Overview of challenges and developments in joining tungsten and steel for future fusion reactors

Joining bulk tungsten and bulk steel directly is a reasonable start for joining experiments, because established processing techniques can be used, only one joining step is carried out (and maybe one PBHT), only one dissimilar materials interface is present, and no materials other than the base materials are required, which helps with nuclear considerations. Solid state bonding techniques, such as uniaxial diffusion bonding (DB) and spark plasma sintering (SPS), have been carried out at 1050 °C, 960 °C, and 800 °C for 0.5–4 h at uniaxial pressures between 15 and 60 MPa, causing approximately 8% creep in the steel part to ensure proper bonding [21–24]. To date, joining tungsten and steel directly has not been successful with any of the suggested techniques and parameters. Combinations of high joining temperatures and long bonding durations result in joints of high as-joint ultimate tensile strength (∼450 MPa) at 650 °C, but due to exceedance of the α–γ transition during joining, Eurofer is not martensitic any more. PBHT causes high residual stresses, so that the joint breaks easily during cross section preparation [23]. This observation conforms to the results of the FE analysis of section 2. Apart from that, brittle intermetallics and carbides precipitate at the direct interface of tungsten and Eurofer [21, 23]. Successful direct joining of tungsten and steel at temperatures below the Eurofer α–γ transition temperature has not been reported, creep seems to be too low here for proper bonding.

Resulting from the unsuccessful direct solid state bonding of tungsten and steel, a variety of attempts use discrete interlayers in the joint. In several publications, the insertion of 1 mm thick vanadium layers between the tungsten and steel parts has been investigated [24–28]. Vanadium is a reasonable candidate because it shows a comparably fast decline of the shut-down dose rate after exposure to a DEMO relevant neutron spectrum [19] and offers a CTE between the one of Eurofer and tungsten (8.4 × 10⁻⁶ ${{\rm{K}}}^{-1}$ at room temperature). Joining has been carried out at 1050 °C, 800 °C, and 700 °C for 1–4 h at uniaxial pressures corresponding to ∼10% creep at the Eurofer side of the joint. According to [27], at 700 °C, this corresponds to approximately 97 MPa bonding pressure. The outcome of this strategy seems promising, particularly when joining at low temperatures (700 °C) and long durations (4 h). No cracks are apparent, and in comparison to higher DB temperatures, this combination forms no FeV intermetallics and only a very thin VC layer (1 μm) at the vanadium-Eurofer interface. At room temperature, the joint exhibits a yield strength of ∼320 MPa and ultimate tensile strength of ∼330 MPa with low elongation at fracture. At 550 °C, yield and ultimate strength decrease by 25% and 12%, but the elongation increases drastically due to plastic deformation of vanadium. The location of fracture depends on the processing conditions and test temperature, and is described in detail elsewhere [27, 28]. Room temperature shear tests, Charpy impact tests at 550 °C, and thermal fatigue tests were carried out, too, proving a shear strength of 200 MPa, absorbed impact energy of 2 J and the endurance of 100 cycles between 350 °C and 500 °C in a vacuum furnace, respectively. After thermal fatigue testing, which took about 130 h, the VC layer increased slightly with some decarburization of Eurofer.

DB has also been carried out with the aid of a 1 mm thick niobium interlayer at 1050 °C for 1 h at uniaxial pressures causing ∼8% creep in Eurofer, with the typical PBHT [29]. Although initial experiments showed reasonable results, the strategy was terminated, probably owing to the excessive activation of niobium during DEMO relevant neutron exposure [19]. Similar to vanadium, niobium is a strong carbide former potentially decarburizing Eurofer.

Another discrete interlayer option, studied in many publications, is titanium. The activation of titanium is acceptable and the CTE equals 8.6 × 10⁻⁶ ${{\rm{K}}}^{-1}$ at room temperature. Joining has been carried out by pulse plasma sintering [30], DB [31, 32], and hot isostatic pressing (HIP) [33–38]. While the Ti interlayer used for pulse plasma sintering was rather thick (1 mm), it varied between 0.1 and 0.6 mm for DB, and between only 0.03 and 0.5 mm for HIP. The joining with pulse plasma sintering was carried out at 1000 °C. Required pressure and time are not provided. DB takes place at 950 °C for 0.5–1 h at 10–15 MPa uniaxial pressure. In contrast, for HIP, the (isostatic) pressure is strongly increased to 100 MPa, where joining is usually done at 900 °C for 1.5 h and combined with a second heat treatment (750 °C, 2 h, 70 MPa). Only two HIP joints have been realized at a temperature (760 °C) below the Eurofer α–γ transition, using higher stresses (150 MPa) and longer joining durations (4 h) [37, 38]. Despite processing differences, some commonalities are found among the aforementioned studies. Except for the pulse plasma sintered joints, no study reports the existence of cracks. In terms of microstructural changes, in [30, 38] the diffusion of iron to titanium is reported, and in [35, 36] the decarburization and ferritization of steel at the Ti-steel interface. After extended joining under high pressure, interdiffusion of titanium and tungsten is observed [37]. Mechanical test results are limited for this type of joint. The shear strength of titanium joints, HIPed at 950 °C and subsequently heat-treated, reaches 89 MPa with intergranular failure in the tungsten part [35]. HIPing at 760 °C yields shear strengths of 50–65 MPa with failure at the Ti-steel interface [38]. After annealing at 700 °C for 5 h, the shear strength increases to 120–130 MPa [38]. An interesting aspect for all types of tungsten-steel joints, which has only been assessed for HIPed Ti joints is the impact of ${{\rm{D}}}_{2}$ on the joint. Depending on the ${{\rm{D}}}_{2}$ partial pressure, the shear strength of the joint decreases dramatically, and at 1 bar ${{\rm{D}}}_{2}$ the titanium interlayer turns into ${\mathrm{TiD}}_{2}$ powder.

Other, less commonly used interlayers are pure iron, pulse plasma sintered at 700 °C and at 1000 °C [30, 39], FeTi [30], and NiFe [40].

Compared to solid state bonding, from a technological standpoint brazing is preferable. No external force is required and the joining of curved surfaces is easier because the liquid braze is tailored to penetrate the parent materials and spreads well.

A large number of high temperature brazing techniques, listed in table 1, has been developed. A dedicated comparison and evaluation is challenging, given significant differences regarding the used brazes, layer thickness, and processing steps. Independent from these differences, brazing tungsten and steel is usually carried out above the Eurofer α–γ transition temperature, at 950 °C–1100 °C, in vacuum furnaces. Some techniques are carried out in flowing argon atmosphere, further facilitating the processes [41–43]. Despite high joining temperatures used, usually no PBHT is carried out to re-establish the original ferritic-martensitic microstructure of Eurofer, although the need is described in [44–47], where it has been done. Altered Eurofer microstructures and mechanical properties should be considered after brazing, if PBHT is neglected. Usually, the braze layers have a low thickness (15–200 μm), which does not allow to account for the mismatch of CTE of tungsten and steel. Hence, some attempts make use of additional interlayers to the braze, consisting of e.g. Cu [26, 41–43, 48], V base alloys [45–47], or Ti [49]. For these interlayers, the same boundary conditions like for indirect solid state bonding have to be fulfilled. With regard to that, the large CTE, low strength and low liqidus temperature of copper used in some braze joints may cause challenges during cyclic high-temperature application. Compared to solid state bonding, with brazing, large reaction layers, often composing of secondary phases, are formed at interfaces already during joining and the reaction layers are decisive for the performance of the joint. While the layers are supposed to enhance the bond between adjacent materials, they are sometimes intrinsically brittle and weak, hence representing the weak spot of the joint. Detailed information about the observed microstructures after brazing may be found in the literature of table 1.

The effort put into testing to evaluate brazed joints differs strongly between the brazing techniques. While microstructural analyses are available for all mentioned joints, mechanical and thermo-mechanical test results exist only for few material systems. Among these, highest room temperature shear strengths are in the range of 230–250 MPa and are reported for the material combinations W–Ni–Cu–Ni–W [43], Eurofer–Ni–Cu–Ni–Eurofer [43], W–(Sn–Fe)–Ti–(Fe–Sn)–CLF-1 [49], and W–Cu–Rusfer [46]. After ageing [43] and thermal cycling [44–46, 49, 53] the shear strength is reported to decrease by ∼35% with Ni–Cu–Ni interlayers [43] and by ∼60% with Cu interlayers. It should be noted that a simple summarization of the test results should be taken with care because different joining geometries cause unequal residual stresses in the joint, which strongly affect the test result. This applies to mechanical and thermo-cycling tests and, of course, also for the comparison with other joining techniques named in this publication.

Besides joining bulk materials, the deposition of tungsten on steel by means of plasma spraying, precisely atmospheric plasma spraying (APS) [56–61] and vacuum plasma spraying (VPS) [13, 58, 60, 62–67], has been studied for fabrication of the FW. The successive deposition of tungsten droplets allows to manufacture thick layers on steel substrates and to handle more complex surface geometries compared to solid state bonding and brazing of bulk materials. Large areas may be produced quickly. On the other hand, spraying of tungsten is challenging because many aspects influence the process. Powder particles injected in the plasma plume need to melt thoroughly without overheating to create dense layers without pores and cracks. This is achieved by carefully optimizing the gun power, powder particle size distribution, particle velocity, spraying distance, feeding distance, substrate temperature, surrounding atmosphere, and multiple other factors [59, 61, 68]. The number of these aspects indicates that layers produced by one machine and operator are difficult to reproduce elsewhere identically. Layer characteristics named in many publications and hence being considered key challenges for the assessment of sprayed plasma-facing tungsten here are porosity, cracks, impurities (e.g. oxides), and adherence of neighbouring particles. Compared to bulk tungsten, which is already brittle below ∼800 °C, sprayed coatings usually contain more flaws and hence have a lower modulus [56, 61, 62], strength [61, 62], fracture toughness [69] and thermal conductivity [56–59, 61, 62, 70]. VPS coatings are usually superior to APS coatings in terms of the aforementioned key challenges. The porosity of VPS tungsten coatings was achieved to be as low as 0.6–0.85 vol% [58, 66], and of APS tungsten coatings 2.0–2.3 vol% [56, 59]. The oxygen amount was measured to 'no remarkable [..] oxides' in VPS layers [58], and to 0.3–5.5 at% [59, 61, 71] in APS layers.

In combination with complex stress profiles in the coating and substrate after spraying [72–74], and low adherence to the substrate, coatings often exfoliate after spraying. Reasons for low adherence are inexistent chemical reactions between the tungsten coating and the substrate (bonding is caused by mechanical clamping mostly), pores, and cracks. The effect of low adherence is often omitted because rather thin tungsten layers (1 mm seems to be the current standard and is used in [13, 56–60, 63, 65, 71, 75]) and small substrates are coated, both of which limit maximum stresses. In contrast, for application as FW armour, tungsten layers of 2 mm thickness are required over large areas. Tungsten layers of this thickness were only produced in [61, 62, 66]. The actual adhesion strength is reported only in [62] and equals 20–22 MPa there. However, failure occurs in the sprayed layer, not at the dissimilar materials interface. Possibly due to an imperfect microstructure of sprayed tungsten and the direct exposure to the fusion plasma, tests are focussed on microstructural analyses, on thermal shock, and on thermo-cycling [58, 60, 62–64]. The tests show that, despite many difficulties related to plasma spraying and imperfect microstructures compared to bulk tungsten, sprayed tungsten coatings can tolerate 1000 heat pulses of 2 MW m⁻² [62], 100 pulses of 4.8 MW m⁻² [63], and 30 pulses of 12 MW m⁻² [64]. In terms of cyclic stability, plasma sprayed tungsten seems to represent a promising FW material.

The combination of plasma spraying with advanced-tungsten developments such as tungsten alloys, which withstand oxidation, and the introduction of fibres for extrinsic ductilization have not yet been studied.

FG tungsten/steel interlayers have been developed to accommodate thermally-induced stresses at the tungsten-steel joint of the FW. In this context, thin-film and thick FG layers were manufactured. While the former exhibit a mixture on the atomic level, the latter usually compose of several sublayers, each representing a MMC of different tungsten/steel composition. The insertion of both thin-film and thick FG layers in the FW achieves a smooth transition from tungsten to steel, macroscopically. Based on rules of mixture considered for each sublayer, it is hypothesised that the smooth transition is adopted by the material properties, particularly the CTE. Taking FG layers with linear and stepwise grading of the material composition (and material properties) into account, multiple FE analyses have shown a positive impact of few millimetre thick FG layers on the stress-strain distribution in the FW during DEMO-relevant heat loads [12–15, 18].

For the fabrication of FG layers, several processing techniques have been studied. They may be divided into two categories. The first comprises approaches to create both the ∼2 mm thick plasma-facing tungsten layer and the FG layer in a single process. Plasma spraying [13, 18, 56, 69, 73, 76–78], laser cladding [76], SPS [76], hot pressing [76, 79], and resistance sintering under ultra-high pressure [80] were studied in this framework. Among these techniques, plasma spraying is most important because, exclusively, it requires no additional step to join the FG layer to structural Eurofer steel. All materials of the first category base on powder metallurgy, and hence many of the challenges for sprayed plasma-facing tungsten named in the previous subsection apply here, too. The second category contains approaches to manufacture FG layers on bulk tungsten substrates, later used as plasma-facing material. Besides powder metallurgical routes, e.g. electro discharge sintering (EDS) [81] and plasma spraying [24, 61], thin-film magnetron sputter deposition has been studied [24]. Although graded thin films are incapable of redistributing thermally-induced stresses, they may increase the adhesion strength of the tungsten-steel joint. For powder metallurgically-made FG layers of the second category, the bond between the tungsten rich end of the layer and bulk tungsten is usually weak. The insertion of thin vanadium foils has been used to improve the bond [24, 26, 81]. The bond between the steel rich end and the steel structure can be created by DB at ∼700 °C.

While microstructural characterizations are available for all FG layers named here, summarizing and comparing other properties is again challenging. FG layers of the first category (mainly VPS layers) contain significant residual stresses [18, 73]. They are tested with respect to toughness [18, 69] and cyclic heat load endurance [78], which is reasonable due to the aforementioned difficulties of sprayed plasma-facing tungsten. For FG layers of the second category, structural-mechanical and thermo-physical properties of individual sublayers are studied in more detail [56, 61, 81–83], because plasma-facing capabilities are not as important when standard bulk tungsten is used and the focus can be centred on stress-relieving abilities.

Despite comparison difficulties, it is shown that thick dense FG layers with porosities as low as 1–2 vol% and no cracks can be produced by EDS, APS and VPS after process optimization [56, 73, 81]. Thin sputtered layers have been manufactured completely dense. Besides porosity, the formation of intermetallic precipitates has a noticeable impact on FG tungsten/steel layers. The binary thermodynamic phase diagram Fe+W exhibits two intermetallic phases, ${\mathrm{Fe}}_{2}{\rm{W}}$ and ${\mathrm{Fe}}_{7}{{\rm{W}}}_{6}$. They are hard, brittle, and feature low thermal conductivity [82]. Precipitates of the two types are seen to form already during production by VPS, hot pressing, SPS, EDS, and APS [24, 61, 76, 81] due to the large contact area between individual tungsten and steel volumes on the microscopic scale. For the implementation of FG layers in FW FE analyses, usually ideal rules of mixtures of tungsten and steel are assumed [12, 14, 15, 18], possibly overrating the impact of the FG layer. The influence of microstructural imperfections like pores, precipitates and cracks are, in contrast, reflected in mechanical and thermo-physical characterizations of individual tungsten/steel composites and sublayers in [56, 61, 76, 81–83], indicating that modelling needs some improvements. Young's moduli of APS layers, measured for individual sublayers at room temperature in [56, 76] vary between 19% and 23% of moduli calculated by a linear rule of mixture for the different sublayers. Higher fractions in the range of 50%–100% were measured for SPS composites in [83], for EDS composites in [81] and for APS layers in [61]. Yield strength, reported for tungsten/steel composites made by EDS and APS exceed 100% of yield strength calculated by a linear rule of mixture [61, 81]. The thermal conductivity is reported to vary only very little with the tungsten/steel ratio in sublayers made by APS and EDS, and is similar to the conductivity of Eurofer. Only the measured CTEs usually correspond well to the linear rule of mixture for composite materials [56, 61, 81]. Additionally, in [69] the interface toughness of VPS layers is studied and reported to equal approximately 250 J m⁻².

With respect to the general performance a FG tungsten/steel joint and to differences between ideal and actually measured FG material properties, figure 6 presents the stress and plastic strain distribution in the top right corner of the FW component presented in section 2 after cooling from 700 °C (case (b)). This time, the distinct tungsten-steel interface is replaced by a 1 mm thick FG interlayer, which ranges 0.5 mm into the tungsten and the steel part each. The interlayer composes of three sublayers (25, 50 and 75 vol% tungsten). Homogenised, isotropic material properties assigned to the individual sublayer continua base on (a) ideal rules of mixture of tungsten and Eurofer, and (b) actually measured properties of Fe/W sublayers as presented in [81]. While it is unlikely that the steel volumes in FG layers will retain original Eurofer properties after fabrication, the interpolation of material properties based on Eurofer and tungsten represents the ideal case.

Zoom In Zoom Out Reset image size

According to figure 6(a), in the component with ideal FG properties a smoother transition of von Mises stresses from the Eurofer part to the tungsten part is seen compared to a component featuring a direct tungsten-Eurofer joint (see figure 4). The von Mises maximum changes from a random position along the tungsten-Eurofer joint to the bottom right corner of the tungsten part and increases by approximately 3% to 1398 MPa. The maximum of σ_yy stresses retains its position, now coincides with the von Mises stress maximum, and decreases by ∼10% to 989 MPa. A beneficial effect of the FG layer is particularly seen in the plastic strain distribution, whose local maximum in the joint zone decreases by 45% from 0.011 to 0.006 mm mm⁻¹. The global plastic strain maximum is still localized at the top left corner of the outermost cooling channel and equals 0.028 mm mm⁻¹ with and without FG layer.

The consideration of actually measured material properties in the FE analysis causes further changes, shown in figure 6(b). The combination of measured low Young's moduli and high yield strengths allows to accommodate more thermally-induced stresses elastically. Although the smooth von Mises stress transition is now absent and higher von Mises stresses are present in the uppermost sublayer of the FG transition (max 1605 MPa), the σ_yy maximum further reduces to 968 MPa and plastic strains to 0.005 mm mm⁻¹. The respective maximum positions do not change. It should be noted here that the Fe/W composites studied in [81] are brittle, potentially resulting from many microstructural interfaces. Hence, the low plastic deformations seen here are not only a benefit but also a prerequisite for the mechanical integrity of the FW component.

Ongoing work will further show that the amount of ${\mathrm{Fe}}_{7}{{\rm{W}}}_{6}$ in Fe/W composites is not constant but increases over time already at the cooling water temperature of WCLL blankets, 300 °C. Taking this behaviour into account, the characteristics of FG tungsten/steel layers depend on their temperature-time history, which should be taken into account for modelling. While after manufacturing, the properties of FG layers may depend on the particular fabrication technique, this link is assumed here to vanish during high temperature application following a growing influence of intermetallics properties on the homogenized material properties. For FW applications, a solution that reliably suppresses the precipitation in FG tungsten/steel layers is required.