see Figure 7) shows that all the samples are in a ductile state. The results for the Pt samples (Table II) confirm metallographic observations and measurements of microhardness, indicating that the metallic matrix had been hardened before test: yield stress is 55 MPa.

Sanding the platinum increases the yield stress by a factor of two (100 MPa), but has practically no effect on the strength and microhardness of the Pt_sanded samples (170 MPa and 8.9 MPa, Table II and Table I, respectively) in comparison with Pt samples. This is normal behaviour for f.c.c. metals, as the volume of a sample generally determines its strength, while the surface makes a significant contribution to the yield stress of a metallic sample. In practice, the sanded surface does not change its morphology up to the fracture area, and the alumina inclusions disappear only in the vicinity of the fracture surface because of local plastic flow of the material, see Figure 6(b). Evaporation of ceramic causes a big rise in the microhardness (9.7 MPa, Table I) and yield stress (130 MPa, Table II) in the Pt+Cer samples, and in doing so the strength of the material increases by ∼ 100 MPa, as in the Pt samples and the Pt_sanded samples, while the micro-hardness of their “platinum side” after mechanical test remains the same (9.7 MPa, Table I). A stabilising anneal decreases the initial microhardness and yield stress for Pt+Cer+An samples (6.9 MPa and 65 MPa, respectively). After tensile testing, this parameter becomes 8.9 MPa and the strength reaches 160 MPa, respectively. Despite differences in some mechanical properties (σ_0.2, σ_B and H_V), the morphology of the fracture and back surfaces of the Pt+Cer and Pt+Cer+An samples does not change.

Observation of the behaviour of the samples at 1300°C in air has shown that they are typical of f.c.c. metals at high temperatures: the applied stress and roughness of the back surface determine the creep- rupture life (τ_life) of the sample under load (11). In practice, at high stress (90 MPa), the τ_life and the morphology of the back surface near the fracture zone do not depend on the type of material (Table II, and Figures 8 and 9). Lowering the applied stress increases the sensitivity of τ_life to the state of the surface layer (Figures 10 and 11). For example, the size and number of neck regions in the samples start to grow at middle and low applied stresses. This causes an increase in the area near the fracture zones where alumina inclusions are absent. The τ_life for Pt at 70 MPa and 50 MPa (4 and 18 MPa, respectively) are more than 1.5–2 times greater than for Pt+Cer and Pt+Cer+An samples though are the same as the Pt_sanded samples which also have rough surfaces with implanted alumina particles.

Thus, the geometrical characteristics of the surface relief and the distribution of alumina inclusions are the same in all three cases. This effect is connected with the structure of the surface layer in the samples, where the deep relief and the second phase inclusions (here, Al₂O³ particles) induce the fracture process, including the localised plastic deformation in the platinum matrix. The difference between Pt+Cer, Pt+Cer+An and Pt_sanded samples may be explained by the fact that in the first two materials (Pt+Cer and Pt+Cer+An) the transition layer is dispersion strengthened platinum, which displays high strength and semi-brittle fracture mode at high temperatures, see Figure 12, whereas in the Pt_sanded material the surface layer only contains mechanically penetrated alumina particles. Indeed, the regime of heating the platinum substrate during evaporation of alumina is similar to that of the temperature treatment of platinum strengthened by oxide particles. Despite this, the fracture surfaces of all the samples investigated at high temperature have the same morphology attested earlier as necking to a line.

Discussion

These experiments have shown that the transition layer between the platinum and the alumina coating (30 µm in depth) is formed in stages by sanding and plasma evaporation of ceramics. The transition layer contains small (∼ 1 µm) and large (∼ 20–50 µm) alumina inclusions and possesses a rough surface on the ceramic side. The Al₂O³ particles have penetrated into the platinum matrix during the sanding stage, but this treatment has not changed the mechanical behaviour of the platinum either at room or elevated temperatures.

Differences begin to show only after the evaporation of alumina, when the sanded surface has been heated to a temperature close to the alumina melting point then cooled very quickly to 300–400°C. Performing a stabilising anneal is important for the mechanical properties of the alumina coating. However, its influence on the morphology and properties of the transition layer is negligible. In performing a stabilising anneal it should be noted that the surface layer, enriched by oxide particles, causes plastic deformation in the recrystallised grains of platinum.