In the pursuit of enhanced performance for gas turbine engines, the composition of nickel-based single crystal superalloys has become increasingly complex, with higher concentrations of refractory elements. Concurrently, the geometry of turbine blade casting parts has grown more intricate. This evolution has made freckle chains a prevalent and costly defect in directional solidification. Once formed, these defects cannot be eliminated by subsequent heat treatment, leading to a severe degradation of mechanical properties and increased scrap rates. The formation of freckles is influenced by a multitude of interlinked factors including alloy composition, the specific geometry of the casting parts, and the parameters of the directional solidification process itself. This investigation delves into the formation mechanisms of these defects, particularly within the context of ultrathin-walled casting parts, where conventional explanations based solely on thermosolutal convection often fall short in predicting the precise location of defect occurrence.
The classical theory for freckle formation hinges on convective instability within the mushy zone. Elements like Tungsten and Rhenium segregate to the dendrite core, while Aluminium and Titanium enrich the interdendritic liquid. This creates a solute-rich, less dense liquid in the mushy zone beneath a bulk liquid of relatively constant, higher density. This unstable “top-heavy” configuration can lead to channel convection, where buoyant plumes rise, fragmenting dendrite arms and carrying them to the casting surface where they grow into freckle chains. However, a critical observation challenges this as the sole mechanism: freckles typically appear only on specific surfaces or edges of a casting part, not uniformly across all surfaces exposed to the melt. This selective occurrence suggests that other factors, such as local variations in solidification sequence and induced interdendritic flow, play a decisive role.
This study was designed to isolate and examine these factors. We focused on the fabrication of ultrathin-plate casting parts with a nominal thickness of 0.6 mm. The extreme thinness of these casting parts simplifies the dendrite array into an almost two-dimensional configuration, allowing for clearer observation of dendrite interactions and defect initiation. The plates were arranged radially within the directional solidification furnace, creating a controlled scenario where one edge of the plate was nearer to the furnace center and the opposite edge was closer to the furnace wall. This setup inherently produces radial thermal gradients, influencing the local solidification sequence. By employing two distinctly different withdrawal rates, we aimed to probe the interaction between solidification morphology, interdendritic feeding flow, and the nucleation of freckle defects.
Experimental Methodology and Characterization
The foundation of this research lies in the precise fabrication and analysis of model casting parts. A third-generation nickel-based single crystal superalloy, comparable to alloys like DD33, was utilized. Its nominal composition, rich in Re, W, Ta, and Co, is representative of modern high-performance alloys prone to segregation-related defects. The chemical composition is summarized in Table 1.
| Ni | Cr | Co | Mo | W | Ta | Re | Al | Ti | Hf | C |
|---|---|---|---|---|---|---|---|---|---|---|
| Bal. | 2.5 | 9.0 | 1.5 | 6.0 | 8.0 | 4.0 | 6.0 | 0.2 | 0.1 | 0.01 |
The casting parts were designed as flat plates with a constant thickness of 0.6 mm but with variations in their planar outline (e.g., rectangular, tapered) to subtly influence the local thermal field. These wax patterns were assembled to project radially from a central pouring cup. After standard ceramic shell investment and dewaxing, the molds were directionally solidified using the High Rate Solidification (HRS) technique with a spiral selector for single crystal seeding. The critical process parameters were the withdrawal rates, set at 1 mm/min and 6 mm/min, with identical superheat and furnace temperature profiles otherwise maintained.
Post-solidification, the shells were removed, and the single crystal plates were sectioned for analysis. Macro-etching revealed the overall dendrite structure and the location of any defects. Transverse cross-sections were taken from regions where abnormal dendrites initiated. These samples were meticulously prepared for metallographic examination using Optical Microscopy (OM) and Scanning Electron Microscopy (SEM). A key metric, the primary dendrite arm spacing (PDAS, λ1), was statistically measured from these images. The formula used for calculating the average PDAS in a given area is:
$$ \lambda_1 = \sqrt{\frac{S}{N}} = \frac{1}{\sqrt{n_1}} $$
where $S$ is the area of the observed field, $N$ is the number of dendrites within that area, and $n_1$ is the dendrite number density.
Furthermore, the volume fraction of γ/γ’ eutectic, a direct indicator of the extent of microsegregation and the composition of the last-to-freeze liquid, was measured using image analysis software on multiple fields from both the “inner” (near furnace center) and “outer” (near furnace wall) regions of the transverse sections. To gain a three-dimensional perspective on dendrite growth and defect evolution, X-ray Computed Tomography (XCT) was performed on needle-shaped samples extracted from defective regions. Finally, the solidification process was simulated using ProCAST software to visualize the temperature field, isotherms, and the shape of the mushy zone for the different casting parts and withdrawal rates.
Microstructural Results: The Influence of Withdrawal Rate
The experimental results revealed a stark contrast in the behavior of the casting parts solidified at 1 mm/min versus 6 mm/min.
At 1 mm/min Withdrawal Rate: All plate casting parts, regardless of their planar outline, exhibited the initiation of abnormal dendrite growth. Consistently, this initiation site was located on the surface of the plate facing the center of the furnace. The defects were not associated with changes in the plate’s outline geometry. The abnormal structures manifested as either severely tilted dendrites that grew into large stray grains or as classical freckle chains composed of equiaxed grains. Measurement of the primary dendrite arm spacing showed relatively large values, characteristic of a low withdrawal rate. Most tellingly, quantitative analysis of the eutectic phase revealed a significant disparity across the thin section. The eutectic volume fraction near the furnace wall was approximately 1%, while near the furnace center it was about 3.6%. This gradient indicates a pronounced difference in the local solidification history and residual liquid composition.
At 6 mm/min Withdrawal Rate: The incidence of defects was drastically reduced. While a very limited number of isolated tilted dendrites could still be found on the inner surface, no fully developed freckle chains were observed in any of the casting parts. The primary dendrite arm spacing was measurably finer, and the secondary dendrite arms appeared more developed. The eutectic distribution was much more uniform, with volume fractions of around 6.0% at the outer edge and 6.5% at the inner edge. This near-uniformity suggests a different interdendritic environment during solidification.
| Withdrawal Rate | Typical PDAS (μm) | Eutectic (Outer Edge) | Eutectic (Inner Edge) | Dominant Defect Type | Defect Location |
|---|---|---|---|---|---|
| 1 mm/min | ~240-260 | ~1 vol.% | ~3.6 vol.% | Freckle chains & tilted dendrites | Exclusively inner surface |
| 6 mm/min | ~235-250 | ~6 vol.% | ~6.5 vol.% | Isolated tilted dendrites (rare) | Inner surface (rare) |
Solidification Simulation and Analysis of the Mushy Zone
The ProCAST simulations provided crucial insights into the thermal conditions responsible for the observed phenomena. For both withdrawal rates, the simulations confirmed that the solidification front (mushy zone) was not planar or uniformly curved in the radial direction. Instead, it was consistently tilted. The portion of the casting part closer to the furnace center entered the cooler zone of the furnace first. However, being shielded by the outer portions of the cluster and having less direct radiative cooling to the chamber wall, it experienced a lower axial thermal gradient. Conversely, the outer edge of the plate, directly facing the water-cooled chamber, experienced a steeper axial gradient. This resulted in a wider mushy zone at the inner edge and a narrower one at the outer edge, creating a radial thermal gradient ($\frac{\partial T}{\partial r}$) and causing the liquidus isotherm to slope from the inner edge (deeper in the hot zone) to the outer edge (further into the cold zone).
The magnitude and distribution of this radial temperature gradient were affected by the local geometry of the casting parts, but its fundamental direction and the consequent tilt of the mushy zone were universal features of the process for these radially arranged thin plates. The simulated radial gradient in the regions where defects initiated was in the range of 6-10 °C/mm for both withdrawal rates, linking the defect sites directly to this specific thermal condition.
Proposed Mechanism: Radial Feeding Flow and Its Consequences
Synthesizing the experimental and simulation results leads to a coherent mechanism for freckle formation in these ultrathin-walled casting parts, which extends beyond the classical buoyancy-driven channel convection model.
The tilted mushy zone establishes a distinct solidification sequence: dendrites at the inner edge of the plate (near the furnace center) begin to solidify and coarsen ahead of those at the outer edge. As solidification proceeds, solidification shrinkage occurs. The still-liquid interdendritic regions in the outer, lagging portion of the mushy zone must feed this shrinkage in the inner, leading portion. In a thin-walled casting part with a limited number of dendrites across its thickness, this creates a radial interdendritic feeding flow, $ \vec{V}_{feed} $, from the outer to the inner region. The driving force for this flow can be related to the pressure drop required for feeding, which is a function of the permeability, $K$, of the mushy zone and the viscosity, $\mu$, of the liquid:
$$ \vec{V}_{feed} \propto – \frac{K}{\mu} \nabla P $$
where $\nabla P$ is the pressure gradient induced by shrinkage in the inner, more solid region.
At a slow withdrawal rate (1 mm/min), the dendrites are coarse, and the secondary arm spacing is large. This results in higher permeability, offering less resistance to this radial feeding flow. The flow can attain significant velocity. This flow has two potent effects:
- Direct Mechanical Impingement: The flowing liquid directly impacts the dendrite trunks and arms in the inner mushy zone. When the shear stress imposed by the flow exceeds a critical value, it can cause local remelting (via solute washing) or even mechanical fragmentation of dendrite arms. The fragmented dendrite arms are then transported by the flow and can become trapped in nearby interdendritic regions where they continue to grow with a different orientation, forming a tilted dendrite or the nucleus of a freckle grain.
- Induction of Localized Thermosolutal Convection: The feeding liquid originating from the outer mushy zone has a different composition—it is closer to the bulk liquid composition and is less enriched in segregating elements than the liquid already present in the deep, inner mushy zone. When this “fresher” liquid intrudes into the solute-rich inner region, it can create localized density inversions that trigger highly focused thermosolutal convection cells. These cells are extremely effective at fragmenting dendrites right at the site where freckles are observed to form.
The significant gradient in eutectic content (1% vs. 3.6%) at 1 mm/min is direct microstructural evidence of this strong, composition-altering radial flow. The three-dimensional XCT observations, showing dendrite fragmentation and tilted growth initiation specifically on the inner side, perfectly corroborate this mechanism.
In contrast, at a high withdrawal rate (6 mm/min), the microstructure is refined. The primary dendrite arm spacing is smaller, and more importantly, the secondary dendrite arms are highly developed, creating a much denser, lower-permeability network. This dense network severely impedes and dampens any radial feeding flow. The permeability $K$ in the Kozeny-Carman type relationship is drastically reduced for a finer structure:
$$ K \propto \frac{\lambda_2^2}{f_l^2} $$
where $\lambda_2$ is the secondary dendrite arm spacing and $f_l$ is the liquid fraction. A smaller $\lambda_2$ leads to a much lower $K$. Consequently, the feeding flow velocity $ \vec{V}_{feed} $ is negligible. Solidification shrinkage is accommodated more locally, leading to the observed uniform eutectic distribution (~6% vs. ~6.5%). Without the strong radial flow, there is no dominant mechanism to fragment dendrites on the inner surface, explaining the near absence of defects.
Conclusion
This investigation into ultrathin-walled single crystal superalloy casting parts has elucidated a critical mechanism for freckle defect formation that is particularly relevant for components with strong radial thermal gradients. The key conclusions are:
- The radial arrangement of casting parts in a directional solidification furnace inherently creates a tilted mushy zone, with the inner region (near furnace center) solidifying ahead of the outer region (near furnace wall).
- This differential solidification sequence drives a radial interdendritic feeding flow from the lagging outer region towards the leading inner region to compensate for solidification shrinkage.
- At low withdrawal rates (e.g., 1 mm/min), the coarse dendritic structure offers high permeability, allowing this feeding flow to become strong enough to fragment dendrite arms through direct impingement and/or to initiate localized thermosolutal convection. The fragments grow into tilted dendrites or freckle chains, exclusively on the inner surface of the casting part. The process is evidenced by a strong radial gradient in eutectic phase fraction.
- At high withdrawal rates (e.g., 6 mm/min), the refined and densely branched dendritic structure presents very low permeability, effectively stifling the radial feeding flow. Shrinkage is accommodated locally, resulting in a uniform microsegregation profile and a very high resistance to freckle formation.
- Therefore, for thin-walled casting parts prone to such radial thermal fields, controlling the withdrawal rate to achieve a fine, low-permeability dendritic structure is a highly effective strategy for mitigating freckle defects, overriding the classic susceptibility linked to alloy density inversion alone. This understanding provides a vital guideline for the process optimization of complex, high-performance single crystal casting parts.