*4.2. Correlation between Weldability Assessment Trials*

In LBW Varestraint test, which is an externally loaded weldability test, cracks were mainly induced on the surface and in FZ since the strain was applied while the melt pool was solidifying and it was forced to pull away when the material did not have minimum strength and ductility to accommodate residual deformations. FZ cracking was highly enhanced in LBW Varestraint test, featuring an elongated V-shape solidification line due to the LBW parameters and energy density which was required to achieve full penetration and minimum weld width requested by industrial quality standards [16]. In this study, the section of Varestraint testing samples had to be reduced to 3.2 mm to deform them by bending due to test bench capability and to adjust them to representative welding applications.

Minimum differences in HAZ cracking were determined between heats with somehow better performance of NP and E. However, the relatively high scattering of TCL va-lues made the comparison between alloys difficult.

Clear differences between alloys were observed in on-cooling hot ductility Gleeble tests. In these tests, only incipient melting of low melting point phases took place. This melting was enough to wet grain boundaries, leading to a brittle fracture without any

area reduction at the test temperatures close to NDT. By reducing on-cooling testing temperature, we were able to study ductility recovery behaviour. On-cooling hot ductility results showed that moulds O and E (without Laves phase in BM at the beginning of the welding test) had shorter BTR and fast ductility recovery rates reaching the previous values before later ductility drop (Table 6). In both moulds, the ductility was effectively recovered at temperatures above 1050 ◦C (Figure 7). Moulds N and NP presented higher BTR (85◦) and very low recovery rate and ductility recovery capability. Coarser γ grain sizes in these samples could be the reason for this behaviour, since large grain sizes promote continuity of liquid resulting from incipient melting at grain boundaries and reduce extension of interfacial area between solid-state γ grains [2].

Mould P with higher Si content in its chemical composition had the largest BTR (105 ◦C) and intermediate recovery rate and capability. Longer BTR can be related to both the higher amount of residual Laves phase in the microstructure and its greater Si content. Both factors will contribute to increasing the volume fraction of intergranular liquid formed at the peak temperature reached in on-cooling test. Consequently, the temperature must be decreased to a lower point to allow full resolidification of the Si-enriched liquid. It must be mentioned that the effect of Si content on 718 alloys has been previously investigated, concluding that HAZ cracking trend is favoured if high Si contents are combined with high Mn or C contents [2,4].

Hot ductility tests give an insight about the response of material to liquation and subsequent resolidification. However, only a very limited amount of material is melted as opposed to real welding applications in which a relatively high amount of material is melted in the FZ and significant microstructural changes occur in this zone.

Indeed, this was observed in the FZ microstructural characterisation of bead-onplate samples. Laves phases were regenerated in FZ, whose chemical composition was completely different from the original Laves phase in HAZ and BM. FZ Laves phase in 9 mm plates were particularly depleted in Nb, Si, and Mo (Table 7). Surprisingly, the number of Laves phases in the less than 3 mm plates were much lower than in the 9 mm plates with faster welding cooling rates.

Whereas 3 mm plates were free of cracks, remarkable microfissures were identified in HAZ along grain boundaries in 9 mm bead-on-plate samples for the five investigated heats. As described previously, these cracks were decorated by continuous film, whose composition matched with Laves phase of FZ (Nb (20.5–22.1 wt %), Mo (4.7–5.0 wt %), and Ti (1.5–1.7 wt %)). Composition of this Laves phase film is quite independent of chemical composition, with only minor deviations in Si content. Laves phase with very similar Nb concentrations in LBW welds were reported by Odabaçi et al. [25].

As can be observed in Figures 12 and 15, corresponding to cross-sections of 9 mm thickness bead-on-plate samples of moulds P and O, respectively, the Laves film was also extended through FZ. This is strong evidence of backfilling mechanism. The current results demonstrate that in 9 mm bead-on-plate samples, the terminal liquid which remains in the melt pool at the end of the solidification process was diffused along grain boundaries, giving rise to a continuous Laves phase film that caused HAZ microfissuration.

Looking at the micrographs, we concluded that this diffusion was completed along base material γ grain boundaries which are perpendicular to fusion line and delimit threepoint intersections. Similar crack morphologies have been reported by Bai et al. [19] in LBW. The area below the nail head is a particular critical point where these microfissures pop up. It is worth mentioning that grain morphology was quite columnar, with grain boundaries elongated along the thickness, particularly in samples cast without ceramic blanket. Therefore, at the nail head area where the fusion line was quite horizontal, there was a high probability of having grain boundaries in the parent material intersecting fusion line, leading to three-point intersection, which is a critical point with high stresses during melt pool solidification [2]. Moreover, the elongated and straight morphology of vertical grain boundaries would be favourable for the described diffusion-based backfilling mechanism. Evidence of comparable backfilling mechanism which promotes HAZ cracking

has been recently reported in Haynes® 282® casting alloy [32]. In this case, cracking was exacerbated by diffusion of B. Other authors cited by [2] also concluded that segregation of B has a detrimental effect of HAZ liquation and FZ solidification cracking. However, this is not an influencing parameter in this work since ingots with the same lean B content of 0.002 wt % were used as raw material for the five heats.

**Figure 15.** (**a**) Film along crack in HAZ from 9 mm thickness sample mould O cross-section. (**b**) Enlargement of crack tip.

Straight bead-on-plate welds were more prone to HAZ cracking phenomena since both total number of cracks observed in cross-sections and TCL of those cracks were comparatively higher than in circular welds (Figure 11). Bead-on-plate samples from NP mould showed a consistently lower TCL than equivalent samples from moulds O and P. Therefore, it can be concluded that the lower aspect ratio of these grains due to slower solidification rates during initial casting reduced the probability of having deleterious three-point intersections along the length of fusion line.

No cracks were observed in less than 3 mm thickness samples welded with comparable LBW parameters. In this case, two remarkable differences were observed when looking at the cross-section of these samples. On one hand, melt pool or FZ had a "bowl shape" without nail head, typical in keyhole mode LBW, which reduces the risk of perpendicular intersection with columnar grains of parent material in comparison with "nail or mushroom" shape observed in 9 mm thickness plates. On the other hand, percentage of Laves phases in FZ was much lower than in 9 mm plates, as observed from Table 7. Both factors were critical to avoid formation of HAZ microfissures observed in thicker plates.

The percentage of Nb content of Laves phase observed in FZ of bead-on-plate samples was close to the eutectic point of the pseudo-binary diagram of alloy 718 (Figure 16a). This means that the terminal liquid would solidify as L → γ + Laves eutectic. It must be noted that most Laves phases observed at high magnification had a eutectic microstructure composed of γ and Laves phase (Figure 16b). This solidification path would be associated with larger volumes of terminal liquid that solidified at lower temperatures (down to 1180 ◦C) [4], since stepped liquid to solid transformation is hampered. If the amount of terminal liquid and its coexistence temperature range are increased, then one should expect a higher grain boundary wetting and infiltration risk in the three-point intersections, particularly if those grain boundaries are long and straight, as in the case of columnar microstructures obtained in the investment casting process.

**Figure 16.** (**a**) Pseudo-binary phase diagram for alloy 718 [33] <sup>1</sup> and (**b**) eutectic Laves phase in FZ (mould NP). <sup>1</sup> Reproduced from [33], with permission from J. Andersson, 2021.

In order to analyse the solidification path and steps, we carried out thermodynamic and diffusion-based simulations with the three heats with different Si contents. Figure 17a shows the calculated solidification path of mould E determined using Scheil simulation, which considers back diffusion of elements in the primary phase at high cooling rate of 100 ◦C/s, representative of welding processes. The simulation predicted the formation of NbC carbides and Laves phase at the final stage of solidification, significantly reducing the solidus temperature compared to solidification under equilibrium condition. In Figure 17b, we can see that at the end of solidification, the liquid enriched in Nb, Mo, Ti, and Si, which allowed the formation of Carbides and Laves phases, lowering the solidus temperature. Table 10 depicts the solidus and liquidus temperatures of the heats with different Si contents. It can be observed that the solidus temperature decreased with increasing Si content and the solidification range increased. As observed experimentally, the percentage of Laves phase and its Si content increased in the as-cast material through increasing the Si content of the alloy. However, in LBW, the cooling rate can be extremely fast [25] and this leads to the limitation of Nb segregation. Therefore, the resulting terminal interdendritic liquid will have a composition close to the eutectic point, and therefore it will have long persistence and relatively high volume.

**Figure 17.** Thermo-Calc simulation of alloy E (standard Si content): (**a**) solidification path and (**b**) microsegregation prediction using the Scheil model.


**Table 10.** Solidus and liquidus temperatures determined by equilibrium and Scheil simulation.
