Metal casting shrinkage is the volume a casting loses on its way from the ladle to room temperature, and it arrives in three stages that are compensated in two entirely different ways: the liquid and solidification stages are fed with metal from risers, and the solid stage is compensated by cutting the pattern oversize.
The number that goes on the pattern order is the second kind. It is a starting value, not a property of the alloy — the same grade contracts differently depending on how the mold and the cores restrain it, and that is why the first article, rather than a table, settles what your dimensions really are.
Three Stages of Casting Shrinkage
A casting contracts in three stages — liquid, solidification, and solid-state — and they differ by an order of magnitude, so the stage a number describes matters more than the number. For A216 WCB low-carbon cast steel, the split looks like this.
| Stage | Temperature range | Volume change |
|---|---|---|
| Liquid | pouring 1530 °C → liquidus 1495 °C | −0.4% |
| Solidification | liquidus 1495 °C → solidus 1430 °C | −3.6% |
| Solid | solidus 1430 °C → 25 °C ambient | −6.0% |
All three are volume percentages, which is where most published shrinkage data goes wrong on the way to a pattern shop. The 6.0% solid contraction is not a 6% pattern allowance; on a linear dimension it works out to 2.04%.
1. Liquid Shrinkage
Liquid shrinkage is the contraction of the melt between pouring temperature and the liquidus, and it is the one stage that costs you nothing. In steel it stays below 1% and is fed from the risers while everything is still liquid.
2. Solidification Shrinkage
Solidification shrinkage is the volume lost as the metal changes state, and it is the stage that produces defects rather than dimensional error — this is the volume the risers exist to replace. In cast steel it runs from 3.16% for pure iron up to about 5% in low-carbon steels, rising with carbon content as the solidus temperature drops.
Cast iron does not follow steel here. Graphite occupies more volume as a precipitate than it does dissolved in the melt, so the net change can be positive: a Bureau of Standards measurement series on gray iron found an expansion of 1.65% through solidification, while a low-carbon, low-silicon iron in the same series contracted 5.85%.
Carbon therefore works in opposite directions in the two families — dissolved carbon in steel deepens solidification contraction, precipitated graphite in iron offsets it. Ductile iron sits in between: graphite expansion cancels part of the contraction but not all of it, so the castings still need feeding unless mold rigidity, pouring temperature and graphite precipitation are deliberately balanced against each other.
3. Solid-State Shrinkage
Solid-state shrinkage is the stage the pattern compensates: contraction from the solidus down to room temperature, and the only stage that shows up as a dimension out of tolerance. Measured on sand-cast iron bars, linear contraction ran from 0.67% to 2.08%, varying inversely with silicon and carbon content and with section size — a heavy section contracts less than a thin one in the same iron.
It is also the stage where a calculated number stops being dependable. Metal is rarely free to contract: cores, mold walls and the casting’s own geometry restrain it, and where the resulting stress exceeds the metal’s resistance the casting yields instead of shrinking. The part can come out larger than predicted, not smaller.
Casting Shrinkage Formula
The pattern is cut oversize by the linear shrinkage allowance S:
Pattern dimension = casting dimension × (1 + S)
S is a linear fraction, so a 2.08% allowance enters as 0.0208. Pattern shops quote it as a shrink rule in inches per foot instead; divide by 12 to convert.
Published contraction data, on the other hand, is usually a volume change, and a volume change is never an allowance. Take the cube root first:
Linear ≈ 1 − (1 − volumetric)^(1/3)
A 6.0% volumetric solid contraction becomes 2.04% linear. Entering the 6% directly puts nearly three times the allowance into the tooling, which is an expensive way to find out the difference.
Once the first article exists, the allowance stops being an estimate. Measure the same features on the pattern and on the casting:
Actual shrinkage = (pattern dimension − casting dimension) ÷ casting dimension
Run it feature by feature rather than once for the part. A long unrestrained dimension and a dimension between two cored bosses will not return the same value, and the difference between them is the pattern correction.
Typical Casting Shrinkage Allowances
Published patternmaker’s allowances for cast steel run from 7/32 to 5/16 inch per foot — 1.82% to 2.60% — and which end you land on is decided by the grade, not by the fact that it is steel.
| Grade | Type | Allowance (in/ft) | Linear |
|---|---|---|---|
| HC (UNS S92605, 26–30 Cr ferritic) | heat resistant | 7/32 | 1.82% |
| HA (9Cr-1Mo, cast A217 C12) | heat resistant | 1/4 | 2.08% |
| CA15 (UNS J91150, 12 Cr martensitic) | corrosion resistant | 1/4 | 2.08% |
| CE30 (UNS J93423, austenitic) | corrosion resistant | 5/16 | 2.60% |
| CF3 (UNS J92500, 304L-type austenitic) | corrosion resistant | 5/16 | 2.60% |
The austenitic grades need roughly 43% more allowance than ferritic HC. Carrying one house number across a stainless conversion is enough by itself to put a long machined face out of stock.
Those per-grade values are published for the high-alloy series only. For plain carbon and low-alloy steel, ductile iron and gray iron there is no equivalent published allowance, and the shrink-rule tables that circulate for them are not traceable to a measurement — treat the contraction data as the anchor instead: 2.04% linear for low-carbon cast steel, and 0.67% to 2.08% for sand-cast irons depending on composition and section.
Treat every one of these as a starting value. SFSA’s own guidance for casting designers, from the body that publishes the per-grade allowances, calls the variability of patternmaker’s shrink “a very important design consideration” and recommends a first article to establish what the dimensions actually are before production. Correcting pattern equipment through one or more reverse-engineering steps is normal tooling practice, not a foundry failure.
What Causes Shrinkage Defects?
A shrinkage defect is solidification contraction that had no liquid metal path to it at the moment it happened. The last metal to freeze in a section pulls on whatever liquid it can reach; when the feeder is already solid or too far away, the missing volume stays behind as a cavity or as porosity spread through the dendrite network.
The most common cause is a feeder that cannot deliver enough. Only 15% to 30% of a riser’s volume is usable feed metal — the rest freezes in place — so a riser that looks generous on the layout can still empty before the casting is solid.
The second is a feeder that freezes too early. Its solidification time must exceed that of the section it feeds, conventionally by a modulus ratio of 1.2, and even at 1.2 soundness is not guaranteed, because interdendritic feeding fails on its own terms once the dendrite network closes. With no effective feeding at all, a steel section can hold as much as 7% porosity.
The third is geometry, and that one is on the drawing rather than in the foundry. Sudden expansions in section volume and heavy junctions create hot spots that freeze last and freeze isolated, and no riser layout fully rescues a part that solidifies away from its feeders.
Hot tearing is a separate failure: it is thermally induced stress tearing metal that is still weak, a restraint problem rather than a feeding one.
How to Reduce Casting Shrinkage Defects
Most of what reduces shrinkage porosity is decided before the foundry quotes — on the part drawing, and in the alloy and molding process you specify. Removing abrupt section changes and tapering sections toward the feeders lets the casting solidify progressively into metal that is still liquid, which eliminates hot spots outright instead of feeding them.
Everything after that is feeding distance: how far from a riser a section still comes out sound. It is not a property of the metal alone. Against a base case of AISI 1025 steel in a furan no-bake mold poured with 60 °C (140 °F) superheat, simulation of steel plates and bars puts the levers at these values.
| Change from the base case | Feeding distance |
|---|---|
| Green sand instead of furan no-bake | × 1.09 |
| Zircon sand | × 0.96 |
| Chromite sand | × 0.88 |
| Superheat raised to 120 °C | × 1.12 |
| Superheat dropped to 30 °C | × 0.94 |
| AISI 4340 instead of AISI 1025 | × 0.86 |
| End chill added | × 1.19 |
A chill and a chilling sand are opposite levers, which surprises most people. A chill placed against one section lengthens the end-effect feeding distance, but switching the whole mold to a high-conductivity medium such as chromite shortens it, and green sand — the least conductive of the four — feeds farther than no-bake.
Superheat behaves the same counterintuitive way. It adds liquid contraction, yet feeds farther, because liquid contraction is the cheap stage to feed and the thermal gradient is what actually decides how far the riser reaches.
Alloy content costs more than most engineers expect. AISI 4340 feeds 14% shorter than 1025 under identical conditions, while the other common low-alloy grades sit within 5% of plain carbon steel. Upgrading a housing from 1025 to 4340 for strength is not a drop-in change: the riser layout that produced sound castings before has to be reworked, and that is a fair question to put to the foundry at quote stage.
Shrinkage in Different Casting Processes
The process does not change how much the alloy contracts — it changes how repeatably that contraction can be predicted and held, which reaches you as the tolerance grade you are able to specify.
| Process condition | Tolerance grade, long production series |
|---|---|
| Sand molding, large castings, all processes capable | CT 12–14 |
| Sand molding, most casting types and processes | CT 10–12 |
| Sand molding, within capability but not for all casting types | CT 8–10 |
| Investment casting | CT 5–7 |
Those grades assume a long production run, where the pattern has been corrected against real castings. A short series runs about one grade looser.
Among the sand processes the ranking depends on casting size. Below 50 lbs, green sand is comparable to the others; above 200 lbs, no-bake typically holds tighter tolerances than green sand; shell molding holds the tightest of any sand process but is limited in the size of casting it can produce. For medium-to-large parts that puts no-bake ahead on dimensional grounds, and investment casting reaches CT 5–7 where the tolerance justifies the tooling.
Whatever the process, the machining allowance absorbs what the shrink allowance misses.