How to Design Magnesium Die Casting Dies: AZ91 and Tooling

A die machined from an aluminum shrinkage pull prints a 250 mm AZ91D housing roughly 0.5 mm oversized, and that single number is enough to scrap a 4-cavity tool's core set. The re-cut runs 3,000 to 8,000 USD plus 2 to 3 weeks of schedule, and it is almost always avoidable. Magnesium die casting dies fail on a short list of numbers that differ from aluminum: a lower shrinkage, a slower gate, a cover gas that leaves fluoride residue, and a die temperature window that has less room than the caster's spray system suggests. This guide sets out those numbers, alloy by alloy decision, for AZ91D hot-chamber tooling.

Every figure below is a typical industry value for the die, and each one needs confirming against the alloy certificate and the caster's machine sheet before a drawing is released.

The Snapshot

• Shrinkage of 0.0105 to 0.0115 mm/mm. AZ91 pulls less than Al-Si at 0.0130 to 0.0145, so a die cut to the aluminum shrinkage runs the magnesium part 0.3 to 0.4 percent oversized.
• Gate velocity of 20 to 40 m/s. Aluminum tolerates 45 to 70 m/s; magnesium turbulence makes MgO skins that end up as 100 to 300 micron inclusions in the casting.
• Die temperature of 180 to 260 °C. A hot spot above 300 °C starts thermal-fatigue cracking that shows at 40,000 to 70,000 shots.
• Tool cost of 45,000 to 120,000 USD and 40,000 to 150,000 shots of life. A 4-cavity AZ91 housing die lands there, with resurfacing every 25,000 to 40,000 shots at 3,000 to 8,000 USD.

Why Magnesium Runs Hot-Chamber

Magnesium is the alloy that gets hot-chamber treatment, and the reason is heat per kilogram, not temperature. AZ91D melts between 596 and 651 °C, against an Al-Si liquidus band of 540 to 620 °C, yet the magnesium melt carries far less heat into every shot. That low heat load is why the gooseneck and plunger tip of a hot-chamber machine survive the alloy, where the same hardware melts into an aluminum bath. The machine comparison itself sits in the hot chamber versus cold chamber guide.

The shot system then sets the die's duty. A typical magnesium fill opens in 3 to 8 milliseconds through the gate, the full casting cycle runs 20 to 60 seconds, and the die sees a heat swing per shot that is a fraction of the aluminum equivalent. On the same H13 tool steel, magnesium dies reach 40,000 to 150,000 shots, while the same geometry cast in aluminum dies at 40,000 to 80,000. The magnesium die is not tougher; the thermal and chemical load on it is simply smaller.

That smaller load removes one famous failure mode entirely. Magnesium carries no meaningful iron pickup, so the steel does not dissolve into the melt and the classic aluminum die soldering and washout are largely gone. Do not read this as a free pass. The failure mode does not disappear, it moves: oxide build-up from the melt film and cover-gas residue on the die face become the two things that eat a magnesium die over its life.

Shrinkage and the Tolerance Stack

The first number a designer must change when moving from an aluminum die to a magnesium die is the shrink rule. AZ91 pulls 0.0105 to 0.0115 mm/mm, while Al-Si alloys pull 0.0130 to 0.0145 mm/mm. Reuse the aluminum pull on a magnesium tool and every linear dimension lands 0.3 to 0.4 percent oversized, which is 0.6 to 0.8 mm of error across a 200 mm housing face.

The tolerance stack then has to be re-derived, not shifted by a constant. Magnesium's lower shrinkage is a gift: it lets you core closer to final dimensions, so machined-after-casting faces can shrink their allowance. But the die temperature window of 180 to 260 °C moves the pull inside its band, so pin the shrink value to the actual running die temperature at the trial, then split the difference across the drawing with GD&T so one number does not carry the whole risk.

Practically, we tool magnesium dies to a plus or minus 0.005 mm working tolerance under our IATF 16949 system, with plus or minus 0.002 mm on the EDM and jig-grind work that makes cores and slides. That die-side accuracy only pays off if the drawing itself states the magnesium pull, the nominal die temperature and the CMM check points on the first-article plan. A die cut to the wrong shrink rule cannot be rescued by any amount of precision machining.

Gating, Runners and Overflows

Gate velocity is where magnesium and aluminum diverge hardest. Aluminum hot- and cold-chamber practice runs gates at 45 to 70 m/s. Magnesium should be gated at 20 to 40 m/s, because magnesium is far more oxidation-sensitive and every bit of turbulence shreds the surface film into the shot. Those MgO skins do not re-melt: they end up as hard inclusions 100 to 300 micron across inside the casting, and they ruin machined threads and pressure-test seals.

Size the gate for the slower velocity, not the faster machine. Gate thickness runs 1.0 to 2.5 mm, the biscuit diameter sits at 40 to 70 mm, and the runner is profiled so the melt accelerates late and decelerates into the cavity instead of jetting. Then oversize the overflows deliberately: 5 to 15 percent of shot weight goes into them, because with magnesium the oxide film has to be trapped ahead of the cavity, not folded into it. The sizing logic carries over directly from the overflow and weir vent design guide.

Venting follows the same oxidation logic. Keep vent land depth at 0.05 to 0.15 mm and widen total vent area before you deepen a single vent. Magnesium oxide freezes fast, so a vent geometry that would work on aluminum plugs with solidified film and causes a burn instead of a vacuum defect. When a vent question turns into a shot question, the answer is usually more area at the same thin land, not a deeper channel.

Finally, gate the shot system itself with discipline. Keep shot size at 60 to 80 percent of pot capacity and plunger-tip speed at 3 to 6 m/s during cavity fill. Most hot-chamber machines carry no intensifier at all, so packing pressure is not available as a rescue. Dimensional accuracy has to come from a stiff die and zero flash, which is a tooling decision made at the drawing stage, not a machine dial turned during production.

SF6 Cover Gas and Fluoride Residues

Magnesium melt is protected from burning by a cover gas, typically SF6 at 0.3 to 0.8 percent in air, with CO2 or SO2-based mixtures as alternatives. That gas is not innocent toward the tool. The fluorinated chemistry leaves sulfate and fluoride residues on every die surface it touches, and those residues are the reason a magnesium die has a maintenance cadence that an aluminum die does not.

Here is the failure sequence with its numbers. A polished core left unprotected accumulates fluoride residue until the chemistry eats into the steel: pitting appears on a mirror core within 5,000 to 20,000 shots if the tool is not cleaned and re-coated on schedule. A white, chalky film on the die face is not a clean tool. On a magnesium die, that white surface is fluoride attack starting, and a buyer who reads it as corrosion-free is looking at the first 5,000 shots of the end of a core face.

So the coating decision is a chemical-resistance decision before it is a release decision. The die coating has to survive fluoride attack, which is why nitrided or PVD-treated core faces outlive bare polished steel on magnesium work. And one shop shortcut has to be called out as forbidden: aluminum-style weld repair consumables of the ER5356 type are not used on magnesium dies, because their chemistry does not match the H13 face. Repair goes the other way, with same-tool-steel laser build-up, and the part is re-hardened and re-coated before it returns to the caster's machine.

Thermal Management and Die Temperature

Magnesium solidifies across only about a 50 °C range and conducts heat less readily than aluminum, so the die runs cooler but the temperature window still decides part quality. Hold the die at 180 to 260 °C with a closed-loop spray system, with thermocouples in the core, cavity and slide zones rather than one probe in the die shoe. A caster running a cold die makes cold shuts; one running hot buys a shorter life.

The upper limit is the expensive one. Local hot spots above 300 °C trigger a die solder of a different kind on magnesium, and they seed thermal-fatigue cracks that show as heat checking at 40,000 to 70,000 shots. That is a quarter to a half of the 40,000 to 150,000 shot life window, so a hot boss can finish a tool mid-contract. For the aluminum comparison on spray and die temperature discipline, the mechanism sits in the die soldering prevention guide.

The tool fix is local conductivity, not more spray. Thick bosses, thread cores and junction masses pull heat slowly through magnesium's lower thermal conductivity, so put high-conductivity inserts where the heat hides. Beryllium-copper inserts on thick bosses pull heat out of the zone at several times the rate of H13, and conformal cooling lines drilled or printed near the hot mass hold the local die temperature inside the 180 to 260 °C band without flooding the rest of the face with spray. Bake the insert choice into the die design; a hot spot found at 50,000 shots costs a re-made core, not a nozzle change.

Die Steel and Surface Treatment

The base material is familiar even if the load is not. Magnesium dies run on H13, grade 1.2344, tempered to HRC 44 to 48 on the working face. That band matters: too hard and the face loses its resistance to thermal fatigue, too soft and slide ways and shut-offs wear early. The alloy-to-steel logic is laid out in the die casting mold steel selection guide, and magnesium simply sits at the low-heat end of every one of those charts.

Surface treatment then does the chemical job that heat would not. Nitriding or a PVD coating on the core face is the fluoride shield, and it should be specified on every magnesium core, cavity and slide that sees the cover-gas atmosphere. One rule has no exceptions: no chrome plating over 0.02 mm on a magnesium die face. Thick chrome spalls under thermal cycling, and the MgO film packs under the lifted flake, scratching every subsequent casting from a pocket you cannot see through the oxide dust.

The pre-hardened versus through-hardened trade-off, and the vacuum heat-treat spec that protects grain size, are purchase-order decisions most buyers never write down. Under our IATF 16949 system the heat-treat report is part of the tool package, with the HRC 44 to 48 band verified on the insert body rather than a witness coupon alone. If your current magnesium tool package has no hardness record, ask for it before the next resurfacing bill arrives.

Magnesium vs Aluminum Tooling Parameters

The table below collects the tooling numbers that change when an aluminum die becomes a magnesium die. Read it as a translation layer: nothing about H13 or the die shoe changes, while almost everything about pull, gate, temperature and maintenance does.

← swipe to scroll →

Table note: bands are typical industry values for AZ91D versus common Al-Si die casting alloys on hot- or cold-chamber tooling. Confirm melt temperature, pull and tonnage against the alloy certificate and the caster's machine datasheet before releasing a die drawing.

Part Design Limits from the Alloy

Some die decisions are forced by the part, and AZ91D writes a stricter part rulebook than aluminum. Minimum wall on magnesium is 2.0 to 2.5 mm, where aluminum thin-wall housings routinely cast at 1.5 mm. The hot-chamber shot and the slower gate velocity cannot reliably fill a 1.5 mm magnesium wall at 20 to 40 m/s without a freeze at the flow tail. Design the wall at 2.0 mm minimum, then let the die worry about nothing smaller.

Rib height stays under 6 times the nominal wall, and the rib root should not exceed 60 to 70 percent of the wall it attaches to, or the rib head sinks and the sink line telegraphs to the cosmetic face. Deep blind pockets need draft and an escape for the slide; a blind pocket with no draft is a die that cannot be made without a lifters-and-slides budget nobody priced. When a magnesium housing needs more stiffness, thicken the section locally instead of adding ribs: AZ91D creeps at a 150 °C service temperature, so a long thin rib carries a creep risk that a chunky local pad does not.

Threaded features deserve one more sentence because magnesium housings are full of them. Cast threads in AZ91D work, but the MgO inclusion problem makes tapped holes the safer default on anything above M6, and inserts molded into the die need a 0.2 to 0.3 mm retention clearance controlled by EDM on the core pocket. Every wall, rib and thread decision that moves a magnesium die into aluminum territory comes back as a trial-day fix, at caster downtime rates.

Defect Triage Mapped to Tool Changes

When a magnesium casting goes wrong at the caster's machine, the fix belongs in the die, and each defect maps to a tool change with a number attached. This is the triage list we work from when a trial part comes back from the partner foundry with a defect report.

MgO stringers across the part mean the melt folded oxide instead of trapping it: enlarge total vent area by 30 to 50 percent and slow the gate, which usually means a thicker gate section that drops velocity back toward the 20 to 40 m/s band. Cold shuts mean the melt froze before the flow met: raise die temperature by 20 °C and thicken the gate, because a hot-chamber machine without an intensifier cannot pack a cold joint shut after the fact. Porosity traced to hydrogen calls for shot-sleeve temperature discipline at the caster and a corrected gate-to-biscuit ratio on the tool, not a new machine.

Flash that starts small and then sticks is a stiffness signal, not a wear signal. Magnesium needs only 3 to 8 MN of clamp per 1,000 cm² of projected area against aluminum's 5 to 12, so casters happily run magnesium on lighter machines with lighter die shoes. That is exactly where the problem lives: the cavity pressure of a 20 to 40 m/s fill pushes the shoe apart by elastic deflection, the flash line grows, the flash wedges the shut-offs and the die starts sticking along the parting line. The tool fix is thicker shoes, side-locks sized to the projected area and shut-off land pressure kept inside the coating spec.

Tool Cost and Die Life

A 4-cavity AZ91 housing die prices at 45,000 to 120,000 USD and delivers 40,000 to 150,000 shots on H13. Those brackets are wide because the cost and the life are set by the same three decisions: shrink-rule discipline at the drawing, coating chemistry against fluoride, and the resurfacing cadence the tool is designed to accept.

The cadence is the number buyers miss. A magnesium die should be pulled, cleaned of fluoride residue, inspected and re-coated every 25,000 to 40,000 shots, and that resurfacing runs 3,000 to 8,000 USD per event. Skip it and the 5,000 to 20,000 shot pitting clock on the core face runs out inside your first production year. Budget a housing die at, say, 120,000 shots over its contract life and you are funding three to four resurfacings, 9,000 to 32,000 USD of maintenance, whether it is scheduled or emergency.

So the real part cost is set by the maintenance plan, not the initial steel. When we make the die, it ships with the coating spec, the shrink rule used, the resurfacing interval and the re-coate vendor path written into the tool package, the same way an IATF 16949 program file would demand. Lead times on that tool run 2 to 5 weeks for the standard components in 3 to 7 days, which is the schedule a prototype-to-first-article loop on a magnesium housing normally fits.

Magnesium Die Design Checklist

Run this list against the die drawing before a magnesium tool is released to the shop floor. Every line is a number, not an intention.

• Shrink rule restated. Die carries 0.0105 to 0.0115 mm/mm magnesium pull, not the aluminum 0.0130 to 0.0145 value, with the stack re-derived against the nominal 180 to 260 °C die window.
• Gate at 20 to 40 m/s. Gate thickness 1.0 to 2.5 mm, biscuit 40 to 70 mm, shot at 60 to 80 percent of pot capacity, plunger-tip speed 3 to 6 m/s.
• Oxide budget. Overflows at 5 to 15 percent of shot weight, vent lands 0.05 to 0.15 mm, and enough total vent area for a 30 to 50 percent growth if MgO stringers appear.
• Thermal plan. Closed-loop spray targets 180 to 260 °C, no local zone above 300 °C, BeCu or conformal-cooled inserts on every thick boss.
• Steel and surface. H13 1.2344 at HRC 44 to 48, nitrided or PVD core faces, chrome nowhere above 0.02 mm, no ER5356-type weld consumables, laser build-up with same-tool steel only.
• Part limits. Wall 2.0 to 2.5 mm minimum, ribs under 6 times wall, local thickening instead of ribs near 150 °C creep service.
• Clamp and stiffness. 3 to 8 MN per 1,000 cm² projected verified against shoe deflection, side-locks sized to area, shut-off lands inside coating thickness.
• Life file. Resurfacing every 25,000 to 40,000 shots at 3,000 to 8,000 USD written into the tool package, with target life of 40,000 to 150,000 shots stated on the quote.

The Bottom Line

A magnesium die is an aluminum die that has been told the truth about its alloy. The pull drops to 0.0105 to 0.0115 mm/mm, the gate slows to 20 to 40 m/s, the die cools to a 180 to 260 °C window, the attack changes from iron soldering to fluoride residue, and the maintenance clock resets to a 25,000 to 40,000 shot resurfacing beat. Get those five numbers onto the drawing and the 45,000 to 120,000 USD tool reaches its 40,000 to 150,000 shots; ignore them and the die is oversized, inclusion-filled or pitted long before the contract ends.

DieStrike builds magnesium die casting dies under an IATF 16949 quality system with plus or minus 0.005 mm mold tolerance, plus or minus 0.002 mm on EDM and jig grinding, HRC 62 capability on components, and tool lead times of 2 to 5 weeks across a shop of 120+ machines, with tier-1 references including TE, Amphenol, Luxshare and Dongshan. Send the part print and the alloy spec through our die casting mold service page, and the DFM review comes back with the shrink rule and gate numbers already changed for magnesium.