I carried out a systematic study of slag inclusion defects in ZM6 magnesium alloy produced by investment casting. My focus was the influence of each component of the gating system on slag entrainment, oxide-film folding, gas entrapment, and final radiographic quality. I selected ZM6 because it is a commercial magnesium-rare-earth-zirconium alloy used for high-strength, heat-resistant, and corrosion-resistant investment casting components. In my work, I designed six gating-system variants, analyzed their flow fields with numerical simulation, compared the predicted behavior with actual pouring trials, and verified the final castings by digital radiography. The results showed that slag inclusion in ZM6 investment casting is not controlled by a single gating element. Instead, it is governed by the coupled action of the pouring cup, sprue geometry, sprue taper, ceramic filter, slag trap, filling direction, and local cooling rate.
In my investment casting trials, I treated slag inclusion as a multi-source defect. The first source is primary slag formed during melting because of dirty charge materials, inadequate fluxing, or incorrect melt handling. The second source is secondary slag formed during pouring because of turbulence, air entrainment, reaction with the shell, or poor gating design. The third source is folded and entangled double oxide films formed when the melt surface is disturbed, broken, and re-entrained. In magnesium investment casting, these films are especially harmful because they are not wetted by the melt and they create sharp discontinuities in the matrix. Under service loading, they act as stress concentrators and can initiate microcracks. Therefore, I considered both the hydrodynamic and thermal conditions of the investment casting process.
Alloy and Process Basis
I used a magnesium-rare-earth-zirconium alloy with the nominal chemistry shown in Table 1. The rare-earth addition improves elevated-temperature strength and promotes a more protective surface film. Zirconium refines the grain structure. However, the alloy still has a wide freezing range, low superheat retention, and relatively low fluidity. These characteristics make investment casting of thin sections sensitive to cold shuts, misruns, and slag entrapment. I therefore designed the gating systems to balance smooth flow against sufficient filling capability.
| Element | Nd | Zn | Dissolved Zr | Total Zr | Al | Fe | Si | Ni | Cu | Be | Single impurity | Total impurity | Mg |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Content, wt.% | 2.0–2.8 | 0.1–0.7 | ≥0.4 | 0.4–1.0 | ≤0.02 | ≤0.02 | ≤0.01 | ≤0.005 | ≤0.03 | ≤0.001 | ≤0.05 | ≤0.2 | Balance |
Table 2 lists the baseline investment casting parameters I used in both simulation and production trials. I kept shell thickness, shell preheat, pouring temperature, interface heat-transfer coefficient, and pouring rate as constant as possible so that the differences between trials could be attributed mainly to gating geometry. In actual production, the furnace was evacuated and then backfilled with a carbon-dioxide and sulfur-hexafluoride mixture. This cover gas reduces magnesium oxidation, but it does not eliminate slag if the flow field is poor.
| Parameter | Value |
|---|---|
| Shell thickness | 7 mm |
| Shell preheat temperature | 400 ± 25 °C |
| Pouring temperature | 750 ± 5 °C |
| Cooling method | Air cooling |
| Interface heat-transfer coefficient | 400 W·m-2·K-1 |
| Pouring rate | 1 kg·s-1 |
| Furnace vacuum before backfill | -0.08 MPa |
| Backfill pressure | -0.02 MPa |
| Cover gas | Air + CO2 + SF6 |
| Pouring height above cup | Approximately 300 mm |
| Shell layers | 6 |
| Boric acid addition to shell | 0.5% |
Physical Criteria for Slag Entrainment
I used several dimensionless and transport equations to interpret the simulation and production results. The Reynolds number indicates whether the melt is likely to remain laminar or become turbulent in a given channel. In investment casting of magnesium, even moderate turbulence can break the protective surface film and entrain it into the bulk melt.
$$
Re = \frac{\rho v D_h}{\mu}
$$
Here, ρ is density, v is velocity, Dh is hydraulic diameter, and μ is dynamic viscosity. I considered a high Reynolds number to be unfavorable because it promotes eddy formation and film folding. The Weber number is also useful because it compares inertial forces to surface-tension forces.
$$
We = \frac{\rho v^2 L}{\sigma}
$$
In this expression, L is a characteristic length and σ is surface tension. A high Weber number means that the melt can easily disrupt its own surface film. For magnesium alloys, the surface film is thin, brittle, and easily folded, so I paid special attention to regions where the melt free surface was unstable.
Continuity requires that a reduction in area increases velocity. This is fundamental to gating design because an oversized sprue may allow the melt to fall as a free stream, while an undersized sprue may accelerate the melt too much.
$$
A_1 v_1 = A_2 v_2
$$
Bernoulli’s equation, written with pressure, velocity, and elevation terms, helped me explain why a sprue with no taper can create negative pressure and gas aspiration.
$$
P + \frac{1}{2}\rho v^2 + \rho g h = \text{constant}
$$
If the melt separates from the sprue wall, the local pressure can fall below the surrounding gas pressure. The approximate pressure deficit is related to the local dynamic pressure.
$$
\Delta P_{\text{gas}} = P_{\text{atm}} – P_{\text{local}} \approx \frac{1}{2}\rho v^2
$$
This pressure deficit drives gas into the melt and increases the probability of porosity, gas entrapment, and oxide-film formation. I therefore preferred sprue designs that remain full or nearly full during filling, but without excessive acceleration.
The filling time is controlled by mold volume and volumetric flow rate. I used the following relation to compare predicted and actual pouring behavior.
$$
t_f = \frac{V}{Q} = \frac{V}{A v}
$$
For gravity pouring, the ideal flow rate through a restriction can be expressed by a discharge equation.
$$
Q = A C_d \sqrt{2gH}
$$
In this equation, Cd is the discharge coefficient, g is gravitational acceleration, and H is the effective metal head. The actual flow rate is lower because of friction, filter resistance, and progressive solidification.
I also considered heat loss during filling. A slow fill is not automatically safe because the melt temperature decreases with time.
$$
T(t) = T_m + (T_0 – T_m)\exp\left(-\frac{h A_s t}{\rho c_p V}\right)
$$
Here, Tm is mold or shell temperature, T0 is initial melt temperature, h is heat-transfer coefficient, As is surface area, cp is specific heat, and V is volume. This equation explains why the strongest slag-control design in my trials also produced severe cold shuts. The filter and choke cup reduced velocity, but they also increased filling time and heat loss.
For the ceramic filter, I evaluated filtration efficiency in terms of inlet and outlet inclusion concentration.
$$
\eta_f = \frac{C_{\text{in}} – C_{\text{out}}}{C_{\text{in}}}
$$
The filter can capture large oxide particles and promote smoother flow, but it cannot remove dissolved gas, very fine films, or inclusions that form after the filter. This is why I observed improvement but not complete elimination of slag in the filtered investment casting trials.
Gating System Design
I designed six investment casting gating systems. The casting module was kept identical, while the pouring cup, sprue, taper, and filter were varied. Table 3 summarizes the six schemes. Scheme 1 was the control. Scheme 2 used a slot sprue. Scheme 3 added a tapered slot sprue. Scheme 4 used a choke-type pouring cup with a recessed buffer. Scheme 5 added a magnesia ceramic filter to the control. Scheme 6 combined the choke cup and the ceramic filter.
| Scheme | Pouring cup | Sprue geometry | Taper | Filter | Slag trap | Main purpose |
|---|---|---|---|---|---|---|
| 1 | Funnel cup, bottom diameter 75 mm, height 160 mm, 10° draft | Square 40 mm × 40 mm | None | None | None | Control condition |
| 2 | Same funnel cup | Slot 40 mm × 20 mm | None | None | None | Reduce sprue void space |
| 3 | Same funnel cup | Slot 40 mm × 20 mm | 5° | None | None | Match falling melt profile |
| 4 | Choke cup with 30 mm high, Φ140 mm recess | Slot 40 mm × 20 mm | 5° | None | Recess and 40 mm entry above bottom | Buffer initial melt and retain front slag |
| 5 | Funnel cup as Scheme 1 | Square 40 mm × 40 mm | None | Magnesia ceramic | None | Filter slag without major cup change |
| 6 | Choke cup as Scheme 4 | Slot 40 mm × 20 mm | 5° | Magnesia ceramic | Recess and filter | Combine buffer and filtration |
The casting part consisted of three plates with nominal dimensions 8 mm × 120 mm × 210 mm. These plates were connected to a vertical runner. Plates 1 and 2 were connected through ingates and were spaced 50 mm from the vertical runner. A wax rod of Φ10 mm was used above them as a discharge and vent path. Plate 3 was connected directly to both the vertical runner and the ingate, so I could compare the effect of filling direction and slag discharge. The bottom runner extended 30 mm beyond the last plate to act as a buffer for the first metal entering the mold. This extension was important because the first melt often carries the highest concentration of oxide films and entrained gas.
All connections were rounded with R5–R15 fillets to reduce sharp changes in flow direction. The shell was built with six layers, approximately 7 mm thick. I added 0.5% boric acid to the shell material to reduce interfacial reaction between the magnesium melt and the shell. I also placed an iron gauze filter at the bottom of the pouring cup to prevent large debris from entering the gating system. The metal was melted in a 150 kg medium-frequency magnesium furnace with flux protection. During pouring, the furnace door was closed, the chamber was evacuated, and then a CO2 + SF6 mixture was introduced. The actual atmosphere therefore remained a mixture of air, CO2, and SF6.
Numerical Simulation of Flow Fields
I used a three-dimensional geometry model and imported it into a casting simulation environment. I generated surface meshes with a nominal size of 3 mm, then created volume and shell meshes. The boundary conditions and process parameters are given in Table 2. I analyzed the filling sequence at several times: 0.4 s, 1.2 s, 1.6 s, 2.5 s, 3.2 s, 3.8 s, and 6 s. These snapshots allowed me to identify where the melt front became unstable, where the sprue remained partially filled, and where the filter or choke cup altered the flow field.
Early Filling Behavior
At 0.4 s, the funnel-cup schemes, namely Schemes 1, 2, 3, and 5, showed a common problem. The first metal that contacted the cup wall splashed and then moved directly into the sprue. In Scheme 1 and Scheme 5, the square sprue of 40 mm × 40 mm was only about one-quarter filled. The remaining void space allowed the melt to fall, swirl, and contact the cover gas. This condition is highly unfavorable for investment casting of magnesium because the surface film can break and fold repeatedly. In Schemes 2 and 3, the slot sprue reduced the void space, and the melt occupied roughly one-half to two-thirds of the channel. This reduced the turbulence potential, but it did not fully eliminate gas contact. In Schemes 4 and 6, the choke cup held the initial splash in the recess, and the melt had not yet reached the sprue. This confirmed that the choke cup functions as an effective buffer and slag trap for the first liquid.
At 1.2 s, the differences became more pronounced. In Schemes 1 and 5, the pouring cup, sprue, and bottom runner were still not full. In Schemes 2 and 3, the cup bottom, sprue, and bottom runner were already full. In Schemes 4 and 6, the cup bottom was full, but the sprue and bottom runner were not. This showed that a conventional funnel cup does not actually slow the melt. When the sprue fills, the cup bottom begins to fill only afterward, so the cup cannot act as a buffer. In contrast, the choke cup forces the melt to rise in the recess before it can enter the sprue. That geometry produces a strong slowing effect and retains the front slag in the cup.
In the casting cavity at 1.2 s, Schemes 1, 2, and 3 had already filled to Plate 3. The liquid surface was steeply sloped, which indicated excessive velocity. The melt rushed to the far side of Plate 3, hit the end, and then returned toward the runner. This return flow generated severe turbulence. Scheme 3 was the fastest because the 5° taper made the sprue act as a positive-pressure channel and accelerated the melt. Schemes 4 and 6 had only reached the bottom runner because the choke cup delayed the flow. Scheme 5 had a filter in the runner, and the melt reached the ingate of Plate 3 more smoothly than in Scheme 1.
Intermediate Filling Behavior
At 1.6 s, the gating system in most schemes was similar to the previous snapshot. In the casting, Schemes 1, 2, and 3 showed a flatter surface in Plate 3, but the melt height in the plate remained higher than that in the bottom runner. This again indicated that the flow had not fully settled and that the plate was being filled by a strong forward jet. Plates 1 and 2 had not yet started to fill. In Scheme 4, the melt had reached the ingate of Plate 3, while the ingates of Plates 1 and 2 remained nearly empty. In Scheme 5, the filter produced a smoother fill in Plate 3, but a measurable slope remained. In Scheme 6, the combination of choke cup and filter reduced the velocity so much that the melt had not yet reached the casting cavity.
At 2.5 s, only the gating system of Scheme 1 remained unfilled. The other five schemes had full gating systems. In Plate 3, Schemes 1 through 4 had filled about one-half, while Schemes 5 and 6 had filled about one-quarter with a nearly flat and bottom-up front. Plates 1 and 2 began to fill. Scheme 3 had already reached the plates, whereas Schemes 4 and 6 had only reached the ingates. This confirmed the ranking of filling speed: Scheme 3 was fastest, and Scheme 6 was slowest. Plate 3 filled earlier than Plates 1 and 2 in most schemes, which is significant because Plate 3 is directly connected to the vertical runner. The direct connection allows some of the early slag to be discharged through the runner, whereas Plates 1 and 2 may retain the early slag that enters their ingates.
At 3.2 s, the gating system of Scheme 1 was still not full. In Plate 3, Schemes 1 and 2 had filled about two-thirds, Scheme 3 about three-quarters, Schemes 4 and 5 about one-half, and Scheme 6 had just started. In Plates 1 and 2, Schemes 1, 2, 3, and 5 filled smoothly. Schemes 4 and 6 had only reached the ingates. The height difference among Plate 1, Plate 2, and Plate 3 was smallest in Scheme 6. This indicated that the strong buffering action of the combined choke cup and filter produced a more uniform filling front, but at the cost of very slow filling.
At 3.8 s, Schemes 1 and 2 had filled about 80% of the plates, and Plates 1 and 2 lagged slightly behind Plate 3. Scheme 3 had completed all three plates. Scheme 4 had filled about one-half of Plate 3, and Plates 1 and 2 had just begun. Scheme 5 had filled about two-thirds, with Plates 1 and 2 again slightly behind Plate 3. Scheme 6 had just started all three plates. These results showed that the gating system controls not only the total filling time but also the sequence in which different casting sections are filled.
Late Filling Behavior
At 6 s, Schemes 1, 2, 3, and 5 had completed filling. Scheme 4 had filled about 80%, with Plates 1 and 2 still behind Plate 3. Scheme 6 had filled only about 25%, and Plates 1 and 2 were slightly behind Plate 3. The simulated filling time for Schemes 4 and 6 was about 7.5 s. However, actual pouring showed a much longer time for Scheme 6. This discrepancy is important because it shows that a constant pouring-rate assumption in simulation does not fully represent the dynamic interaction between the crucible, the choke cup, and the filter.
| Time | Scheme 1 | Scheme 2 | Scheme 3 | Scheme 4 | Scheme 5 | Scheme 6 |
|---|---|---|---|---|---|---|
| 0.4 s | Square sprue about 1/4 full; splash enters sprue | Slot sprue about 1/2–2/3 full; less void | Slot sprue about 1/2–2/3 full; taper begins to accelerate | Melt held in cup recess; not yet in sprue | Square sprue about 1/4 full; filter not yet active | Melt held in cup recess; filter not yet active |
| 1.2 s | Cup, sprue, and bottom runner not full; Plate 3 strong jet | Cup bottom, sprue, and bottom runner full; Plate 3 strong jet | Fastest fill; Plate 3 strong jet | Cup bottom full; sprue and runner not full; no casting fill | Filter slows melt; Plate 3 ingate reached | Very slow; no casting fill |
| 1.6 s | Plate 3 flatter but still high velocity; Plates 1–2 empty | Plate 3 flatter; Plates 1–2 empty | Plate 3 flatter; Plates 1–2 empty | Plate 3 ingate reached; Plates 1–2 empty | Plate 3 smoother; some slope remains | Casting cavity not yet reached |
| 2.5 s | Gating still not full; Plate 3 about 1/2 | Gating full; Plate 3 about 1/2 | Gating full; Plate 3 about 1/2; Plates 1–2 begin | Gating full; Plate 3 about 1/2; Plates 1–2 at ingate | Plate 3 about 1/4; bottom-up smooth fill | Plate 3 about 1/4; bottom-up smooth fill |
| 3.2 s | Sprue still not full; Plate 3 about 2/3 | Plate 3 about 2/3; Plates 1–2 smooth | Plate 3 about 3/4; Plates 1–2 smooth | Plate 3 about 1/2; Plates 1–2 at ingate | Plate 3 about 1/2; Plates 1–2 smooth | All plates just beginning; smallest height difference |
| 3.8 s | About 80%; Plates 1–2 slightly behind Plate 3 | About 80%; Plates 1–2 slightly behind Plate 3 | All plates complete | Plate 3 about 1/2; Plates 1–2 begin | About 2/3; Plates 1–2 slightly behind Plate 3 | All plates just beginning |
| 6.0 s | Complete | Complete | Complete | About 80%; Plates 1–2 behind | Complete | About 25%; Plates 1–2 behind |
Interpretation of Simulation Results
From the simulation, I reached several conclusions. First, the conventional funnel cup did not buffer the initial melt. It allowed the first splash and its associated oxide film to enter the sprue directly. In the choke cup, the recess acted as a small reservoir. The front metal had to rise before it could enter the sprue, so the cup retained a significant portion of the early slag. This is why the choke cup improved the radiographic results in actual investment casting.
Second, the square sprue of 40 mm × 40 mm was too large for the actual flow rate. It remained partially filled for much of the pour, creating a free-fall condition. The melt contacted the cover gas, generated new oxide films, and produced vortices along the sprue wall. The slot sprue of 40 mm × 20 mm reduced the void space and promoted a fuller channel. However, the improvement in final slag defects was not as large as I expected. This suggests that once the melt has been contaminated in the cup and early sprue, a smaller sprue alone cannot remove the existing slag.
Third, the 5° taper was excessive. In principle, a tapered sprue can match the acceleration of a falling stream and keep the sprue full. In my trials, however, the taper increased velocity and made the melt enter the casting with a strong forward jet. The result was severe turbulence in Plate 3 and a higher risk of entrainment. Therefore, sprue taper must be optimized together with pouring rate and choke area.
Fourth, the magnesia ceramic filter had a strong effect on the flow. It reduced the velocity after the filter and produced a smoother filling front in the casting. It did not significantly increase the total filling time in Schemes 1 and 5, which is favorable. In Scheme 6, however, the combined choke cup and filter reduced the velocity so much that the melt lost heat and the filling time increased dramatically.
Fifth, the filling direction was critical. Plate 3, which was directly connected to the vertical runner, filled earlier and faster than Plates 1 and 2. At first, this seems unfavorable because fast filling can cause turbulence. However, Plate 3 also had a direct discharge path. The early slag could be carried through the runner and out of the plate, so Plate 3 ended up with fewer defects. Plates 1 and 2 filled more smoothly, but they had no direct discharge path for the front metal. Therefore, smooth flow alone does not guarantee low slag. The gating system must also provide a path for the early contaminated metal to leave the casting cavity.
Based on these observations, I predicted the slag severity ranking as follows: Scheme 1 was worst, followed by Scheme 3, Scheme 2, Scheme 5, Scheme 4, and Scheme 6 as best. For the individual plates, I predicted that Plates 1 and 2 would have similar defect levels and that Plate 3 would have the least slag.
Production Trials and Radiographic Verification
I then produced investment casting modules according to the six schemes. The more complex modules, especially Schemes 4, 5, and 6, were assembled with the designed cups, sprues, filters, and slurry shells. After shell building, the modules were poured under the conditions in Table 2. I recorded the actual pouring time and compared it with the simulation. The actual filling times are shown in Table 4.
| Scheme | Simulated filling time | Actual filling time | Observation |
|---|---|---|---|
| 1 | About 6 s | 6 s | Consistent with simulation |
| 2 | About 6 s | 6 s | Consistent with simulation |
| 3 | About 6 s | 6 s | Consistent with simulation |
| 4 | About 7.5 s | 11 s | Choke cup increased actual time |
| 5 | About 6 s | 6 s | Filter did not change total time significantly |
| 6 | About 7.5 s | 40 s | Combined buffer and filter caused severe slowdown |
The actual pouring time for Schemes 1, 2, 3, and 5 agreed well with the simulation. Scheme 4 took 11 s instead of 7.5 s. Scheme 6 took 40 s, far longer than the simulated 7.5 s. I attributed this to the dynamic interaction between the melt supply and the filter. In simulation, I imposed a constant pouring rate. In actual investment casting, the melt first fills the cup and the region above the filter. As the filter restricts flow, the melt level above the filter rises, and the effective head may change. At the same time, the melt loses heat to the shell and the filter. The increased viscosity and the onset of solidification further reduce the flow rate. In Scheme 6, the choke cup and the filter together created a double buffering effect. The melt moved very slowly, and by the time it reached the thin plates, its temperature had dropped below the level required for proper fusion.
After pouring and shell removal, I cut the test plates and inspected them by digital radiography. The radiographic results can be summarized as follows. Schemes 1, 2, and 3 showed large slag inclusions and gas entrapment. The defects were not only oxides but also folded films and mixed oxide-gas complexes. Scheme 4 showed smaller slag defects and smaller gas entrainment. Scheme 5 showed still smaller slag defects and no measurable gas entrapment. Scheme 6 showed no slag or gas entrapment, but Plates 1 and 2 had severe cold shuts and were completely broken in the middle. These results confirmed my prediction that strong buffering and filtration improve slag control but can create cold-shut defects if the filling time becomes excessive.
| Scheme | Radiographic slag | Gas entrapment | Cold shut | Overall assessment |
|---|---|---|---|---|
| 1 | Large and numerous | Severe | None | Worst slag condition; control condition |
| 2 | Large; slightly fewer than Scheme 1 | Severe | None | Slot sprue alone insufficient |
| 3 | Large; similar to Scheme 2 | Severe | None | Excessive taper accelerated melt |
| 4 | Smaller area and smaller individual defects | Reduced | None | Choke cup beneficial but slow |
| 5 | Further reduced | None observed | None | Filter beneficial without excessive time penalty |
| 6 | None observed | None observed | Severe; Plates 1 and 2 broken | Best slag control but unacceptable cold shut |
In all schemes, Plate 3 had fewer defects than Plates 1 and 2. This was consistent with the simulation. Plate 3 was directly connected to the vertical runner, so the early contaminated metal could be discharged through the runner rather than being trapped in the plate. Plates 1 and 2 had similar defect levels. Their ingates were located away from the vertical runner, and their filling path did not provide a direct discharge for the front metal. As a result, the early oxide films and gas bubbles remained in these plates.
Discussion of Slag Formation Mechanisms
My results support a multi-mechanism view of slag inclusion in ZM6 investment casting. The first mechanism is splash-generated oxide film in the pouring cup. When the melt first strikes the cup, the free surface breaks and new oxide forms. If the cup is a simple funnel, this oxide enters the sprue. If the cup has a recess or choke, the oxide can be retained. This is why the choke cup improved the radiographic quality.
The second mechanism is free-fall and gas aspiration in an oversized sprue. In Scheme 1, the square sprue remained partially empty. The falling stream created a low-pressure region and pulled cover gas into the melt. The gas reacted with magnesium to form oxide and possibly fluoride-containing films. These films then folded and became inclusion defects. The slot sprue reduced this effect, but it did not eliminate it because the cup already contained contaminated metal.
The third mechanism is excessive acceleration. In Scheme 3, the 5° taper made the sprue behave as a positive-pressure channel. The melt velocity increased, and the jet entered Plate 3 with high momentum. This caused surface folding and entrainment. A taper is beneficial only when it matches the natural acceleration of the stream and keeps the sprue full without generating an excessive jet.
The fourth mechanism is filtration. The magnesia ceramic filter captured large inclusions and promoted smoother flow after the filter. The filter did not react with the magnesium melt and had relatively good wetting behavior. However, it could not remove fine oxide films that formed downstream of the filter or dissolved gas that came out of solution during solidification. This explains why Scheme 5 had reduced slag and no gas entrapment in the radiographs, but still was not completely defect-free.
The fifth mechanism is heat loss and premature solidification. Scheme 6 had the best slag control because the melt was strongly buffered and filtered. However, the filling time increased to 40 s. The thin plates cooled rapidly. When the melt finally reached Plates 1 and 2, the front metal had already solidified or become too viscous to fuse with the later metal. The result was a cold shut. This shows that slag control cannot be optimized independently of filling capability.
The sixth mechanism is slag discharge path. Plate 3 had fewer defects because it was directly connected to the vertical runner. The early melt could pass through the plate and into the runner or the extension. In contrast, Plates 1 and 2 were fed through ingates. The first metal entering those plates had no easy exit, so inclusions remained. This finding suggests that an effective investment casting gating system should include a dedicated slag trap or discharge path near the front of the casting cavity.
Implications for Investment Casting Gating Design
Based on my experiments, I recommend that ZM6 investment casting gating design should consider the following coupled requirements. The pouring cup should buffer the initial melt without causing excessive heat loss. The sprue should remain full or nearly full, but it should not accelerate the melt to a velocity that causes surface folding. The filter should be placed where it can capture large inclusions without creating an excessive pressure drop. The casting cavity should be oriented so that the front metal does not enter critical thin sections first. A slag trap or discharge extension should be provided near the first filling region.
The conventional funnel cup is simple and economical, but it does not retain the front slag. The choke cup is effective for slag retention, but it increases metal consumption and can cause cold shuts. The slot sprue reduces void space, but by itself it does not solve the problem. The 5° taper used in my trials was too large. A smaller taper, perhaps combined with a different choke area, may provide a better balance. The magnesia ceramic filter was the most effective single addition for reducing slag and gas entrapment without significantly increasing total filling time.
For an economically balanced investment casting design, I would combine a funnel cup with a tapered slot sprue and a magnesia ceramic filter. However, I would still add a slag trap at the bottom of the cup and a local slag-collection feature in the sprue. This combination should provide moderate buffering, reasonable filling time, and effective filtration. The exact taper and choke dimensions should be optimized by simulation and validated by pouring trials because the optimal values depend on alloy superheat, shell temperature, casting thinness, and filter permeability.
| Design element | Benefit | Risk | Recommended use |
|---|---|---|---|
| Funnel pouring cup | Simple, low metal consumption, fast fill | Poor initial slag retention; splash enters sprue | Use with downstream filter and slag trap |
| Choke pouring cup with recess | Strong buffering; retains front slag | Increased metal consumption; longer fill; cold-shut risk | Use for thick sections or when slag risk is dominant |
| Square sprue | High flow capacity; simple pattern | Partial filling; gas aspiration; vortex formation | Avoid for thin-section magnesium investment casting |
| Slot sprue | Reduced void space; fuller flow | Still insufficient alone; may increase velocity | Use with controlled taper and filter |
| 5° sprue taper | Can keep sprue full in principle | Excessive acceleration in my trials | Reduce taper; optimize with simulation |
| Magnesia ceramic filter | Captures large slag; smooths flow; no major time penalty in simple layout | Cannot remove fine films or dissolved gas; can cause excessive slowdown when combined with choke cup | Use in a balanced gating system |
| Bottom runner extension | Buffers first metal; provides discharge volume | Increases yield loss if too large | Use a moderate extension |
| Direct connection to vertical runner | Provides slag discharge path | Can increase local filling velocity | Use for non-critical or thick plates; shield thin sections |
Quantitative Summary of Observations
To compare the six schemes more systematically, I assigned qualitative indices for slag severity, gas entrapment, filling safety, and overall suitability. The ranking in Table 6 reflects the combined simulation and radiographic evidence. A lower number indicates a better result.
| Scheme | Slag severity rank | Gas entrapment rank | Filling safety rank | Overall rank |
|---|---|---|---|---|
| 1 | 6 | 6 | 5 | 6 |
| 2 | 5 | 5 | 4 | 5 |
| 3 | 4 | 4 | 3 | 4 |
| 4 | 3 | 3 | 2 | 3 |
| 5 | 2 | 2 | 6 | 2 |
| 6 | 1 | 1 | 1 | 1 for slag, worst for cold shut |
The overall ranking is not a simple single number because Scheme 6 had the best slag and gas results but the worst filling safety. In investment casting production, a cold shut is a fatal defect. Therefore, Scheme 6 is not acceptable despite its excellent slag control. Scheme 5 provided the best compromise in my trials because it reduced slag and eliminated measurable gas entrapment without a significant increase in filling time. Scheme 4 also improved slag but was slower and used more metal. Scheme 1, Scheme 2, and Scheme 3 were unacceptable for thin-section ZM6 investment casting because they produced large slag and gas defects.
Thermal and Fluid Coupling
The interaction between fluid flow and heat transfer was one of the most important findings. A gating design that is excellent from a fluid-mechanics standpoint may fail because it removes too much heat. I used the following thermal criterion to interpret the cold-shut risk.
$$
t_{\text{flow}} < t_{\text{solid}}
$$
Here, tflow is the time required for the melt to fill a thin section, and tsolid is the time available before the melt front becomes too viscous or solid to fuse. In Scheme 6, the actual filling time reached 40 s, which exceeded the solidification time of the thin plates. The result was a complete break in Plates 1 and 2. This confirms that slag-control measures must be evaluated together with filling capability.
I also considered the local cooling rate in the thin plates. The plate thickness was only 8 mm, so the surface area to volume ratio was high. Heat loss to the shell was rapid. When the melt velocity is high, the residence time in the gating system is short, and the melt arrives at the plate with sufficient superheat. When the melt velocity is too low, the melt loses superheat before it reaches the plate. Therefore, the optimal gating system must provide enough velocity to fill the thin sections before solidification, while avoiding the high velocity that causes turbulence and oxide entrainment.
This balance can be expressed as a window of acceptable flow rates.
$$
Q_{\min} < Q < Q_{\max}
$$
The lower limit is set by filling and cold-shut requirements. The upper limit is set by turbulence, gas entrainment, and oxide-film folding. Investment casting of ZM6 is particularly challenging because the acceptable window is narrow. The alloy has a wide freezing range and relatively low fluidity, so the lower limit is high. At the same time, the surface film is fragile, so the upper limit is low. A successful gating system must operate within this narrow window.
In my trials, Scheme 1 operated above the upper limit. The melt was fast and turbulent. Scheme 6 operated below the lower limit. The melt was slow and cold. Scheme 5 operated closer to the middle of the window. The filter reduced velocity after the filter, but the filling time remained acceptable. This is why Scheme 5 gave the best overall result.
Slag Trapping and Discharge
I also examined the role of slag traps and discharge paths. The choke cup acted as a trap for the initial splash. The bottom runner extension acted as a buffer for the first metal entering the mold. The direct connection of Plate 3 to the vertical runner acted as a discharge path. These three features worked together to reduce slag in Plate 3. However, Plates 1 and 2 still suffered because their ingates did not provide a direct discharge.
I therefore recommend that future investment casting designs include a dedicated slag trap at the end of the bottom runner and a small discharge channel from the first-filled casting section. The discharge channel should be designed so that it fills early and freezes last, or at least remains open long enough to carry the front metal out of the critical section. The cross-section of the discharge channel should be large enough to pass the expected oxide films but not so large that it reduces the filling rate of the casting.
An approximate criterion for slag discharge is that the discharge path should have a lower flow resistance than the critical casting section during the early filling stage.
$$
R_{\text{discharge}} < R_{\text{casting}}
$$
Here, R represents hydraulic resistance. If the discharge path has lower resistance, the first metal will preferentially enter the discharge path and carry the slag away from the casting. Once the discharge path fills or freezes, the main casting continues to fill. This concept is consistent with my observation that Plate 3, which had a direct connection to the runner, had fewer defects than Plates 1 and 2.
Filter Behavior in Magnesium Investment Casting
The magnesia ceramic filter performed well in my trials. It did not react with the magnesium melt, and it provided a large surface area for capturing oxide films. The filter also changed the flow pattern. After the filter, the melt velocity decreased and the filling front became smoother. This reduced the tendency for surface folding and gas entrainment. In Scheme 5, the filter reduced slag and eliminated measurable gas entrapment. The total filling time was essentially unchanged compared with Scheme 1. This is a significant advantage because it means that filtration can be added without automatically causing cold shuts.
However, the filter had limitations. It could not capture fine inclusions that passed through the filter or formed downstream. It also could not remove dissolved gas that evolved during solidification. In addition, when the filter was combined with a choke cup in Scheme 6, the pressure drop became too large. The melt accumulated above the filter, lost heat, and eventually filled the mold very slowly. This shows that filter selection and placement must be matched to the available metal head and the required filling time.
I used the following expression to estimate the pressure drop across the filter.
$$
\Delta P_f = \frac{\mu v L_f}{K} + \rho C_f v^2
$$
Here, Lf is filter thickness, K is permeability, and Cf is a loss coefficient. A low-permeability filter or a thick filter increases pressure drop. If the pressure drop is too high, the flow rate falls below the filling requirement. Therefore, the filter should be selected with sufficient permeability, and the gating system should provide enough head to drive the melt through the filter without excessive slowdown.
Practical Design Guidelines
From my study, I developed the following practical guidelines for ZM6 investment casting. First, avoid a simple funnel cup without a downstream slag trap. The initial splash is a major source of oxide films. Second, avoid an oversized sprue that allows free fall and gas aspiration. A slot or reduced-area sprue is better, but it must be combined with a properly designed pouring cup. Third, optimize the sprue taper. A taper that is too small may not keep the sprue full; a taper that is too large accelerates the melt excessively. Fourth, use a magnesia ceramic filter when possible. It provides effective filtration and flow smoothing without a major time penalty in a balanced design. Fifth, provide a slag discharge path from the first-filled section. This path should be designed to carry the front metal out of the critical casting region. Sixth, evaluate the thermal budget. Any measure that slows the fill must be checked against the solidification time of the thinnest section. Seventh, validate the design by both numerical simulation and radiographic inspection.
| Guideline | Reason | Expected effect |
|---|---|---|
| Use a pouring cup with a buffer or a downstream trap | Retain initial splash and oxide film | Reduces large slag defects |
| Avoid oversized square sprues | Prevent partial filling and gas aspiration | Reduces gas entrapment and folded films |
| Optimize sprue taper rather than using a fixed large angle | Avoid excessive acceleration | Reduces jet-induced turbulence |
| Place a magnesia ceramic filter in the runner | Capture large inclusions and smooth flow | Reduces slag and gas defects |
| Provide a slag discharge path near the first-filled section | Remove front metal from critical regions | Reduces slag in plates fed through ingates |
| Check filling time against thin-section solidification time | Prevent cold shuts | Ensures complete fusion |
| Use simulation with actual pouring dynamics | Constant flow rate may not represent filter interactions | Improves prediction accuracy |
Comparison of Simulation and Production
The agreement between simulation and production was good for Schemes 1, 2, 3, and 5. The filling sequence, hot spots, and general defect locations matched the radiographic results. The simulation correctly predicted that Plate 3 would fill first, that Plates 1 and 2 would fill later, and that the filter would smooth the flow. The simulation also correctly predicted that the choke cup would slow the melt and reduce slag.
The main discrepancy occurred in Scheme 6. The simulated filling time was about 7.5 s, but the actual filling time was 40 s. This discrepancy arose because the simulation used a constant pouring rate, while the actual process experienced a dynamic decrease in flow rate as the melt accumulated above the filter and lost heat. This finding shows that investment casting simulation should include the interaction between the crucible, the pouring cup, the filter, and the shell. A constant inlet velocity boundary condition may be adequate for simple gating systems, but it is insufficient for strongly buffered systems with high pressure drop.
Despite this discrepancy, the simulation was valuable. It identified the mechanisms responsible for slag formation and helped me interpret the radiographic results. It also showed why some design changes had little effect and why others had a strong effect. I therefore consider numerical simulation to be a necessary tool for investment casting gating design, but it must be validated and calibrated with actual pouring trials.
Conclusions
1. I found that slag inclusion in ZM6 investment casting is controlled by the coupled effects of pouring cup design, sprue geometry, sprue taper, ceramic filtration, filling direction, and thermal conditions. No single gating element can eliminate slag.
2. The choke-type pouring cup provided strong buffering and retained the initial splash and oxide film. It reduced slag defects in the final castings. However, it increased metal consumption and slowed the fill, which increased the risk of cold shuts and misruns.
3. The slot sprue reduced the void space in the sprue and decreased free-fall gas aspiration. However, when used alone, it did not significantly improve the final slag defects. The 5° taper used in my trials was too large and accelerated the melt, causing severe turbulence and slag entrainment.
4. The magnesia ceramic filter was effective. It captured large inclusions, smoothed the flow, and reduced gas entrapment. In a balanced gating system, it did not significantly increase the total filling time. However, it could not remove all fine oxide films or dissolved gas.
5. The filling direction was critical. The plate directly connected to the vertical runner had fewer defects because the front metal could be discharged through the runner. Plates fed only through ingates retained more slag and gas defects. A dedicated slag discharge path is therefore recommended.
6. The combination of a choke cup and a ceramic filter produced the best slag and gas results but caused severe cold shuts because the filling time increased to 40 s. This showed that slag-control measures must be balanced against filling capability and thermal safety.
7. The numerical simulation agreed well with production for most schemes. The main discrepancy occurred when the actual flow rate decreased dynamically because of filter resistance and heat loss. Future investment casting simulations should include this coupling for strongly buffered systems.
8. For an economical and robust ZM6 investment casting design, I recommend a funnel-type pouring cup with a local slag trap, a tapered slot sprue with an optimized taper, a magnesia ceramic filter in the runner, and a slag discharge extension near the first-filled section. This combination should provide reasonable slag control without the severe cold-shut risk observed in the fully buffered design. The exact dimensions should be optimized by simulation and verified by radiographic inspection.
| Final recommendation | Function | Reason |
|---|---|---|
| Funnel cup with bottom slag trap | Balance fill speed and initial slag retention | Reduces metal consumption compared with a full choke cup |
| Tapered slot sprue with reduced taper | Keep sprue full without excessive acceleration | 5° taper was too aggressive; optimize smaller angle |
| Magnesia ceramic filter in runner | Capture large inclusions and smooth flow | Effective and compatible with magnesium melt |
| Bottom runner extension | Buffer first metal | Protects casting from initial contamination |
| Slag discharge path near first-filled section | Remove front metal | Plate 3 had fewer defects due to direct discharge |
| Simulation with dynamic inlet condition | Predict actual filling time | Constant flow rate underestimated Scheme 6 time |
My study demonstrates that investment casting of ZM6 magnesium alloy requires a carefully balanced gating system. The goal is not simply to slow the melt or simply to filter it. The goal is to keep the melt within a narrow window of velocity and temperature so that it fills thin sections completely while avoiding turbulent surface folding and oxide entrainment. When this balance is achieved, slag inclusion defects can be significantly reduced, and the reliability of magnesium investment casting components can be improved.