The seed-to-mould gap as a source of stray grains at the base of a single crystal

A seed is the only way to set the crystallographic orientation of a single-crystal casting precisely, both axially and azimuthally. A grain selector picks a grain by competitive growth, giving an orientation within tolerance but not a chosen one. A seed passes its own lattice to the casting: it is partly melted back from the top, the melt grows on it epitaxially, and the single crystal continues upward.

A seed cannot be fitted without a gap, though. In practice about 0.1 mm is left between the seed and the cavity wall for assembly and thermal expansion. What happens in that gap during preheating and pouring, and whether it can spoil the single crystal, has had little study. A similar chilled layer on a helical selector is known to have caused stray grains; for seeds there were no systematic data.

A group at Central South University in Changsha tested it directly: they enlarged the gap on purpose and looked at what grows at the base of the casting.

Why the gap matters at all

In seeding, what matters is not only the seed's orientation but how it is melted back. The seed stands on the water-cooled chill plate: its bottom stays cold and does not melt, while its top is melted back by heat from the mould and the poured metal. The melt-back boundary is where the single crystal continues. If metal that froze quickly, with no link to the seed lattice, appears next to that boundary, it becomes a source of new grains.

The authors also recall the chemistry: in single-crystal alloys the γ′-forming elements (Al, Ti, Ta) accumulate in the interdendritic regions, and that is where γ/γ′ eutectic forms. Its amount depends strongly on composition – even a small difference in the sum of γ′ formers changes the eutectic fraction, and rhenium and hafnium increase it. These regions melt before the rest of the metal, so on preheating they are the first to turn liquid.

How the trial was run

The material was DD419, a second-generation single-crystal alloy: Cr 6.56%, Co 9.38%, W 6.27%, Re 2.94%, Al 5.56%, Ti 1.08%, Ta 6.47%, Hf 0.13%, balance nickel; the sum Al + Ti + Ta is 13.11%. Both seeds and castings were made of it. The seeds were 7 mm single-crystal rods oriented along [001] (deviation under 10°).

On two sets of seeds a flat was cut along the axis by wire EDM so that, once in the cavity, a gap of about 0.3 mm (seed A) or about 1.0 mm (seed B) remained. The surface was then ground to remove oxide and machining marks – as it would be before a normal production pour – so that the surface condition matched practice. Five seeds of each kind were set in one shell and fixed at the bottom with a thin layer of ceramic slurry of the shell's own composition, which did not hinder heat flow to the chill plate and did not run into the gap.

The trial

0.1 mmthe usual production seed gap, by the authors' estimate

0.3 and 1.0 mmgaps created on purpose to make the effect visible

1550 °Cmould preheat and pouring temperature, 15-minute hold

5 + 5seeds of each kind in one mould

The mould was set on the water-cooled chill plate, raised into the hot zone, preheated to 1550 °C, held for 15 minutes, filled with metal at the same temperature and withdrawn. After knock-out and removal of the gating, the seeding zone was etched and examined on longitudinal and transverse sections by optical and electron microscopy; zone compositions were measured by EDS, and eutectic fraction and particle size by image analysis, on at least three fields per specimen.

What is seen from outside

Both specimens carry a layer about 30 mm long on the seed, with fine equiaxed grains on its surface. The authors explain its origin this way: on preheating, metal from the upper part of the seed melts and runs down into the gap, while the bottom of the seed sits on the chill plate and does not melt, so the running metal freezes at some height before reaching the bottom.

Then comes the decisive difference. In specimen A normal dendritic growth continues above the layer: the directionally solidified matrix is barely affected, and only one slightly deviated grain grew into the casting, with a low-angle boundary. In specimen B several high-angle grains grow from the layer into the directionally solidified zone, cover the surface and run through the whole casting – the result is polycrystalline instead of single-crystal.

Structure of the layer

In specimen A the layer is about 200 µm thick and splits into two zones. The lower one is eutectic, with a lot of γ/γ′ eutectic. The upper one is equiaxed: γ grains with as much eutectic as the bulk of the casting.

In specimen B the layer is about 1 mm thick and the boundary between the zones is clear. The very bottom of the layer hangs as a droplet that does not touch the seed: as the metal ran down the gap it cooled and lost fluidity, and gravity and surface tension froze the front as a droplet. The eutectic here is very coarse – the droplet lost heat slowly because it did not touch the seed metal. Higher up the layer is tight against the seed and the eutectic is clearly finer, and finest right at the seed: the seed acts as a chill. The authors cite calculations showing that before directional solidification starts, the seed is colder than the mould.

In numbers: the γ/γ′ eutectic fraction is 52.7% in the lower part of the eutectic zone and 61.4% in the upper part; in the equiaxed zone it is 5.0%, as in ordinary DD419 castings. γ′ particles in the eutectic zone are coarse and irregular – 1.5 and 1.2 µm, as in the interdendritic regions of an ordinary casting; in the equiaxed zone they are fine, cuboidal and uniform – 0.28 µm.

Where the two zones come from

Composition gave the answer. The sum of γ′ formers Al + Ti + Ta is 17.22% in the eutectic zone, 12.45% in the equiaxed zone and 13.11% in the master alloy. Tantalum is 10.66% in the eutectic zone against 7.52% in the equiaxed zone. The sum W + Re, conversely, is 5.22% in the eutectic zone against 8.46% – about 40% lower. That is what interdendritic metal looks like: rhenium and tungsten go to the dendrite cores, aluminium, titanium and tantalum to the last liquid.

Hence the sequence. On preheating, the eutectic-rich interdendritic regions of the seed melt first – their melting point is lower. That liquid is the first to run into the gap and freezes quickly, forming the eutectic zone with ten times the usual eutectic. The charge melt poured afterwards, not enriched in those elements, fills the rest of the gap. Against the cold seed it undercools, partly fuses with it and renucleates as randomly oriented γ grains – the equiaxed zone. Once withdrawal starts, these grains grow upward and become stray grains that destroy the single crystal.

The wider the gap, the thicker the layer and the more equiaxed grains it holds that can grow into the casting. At 0.3 mm they produced one slightly deviated grain; at 1.0 mm, several highly misoriented ones.

Limits of the study

The 0.3 and 1.0 mm gaps are deliberately larger than usual: the authors amplified the effect to see and analyse it. Where the threshold between "nearly harmless" and "polycrystalline" lies, and how it depends on alloy, preheat cycle and seed cavity geometry, the paper does not show. It covers one alloy, straight cylindrical seeds and five specimens per variant, with no yield statistics. The withdrawal rate is given in the paper as 3 mm/s – most likely a typo, so we leave it aside.

The mechanism, on the other hand, is described in detail and backed by both structure and composition: a two-zone layer with eutectic below and equiaxed grains above is a recognisable signature to look for on your own sections.