Investigation of the effect of inert inclusions on densification during solid-state sintering of metal matrix composites

1 Introduction

Nowadays, technology is evolving faster than in the past decades, thanks to the possibility of obtaining new materials that offer special characteristics or properties that allow designers to improve and create new devices. Composites are actually a great alternative to improve or to combine some properties between two or more monolithic materials. To fabricate those kinds of materials, powder metallurgy is an appropriate technique. It consists of consolidating a powder mixture by successive pressing and sintering, or both at the same time. Sintering can be achieved either in solid or liquid state. In those processes, densification is the most important parameter to control in order to obtain the desired properties. The densification of a metal-ceramic powder mixture during hot pressing has been extensively studied [1–3]. Infiltration of a melting metal is another technique that can be used to fabricate composites [4]. Mechanical alloying has also been used to produce composites [5, 6], and some other new techniques like the field-assisted sintering technology process [7]. However, most composites are fabricated in the simplest way: sintering a mixture of powders poured into a mold or cold pressing them to get a green body [7–13]. We are interested here in the free sintering of powder mixtures, i.e. without pressure, which is the most used process in powder metallurgy, though it rarely leads to fully densified parts. As the final properties of the composites are strongly influenced by both densification and volume fraction of the reinforcing particles, it is important to obtain the better combination of both.

Experiments on the densification of powder mixtures during free sintering have been performed by different authors [8, 11–16]. In most cases, the materials were composed of ceramic matrix particles mixed with much larger ceramic inclusions. These studies demonstrate that even a very small amount of inclusions (typically 3% vol.) can significantly affect densification. This effect has been attributed to different phenomena such as inclusion-induced stress [17, 18], initial packing heterogeneities [19], and network formation [14–20]. Analytical models treating the densification of a matrix containing inert inclusion have been proposed [18, 20–23]. For example, from analytical calculations, Lange [20] concluded that densification was inhibited when the inclusions formed a percolative network. Percolation occurred when the fraction of spherical inclusions was around 0.16 for the total volume [24]. Postrach and Pötschke [11] reported heterogeneous sintering with grain growth of matrix particles in zones located between inclusions, and “de-sintering” in other zones due to internal tensile stress. Huang and Pan [25] developed a model that takes into account neck breaking based on the strain state, and they also found that the remaining porosity between inert particles has a significant effect. Olmos et al. [26] performed numerical simulations based on the discrete elements method (DEM), concluding that densification rate is practically stopped when 30% vol. of inclusions is used. Recently, Yan et al. [27] studied the effect of the size and homogeneity of rigid inclusions by DEM, and found that smaller particles and nonhomogeneous distribution significantly reduced densification. Numerical models based on the finite element method have also been used to study densification from the viscous point of view [28, 29]. Those models consider matrix particles as a continuum, which is relevant when the inclusions are much larger than the matrix particle. Hong and Dharani [29] concluded from their simulations that densification rates were reduced by the agglomeration of inclusions.

This paper is focused on the study of free sintering and the effect generated by the addition of reinforcing particles that are still inert with respect to the matrix during the whole sintering process. As the consolidation should be mainly due to the sintering of metal particles, reinforcing particles will be referred to as inclusions. We were interested in evaluating the effect of introduced irregularly shaped particles into a matrix of spherical particles with different particle size distribution. For such a study, spherical copper powders were used as a matrix of the composite, and tungsten carbide (WC) and tungsten (W) particles were used as inclusions. We also evaluated the effect of the volume fraction of the inclusions from 5% up to 30%. To analyze the densification behavior of such mixtures, we performed dilatometry experiments and we followed the sintering progress by measuring the densification rates of mixtures with various fractions of inclusions from the data obtained in the dilatometry tests. We also observed the microstructure of composites after sintering by scanning electronic microscopy (SEM). Additionally, tests of microhardness were carried out and the results discussed as a function of densification.

2 Materials and methods

Spherical copper (Cu) powders (ECKA Granules, Fürth, Germany) with a particle size distribution of <40 μm were used as a matrix of the composite, and two different kinds of particles as inert inclusion were used. The first were irregular WC powders with a particle size distribution between 40 and 80 μm. The second ones were polygonal W powders with a different particle size distribution inferior to 40 μm (Figure 1). W and WC powders were produced by Eurotungstene metal powders (Grenoble, France). The particle size distribution was calculated by measuring the Ferret’s diameter of around 1000 particles from several SEM pictures with the aid of Image J software which is free software (website: https://imagej.nih.gov/ij/index.html) (Figure 1D). We notice that the particle size distribution of Cu and W particles are very close, and the d₅₀ is 27 μm for both of them with the difference of Cu particles being spherical. On the other hand, WC particles have a d₅₀ of 72 μm. In order to get statistical information from the wide distribution of particles, the span value was calculated. It was defined as (d₉₀-d₁₀)/d₅₀. For span values close to 0, it is expected that the size distribution will be concentrated in a narrow band. Here, we obtained values of 0.85, 0.96, and 0.33 for the Cu, W, and WC, respectively. We notice that the W particles have a wider distribution.

Figure 1:

SEM micrographs of initial powders, (A) copper matrix, (B) WC, (C) W, and (D) particle size distribution of particles.

We fabricated composites by mixing copper powders with 5%, 10%, 15%, 20%, and 30% vol. fractions of the inert inclusions. Volume fractions were calculated by assuming that the copper particles had a bulk density of 8.9 g cm^-3, as pure copper and inclusion powders of WC and W have bulk densities of 14.95 and 19.25 g cm^-3, respectively. The mixtures were dry mixed by using a turbula for 30 min. Then, the mixed powders were poured into an alumina crucible and tapped before they were introduced into a cylindrical furnace, which was heated at 450°C under reducing atmosphere (N₂-5% H₂) for 30 min. This was achieved with the aim of, first, eliminating the oxygen content inside the copper powders that causes swelling of the samples, as it was reported before by Upadhyaya and German [30], and, second, avoiding the friction that could occur within the crucible walls. Next, samples were taken from the crucible and directly sintered in a vertical dilatometer Linseis L75 (Linseis, Robbinsville, NJ, USA), at 1050°C under reducing atmosphere with a heating rate of 30°C min^-1. The weight density of the cylindrical samples was calculated before and after sintering by simply measuring and weighing them. Relative density was calculated by dividing the weight density by the theoretical density, the latter being deduced from the simple law of mixture: ρ_t=ρ₁f₁+ρ₂f₂, where ρ_i and f_i are the theoretical density and the volume fraction of component i. After that, the microstructure of the sintered samples was analyzed by SEM. For that, samples were cut and polished to observe their inner microstructure with a field emission scanning electron microscope (TESCAN, Kohoutovice, Czech Republic). Finally, microhardness tests were performed on the polished surfaces by using a microhardness tester (Mitutoyo MVK-HVL, Mitutoyo America Corporation, Chicago, IL, USA) by using a load of 50 g.

3 Results and discussion

The relative density of pure copper and mixtures during the whole sintering cycle is plotted in Figure 2 as a function of the temperature. A small reduction in the initial relative density of the composites with respect to the sample without inclusions is noticed. That reduction is mainly attributed to the differences in size and shape between matrix and inclusion particles, which lead to a less dense packing. Nevertheless, when 15% vol. is used, the initial green density is very close to that of the sample without inclusions, for both kinds of inclusions. We assume that this finding could be a signal of the better distribution of the inclusions between the spherical copper particles because the same procedure of filling was performed for all samples. Sintering is inhibited by the addition of the inclusions; however, inclusions of WC have a major effect on 5% vol. of inclusions contrary to the inclusions of W, which have the same particle size distribution as the Cu powders. It is observed that the relative density at the beginning of the isothermal sintering is different for all samples; this could be due to both lower initial densities and/or addition of the inert inclusions. Nevertheless, the addition of the inclusions has no effect on the temperature at which sintering is activated, around 850°C, which has been found in ceramics systems [31], except for the sample with 5% vol. of inclusions for both kinds of inclusions. This behavior is difficult to explain, and it has an effect on the final relative density; we assume that it might be generated by the rearrangement of the particles during the heating stage because the volume fraction of inclusions is low; however, it is only an assumption.

Figure 2:

Relative density as a function of the temperature during the whole sintering cycle: (A) Cu-WC and (B) Cu-W.

In order to compare densification during the sintering plateau, we depicted densification (D-D₀/D₀) during the isothermal temperature as a function of time (Figure 3). To obtain a better comparison, the initial relative density (D₀) is the relative density reached by each sample at the beginning of the sintering plateau. Densification is more affected when WC particles are used, particularly for <10% vol. of inclusions. When 5% vol. of inclusions is used, the densification is 34% lower than that obtained for the matrix without inclusions. Then, the densification of composites diminished as the volume fraction of inclusions increased. This is consistent with the findings from other authors [25, 26]. The worst case is when 30% vol. of inclusions is used, reducing densification by 67% (Figure 3A). On the other hand, when W particles are used for lower volume fractions, there seems to be a less detrimental effect as densification is only reduced by 19% for 5% vol. of inclusions (Figure 3B). That value is half of that found when the same quantity of particles of WC is used. Nevertheless, for volume fraction of inclusions ≥20%, the reduction in densification is nearly the same in both cases. For both kinds of inclusions, it can be observed that reduction of the densification starts from the very beginning of the isothermal plateau and then densification is reduced as much as the volume fraction of inclusions increases. This behavior might be due to the formation of a rigid skeleton by the inclusions when they are in contact with one another, as pointed out by Lange [20]. Nevertheless, when lower volume fractions of inclusions are used, densification might be slower because of dynamic processes occurring during sintering, like rearrangement and new contacts creation, are inhibited. In particular, new contacts between Cu particles (sintering particles) are reduced by the substitution of inclusions (non sintering particles), as was observed by Olmos et al. [32, 33]. The maximal densification will depend on the quantity and size of particles, observing that for <15% vol. of inclusions, particles larger than the matrix are more detrimental and for higher volume fractions smaller particles have a stronger influence. This observation is contrary to that established by other authors [34].

Figure 3:

Isothermal densification as a function of the time during the sintering plateau: (A) Cu-WC and (B) Cu-W.

The densification rate during the sintering plateau was calculated from the dilatometry data and is plotted as a function of the densification during the isothermal plateau (Figure 4). In all cases, a maximum value at the beginning of the plateau is observed, and then the densification rate decreases as the densification is achieved. When WC particles are used, the densification rate is reduced as the volume fraction of inclusions increases, as expected. Curves related to the composite samples are parallel to those of the matrix without inclusions (Figure 4A), and the densification is reduced 10 times when the greater quantity of inclusions is used. Concerning the W inclusions, the same behavior is observed (Figure 4B). Nevertheless, when the quantity of inclusions is 20% or 30% vol., the densification rate shows a sharp drop at early densification. Moreover, for the sample with 30% vol. of inclusions, the densification rate is 13 times lower than the one without inclusions. That means it is 1.3 times slower than that obtained for the same quantity of WC particles. This observation suggests that there is a densification value where the rigid skeleton is formed and densification is no longer allowed. The behavior found for particles two times larger than the matrix has been observed by Olmos et al. [26] when spherical alumina particles were used as inclusions. On the other hand, when W particles are used, we notice that the same behavior of sintering is found; however, a slower densification rate is obtained for composites with >20% of inclusions, which could be linked to the size and shape of inclusions.

Figure 4:

Densification rate as a function of densification during the sintering plateau: (A) Cu-WC and (B) Cu-W.

With the aim to compare the effect caused by the addition of the two different inclusions, the final relative density is depicted as a function of the percentage volume of inclusions (Figure 5). It was found that lower values of the relative density are attained when WC particles are used as reinforcing particles except when 30% vol. of particles are used. In that case, the effect of W particles is higher than that introduced by the WC particles. We estimated the difference of the final relative density reached by Cu-W and Cu-WC composites for every volume fraction of inclusions. We found that the difference is reduced as the volume fraction is increased until it was reversed for 30% of inclusions. This could be associated to the number of inclusions inside the matrix as more particles will be needed for the same volume fraction when particles are smaller, which could cause a reduction on the coordination number of Cu particles.

Figure 5:

Evolution of the relative density as a function of the volume fraction of inclusions for the two kinds of inert inclusions used in this experiment.

Homogeneous distribution of the inclusions inside the matrix is observed for both kinds of inclusions (Figures 6 and 7). It is observed for Cu-WC composites with 5% vol. that most of the particles are isolated and surrounded by the matrix. As the volume fraction of inclusions increases, small agglomerates of inert particles appear and larger pores are observed (Figure 8). Porosity is not easily observed from the polished surface, as the copper material is smeared to fill in the pores; nevertheless, larger pores can be detected close to a group of inclusions, which is logical as copper matter is not able to fill those pores because they are located between particle inclusions.

Figure 6:

SEM micrographs of the internal microstructure of the composite Cu-WC with different volume fractions of reinforcing particles: (A) 5%, (B) 10%, (C) 15%, (D) 20%, and (E) 30%.

Figure 7:

SEM micrographs of the internal microstructure of the composite Cu-W with different volume fractions of reinforcing particles: (A) 5%, (B) 10%, (C) 15%, (D) 20%, and (E) 30%.

Figure 8:

SEM micrographs of the internal microstructure of the composite Cu-WC with different volume fractions of reinforcing particles: (A) 5%, (B) 10%, (C) 15%, (D) 20%, and (E) 30%.

In the case of composites containing W particles, which are the same size as the matrix powders, the formation of small agglomerates from the 10% vol. is observed. In spite of that, dispersion of W particles inside the matrix is good. Larger holes observed on the microstructures of composites of Cu-W containing >15% vol. of inclusions are due to the fact that some particles were torn out from the matrix during the polishing step, showing that they are not well trapped between copper particles, contrary to what is found when WC particles are used. We notice that torn particles from the matrix during the polishing stage are mainly located in agglomerates, although a few isolated particles have also been torn out (Figure 9B–E); otherwise, most of the isolated particles can be well surrounded by the matrix (Figure 9A). The inert behavior of the inclusions is verified in Figures 8 and 9, as no reaction or bonding is observed between the matrix and the reinforcing particles, which implies that densification is driven only by the copper powders during solid-state sintering. It also should be pointed out that the wettability between matrix and inclusions is practically negligible.

Figure 9:

SEM micrographs of the internal microstructure of the composite Cu-W with different volume fractions of reinforcing particles: (A) 5%, (B) 10%, (C) 15%, (D) 20%, and (E) 30%.

In order to evaluate the sintering progress between matrix particles, we performed microhardness Vickers tests measuring only on the matrix surface by using the lowest load, as copper is a soft material (Figure 10). It was found that the value of hardness increases to reach a maximum value for 15% vol. of inclusions for both kinds of inclusions used. Then, the value decreases for volume fractions of inclusions >20%. However, the final value is 1.3 times higher than that obtained from the sample without inclusions. The composites using WC as inclusions show a somewhat higher value of hardness than those using W, except for 10% vol. of inclusions. The maximum value of hardness is almost two times higher than that obtained by the matrix of copper without inclusions. The higher values of hardness found in samples with inclusions indicate that the net of necks becomes stronger during sintering in spite of reached lower relative densities, which means an increase in the porosity of the sample. The strengthening mechanisms of the matrix could be due mainly to both the effect of the transmission of load from the matrix to the inclusion particles and the mismatch in the coefficient thermal expansion, as it was pointed out by Casati and Vedani [35]. The first one is attributed to the volume fraction of inclusions that have a higher stiffness than the matrix particles; thus, the charge is transmitted from the matrix to the W and WC particles when the matrix is completely interconnected. The second one is mainly because during cooling, the CTE is completely different between the matrix and the hard inclusions; therefore, internal stresses are generated in the matrix causing hardening. As the CTE is three times larger for the matrix than that for hard inclusions, this mechanism should play a major role that can be noticed in Figure 2, where the slope during the cooling is different between the sample without inclusions and the composites. The small difference between the values of microhardness found in both kinds of composites could be attributed to the higher mechanical properties of the WC. The reduction of the microhardness values when a volume fraction of 15% or higher is used could be because the interconnectivity of the matrix is reduced as the volume fraction of inclusions is increased, which is detected by a reduction on the relative density. It was also observed that the shape of particles plays an important role in the internal retention of the matrix particles; thus, when particles have irregular and high aspect ratios, they can be trapped by the matrix during densification, as no chemical bond is developed during the thermal cycle, as it has been shown in the graphical abstract. To the contrary, more rounded particles can easily be torn out of the matrix. It can be observed in Figures 7 and 9 that particles of W were torn out when they were both isolated and forming agglomerates. On the contrary, the WC particles were only torn out when they form agglomerates because irregular particles can be more easily wrapped between the matrix particles during sintering. On the other hand, the remaining porosity after sintering is very similar in both kinds of composites as we can notice in Figures 8 and 9. Thus, the effect generated on the hardness of the matrix could be similar in both cases. However, the values are very close and the most important conclusion from those results should be that 15% vol. of inclusions is the maximum value to improve the mechanical properties.

Figure 10:

Hardness as a function of the volume of reinforcing particles for both kinds of reinforcing particles used.

4 Conclusions

Two different kinds of inert inclusions (WC and W) were introduced into a copper matrix, and their effect on the sintering behavior and mechanical properties were evaluated. From dilatometry tests, it was assessed that densification was reduced as much as the volume fraction of inclusions increased. It was found that 15% vol. of inclusions was the maximum quantity that could be used to improve the mechanical properties even though the densification during the sintering plateau was decreased by 50%. The physical characteristics of inclusions had an effect on the densification, with smaller and regularly shaped particles being more favorable when <20% vol. of inclusions were used. The inert relationship between copper and WC and W particles was observed in SEM images, and it was found that irregular and larger WC particles were the best option. Comparison with other literature results should be made with care due to the high porosity of materials in the present study.