Porosity, Shrinkage and Hot Tears: One Root Cause
Why porosity, shrinkage, hot tears and leaks often share one origin — oxide bifilms entrained during filling — and how to find them.
Xinyouyuan Engineering Team
10/22/202610 min read
Four defects, four investigations, no progress
A stainless steel valve body starts coming back from the customer's leak test. The foundry opens the reject. There is a cluster of round voids near the inlet gate. Another batch shows a spongy region on the heavy flange. A third has a hairline crack that only appears after the parts are machined. The quality manager files four corrective actions: one for porosity, one for shrinkage, one for hot tearing, one for the leak.
Six months later, four sets of parameters have been adjusted, two gating layouts have been tried, and the rejection rate has not moved. This is one of the most common and most expensive patterns in investment casting, and it usually has a single explanation.
The four findings are not four problems. In a large share of cases they are four visible expressions of one thing that happened in the first two seconds of pouring.
What the four defects have in common
Porosity, shrinkage, hot tears and leaks are taught, measured, specified and corrected as separate phenomena. In a plant they live in four different departments: melting owns gas, gating owns shrinkage, metallurgy owns cracking, and the customer owns the leak. That organisational split is exactly why the root cause survives for years.
There is a single physical object that can produce all four. It is not a gas bubble, not a shrinkage cavity and not a stress concentration. It is a folded oxide film that was dragged into the metal while the mould was filling.
What an oxide bifilm actually is
Every liquid metal surface in air carries a (semi-)solid oxide skin. It forms in milliseconds and it behaves like a membrane, not like a liquid. When that surface folds over on itself, or when a droplet falls back into the melt, two dry oxide faces come into contact and bond to each other. What results is a double film: an outer face that is atomically bonded to the surrounding metal, and an inner interface that is not bonded at all.
That inner interface is, in mechanical terms, a crack. Not a crack that will grow into a defect later — a crack that already exists, in the liquid, before solidification begins.
Four consequences follow directly, and each one matters on the shop floor:
The thickness is nanometric and the area can be macroscopic. Radiography, ultrasonic testing and penetrant inspection routinely miss them. "We inspected it and found nothing" is not evidence that the metal is clean.
It does not need to become a defect; it only needs to be opened. Everything that happens afterwards — gas diffusion, shrinkage, thermal contraction, service stress, corrosion — acts on a pre-existing separation.
The outer face wets perfectly. Because it is atomically continuous with the matrix, it becomes a preferred nucleation site for gas pores, shrinkage, second phases and dendrites. This is why some castings show second-phase morphology that classical nucleation theory cannot explain.
It cannot be removed. Remelting, filtering and long holding reduce the population; HIP and forging can close a film but cannot dissolve the interface. The only reliable strategy is not to create it.
One additional mechanism deserves separate mention. A bubble rising through liquid metal leaves a wake, and that wake is an oxide tube. The bubble itself may escape and never appear on a radiograph, while the tube it dragged behind it stays in the casting. This is why "we degassed and still have porosity" is such a common report: the gas left, the damage stayed.
One film, four different outcomes
The same entrained film opens in different ways depending on what acts on it afterwards. This table is the diagnostic core of the model:
Gas diffuses into the gap — hydrogen, nitrogen or water vapour expands the interface into a rounded cavity. The shop calls it porosity, pinholes or "gas porosity".
Hydrostatic pressure and feeding demand pull it open — the film is drawn apart by volume loss during solidification. The shop calls it shrinkage porosity, or micro-shrinkage when it is fine and diffuse.
Uniaxial tensile strain during solidification — restrained contraction opens the film into a straight separation along the interface plane. The shop calls it a hot tear, and later a cold crack.
Service or machining stress — the film continues to propagate. The shop calls it fatigue failure, brittle fracture or delayed cracking.
Corrosive medium enters the channel — the film's inner interface is a connected micro-channel. The shop calls it pitting, filiform corrosion or stress corrosion cracking.
The practical value of this list is not theoretical elegance. It is that a part showing dispersed pinholes plus local shrinkage plus a tear at the gate plus a failed pressure test should not be treated as four problems. It should be treated as one filling problem with four symptoms.
The numbers that decide whether you entrain air
Entrainment is not a matter of operator skill or judgement. It is governed by thresholds, and the thresholds are uncomfortably low.
Critical meniscus velocity is about 0.5 m/s for most liquid metals. Above that, surface tension can no longer hold the front together and the surface starts to fold, break and entrain.
Critical free-fall height is about 10 mm. Set v = √(2gh) with v = 0.5 m/s and h comes out at roughly 12 mm. In other words, any free fall inside a mould exceeds the limit. There is no such thing as a safe drop.
Damage scales with v². A 200 mm drop reaches about 2 m/s; a 1 m drop reaches about 4.4 m/s. Against the 0.5 m/s threshold, that is roughly 16 times and 78 times the kinetic energy per unit mass.
A conventional trumpet pouring basin behaves like an air pump. Under a bottom-pour ladle stream it can entrain at least 50% and up to 80% air by volume, producing a metal-air emulsion that then has to separate inside your mould.
Entrainment, not raw material, dominates inclusion content. Svoboda's统计 of macro-inclusions in carbon and low-alloy steels found roughly 83% reoxidation products and 14% mould-material debris, leaving only about 4% for genuine external inclusions such as slag, refractory and deoxidation products. Around 96% of what you find was picked up after the metal left the furnace.
That last number reframes the whole quality discussion. Furnace practice can be close to perfect. The defects are made in the transfer, the pour and the fill.
Why "bigger runner, slower flow" makes it worse
The intuitive response to turbulence is to enlarge the runner so the metal moves more slowly. In practice this often increases defects, and the reason is geometric.
Enlarging a cross-section reduces velocity only where the passage stays full. A larger passage under the same head tends to run partly empty: the stream separates from the roof, falls, and meets air over a large free surface. You have replaced fast metal with falling metal, and falling metal entrains far more than fast metal. The v² term does the rest.
The correct direction is the opposite one: reduce the section, keep the system full, never let it fall, never interrupt it. In gating terms this means a naturally pressurised system — a tapered sprue that narrows downward to match the accelerating stream, and runners and gates that only ever decrease in area from the choke to the ingate.
The meniscus must never stop
Velocity is only half of the problem. The other half is stalling.
Whenever the liquid front stops advancing — pouring interrupted, section suddenly enlarged, metal falling in a waterfall, a long horizontal run that pools — the oxide skin on that front thickens and stiffens in place. When metal advances again it folds that stiffened skin into the bulk. Slow filling is therefore just as dangerous as fast filling, and in some geometries more so.
The practical conditions that follow are strict: fill from the bottom up, allow no falling or sliding down an inclined surface, allow no long horizontal run, never interrupt the pour, and let the front expand continuously as it rises. These conditions conflict directly with top pouring, which is otherwise attractive for feeding. That conflict is real and has to be resolved deliberately rather than ignored.
Translating this into investment casting practice
Investment casting starts with an advantage: it is normally bottom-gated already, and the ceramic shell can hold a smooth, tapered passage that sand moulding cannot. That advantage is frequently thrown away in the wax pattern design. The measures that matter most, roughly in order of return:
1. Control the sprue with a taper
The sprue should narrow from top to bottom so that the stream stays full and the velocity stays roughly constant. A straight cylindrical sprue in a tall mould accelerates all the way down and then discharges at several metres per second into the runner.
2. Do not enlarge the gate downstream of the choke
Every expansion is a place where the stream separates from the wall and air has somewhere to sit. Keep the ingate as the smallest section in the system and size it from the required fill rate, not from habit.
3. Remove the deep pouring cup
A deep basin with a large free surface is the pump described above. Shorten it, keep it coaxial with the sprue, and reduce its free surface. Where the ladle allows, direct contact pouring — the ladle lip meeting the cup with no exposed falling stream — is the single highest-value change available, and it can reduce entrainment damage by an order of magnitude.
4. Place the filter low and close to the sprue-runner junction
Filter placement matters more than filter presence. A filter sitting high in the cup with an air space behind it becomes an air reservoir. Place it low so that metal immediately covers its downstream face, and provide a small trap ahead of it for the first contaminated metal.
5. Use the shell to fix the fill, and rotation to fix the feed
Bottom filling satisfies the no-fall condition but works against gravity feeding. The elegant resolution, where part geometry and handling allow, is to fill from the bottom and then rotate the mould so that gravity takes over the feeding direction.
6. Treat shell permeability as a real variable
A ceramic shell is thin but it is not permeable in the way a sand mould is. Back-layer formulation, stucco grading, venting wax and complete burnout all affect whether gas can leave the mould cavity as fast as metal enters it. A shell that cannot vent will produce porosity that no amount of gating redesign will fix.
A four-question test: one problem or four?
Before opening four corrective actions, run this test. If two or more answers are yes, treat the defects as one entrainment problem.
Are the defects co-located? Do the pores, the shrinkage and the tear cluster near the ingate, or along the last-filled region?
Do they appear together in the same batch? Not necessarily in the same part — batch-level co-occurrence is the signal.
Did the customer's functional test fail without a visible defect? A pressure or leak failure with a clean radiograph is a classic film signature.
Did parameter changes in four different directions produce no improvement? Independent problems usually respond to at least one of four independent fixes. A shared cause does not.
Two additional confirmations are worth the effort when they are available. Fractography on a failed part: a film-origin fracture surface is smooth and featureless at the origin, without the dimples or cleavage steps a metallurgical crack would show. And sectioning the gate region rather than the defect itself — films concentrate where the metal turned.
Making it measurable
An argument that cannot be measured will not survive a production meeting. Two measurements make entrainment visible:
Reduced pressure test (RPT). Solidify a small sample under roughly 0.1 atm and compare its density with one solidified at atmosphere. The density index, (ρatm − ρvac)/ρatm × 100%, responds to gas content and to film opening.
Bifilm index. Section the RPT sample and measure the total length of crack-like film traces on the section, in millimetres. This converts "melt cleanliness" into a number that can be tracked batch to batch.
Use them together. If density index falls but bifilm index does not, the problem is not in the furnace — it is in the transfer and the pour. Set your own baseline from your own history rather than importing a threshold from a paper; the absolute values depend on alloy, section and sampling method.
The honest cost side
None of this is free, and a proposal that ignores the cost will be rejected by whoever owns the yield number.
Contact pouring requires ladle geometry and operator discipline that many shops do not have today.
Trident-type gating with dual filters and a bubble trap can raise the metal poured per good casting to nearly twice that of a basic system. Yield falls; casting quality rises. That trade must be made explicitly, part by part.
Tapered sprues and reduced gate areas change wax tooling, and that is a capital item.
Low-pressure or counter-gravity filling solves the entrainment problem elegantly and costs a machine.
The sensible sequence is to fix the cheap things first: shorten the pouring cup, lower the filter, verify the sprue taper, stop interrupting the pour, and check shell venting. Those cost almost nothing and address the largest share of entrained damage. Expensive geometry changes come after the free ones have been measured.
Where the model has limits
This framework comes from the Campbell school of casting science and it is influential rather than universally accepted. Published commentary has challenged the difficulty of quantifying bifilms, the absolute claim that metals have no intrinsic crack nucleation mechanism, and the tendency to attribute complex failures to a single mechanism. It should be presented as a model that generates good priorities, not as settled consensus or as a standard.
It also weakens in specific regimes: very heavy sections where convection and segregation dominate, highly reactive high-temperature alloys whose surface films behave differently, and applications where dimensional precision rather than structural integrity is the governing requirement. In those cases the entrainment argument still applies, but its weight in the decision is lower.
And it cannot predict a rejection rate. It tells you what to stop doing and in what order. It does not replace trial pours, sectioning and statistics.
What to change first
If a plant recognises its own four-defect pattern in this article, the first month should look like this:
Stop interrupting the pour. Measure pour time per mould and hold it inside a fixed window.
Photograph and measure the pouring cup and the sprue top. Shorten the cup, verify the taper.
Move the filter down and check that metal covers its downstream face immediately.
Add one RPT sample per shift and start a bifilm index trend, even if the first twenty numbers mean nothing.
Do not change four parameters. Change one, in the direction of less entrainment, and measure the result on the same part number.
Most foundries that make this shift find that the four corrective actions they had been running for months were four attempts to treat a symptom that was created before the metal reached the cavity.
Need a second opinion on a specific defect pattern?
If you are dealing with porosity, shrinkage, hot tearing or a pressure-test failure that has resisted several rounds of parameter changes, send us the part drawing, the alloy, the gating layout and a photograph of the defect location. We will tell you whether the pattern looks like a single entrainment problem or something else, and what we would check first.
Start a quote or send a technical question — or write to [email protected].
Quality
Precision manufacturing for diverse industrial applications.
Services
Products
+8615912702921