Development of the Structure of Cemented Carbides during Their Processing by SLM and HIP

Abstract

The study focuses on microstructural evolution in a WC-Co powder mixture during Selective Laser Melting (SLM) and hot isostatic pressing (HIP) processing. This powder mixture contained a 13 ± 0.6% weight fraction of Co binder and WC particles of mean size of 3.0 ± 1.9 μ$\mathrm{m}$. SLM of the mixture produced samples of various densities, depending on the volumetric energy density (VED) applied. High VED levels led to densities of up to 88%. The aspects affected by changes in VED included the pore density as well as the resulting types of phases and the size of WC phase particles. At high VED, the material began to develop cracks due to embrittlement. This had multiple causes: coarsening of $\alpha$-phase (WC), evaporation of $\beta$-phase (Co binder), and precipitation of $\eta$-phase. At low VED levels, pores formed, typically of nonsymmetric shapes, with sizes larger than 500 μ$\mathrm{m}$. Subsequent HIP processing led to an increased density, up to 96% of solid material. Contributions to this increased density were provided by structure transformations, namely, coarsening of $\alpha$-phase by up to 1300% when compared to the powder grain size, and formation of $\eta$-phase. The results provided a basis for steering further research to explore to a greater depth the SLM and HIP processing of selected WC-Co powder mixtures with as yet unused ranges of process parameters.

1. Introduction

The structure of cemented carbides (CC) substantially depends on their manufacturing route. It may involve various techniques which include production of powder mixtures, their compaction, and sintering. The typical resulting microstructure of a cemented carbide consists of hard WC carbide particles ($\alpha$-phase), tough metallic binder of Co ($\beta$-phase) and certain other carbides ($\gamma$-phase) with tetragonal crystal structure. These constituents and their size and amount in the structure govern the mechanical and physical properties of CC and its intended use [1,2,3,4,5].

Typical applications of CCs include cutting tools and parts subjected to wear, such as nozzles, throwing blades, and molds for large-series production [6,7]. All these parts share one feature: very simple shapes of their outside and interior surfaces. To some extent, these surfaces can be adjusted for a specific application by subsequent processing, which may include grinding, tumbling and laser ablation [8,9,10]. However, these processes may not necessarily improve the product properties. An option which can deliver improvement is additive manufacturing (AM) [11,12].

Several AM techniques are available for making cemented carbide products [13,14]. One of them is the SLM process. In SLM, a machine part, for instance, is made by deposition of consecutive powder layers. After each layer is deposited, some regions in this layer on the powder bed are melted by laser. In this manner, the part is gradually built [15,16].

Creating CC parts by this process is rather demanding. Its feasibility depends on the properties of the powder mixture and the process conditions. Powder mixtures for SLM are required to consist of spherical particles with a certain distribution of sizes 10–50 $\mathsf{\mu }$$\mathrm{m}$ and surface properties which facilitate uniform energy distribution across the powder bed region to be melted by laser [17,18]. With steel and non-ferrous powders, the desired state of the particle surface is attained by various techniques, typically by atomization in inert gas or by plasma application [19,20].

Cemented carbide powders are produced using a different method which involves mixing the constituents in attritor or ball mills followed by drying the mixture in inert gas or vacuum. The particles may eventually be passed to granulation [1,2].

Conventional techniques may yield a relatively wide range of WC-Co particle geometries with various bonds between the constituents. The powders may therefore differ in their flowability and ability to produce a uniform layer on the build plate in an AM machine. For the SLM process and for achieving the right energy distribution across the powder bed, it is important that each powder layer is flat and continuous.

These carbide mixtures also show another difference from conventional powders. Their multiphase structure comprises no fewer than two constituents. One of them is the $\alpha$-phase whose melting temperature is 2720 °C and whose evaporation temperature is 6000 °C. The other is $\beta$-phase with melting and evaporation temperatures of 1495 °C and 2870 °C, respectively, provided that $\beta$-phase is cobalt.

Cobalt is the most frequent choice because it has excellent wettability with WC grains of $\alpha$-phase. Thanks to this property, it covers the $\alpha$-phase grains entirely during sintering and forms a skeleton which lends toughness to CCs [1]. As opposed to conventional sintering, SLM produces connection between particles over a very short time, typically under 0.1 s [21]. The growth of these initial joints between powder particles causes gradual decrease of porosity in the CC structure.

The final degree of consolidation of the powder mixture primarily depends on how stable the viscous flow of the melted binder is while it fills the voids between the tungsten carbide. Scherer found, in 1977, that this degree is in fact inversely proportional to the viscous flow and can be described using the Equation (1) [15].

$$x=\frac{3(1-\rho )\pi \gamma a^2}{24{\eta }_m{\rho }^3{r}^3}\left[r-(1-\xi )x+\left(x-\left(xi+\frac{1}{3}\right)r\right)\right]\frac{9(x^2-r^2)}{18rx-12r^2}$$

(1)

where $\rho$ [kg/m${}^3$] is the apparent density, $\gamma$ [N/m] is the particle surface tension, $\xi$ is a parameter with its magnitude between 0 and 1, and ${\eta }_m$ [Pa s] is the viscosity of the solid liquid mixture [15].

Here, the viscosity of the mixture of the liquid binder and the solid phase (WC grains) can be expressed by the Equation (2) [15].

$${\eta }_m={\eta }_l{\left(1-\frac{V_s}{V_m}\right)}^{-2}$$

(2)

where ${\eta }_l$ [Pa s] is the viscosity of the liquid phase, $V_s$ [m${}^3$] is the fraction of the solid phase, and $V_m$ [m${}^3$] is the critical volume fraction above which the mixture exhibits essentially viscosity [15].

Another factor at play when it comes to the final density of the CC structure is the mutual solubility of the phases and their mixing [15,22]. Mixing of the structural phases is driven by temperature gradients in the powder bed. They generate thermocapillary flow known as Marangoni convection. It reduces the temperature gradients in the powder bed and can be expressed in terms of the Equation (3) [22].

$$M_a=\frac{\left(\frac{d_y}{d_x}\right)\left(\frac{W}{\eta }\right)}{\left(\frac{\kappa }{W}\right)}=\frac{\left(\frac{d_y}{d_T}\right)\left(\frac{d_T}{d_x}\right)\left(\frac{W}{\eta }\right)}{\left(\frac{\kappa }{W}\right)}$$

(3)

where W [m] is the linear size of the pool, $\frac{d_T}{d_x}$ [K/m] is the temperature gradient, $\kappa$ [m${}^2$/s] is thermal diffusivity, and $\eta$ [Pa s] is the viscosity [22]

The magnitude of temperature gradients in the powder bed substantially depends on AM parameters. The layer thickness, laser power and scan speed and other parameters dictate the amount of energy available for melting the material in the powder bed, and therefore the structure of the resulting CC. For instance, the volumetric energy density (VED) can be described using the Equation (4) [21,23,24].

$$E=\frac{P}{v·h·t}$$

(4)

where P [W] is the power of the laser, v [m/s] is the motion speed of the laser spot, h [m] is distance among the motion vectors, and t [m] is the thickness of the powder layer.

According to available studies, the final structure of SLM-produced cemented carbide comprises a certain amount of pores, coarse grains of $\alpha$-phase (WC) and $\eta$-phase which forms in the process, basically regardless of the AM conditions used for the powder mixture [21,25,26,27,28,29,30,31,32]. The formation of $\eta$-phase is related to the cooling rate from the melting temperature. In the powder bed, the material only becomes melted for a very short time. Therefore, the matter (atoms, molecules, and ions) can only travel across short distances, which leads to regions depleted of some elements, such as carbon. As a result, the distribution of elements becomes nonuniform.

This, combined with the effects of cooling rate, then leads to formation of complex W-Co-C carbides, referred to as $\eta$-phase, in carbon-depleted regions. When it comes to altering the structure produced by SLM, one of the available options is hot isostatic pressing (HIP). It involves application of hydrostatic pressure up to 400 MPa and high temperatures, up to 2000 °C, and it can profoundly change the structure of the workpiece—not just in terms of porosity, but also phase composition [33,34,35].

The focus of this study was to use the SLM technology with a chosen WC-Co mixture and produce a cemented carbide structure which is suitable for subsequent HIP processing. For this purpose, the powder mixture was converted into several samples using various SLM conditions. The goal was to create samples that exhibit various amounts and types of porosity and different proportions of phases. The samples were then reprocessed using HIP.

After each of the processes, the microstructure of each sample was studied and evaluated. Based on the findings, the most appropriate SLM parameters for subsequent HIP were identified for the powder mixture. In addition, its microstructural evolution induced by SLM and HIP was mapped.

2. Materials and Methods

2.1. Experimental Material

The experimental material was a WC-Co powder mixture under the brand name AM WC701(AW701) marketed by the company Global Tungsten & Powders Corp., Bruntál, Czech Republic. Prior to AM, the powder was examined while considering the requirements of the process. The size distribution and shapes of the powder particles were determined. Their flowability, bulk density, and ability to spread uniformly on the build plate were determined. These properties were then compared to those found in the commercial powder of PH1 precipitation-hardenable steel which has been designed for this AM process. This reference powder was designed by EOS GmbH and supplied by 3R Systems s.r.o., Jaroměř, Czech Republic. Figure 1 shows comparison of analyzed powder particles. Particles of the PH1 stainless steel powder had smooth surface with small amount of defects. Their surface contained mainly defects called satellites. In the case of AW701 powder, more types of surface defects were observed. Apart from satellites, craters and cavities were identified in a large amount.

The AW701 powder particles also vary in their shape. Generally, the shape is spherical but more irregular than the PH1 powder particles, see Figure 2. In addition, a difference in the particle size profile of the two compared powders was noted, see Figure 3.

Recorded differences in the surface quality of the powder particles and the differences in their shape and size significantly affect the ability of the powder to form a uniform layer. This fact was verified by the spread test during which the layer is applied by recoater blade on the build plate of the used AM machine (Figure 4).

The relatively lower fraction of fine powder particles indicated in the plot in Figure 3 had a favorable impact on the flowability and other technological properties of the powder, as evidenced in

Table 1. Technological properties of AW701 powder mixture and PH1 powder.

Powder Type	AW701	PH1
Flowability [s/0.2 kg]	2.7 ± 0.2	3.7 ± 0.1
Bulk density [kg/m${}^3$]	5970 ± 900	4200 ± 600
Tap density [kg/m${}^3$]	7290 ± 110	4600 ± 700
Mean spread powder width [mm]	4.6 ± 0.5	3.5 ± 0.5

. The reason is that the capacity for friction between particles was reduced.

The phase composition of the virgin powder mixture was determined by means of X-ray diffraction in order to obtain reference values for evaluating structural changes upon SLM. Its chemical composition was measured by EDS. The virgin grain size of $\alpha$-phase was found by metallographic analysis. The results are listed in Figure 5.

Powder Mixtures

The size distribution and shapes of particles of the powders were determined using metallographic techniques. Powder samples were mounted and metallographic sections were prepared and examined using PHILIPS XL30ESEM scanning electron microscope (FEI Company, Hillsboro, OR, USA) and imaging with SE and BSE. The magnification was 100–1500× and the micrographs were taken using AxioVision (Carl Zeiss s.r.o., Prague, Czech Republic) and NIS ELEMENTS (Laboratory Imaging s.r.o., Prague Czech Republic) software tools. Chemical composition was measured by EDS. The technological properties of the powders (flowability, bulk density) were determined with reference to EN ISO 4490 [38]. PANalytical X´Pert PRO automatic powder diffractometer (Malvern Panalytical, Eindhoven, Netherlands) with Bragg-Brentano geometry was employed to measure the phase composition of the powders. The X-ray source had copper anode CuK${}_{\alpha }$ (${\lambda }_{k_{\alpha }}=0.15405980$ nm).

2.2. Selective Laser Melting Technology

Selective Laser Melting is one of powder bed manufacturing techniques. Thin powder layers are deposited across the entire build area and selectively melted in locations defined in the input CAD data. After each cycle, the build platform descends by a thickness of a single layer and another cycle is performed (Figure 6). Today, it can be used with various metallic materials based on iron, nickel, aluminum, and titanium.

One of the main fields of application of this technology involves process trials on new materials. In most cases, the goal is to find parameters which lead to the least porosity and superior appearance of the outer surfaces [39,40,41]. Typically, four parameters are reported in professional publications as those which dictate the quality of the product. They are the laser beam power P [W], scan speed v [mm/s], layer thickness t [mm], and the spacing between laser paths h [mm]. These parameters are incorporated in the volumetric energy density constant E [J/mm${}^3$] (Equation (4)) [24]. Once the appropriate set of process parameters are found, the volumetric energy can be used for altering the process, for instance, to improve productivity by changing the scan speed.

2.3. The Methodology of Testing Powder Mixture AW701

The additive manufacturing sequence was carried out in SLM 280HL equipment (SLM Solutions Group AG, Lübeck, Germany) which features a single Yb:YAG (IPG Photonics) fiber laser of 400 W nominal power. Its maximum build space size is 280 × 280 × 350 mm. For these experiments, the reduced space size was 100 × 100 × 120 mm.

After considering the particle size distribution of the AW701 mixture, a layer thickness of $t=$ 40 um was chosen for the AM process. The laser path spacing was fixed at $h=0.12$ mm. It mainly reflects the size of the laser sport which produces the molten pool. Variable parameters for the AW701 powder mixture were then the laser power P and scan speed v.

Table 2. Process parameters for AW701 powder mixture.

Laser Power	Scan Speed	Volumetric Energy Density	Note
P [W]	v [mm/s]	E [J/mm${}^3$]
40	152	54
40	140	60
40	128	65
40	111	75
40	100	83
40	83	100
40	55.5	150
40	35	238
80	175	95	Stopped
80	166	100	Stopped
80	145	114	Stopped
80	128	130	Stopped
80	111	150	Stopped
80	100	166
80	95	175
80	90	185
80	85	196
80	80	208
80	75	222
80	70	238
80	49	340
80	32	520
100	209	99
100	181	115
100	139	149	Stopped
100	87.5	238
100	61.2	340
100	40	520

provides a summary of parameters for all the builds.

The scanning strategy involved hatching with 90° rotation between powder layers. In the building strategy, a single contour scan was followed by interior hatching. The protective atmosphere was nitrogen gas of 97–98% purity from an internal nitrogen generator. Build plates were made of S235 steel and their size was 100 × 100 mm. The builds were 10 mm cubes placed directly on the build plate with no additional support structure.

One of the crucial deficiencies was inadequate adhesion of the first layers to the build plate (Figure 7). As a result, several builds had to be scrapped in the process because they rose above the current level and threatened to cause collision with the recoating device. It was remarkable that at 80 W laser power, the builds created with the lowest volumetric energy density had to be scrapped.

The subsequent HIP process on selected builds was carried out with standard parameters of sintering process at 1390 °C in an argon atmosphere at 40 bar and with a holding time at the sintering temperature 60 min in cooperation with Dormer Pramet s.r.o, Šumperk, Czech Republic. The builds selected for the HIP process were those with low (E = 60 J/mm${}^3$) and high (E = 520 J/mm${}^3$) energy densities.

Samples after SLM Process and HIP Process

Metallographic specimens were examined using PHILIPS XL30ESEM (FEI Company, Hillsboro, OR, USA) and Tescan MIRA3 (TESCAN ORSAY HOLDING, a.s., Brno, Czech Republic) scanning electron microscopes and a CarlZeiss Observer Z1m optical microscope (Carl Zeiss s.r.o., Prague, Czech Republic). The metallographic sections were oriented in the transverse direction to the building direction. The features of interest were the pore structure, microstructural phases and changes in the chemical composition (determined by EDS). The microstructures were revealed by chemical etching with reference to ASTM B657 [42].

AxioVision (Carl Zeiss s.r.o., Prague, Czech Republic) and NIS ELEMENTS (Laboratory Imaging s.r.o., Prague Czech Republic) software tools aided in the evaluation of micrographs. Changes in the specimen shape caused by the SLM and HIP processes were measured using the Zeiss Prismo 7 Navigator machine (Carl Zeiss s.r.o., Prague, Czech Republic) with Zeiss VAST XXT measuring head. The size of the ruby probe tip was 1.5 mm and its length was 50 mm. The probe approached from the top along Z axis. Using X-ray diffraction, changes in the phase composition and structure were mapped.

The automatic powder diffractometer PANalytical X’Pert PRO (Malvern Panalytical, Eindhoven, Netherlands) had a copper source (${\lambda }_{k_{\alpha }}=0.154187$ nm) with ultra-high speed solid state PIXcel detector. The data was stored in a computer memory. For the sake of comparison, the measurement settings were identical for all the specimens. The aperture for the vertical direction was set at 20 mm, the same value as for the software-controlled horizontal aperture. The measurement range was 30 to 100° 2θ using the symmetric Bragg-Brentano geometry.

3. Results

Samples after SLM Process and HIP Process

In the builds produced by SLM, low energy levels already led to creation and gradual expansion of connections between particles, as seen in Figure 8. With increased energy densities, the volume of these joints and their density grew.

The best density achieved in these experiments was 88%. As density increased, the size and shape of pores changed, see Figure 9 and Figure 10.

The initial interconnected pores sized 500 $\mathsf{\mu }$$\mathrm{m}$ and larger which formed at energy densities of less than approximately 100 J/mm${}^3$ became closed pores as the energy density was increased. When VED exceeded approximately 200 J/mm${}^3$, cracks developed. Cracking is associated with temperature gradients which exist in SLM builds [27,29,43].

The pores fall into several groups. In the first group, the pores form due to nonuniformities in the powder bed or low VED values resulting from low power P [W] setting or high speed of laser movement. As a result, the material in the powder bed was only melted for very short periods of time and in narrow locations, and the molten pool was unstable [44]. The result was pores with non-uniform shapes and sizes which sometimes exceeded the layer thickness. The second group comprises round pores. Most of them are smaller than the powder layer thickness. Their origin is related to the reaction between the melted powder bed and the gas phase. It may be an interaction between the inert atmosphere and the powder bed or a reaction caused by residual moisture in the powder bed. As a result, partial pressure of gas occurs at the interface with the melt.

The amount of pressure substantially depends on the gas type. The subsequent HIP process eventually eliminates the cracks as well as the closed pores which formed due to interaction between the melt and the gas phase. As a result, the relative density of the specimens made at high VED increased to 96% (Figure 10). The final density in specimens with low VED levels was 90%.

In specimens created at low VED, the density was considerably improved by gradual elimination of the connected pores. The chosen scanning strategy contributed to this elimination as it helped create a skeleton which isolated the individual pores. Another contributing factor was the presence of unsintered particles which helped gradually shut the pores during HIP.

Cemented carbides are composite materials that contain phases whose solidus, liquidus, and evaporation temperatures differ, which means that they evaporate during processing. The amount of evaporated material depends on VED, as illustrated in Figure 11.

As partial pressure of gas increases with the volume of evaporated material, round pores form at higher energies and take up a larger volume. As more binder evaporates, the cemented carbide becomes more brittle which, along with temperature gradients, contributes to cracking at higher VED.

The mechanisms which contributed to densification of AW701 specimens included coarsening and changes in the morphology of $\alpha$-phase particles, as well as phase transformations caused by the diffusion of interstitial and substitutional elements across the melted portion of the powder bed in SLM and during HIP.

Figure 12 documents the changes in $\alpha$-phase in specimens processed at high VED. The micrographs show the directional growth of particles of this phase. The growth depends on the SLM parameters. Reports [23,26] characterize the process as multi-laminated growth mechanism. In addition, certain grains of $\alpha$-phase with favourable orientations coalesce in the process. Changes in the morphology of this phase are aided by variation in the carbon potential which occurs at the WC-Co interface during interaction with the laser beam.

The growth of WC grains takes place in the direction of heat dissipation. In the subsequent HIP process, these grains continue to grow through Ostwald ripening [45,46]. As a result, HIP-processed specimens contained grains with sizes up to 40 $\mathsf{\mu }$$\mathrm{m}$. This is more than 1300% increase over the initial size.

Metallographic analysis revealed coarsening of WC grains and the presence of $\eta$-phase, both in SLM specimens and those processed with SLM and HIP. The volume fraction of this phase, as well as the final size and shape of WC grains, was controlled by the process parameters.

Figure 13 shows the different sizes of $\alpha$-phase domains, in which coherent diffraction occurs. This is due to variation in SLM process parameters for AW701 powder mixture. The figure also presents changes in micro-stress in the crystal lattice.

The graph in Figure 13 shows that the final crystallite size depends on VED. At higher values, the $\alpha$-phase tends to coarsen more severely in the course of the subsequent HIP process. In specimens prepared at high VED, the mean post-HIP crystallite size was 620 ± 260 nm. In those prepared at low VED, the size was 370 ± 50 nm. The graph also shows that HIP had a beneficial effect on the distribution and magnitude of micro-stress in the crystal lattice that existed after SLM. The micro-stress were reduced and became more uniform.

X-ray diffraction of phase composition confirmed that $\eta$-phase had formed due to SLM processing. Several types of $\eta$ phase were identified in the specimens, depending on their process parameters. Details are given in Figure 14 and

Table 3. Calculated volume fractions of phases in specimens made at low and high VED prior to and after HIP.

	SLM	HIP	SLM	HIP
Phase	P = 100 W v = 40 mm/s	P = 100 W v = 40 mm/s	P = 40 W v = 137 mm/s	P = 40 W v = 137 mm/s
WC	87.5	92.4	25.3	97
W${}_2$C	8.5	–	70	–
Co	0.8	0.3	4.3	2.4
Co${}_3$W${}_3$C	0.3	7.3	–	–
CO${}_6$W${}_6$C	–	–	0.4	0.6
CO${}_3$W${}_9$C${}_4$	2.9	–	–	–

.

In SLM specimens, the level of W${}_2$C phase, which has relatively poor stability, was higher than in virgin powder. This phase decomposed upon HIP, as did the high-temperature $\eta$-phase which had formed in specimens built at high VED levels. It was mainly due to carbon diffusion which relieved its lattice stress and enabled its transformation into more closely-packed hexagonal WC phase whose volume was larger after HIP. Free carbon then contributed to formation of the low-temperature $\eta$-phase whose volume increased after HIP as well.

4. Conclusions

In the present experimental study, the effect of volumetric energy density (VED) in SLM processing of WC-Co powder mixture on the resulting microstructure was explored. Prior to AM, the powder mixture was examined and its properties compared with those of a powder steel tailored to SLM. The findings from all the experiments can be summarized as follows.

• The size distribution and shape of WC-Co particles are equal to those of the reference steel powder. However, the fraction of microscopic particles which cause higher friction is lower. Consequently, the WC-Co powder mixture shows better flowability and wider spread on the build plate than the reference powder.
• By varying the amount of VED, one can control the phase structure and type of pores in SLM builds from WC-Co. By this means, interconnected irregular-shaped pores produced with VED under 100 J/mm${}^3$ can be changed into closed pores and cracks at higher VED above approx. 200 J/mm${}^3$. The formation of cracks was associated with the loss of Co binder—the $\beta$-phase was due to evaporation, the coarsening of $\alpha$-phase WC, the precipitation of $\eta$-phase, and stresses induced by temperature gradients. The process led to precipitation of $\eta$-phase and W${}_2$C. The type of $\eta$-phase depended on the VED value. All the microstructural changes played a role in obtaining the post-SLM relative density which was 65–88%, depending on VED.

The subsequent HIP process eliminated pores in all specimens, regardless of the VED value. The final relative density was 90% in specimens made at low VED and 96% in those produced with high VED levels. The mechanisms which contributed to densification in HIP included the increase in the volume of $\eta$-phase which was more pronounced in specimens produced at high VED, and coarsening of $\alpha$-phase, sometimes by 1300% over its initial size. The primary cause of its coarsening was the repeated decomposition of W${}_2$C phase during sintering. HIP led to less and more uniform micro-stress in the crystal lattice thanks to diffusion.

One can conclude from the above findings that appropriate settings of the additive manufacturing process can lead to specimens with interconnected pores which can be eliminated by subsequent HIP processing. One can therefore employ higher scanning speeds and lower laser power in the building process. This technology is capable of producing structures with less brittle phase ($\eta$-phase), finer $\alpha$-phase, and lower amounts of evaporated material, namely $\beta$-phase, which causes further embrittlement.