PLASMA ENAMELLING: EFFECT OF FEEDSTOCK COMPOSITION ON THERMAL STRESS
Y Bao¹ ,T Zhang² and D T Gawne^2
1School of Engineering, London South Bank University, London, SE1 0AA, UK.
²School of Engineering, Kingston University, London, SW15 3DW, UK.

Abstract
Experimental measurements have been carried out with the aim of investigating the thermal stresses generated during plasma-spray deposition of enamels on steel. The research shows that the behaviour of these materials is fundamentally different from metals and ceramics: the quench stress in the enamel can be eliminated by control of the plasma-scanning action. The effect is unique to enamel because its low glass-transition temperature in relation to the plasma-spray thermal cycle enables the quench stress to be relaxed. However, significant thermal mismatch stresses due to the expansion mismatch between the enamel and the steel substrate can stilldevelop on subsequent cooling to the detriment of the properties of certain enamels. The work shows that the addition of a ceramic, such as alumina, as a second phase allows this to beminimized. Specifically, the thermal stress in enamel can be reduced to zero by control of the volume-fraction of the second-phase in the feedstock powder. A theoretical model is developed for the prediction of thermal stresses generated in enamel during plasma-spray deposition.

Introduction
Plasma enamelling is a new enamelling process under development [1]. The process is illustrated schematically in Fig.1. It consists of injecting powdered frit into a plasma jet in which it is melted and projected onto an article to form an enamel coating. The frit particles melt in the plasma; accelerate to velocities of 100- 200 m/s and impact with the substrate surface to form disc-shaped splats. The splats flow, accumulate and compact into a dense deposit as a result of the high impact velocity. The spray gun scans across the substrate surface and then repeats the scanning action until a coating of the required thickness is obtained. Control of the heat transfer from the plasma gas to the particle interior, the particle trajectory and consequent flow on impact enables dense enamel to be produced. However, the cooling rate of the splats is of the order of 10⁶ oC/s and this in combination with the thermal mismatch gives rise to thermal stress. The mismatch is due to the difference in thermal expansion coefficients between the enamel and the substrate materials.
Thermal stress can lead to deterioration in the properties of enamel and in extreme cases result in cracking and delamination. This paper investigates the mechanism and development of thermal stress in plasma-sprayed enamel on steel substrates.

Thermal Stress
The thermal stress, σt, in a plasma-sprayed coating may be expressed as the sum of quench stress (σq) due to the contraction of coating, and the mismatch stress (σ_m) due to the difference in the coefficients of thermal expansion between the coating material and the substrate [2].

where T_m, T_s and T₀ are the melting temperature of the coating material, the temperature of the substrate during spraying and room temperature respectively; α and E are coefficient of thermal expansion and Young’s modulus. The subscripts d and s denote the deposit and substrate respectively. Since coatings are an accumulation of individual splats and the Young's modulus of ceramics is high, stress relaxation in ceramic coatings may only take place by sliding between splats. As a result, the quench stress and the stress due to the mismatch between thermal expansion coefficients may only be partly relaxed at any stage. For glass materials, the melting temperature should be replaced with the glass transition temperature, T_g. At temperatures above T_g, the quench stress may be relaxed. The thermal stress is then proportional to: (a) the difference in thermal expansion between the coating and the substrate, and b) the glass transition temperature of the coating.
This analysis suggests that there are two methods to reduce the thermal stress in enamel: (a) selecting enamel materials with low glass-transition temperatures and (b) minimising the difference in the thermal expansion coefficients between the enamel and the substrate. In practice, the former may lead to coatings with low strengths and low service temperatures, which may limit their application. This difficulty may be avoided by employing method (b) using a novel approach available in plasma spraying: adding a second phase to the enamel with an expansion coefficient that compensates for the mismatch. In this investigation, the enamel had a higher expansion coefficient than the steel substrate and so alumina was selected because this has a much lower expansion coefficient than steel, thereby compensating for the enamel and minimizing the mismatch. The reason why this is particularly effective in plasma spraying is that the alumina melts and forms a smooth, well-bonded second phase. The addition of alumina to conventional enamel, on the other hand, does not result in the alumina melting and produces sharp angular particles that may initiate cracks and form an aggressive surface. In plasma enamel, the reduced residual stress combined with the increased strength of the ceramic has the potential of improving the mechanical properties.

Experimental details
Plasma spraying was undertaken using a Sulzer-Metco plasma spray system with an MBN torch and MCN control unit. A 4MP powder feed unit and fluidised bed hopper was used to feed the powder externally into the plasma jet. The feedstock powders were sprayed onto the steel substrate using the plasma arc power of 27 kW and the plasma gas mixture of argon plus 7 % hydrogen.
The enamel frit was supplied by Escol Products Ltd (Huntingdon, UK), milled and classified into a size range of 32-45 μm. Its softening temperature and linear coefficient of thermal expansion are 420 °C and 14.47×10^-6 K^-1 respectively. Alumina powder with average particle size of 20 μm (Sulzer-Metco) was mixed with the enamel powder by ball milling. The compositions of the mixtures are given in Table 1. The additions of alumina were 20, 40 and 60 % by weight. The substrate used was low-carbon sheet steel supplied by Corus PLC (Port Talbot, UK). The steel was degreased and surface roughened with grit blasting immediately before plasma spraying.
The stress measurement was based on the deflection of a cantilever strip during the deposition and subsequent cooling. The experimental set-up is shown in Fig.2, in which a 120 x 20 x 1.2 mm steel strip was clamped at one end and the deflection of the strip during deposition measured using a linear variable displacement transducer (LVDT) attached to the back of the strip. The LVDT was connected to a signal amplifier and the amplified displacement signal was recorded. The temperature of the strip was monitored using a thermocouple attached to the back of the strip.
Computer simultaneously recorded the temperature and the displacement of the sample during spraying

Fig. 2 Schematic of stress-measurement apparatus.

The stresses in the deposit were then calculated with the theory developed for the bimetallic strip in which the stress in the coating is [3]:

where σ is the stress at the top surface of the thin layer coating; δ is the deflection of the strip; subscripts d and s denote the deposit and substrate respectively; a is the thickness of coating and h is the total thickness of coating and substrate; L is the length of the strip; I is the second moment of inertia which equals a³/12 in this case for unit width.

Results and discussions
Structure of the enamel: The microstructures of pure enamel coating, the composite coating and the alumina coating are shown in Fig. 3. Fig. 3a shows that the pure enamel is a continuous singlephase coating. No visible boundary between the splats can be detected under the microscope, which suggests that during deposition the splats are fused together. In the composite coating, the alumina appears as a white phase and the enamel as a grey phase (Fig.3b). The enamel forms the continuous matrix phase and the alumina a plate-like secondary phase. This indicates that the alumina particles were fully melted before impact with the substrate. It is particularly interesting to observe that although alumina has a much higher melting temperature than that of the enamel matrix (2050 oC and 420 oC respectively), its melt-viscosity is much lower. As a result, the alumina particles flow extensively and form thin splats during impact, while the enamel forms substantially thicker splats due to its reduced flow. The structure of the coating is a plate-reinforced enamel matrix composite with the major plate surface aligned parallel to the surface of the substrate. The microstructure of the pure alumina coating is shown in Fig.3c and is significantly different from that of the enamel coating: splat, grain boundaries and porosity are evident. The enamel is clearly the much denser coating.

Fig. 3 Cross-sections of (a) enamels coating, (b) composite coating with 20 % alumina, (c) alumina coating.

Alumina content: The volume fraction of alumina in the enamel was determined by area measurement using optical microscopy. The volume fractions of alumina in the feedstock can be calculated with the equation:

(3)

where W and ρ refer to the weight fraction and density while the subscripts a and e denote alumina and enamel respectively. The calculated results are given in Table 1. The measured volume fractions of alumina in the coatings were similar to those in the feedstock powders. The volume fractions in the feedstock powder were, therefore, used to calculate the stresses in the coating during deposition.
The Young’s modulus, E_c, and thermal expansion coefficient, α_c, of the composite coatings may be calculated using the rule of mixtures:

(4)

where V_i, E_i and α_i are the volume fraction, Young’s modulus and expansion coefficient of component i respectively. The compositions of the five-feedstock powders, the calculated expansion coefficient (α_c) and Young’s modulus (E_c) of the resulting coatings are given in Table 1.

The thermal expansion coefficient of the pure enamel was much higher than that of steel (14.5× 10^-6 °C^-1 compared with 11.7×10^-6 °C^-1 ). For the pure enamel coating, the mismatch of thermal expansion coefficient with the steel was therefore 2.8×10^-6 °C^-1 and cracks were observed. As the alumina is added to the enamel, the thermal expansion coefficient of the composite coating decreases and at larger additions approaches that of the steel substrate. Cracks were no longer observed in the enamel as a consequence of this effect.

Stress relaxation: Fig. 4 gives the deflection measured in the coated steel strips as they cool down towards the end of deposition; time moves from right to left on the diagram as the cooling process proceeds. There is no deflection of the strips above a temperature of 450 °C. On further cooling below this temperature, however, substantial deflections occur. The observation that the deflection and therefore the thermal stress are zero above 450 °C is highly significant. It may be understood by a consideration of the unique properties of enamels in the context of plasma spraying.
During deposition, an individual frit particle melts in the plasma jet, strikes the substrate and flows into a disc-shaped splat. At this stage, the splat will contain a substantial quench stress. As the plasma flame continues to scan the surface, however, it will heat the solidified splat and provide the possibility of stress relaxation. At elevated temperatures, enamel will behave as a viscoelastic material in which the stress will decay or relax in a manner approximated by the Maxwell model:

(6)

where E and η are the Young’s modulus and viscosity, t is the time, σ₁ and σ₀ are the stresses at time t and at time zero (the initial stress). This demonstrates that the stress decays exponentially with a characteristic time constant τ = η/E. τ may be considered as a relaxation time. As the temperature of the enamel rises as the plasma scans the surface and approaches the glass transition temperature, T_g, the viscosity decreases rapidly and above T_g, it falls to a low value. As a result, the relaxation time τ becomes very short in the vicinity of T_g and the rate of stress relaxation becomes high. In this investigation, the data in Fig. 4 show that complete relaxation of the quench stress took place. In physical terms, the stress generated by the rapid cooling was accommodated by the plasma scanning action enabling the enamel to flow and flow for enough time to fully eliminate the stress. The extent of stress relaxation will depend on the temperature-time profile of the enamel during deposition and, in practice, on the operating parameters and the particular component under consideration.

Effect of alumina on thermal stress: The data in Fig. 4 show that the presence of alumina in the enamel has a major effect in the deflection of the samples during cooling. For the pure enamel coating (A0), the strip bends towards the coating (+ ), which indicates that the coating is under tensile stress. This is because the coefficient of thermal expansion of the coating is greater thanthat of the substrate. With the increasing alumina content, the deflection reduces. For alumina contents greater than 48-vol % (A60), the strip bends backwards (- ), indicating that the coating is now under compression.

Fig. 4 Deflection of enamel during cooling as a function of alumina contents.

Fig 4 also shows that the minimum temperature for which the strip remains at the zero position during cooling varies with the composition. With increasing alumina content in the coating, the zero-position temperature increases. It is a significant finding as it implies that stress relaxation becomes more difficult with high alumina loadings in the enamel. It is known that a solid phase effectively thickens a liquid and produces an increase in its viscosity [4]. This effect has not been recorded in enamel but it is expected that a solid alumina phase will increases the viscosity of the enamel melt. This will restrict the ability of the coating to flow and accommodate the thermal stresses. The increase in relaxation temperature with alumina loading can therefore be understood on the basis of the effect of a solid phase on enamel viscosity.
The residual stress in the coating may be estimated with Eq. 1 or Eq. 2. The theoretical Young's modulus of the composites increases with increasing the volume fraction of alumina. According to Eq.1, the reduction in the thermal expansion mismatch is compensated by the increase in Young's modulus. The deflection of the beam should not change substantially, therefore, due to the addition of the alumina phase. Nevertheless, Fig. 4 shows that increasing the alumina content does result in major differences in behaviour and this is related to stress relaxation effects. The ceramic splats may be able to slide at their boundaries within the enamel matrix. Since the degree of the relaxation is unknown, Eq. 1 cannot be used to predict the stress in the coating. However, Eq. 2 is based on the measured deflection of the sample and the effect of stress relaxation is included in the experimental results. In this study, therefore, Eq. 2 is used to calculate the stress in the coatings during cooling.

The calculated thermal stress of the pure enamel coating is given in Fig. 5, which shows that the stress is negligible at temperatures above 300 °C. During cooling, it rapidly rises to near 30 MPa in tension. This is expected, since the enamel has an expansion coefficient higher than that of the steel substrate and its contraction is restricted by the substrate. The presence of substantial tensile stresses in the enamel is expected to result in a tendency for crack formation in the plasma sprayed coatings. This prediction was confirmed by experiment. The maximum thickness of coating that can be produced with this material without cracking (the critical thickness) was found to be only 120 μm.

Fig. 5 Thermal stress in pure enamel (A0 200 μm) during cooling.

Figs. 6. and 7 give the corresponding thermal stresses of the composite coatings calculated usingEq. 2. The addition of alumina to the enamel has clearly been successful in reducing the residual stress. With increasing alumina content, the deflection decreases and changes direction from towards the coating to towards the substrate indicating that the stress changes from tensile to compressive. For the composites containing 40 % and 60 % alumina, the residual stresses are less than 7 MPa. This is because the difference of the coefficients of thermal expansion of coating and substrate are reduced from 2.8×10^-6 K^-1 to 0.6 and -0.9×10-6 K^-1 respectively. These results are consistent with the theoretical model. Thick coatings can be produced with these composite powders without cracking.

Fig. 6 Thermal stress in composite enamel (A40, 40 % alumina, 180 μm) during cooling.

Fig. 7 Residual stress in composite enamel (A60, 60 % alumina, 240 μm) during cooling.

Conclusions
1 The quench stress in alumina-enamel can be completely relaxed when the temperature remains in the vicinity of the glass-transition temperature for sufficient time during deposition. Under these conditions, the only significant source of thermal stress derives from the thermal expansion mismatch between the enamel and the substrate.
2 The addition of a ceramic second phase to form an enamel-ceramic composite is an effective method of decreasing the mismatch in thermal expansion coefficient and reducing the thermal stress in the coating.
3 A theoretical model has been developed for the prediction of thermal stresses generated in enamel during plasma-spray deposition.

References
1. D.T.Gawne, Y.Bao and T. Zhang: Smalto Porcellanato, 1 (2002), pp. 22-32.
2. S. Kuroda and T.W. Clyne: Thin Solid Films, 200 (1991), 49-66.
3. S. Timoshenko: J. Optical Society of America, 11 (1925), pp. 233-255.
4. J.S. Chong, E.B. Christiansen and A.D.Baer: J. Appl. Polym. Sci., 15 (1971), pp. 2007-21.