Thin-walled nickel-base superalloy castings are central to aircraft engine swirlers, combustor hardware, and other hot-section components where low mass, complex internal geometry, and oxidation resistance must be combined in a single part. Investment casting is well suited to this class of components because it can produce fine wall sections, integrated blades, and internal passages with limited machining. The same geometry that makes investment casting attractive also makes defect control difficult: thin sections cool rapidly, feeding paths are narrow, and sharp thickness transitions generate thermal gradients that promote shrinkage porosity and hot tearing.
K4648 is a high-chromium, solid-solution-strengthened nickel-base superalloy developed for service below about 1000 °C. Its high Cr content provides oxidation and hot-corrosion resistance, while the Ni-Cr-W-Mo matrix supplies moderate high-temperature strength and fatigue resistance. The alloy is used for swirlers and other static hot-section parts that experience thermal cycling, oxidation, and moderate stress. Its casting behavior, however, is not forgiving. The solidification range is wide, the melt has a strong shrinkage tendency, and the microstructure contains blocky primary α-(Cr,Ni) and MC carbides that are hard and brittle. These phases can reduce ductility when they form continuous networks along grain boundaries or interdendritic regions.
Component Geometry and Casting Requirements
The swirler considered here is a small, thin-walled ring-and-blade structure. Its maximum outer diameter is 37 mm and its overall height is 28.4 mm. It contains three non-circular ring walls—inner, middle, and outer—with thicknesses of 1.5 mm, 1.25 mm, and 1.5 mm, respectively. The average wall thickness is about 1.3 mm and the minimum wall thickness is 1.25 mm. The radial gaps between the ring walls are 2.75 mm and 3 mm, and twelve small blades are distributed around each gap. The blades have an average thickness of about 1.2 mm. The minimum dimensional tolerance is 0.05 mm.
Because the part is assembled into a precision flow system, the drawing requires near-net-shape casting with no allowance for subsequent machining on the functional surfaces. The acceptance requirements include complete filling, no surface cracks or missing metal, and strict internal quality. The drawing limits individual internal shrinkage indications to 0.05 mm and allows no more than three such indications. These requirements place the process in a narrow window: the gating system must fill 1.25–1.5 mm walls without misrun, must feed shrinkage during a wide freezing interval, and must avoid hot tears at the ring-to-blade junctions.
The alloy composition is given in Table 1. The high Cr level, together with W, Mo, Nb, and Ti additions, produces a complex solidification path and a relatively wide freezing range. Thermodynamic calculation with JMatPro gives a solidus of about 1261 °C and a liquidus of about 1340 °C, a freezing interval of about 179 °C. A wide interval means that a mushy zone can persist across thin sections, isolating liquid pockets that are difficult to feed.
| Element | Ni | Cr | W | Mo | Al | Nb | Ti | C | B | Ce | S | Y |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| wt% | Balance | 33.66 | 4.86 | 2.90 | 0.94 | 0.90 | 0.92 | 0.067 | ≤0.008 | ≤0.03 | ≤0.01 | ≤0.04 |
Thermal-Physical Inputs and Defect Criteria
Finite-element simulation in ProCAST 2016 was used to evaluate filling, solidification, shrinkage porosity, and hot-tearing tendency. Because K4648 was not available in the standard material database, its measured composition was imported and its thermophysical properties were calculated and corrected against handbook values. The shell was modeled as a 7 mm mullite-based investment shell. Gravity casting was applied in the +Z direction at 9.8 m/s². The interfacial heat-transfer coefficient between the nickel-base melt and the silicate-bonded shell was selected after iterative comparison of 300, 500, 750, and 900 W/m²·K; 750 W/m²·K gave the most consistent solidification behavior.
Shrinkage porosity was assessed with the Niyama criterion, which is widely used for microshrinkage in cast alloys. The criterion is expressed as
$$
\frac{G}{\sqrt{R}} \le M
$$
where \(G\) is the local temperature gradient, \(R\) is the cooling rate through the freezing interval, and \(M\) is an alloy-dependent critical value. A low gradient and a high cooling rate reduce the Niyama value and increase the probability of interdendritic shrinkage. The simulated porosity volumes discussed below are outputs at a 1% porosity cutoff, not direct measurements of casting density.
Hot tearing was evaluated with the hot-tearing indicator (HTI) from the elastic-plastic stress model. The HTI combines local strain and stress accumulation in the late stage of solidification, when the alloy has low strength and low ductility. A higher HTI indicates a greater tendency for cracking. The indicator is comparative rather than an absolute failure criterion, but it is useful for ranking gating designs and process windows.
Initial Gating: Complete Fill but Poor Feeding
Top pouring was selected because it provides a direct pressure head for riser feeding and allows the melt to fill the thin upper ring walls before the lower sections solidify. The initial cluster used a central sprue, a circular runner, and gates connected to the middle ring wall. The sprue cross-section was approximately rectangular with an average area of about 240 mm², and the circular runner had an average cross-section of about 130 mm². A 50 mm riser height was used to maintain a residual pressure head. The initial process conditions were a pouring temperature of 1480 °C, a pouring time of 3 s, and a shell preheat temperature of 1180 °C.
The initial design filled completely. At about 90–92% fill, all castings were filled and no misrun was observed. The filling sequence showed some turbulence in the runner system, but the disturbance remained within the gating system and did not produce visible surface defects on the castings. Solidification began at the thin outer ring walls, where heat extraction was fastest, and proceeded inward. The complete cluster solidified in about 2537 s.
Despite complete filling, the initial design was not acceptable. Simulated shrinkage reached about 0.278 cm³, concentrated in the middle ring wall. A section through the defective region showed an isolated liquid zone: the outer ring wall was near 1200 °C and in a solid-liquid state, while the inner ring and blade region remained liquid at about 1382 °C. This temperature separation produced a late-fed mushy zone and a high probability of shrinkage porosity. The hot-tearing tendency was also highest at the inner ring wall, with an HTI of about 0.0056, and elevated values appeared near the blade connections. The combination of a thin wall, a thickness transition, and a late-solidifying ring section made the initial gating inadequate.
Direct-Fed Gating and Reduced Shrinkage
The revised design removed the circular runner and changed the cluster to direct sprue feeding. Six sprue branches were arranged around the cluster, each feeding castings through gates connected to the middle and inner ring walls. Two additional symmetric gates were placed on the outer ring wall. This arrangement allowed the melt to enter the middle ring, inner ring, and outer ring more simultaneously, reducing the isolated liquid zone that had formed in the initial design. The sprue remained filled with liquid metal during solidification, so its latent heat and pressure head helped feed the lower ring sections.
The revised filling sequence was more stable. At 0.76 s, the melt had entered the runner system without splashing. At 0.98 s, the metal rose smoothly in the casting, with no visible gas entrapment. At 1.14 s, most of the casting was filled, and at 2.9 s the cavity and pouring cup were completely filled. The solidification sequence started at the large outer thin wall, moved to the inner and middle ring walls, and then progressed to the blade connections. The last regions to solidify were the casting-to-gate junctions, which were kept under feeding pressure from the sprue. The revised cluster solidified in about 4095 s.
The change in shrinkage was substantial. The revised design produced only small, dispersed shrinkage indications in the ring walls; the large concentrated shrinkage in the middle ring disappeared. The remaining indications were scattered and could be removed or tolerated within the downstream process window. The result shows that for this thin-walled swirler, the dominant geometric problem was not filling but feeding: a gating system that fills completely can still fail if it creates an isolated liquid zone during solidification.
Pouring and Shell Preheat Windows
The pouring temperature and shell preheat temperature were evaluated separately with the revised gating system. The pouring time was fixed at 3 s. Table 2 shows the effect of pouring temperature at a shell preheat of 1180 °C. Increasing the pouring temperature from 1480 °C to 1540 °C reduced the simulated shrinkage volume from 0.0747 cm³ to 0.0043 cm³, a reduction of about 94.2%. At 1570 °C, the shrinkage volume rose slightly to 0.0082 cm³. The HTI at the inner-ring/blade junction did not follow the same monotonic trend: it was 0.0048 at 1480 °C, 0.0055 at 1510 °C, 0.0041 at 1540 °C, and 0.0065 at 1570 °C.
| Pouring temperature (°C) | Pouring time (s) | Shell preheat (°C) | Simulated shrinkage (cm³) | HTI at region A |
|---|---|---|---|---|
| 1480 | 3 | 1180 | 0.0747 | 0.0048 |
| 1510 | 3 | 1180 | 0.0295 | 0.0055 |
| 1540 | 3 | 1180 | 0.0043 | 0.0041 |
| 1570 | 3 | 1180 | 0.0082 | 0.0065 |
The shrinkage reduction with increasing superheat is consistent with improved melt fluidity and a longer feeding time before the thin walls establish a solid network. Above 1540 °C, the benefit is offset by a larger thermal gradient and greater solidification contraction, which can increase both shrinkage and hot-tearing tendency. The non-monotonic HTI reflects a competition between improved feeding and higher thermal strain. For this geometry, 1540 °C is the best compromise among the tested pouring temperatures.
Table 3 shows the effect of shell preheat at a fixed pouring temperature of 1480 °C. At 1150 °C, the simulated shrinkage was 0.1300 cm³. At 1180 °C it fell to 0.0747 cm³, and at 1210 °C it dropped to 0.0002 cm³, a reduction of about 99.8% relative to 1150 °C. At 1240 °C, the shrinkage rose slightly to 0.0004 cm³. The HTI remained in the narrow range of 0.0040–0.0046 across the shell temperatures, indicating that shell preheat has a much weaker effect on hot tearing than on feeding.
| Shell preheat (°C) | Pouring temperature (°C) | Pouring time (s) | Simulated shrinkage (cm³) | HTI at region A |
|---|---|---|---|---|
| 1150 | 1480 | 3 | 0.1300 | 0.0045 |
| 1180 | 1480 | 3 | 0.0747 | 0.0046 |
| 1210 | 1480 | 3 | 0.0002 | 0.0040 |
| 1240 | 1480 | 3 | 0.0004 | 0.0041 |
A hot shell reduces the thermal gradient between the melt and the mold, slows cooling in thin sections, and extends the time during which the ring walls can be fed. The effect is strongest when the shell temperature approaches the solidus; at 1210 °C, the shell is only about 50 °C below the calculated solidus of 1261 °C. Excessive preheat, however, can coarsen the grain structure, prolong solidification, and increase dispersed porosity. The slight increase at 1240 °C is consistent with this trade-off.
Final Process Selection and Trial Validation
Four combinations of the most promising pouring and shell temperatures were compared at a fixed pouring time of 3 s. The results are summarized in Table 4. The lowest shrinkage, 0.0002 cm³, was obtained with a pouring temperature of 1540 °C and a shell preheat of 1210 °C. The HTI values of the four combinations were nearly identical, between 0.0047 and 0.0049, so the final selection was governed mainly by shrinkage control. The selected process was therefore: pouring temperature 1540 °C, pouring time 3 s, and shell preheat 1210 °C.
| Pouring temperature (°C) | Shell preheat (°C) | Pouring time (s) | Simulated shrinkage (cm³) | HTI at region A |
|---|---|---|---|---|
| 1540 | 1210 | 3 | 0.0002 | 0.0047 |
| 1540 | 1240 | 3 | 0.0028 | 0.0048 |
| 1570 | 1210 | 3 | 0.0088 | 0.0047 |
| 1570 | 1240 | 3 | 0.0176 | 0.0049 |
Trial castings produced with the selected parameters were inspected visually, dimensionally, by fluorescent penetrant testing, by sectioning and optical microscopy, and by X-ray radiography. Visual inspection showed complete filling, clear outlines, smooth surfaces, and no sand adhesion, swelling, misrun, or visible cracks. Dimensional inspection with a coordinate measuring machine and calipers confirmed that the critical features were within tolerance. Table 5 summarizes the dimensional results for five castings.
| Feature | Drawing requirement | Maximum measured deviation | Parts inspected | Result |
|---|---|---|---|---|
| Outer ring outer diameter | ∅37 mm, +0.5/0 mm | 37.1 mm | 5 | Qualified |
| Outer ring inner diameter | ∅34 mm, +0.5/0 mm | 34.2 mm | 5 | Qualified |
| Inner ring outer diameter | ∅20 mm, +0.5/0 mm | 20.0 mm | 5 | Qualified |
| Inner ring inner diameter | ∅17 mm, +0.5/0 mm | 17.1 mm | 5 | Qualified |
| Tail cast draft | 70° ± 30′ | 70°08′ | 5 | Qualified |
| Tail vent diameter | ∅6.6 mm | ∅6.6 mm | 5 | Qualified |
| Tail vent wall thickness | 2 mm | 2 mm | 5 | Qualified |
Fluorescent penetrant inspection found no clustered indications and no linear indications. Under the applicable acceptance criteria for this casting, isolated non-linear indications with a longest dimension of 0.3 mm or less, or indications that could be removed by machining, were not counted as defects. No such rejectable indications were observed. Sectioning of blade-to-ring joints and optical microscopy found no microshrinkage. X-ray radiography was performed with an XYD3010 system using N-III film, a 3 min exposure, a 1000 mm focal distance, 80 kV, and 5 mA. The radiographs showed no shrinkage porosity or inclusions in the ring walls or blade joints. The inspection results were consistent with the simulation prediction of low porosity and supported the validity of the selected gating and process window.
Secondary Solution Treatment and Microstructural Evolution
The as-cast microstructure contains a γ matrix, blocky and chain-like primary α-(Cr,Ni) phase, and gray MC carbides. The primary α phase is a Cr-rich BCC phase with a Ni molar fraction above 30%. It forms partly because of Cr segregation during solidification and is distributed along grain boundaries and interdendritic regions. It is a hard, brittle phase; its Vickers hardness is about 6.3 GPa under a 0.2 N load. MC carbides are also brittle and can decompose to M23C6 and M6C during prolonged thermal exposure. Both phases can reduce ductility when they form continuous networks.
A primary solution treatment of 1180 °C for 4 h, air cooling, and aging at 900 °C for 16 h produced a microstructure with granular and rod-like precipitates, blocky M23C6 at grain boundaries and within grains, and fine granular γ′ around the larger blocky phases. The secondary treatment was then applied at 1160, 1200, or 1220 °C for 4 h, followed by air cooling and aging at 900 °C for 16 h. The secondary solution step was intended to dissolve or redistribute part of the primary α and MC phases that are difficult to remove in a single solution treatment, and to modify the size and distribution of M23C6 and γ′.
After secondary solution treatment, the alloy contained larger needle-like α phase, while M23C6 became smaller, more numerous, and more continuously distributed along grain boundaries. The needle-like α phase grew with increasing secondary solution temperature. Its average size was about 3.5 μm at 1160 °C, about 6 μm at 1200 °C, and about 18 μm at 1220 °C; at 1220 °C, some needles extended across entire grains. The M23C6 particles formed at 1160 °C were larger than those formed at the other secondary temperatures. The amount of fine granular γ′ was greatest at 1200 °C. Elemental mapping after the 1200 °C secondary treatment showed a relatively uniform distribution of Cr, Ni, C, Ti, Mo, and W, indicating that the second solution step improved chemical homogeneity.
The higher secondary solution temperatures also increased the number of dispersed micropores inside grains. This is consistent with Kirkendall porosity: during prolonged high-temperature solution treatment, different elements diffuse at different rates, and unequal atomic fluxes can create vacancies that condense into micropores. Because the secondary treatment adds a second high-temperature hold and, at 1220 °C, exceeds the primary solution temperature, the pore density is higher than after the single solution treatment. These micropores become preferential sites for microcrack initiation during tensile loading.
Hardness, Tensile Behavior, and Fracture
Hardness increased after heat treatment. The as-cast alloy had an average Vickers hardness of 215.9 HV. After primary solution plus aging, the average hardness was 252.8 HV. After secondary solution at 1160 °C plus aging, the average hardness reached 305.6 HV, which is about 41.5% higher than the as-cast condition and about 20.9% higher than the primary-treated condition. The hardness peak at 1160 °C is consistent with the larger M23C6 particles formed at that temperature: blocky carbides are hard phases and contribute strongly to indentation resistance, even when their size and continuity are unfavorable for ductility.
Room-temperature tensile testing at 25 °C and a crosshead rate of 0.1 mm/min showed a different optimum. The primary solution plus aging condition had an ultimate tensile strength of 583 MPa and an elongation of 17.2%. After secondary solution at 1200 °C plus aging, the ultimate tensile strength increased to 810 MPa, an increase of about 38%, while elongation decreased slightly to 15.3%. Among the secondary solution temperatures, 1200 °C provided the best combination of strength and elongation. The 1220 °C condition had a lower tensile strength because the higher density of micropores promoted microcrack formation and linkage before the alloy could reach its full strengthening potential.
Fracture surfaces of the as-cast and primary-treated specimens showed necking, fine dimples, and larger smooth dimples, consistent with a mixed intergranular and transgranular ductile fracture mode. No secondary cracks or large voids were observed in those conditions. After secondary solution treatment, secondary cracks appeared, and the fracture surface was smoother in some regions. At 1220 °C, numerous microcracks were found in the necked region, originating from micropores at grain interiors and near boundaries. When a crack is oriented at a large angle to the stress axis, its propagation is resisted and the crack width remains small; when the crack is nearly perpendicular to the stress axis, propagation is easier and the crack opening is larger.
The continuous M23C6 network along grain boundaries has a dual effect. It can pin grain-boundary sliding and accumulate dislocations, which raises tensile strength. At the same time, a continuous brittle carbide network weakens the boundary. After yielding, microcracks can propagate along the boundary, and the combination of continuous M23C6 and pre-existing micropores can become a damage source. The needle-like secondary α phase is also hard and brittle; a moderate amount can hinder grain-boundary sliding and dislocation motion, but excessive coarsening reduces ductility. The best balance was obtained at 1200 °C, where the microstructure retained useful γ′ precipitation and carbide strengthening without the severe porosity and needle coarsening found at 1220 °C.
Process Implications
The K4648 swirler case shows that investment casting of thin-walled superalloy parts is governed by a coupled set of geometric, thermal, and microstructural constraints. A gating system that fills completely can still produce unacceptable shrinkage if it creates isolated liquid regions during solidification. Direct sprue feeding, symmetric gates on multiple ring walls, and a hot shell are effective because they extend the feeding time and reduce the temperature separation between thick and thin sections. Pouring temperature has a narrower optimum: enough superheat is needed to avoid misrun and to feed shrinkage, but excessive superheat increases thermal gradients and hot-tearing tendency.
The heat-treatment results add a second constraint. Secondary solution treatment can improve strength by refining and redistributing M23C6, increasing γ′ precipitation, and modifying the α phase. However, higher temperatures also coarsen the needle-like α phase and generate Kirkendall micropores. The optimum is therefore not the highest hardness or the highest solution temperature, but the condition that balances precipitate strengthening against porosity and boundary embrittlement. For this K4648 casting, the best combination was a 1200 °C secondary solution treatment followed by 900 °C aging, which raised the tensile strength from 583 MPa to 810 MPa while limiting the elongation loss from 17.2% to 15.3%.
For production, the validated process window is a pouring temperature of 1540 °C, a pouring time of 3 s, and a shell preheat of 1210 °C with the direct-fed gating design. This window produced complete filling, dimensional conformity, no rejectable surface indications, and no radiographic shrinkage in the trial castings. The agreement between simulation and inspection supports the use of coupled filling-solidification models for thin-wall investment castings, provided that the interfacial heat-transfer coefficient and thermophysical data are calibrated for the actual shell and alloy. The remaining development work lies in tightening the secondary solution window and reducing dispersed porosity, since those features control the strength-ductility balance of the finished swirler.