How to Stress-Relieve Mold Steel Before and After Machining

A 1.2344 core block comes off the roughing mill on Friday at 28 HRC pre-hard condition. On Monday morning the CMM shows the cavity wall has moved 0.04 mm. Nothing touched it. No crane, no clamp, no cut. The steel moved by itself, because rough milling unbalanced residual stress that was locked inside the block since forging and rolling. That 0.04 mm is gone from your tolerance budget before finishing even starts, and on a mold holding ±0.002 mm on part geometry, it is 20 times the number you are allowed to play with.

Stress relief is the cheapest insurance in mold making. One furnace cycle costs 1 to 3 days and a few hundred dollars. One scrapped hardened core block costs 2 to 4 weeks and thousands. This guide covers the full process window: soak temperatures by steel grade, heating rates, soak time curves, the two relief cycles before hardening, EDM and grinding stress afterwards, measured dimensional change numbers, and the exact conditions where you can skip the furnace without gambling.

The Snapshot

• Stress relief soak runs 560-650°C for annealed tool steels and 480-560°C for already-hardened steels, always at least 30-50°C below the last tempering temperature.
• Standard soak rule: 2 hours minimum, plus 1 hour per 25 mm of section thickness, at a heating rate under 150°C per hour.
• Typical cavity movement from stress relief after roughing: 0.01-0.05 mm on blocks 200-400 mm long, per industry figures.
• Vacuum hardening plus tempering moves a block 0.02-0.10 mm, which is why it must happen before semi-finishing dimensions are cut.
• EDM leaves a recast white layer 5-50 µm thick with tensile stress; a 500-520°C secondary temper removes most of it before the layer turns into grinding cracks or in-service checking.

Why Mold Steel Moves After Machining

Every mold block is a stressed body in equilibrium. Forging, rolling, normalizing and the mill's own rough passes push stress into the steel, typically 200-400 MPa of residual stress in hot-rolled tool steel bars, per industry figures. As long as the cross section stays symmetric, the stresses balance each other and the block sits still.

Machining breaks that balance. When you pocket out 40 percent of a block's mass to form a cavity, you remove material that was carrying part of the internal load. The remaining steel redistributes the stress, and redistribution shows up as movement. The effect is not instant. Most of the shift happens within 24 to 72 hours after cutting, then it creeps for weeks.

Movement is time-dependent, not load-dependent. A roughed block sitting on a shelf distorts the same way as one in the machine. That is why shops that skip stress relief often discover the movement at the worst moment: during finishing, when the cutter follows coordinates that no longer match the steel. Typical relief of residual stress by heating: about 80-90 percent of internal stress relaxes during one proper soak cycle above 550°C.

The scale of the problem depends on the tolerance you chase. A prototype mold at ±0.05 mm rarely notices. A connector cavity mold for automotive OEMs like TE Connectivity or Amphenol runs ±0.005 mm on critical mold dimensions and ±0.002 mm on part geometry. At that level, a single skipped stress relief cycle consumes the entire budget many times over.

Where Residual Stress Comes From

You cannot manage stress you have not located. Mold steel carries residual stress from five distinct sources, and each one calls for a different remedy stage.

1. Upstream metallurgical stress. Solidification, forging and rolling leave macro-stress in the bar. Normalizing or annealing at the mill reduces it, but heavy sections can still carry 100-300 MPa. This is the stress your first relief cycle targets, and it is the reason pre-hardened vs through-hardened mold steel behaves so differently in the shop.

2. Quench and temper stress from the mill. Pre-hardened grades like P20 at 28-32 HRC were quenched and tempered once already. Their internal stress is lower but not zero, especially in blocks over 200 mm thick where the core cooled slower than the surface.

3. Machining stress. Rough milling with heavy depth of cut plastically deforms the surface layer 0.1-0.5 mm deep. Aggressive feeds on worn inserts add more. This layer holds fresh tensile stress that feeds distortion during later heating or even at room temperature.

4. EDM recast stress. Sinker and wire EDM melt and re-solidify a white layer 5-50 µm thick. That layer is hard, brittle and under high tensile stress, up to the yield point of the steel. It is a crack initiator, and it sits exactly where the mold surface must be perfect.

5. Grinding stress. Grinding hardened steel at high wheel pressure generates local heat. If the surface temperature spikes past the tempering temperature, you get re-hardened spots and fresh tensile stress. This is the classic source of grinding cracks on HRC 60+ surfaces.

Sources 1 through 3 are handled before or between machining stages. Sources 4 and 5 need a low-temperature secondary temper after the process, because you cannot stress-relieve away a white layer without tempering the whole hardened part. The routing table later in this guide assigns each source to its cycle.

Stress Relief Temperature by Steel Grade

Temperature selection follows one hard rule: stay below the steel's last tempering temperature, or you drop hardness. For annealed steels heading to hardening, you have freedom up to 650°C. For pre-hardened or through-hardened steels, the ceiling is 30-50°C below the temper that set the current hardness.

← swipe to scroll →

Table note: typical industry figures for atmospheric or vacuum stress relief cycles. Always confirm against your steel supplier's datasheet for the exact heat number.

Two traps hide in this table. First, pre-hardened P20 relief above 620°C partially re-tempers the steel and you lose hardness at the cavity surface, exactly where wear resistance matters. Second, on hardened hot-work steels used for die casting, going too hot on the secondary temper drops you from 50 HRC toward 44-46 HRC. That is fine for some die casting applications, but it must be a decision, not an accident.

Atmosphere matters too. Relief in air at 600°C grows scale 0.05-0.15 mm deep, which eats stock and can oxidize finished faces. Vacuum relief at 10-2 to 10-3 mbar avoids scale completely and is the standard route for hardened blocks and polished surfaces. DieStrike runs hardened cavity work through vacuum cycles on a 120+ machine floor that keeps heat treatment routing inside one quality system.

Soak Time, Heating Rate and Cooling Curves

A stress relief cycle has three phases, and shops get fired on the first and third more often than the soak itself.

Heating: under 150°C per hour. Fast heating creates a temperature gradient between surface and core. A 300 mm block heated at 300°C per hour can show a 60-80°C lag from skin to center. That gradient is new thermal stress, the exact thing you are trying to remove. Practical rule: heat at 100-150°C per hour up to 400°C, then you may accelerate to 200°C per hour through the last stretch, or simply stay slow. Equalizing soaks at 300°C and 450°C help heavy sections catch up before the final ramp.

Soak: 2 hours plus 1 hour per 25 mm. Soak time starts when the coldest point of the load reaches temperature, not when the furnace setpoint does. A 100 mm thick plate needs 2 + 4 = 6 hours. A 250 mm block needs 12 hours. Thermocouples attached to the thickest part of the load are the only honest way to time this. Stress relaxation is a creep process, and creep needs time at temperature: 90 percent of the achievable relief in plain carbon and low-alloy steels happens in the first 2-4 hours, with diminishing returns after.

Cooling: slow to 300°C, then air. Pulling a hot block out at 600°C and letting it cool in still air creates a fresh surface-to-core gradient of 50-100°C and re-introduces stress, sometimes 30-50 percent of what you just removed. The standard curve: cool in the furnace or under insulating blanket at 50-100°C per hour until the block is below 300°C, then air cool. Below 300°C the thermal gradients are small enough that air cooling is safe for sections under 200 mm.

One more variable: part stacking. Blocks touching each other in the furnace shadow heat flow and extend equalization time by 1-3 hours. Spacers of 25-50 mm between plates, load thermocouples on the center and the most shadowed part, and the curve behaves as planned. For a full spec sheet covering atmosphere, fixture and acceptance checks, see the guide on how to specify heat treatment for mold steel.

Stress Relieving After Rough Machining

The first relief cycle is the one that pays for itself. Rough machining removes the most material and unbalances the most stress, so it also triggers the largest single movement event of the whole mold build.

Timing: within 24 hours of roughing, at 60-70 percent of final stock removed. The sequence that works: saw cut the block, mill all six faces to 3-5 mm stock over final, machine rough cavity and cooling line bores, then straight to the furnace. Leave that 3-5 mm skin everywhere, because the relief cycle itself moves the block 0.01-0.05 mm and the skin is your correction allowance.

Why 24 hours? Because the stress redistribution after cutting is fastest in the first day. A block that waits a week on the shop floor has already moved most of its 0.05 mm in random directions, and the movement continues slowly after the relief cycle. Furnace relief takes the process to 80-90 percent completion in one controlled event instead of letting it dribble out during finishing passes.

For pre-hardened P20 at 28-32 HRC, the cycle runs 560-600°C, which keeps hardness loss under 1 HRC when the mill temper was above 600°C. For annealed H13 or D2 blocks destined for hardening, run 620-650°C and recover even more of the forging stress, since there is no hardness to protect yet.

What you gain, measured. Shops that CMM the same block before roughing, after roughing plus relief, and after finishing report that the post-relief movement during finishing drops to 0.005-0.015 mm. Without the cycle, finishing-stage movement of 0.03-0.08 mm on blocks 300 mm and larger is common, per industry figures. On a ±0.005 mm mold dimension that is the difference between one finishing pass and a scrap ticket.

The Second Relief After Semi-Finishing

One cycle is not always enough, and precision molds rarely stop at one. Semi-finishing cuts another 1-3 mm of stock, and every cut unbalances stress a second time. On blocks with deep pockets, long slender cores or asymmetric cavities, the second relief is what protects the final dimensions.

When the second cycle is mandatory:

• Section thickness ratio in the block exceeds 2:1, such as a thin rib wall next to a thick bolster.
• Pocket depth exceeds 60 percent of block thickness, leaving a thin floor that flexes with residual stress.
• Tolerance on the finished geometry is tighter than ±0.005 mm.
• The block is over 300 mm in any direction, where core stress gradients are largest.

The second cycle runs 20-30°C lower than the first and shorter: 550-580°C on pre-hardened steels, 600-620°C on annealed ones, with soak of 2 hours plus 1 hour per 25 mm. Lower temperature, because less stress remains and you want zero risk of softening before hardening or of touching the mill temper a second time.

Between semi-finishing and hardening, the block should sit at final stock plus 0.5-1.5 mm. That thin allowance absorbs the dimensional change of the quench itself, covered next. Routing a second relief adds 1-2 days to the schedule. On a 2-5 week injection mold lead time, that is 10-20 percent of the timeline spent to keep the finishing numbers honest.

Hardening and Tempering: The Biggest Move

Through-hardening a block to HRC 48-62 is the most violent dimensional event in mold making. Two mechanisms act at once. The austenite-to-martensite transformation expands the lattice, adding 0.1-0.4 percent volume, which on a 300 mm block means 0.03-0.12 mm of growth. Meanwhile thermal gradients during quench pull the block out of flat by 0.02-0.08 mm unless the cycle is slow and controlled.

Vacuum hardening with high-pressure gas quench (6-10 bar nitrogen) keeps total movement smaller than oil quench, typically 0.02-0.06 mm on blocks up to 300 mm, because gas quench cools uniformly in all directions. Oil quenching faster at the surface than the core can double those numbers on thick sections. For cavity blocks, inserts and cores destined for HRC 50+ surfaces, vacuum hardening plus at least double tempering is the industry default.

Sequence rule: harden before semi-finish dimensions, never after. The proven routing on a through-hardened cavity block is:

• Rough machine to 3-5 mm stock.
• First stress relief at 600-650°C.
• Semi-finish to 0.5-1.5 mm stock.
• Second stress relief at 550-580°C, optional but recommended over 300 mm sections.
• Vacuum harden to 48-54 HRC, temper twice at 550-600°C for H13 or per grade datasheet.
• Finish grind or hard mill to final size, then polish.

Tempering deserves one more note. Single tempering leaves retained austenite, which can transform later at room temperature and move the block 0.005-0.02 mm weeks after delivery. Double tempering, with the second temper 20-30°C below the first, converts most of that austenite up front. On hot-work steels at HRC 50, that is the difference between a mold that holds size in production and one that drifts during T1 sampling.

EDM White Layer and Grinding Stress

After hardening, the finishing routes each leave their own stress signature. Sinker EDM melts material with sparks at 8,000-12,000°C local temperature, then the dielectric quenches it instantly. The result is a recast white layer: re-solidified steel, untempered martensite plus retained austenite, 5-50 µm thick depending on the last pass energy, sitting in high tensile stress on a hardened surface.

That layer is a crack nucleation site. In injection molding it feeds corrosion pitting along the tensile network. In die casting, where the surface cycles between 200°C and 600°C every 40-60 seconds, it accelerates heat checking. A white layer over 10 µm on an HRC 52 cavity face should never survive to production.

The fix: remove or temper it, in that order of preference.

• Finish with low-energy EDM passes (roughing 12-20 A, finishing under 3 A) to keep the white layer under 10 µm, then remove the remaining layer by polishing or light grinding with 0.01-0.02 mm total stock.
• If the geometry blocks grinding, run a secondary temper at 500-520°C for 2-4 hours. This tempers the untempered martensite in the white layer and relaxes most of the tensile stress, though it cannot remove the brittle phase entirely.
• Shot peening or roller polishing puts the surface into compression and raises fatigue resistance, useful on die casting dies that see 100,000+ thermal cycles.

Grinding leaves a milder but real signature. Dull wheels, high pressure and low coolant flow push surface temperature past the tempering range, producing localized re-hardening and tensile stress that shows up as hairline grinding cracks after only 2,000-5,000 cycles in production. Keep wheel speed and feed moderate, flood coolant at 20-40 L/min on the contact zone, and spark out with light passes of 0.005 mm or less. A temper burn check with acid etch (per ASTM E2586 practice) catches it on critical faces before the mold ships. Choosing between the two finishing routes at all is its own decision, covered in hard milling vs EDM for mold finishing.

How Much Does a Cavity Actually Move?

Buyers ask one question before they pay for extra furnace cycles: how big is the movement really? The honest answer is a range, because distortion scales with block size, asymmetry, prior stress and how much material the cut removed. The figures below are typical industry measurements, useful for budgeting allowances.

• Small insert, 50-100 mm, pre-hardened P20, relief after roughing: 0.005-0.015 mm.
• Standard cavity block, 200-300 mm, P20 or H13, relief after roughing: 0.01-0.05 mm.
• Large plate, 400-600 mm, forged and flame-cut from plate: 0.05-0.15 mm, and occasionally more if the block was cut from the plate corner.
• Vacuum hardening plus double temper, 200-300 mm H13 block: 0.02-0.06 mm growth and out-of-flat combined.
• Wire EDM of a hardened core after all heat treatment: shape deviation from released stress of 0.005-0.02 mm on slender sections.

Two patterns repeat across these numbers. First, flame-cut or sheared blocks carry far more stress than sawn or milled ones, because the cutting torch adds a heat-affected zone with fresh residual stress 2-5 mm deep. Always machine away flame-cut surfaces before the relief cycle, or add a deeper stock allowance. Second, asymmetric geometry moves more than symmetric. A cavity offset to one side of the block acts like a bimetal strip once stress is released, bowing toward the heavier section by up to 0.05 mm on 300 mm spans.

Measure it, do not assume it. A CMM check at three gates (after rough relief, after hardening, after finishing) costs 2-4 hours of CMM time per gate and produces the data trail an IATF 16949 quality system expects. When you know your steel supplier's actual movement signature, you can shrink the stock allowances and reclaim cycle time on every future block. Reading the dimensional report correctly is its own skill, covered in how to read a CMM report.

When You Can Skip Stress Relief

Every furnace cycle adds 1-3 days and $100-500 per load. There are real cases where skipping is the correct engineering call, not just a budget call. Here is the decision list that DieStrike applies on DFM review.

Skip when all four are true:

• Tolerance on the machined feature is ±0.05 mm or looser, at least 20 times bigger than typical stress movement.
• The block is small, under 150 mm in every direction, and sawn from certified annealed or pre-hardened stock.
• Geometry is symmetric, with no deep pockets over 60 percent of thickness and no section ratio worse than 1.5:1.
• No hardening follows, so the quench distortion risk does not exist.

Prototype molds, soft tooling in aluminum 7075, ejector plates, spacer blocks and clamp plates usually land in this bucket. A mold base plate holding ±0.1 mm on bolt patterns does not need a 12-hour soak.

Never skip when any one is true:

• Through-hardening follows. The quench will amplify whatever stress is already in the block, and distorted hardened steel cannot be straightened without risk of cracking.
• Tolerance is ±0.01 mm or tighter anywhere on the block.
• The steel is flame-cut plate, welded (repair welded blocks carry extreme localized stress), or of unknown provenance.
• The part is a die casting die, forging die or any hot-work tool. Heat checking starts at stress concentrations, and residual stress is a free head start for cracks.
• The mold runs a glass-filled or corrosive resin where surface cracking means early polish failure.

The middle ground deserves a sentence: on pre-hardened P20 molds at ±0.02-0.05 mm with moderately deep pockets, some shops substitute a rough-cut, rest 48-72 hours, then re-measure and re-zero the work offsets before finishing. Natural aging recovers maybe 50-70 percent of what a furnace cycle does. It works on small symmetric blocks. On anything larger, it is a gamble with the finishing pass as the stake.

Stress Relief and the ±0.002 mm Budget

Precision connector and optical molds routinely hold ±0.002 mm on part-forming geometry. That number is not achieved by a great finishing pass. It is achieved by a stack of controlled contributions where stress movement is the largest single variable. Here is how a realistic budget looks on a hardened H13 insert at ±0.002 mm final:

• Movement left in the steel after proper double relief and vacuum hardening: 0.005-0.015 mm, removed entirely by the finishing allowance. Zero net cost if the routing is right.
• Finishing machine positioning and thermal growth: 0.001-0.002 mm, managed with cooled machines and in-process probing.
• Measurement uncertainty of the CMM at 20°C ± 1°C: 0.0005-0.001 mm.
• Polish stock removal: 0.003-0.008 mm if uncontrolled, near zero with timed and weighed polish routines.
• Assembly fit and thermal expansion in service at 80°C mold temperature: 0.002-0.004 mm on 50 mm features in hardened steel.

Add the squares and the statistical stack lands around 0.003-0.005 mm, which is why ±0.002 mm on a single feature requires every contributor disciplined, and why one unrelieved block at 0.05 mm of movement breaks the chain completely. This is exactly the arithmetic covered in how to calculate a mold tolerance stack-up, with stress relief as the entry you control at the furnace instead of at the drawing board.

Two budget tactics close this section. First, put the tightest tolerances on inserts, not on the big plate. A 50 mm insert moves less than a 400 mm plate and is cheaper to re-cut if it goes wrong. Second, freeze dimensions in stages: rough, relieve, semi-finish, relieve or harden, finish, measure. Each freeze converts unknown future movement into a known past number. DieStrike runs this staged routing as standard on ±0.002 mm work for connector OEMs including TE Connectivity, Amphenol and Luxshare, inside an IATF 16949 system that records every heat treatment lot.

FAQ

Q1. Can you stress relieve P20 without losing hardness?

Yes, if you stay at 560-600°C and the mill tempered the steel above 620°C. P20 at 28-32 HRC loses roughly 1 HRC or less at 580°C for 4 hours. Going to 650°C drops it 2-4 HRC and softens the cavity surface where wear resistance matters most.

Q2. Is stress relieving the same as tempering?

No. Tempering follows quenching and sets hardness by transforming martensite, usually at 500-650°C for hot-work steels. Stress relief relaxes residual stress without intentionally changing microstructure or hardness. They overlap in temperature range, which is why the rule exists: relief temperature must stay 30-50°C below the governing temper.

Q3. How long after rough machining should the relief cycle run?

Within 24 hours is best, and 72 hours is the practical limit. Movement is fastest right after cutting, so an early cycle captures the redistribution in one controlled event. Waiting two weeks lets the block move randomly, and the relief then adds its own shift on top.

Q4. Does wire EDM need a stress relief afterwards?

On hardened steel, no full relief is possible without touching hardness, but a secondary temper at 500-520°C for 2-4 hours removes most of the recast layer tensile stress. On die casting dies and high-cycle injection molds it is recommended practice. On soft tooling below 40 HRC it is usually unnecessary.

Q5. What does a stress relief cycle add to mold lead time?

One cycle costs 1-2 days including furnace queue, heating, soak and slow cooling. A full double-relief plus vacuum hardening routing adds 4-6 days total. DieStrike still delivers injection molds in 2-5 weeks with these cycles inside the schedule, because they are planned at DFM stage, not added as surprises.

The Bottom Line

Unrelieved mold steel moves 0.01-0.15 mm on its own schedule and eats tolerance budgets that finishing passes cannot give back. Route the cycles early: after roughing, optionally after semi-finishing, always before hardening, and temper away EDM white layers after. Skip only on small, symmetric, loose-tolerance blocks. On ±0.002 mm work, the furnace is part of the machine tool. If you want your steel routing reviewed before cutting starts, send the drawing through our contact page for DFM feedback within 24 hours.