Self-Demolding Geometry in Flexible Silicone Molds

If you're designing a flexible RTV silicone mold, design the peel path before you design the cavity.

A Shore 10A–20A glove or two-part block can release a rigid casting that a steel tool would trap. That is self-demold: the *mold* stretches off a master or off the dump. A flexible silicone *part* stripping off steel is the opposite interface, and it is a different RFQ. Mixing them is how a glove mold gets drafted like an LSR insert, or a 10A food tray gets asked to invert like 70A.

This article is the RTV mold case: platinum or tin skin, often with a rigid jacket, peeled or inverted off a CNC, SLA, or metal master — or off chocolate, gummy, resin, or plaster. Geometry features overlap with part-tooling. The hardness sits on the mold side of the interface.

Peel path first

Draw the motion, not the pretty section.

1. Where does the rubber start to leave? (a lip, a corner, a cut edge)
2. What feature does it have to climb over next?
3. What is the extra width of that feature versus the opening, and how thick is the rubber that must stretch?
4. What is the first place a tear would start? (a knife corner, a thin fin of rubber, a 90° inside corner in the mold)

If you cannot sketch that sequence on a section, you don't have self-demold. You have hope.

If you ignore this: the shop “just uses 10A” and the tear still starts at the same square undercut because the path asked a 2 mm rubber wall to double its length at a sharp edge.

Common mix-up: copying a glove undercut into a stiff block

A brand wants a 12-cavity bear, 18 mm tall, a modest return flange under the chin, food-contact, about 2,000 dumps a month.

Self-demold that works: filleted bead, peel start at the open back, 10A–20A platinum skin, jacket so the sheet does not sag 0.5–1.5 mm. The rubber rolls off the dump.

Self-demold that pretends: the same flange copied into a 30A one-piece block with no peel edge, or into an LSR steel cavity as if Shore 70A will invert like 10A. First pull needs two people and a spatula.

Both quotes can say “flexible mold.” Ask which object stretches: the cavity, or a molded silicone *part*.

Negative draft you can still strip

RTV self-demold works when the mold is the elastomer. Typical pattern: Shore 10A–20A platinum skin, mix 1:1 or 10:1, vacuum degas, often with a rigid shell.

Features that usually self-release in a high-elongation, high-tear skin:

• Rounded beads and barbs with gentle entry angles
• Recessed logos with filleted letter sidewalls
• Modest return flanges if the flange thickness and the skin thickness are both honest
• Through-holes in the casting that are formed by pegs the rubber can stretch around

Features that pretend to self-release and then don't:

• Square, deep undercuts (a dovetail in the master)
• Thin rubber webs that must stretch over a wide lock (a 1 mm skin over an 8 mm bulb)
• Sharp internal corners in the mold (stress risers)
• Hard jackets (30A–40A) asked to invert like 10A
• One-piece blocks with no peel start (you need an edge you can get a fingernail under)

Fillet the lock. An R1 mm on a bead changes the strain concentration more than dropping 5A of hardness. Tear grades matter; a soft rubber with poor tear will split on the first climb.

Tin-cure versus platinum does not change the geometry rule. It changes shrink (~0.1–0.3% platinum after full cure; tin keeps moving) and inhibition. Don't pick tin so the mold “releases better.”

Holes, lips, and return flanges

Holes in the casting, formed in a one-piece RTV mold. The rubber fills the hole as a peg. On demold the peg has to stretch out. Draft the peg (the hole), fillet both faces, keep the hole short. Long, parallel holes in a stiff casting will tear the peg out of the mold and leave a rubber scrap in the dump.

Lips and wiper edges. A thin lip pointing against the pull is a blade. Give it a peel chamfer on the trailing side, or split. Self-demold lips want to be short and radiused.

Return flanges. Treat as a calculated stretch. If the flange returns 3 mm on a 40 mm opening, a 10A glove with a 4 mm skin may roll off. A 30A block of the same geometry will take the flange with it or tear. Shell the soft skin so you are rolling rubber, not lifting a floppy tub.

ISO 3302-1 can class a rubber *article* you cast. It does not tell you whether a flange will strip, and it is not another word for precision.

What to ask when you're buying

• Peel direction, named, with a section of the worst lock?
• One-piece open-back or a keyed split? Bail, negative-draft letter, two-sided logo almost always need a split.
• Skin thickness at the lock, and Shore 10A–20A with a jacket — or a 30A block?
• Coupon of the worst lock at real size before deleting the extra mold piece?
• Platinum or tin, and why (size hold vs dirty-master forgiveness)?
• Recast of the RTV skin included after a tear at the lock?

Worked example: 18 mm bear with a chin return

A 12-cavity 18 mm gummy bear, 4.0 g target, about 2,000 dumps a month, modest chin return. We sketched the peel from the open back over a filleted bead, not a square undercut. Platinum 10A–20A skin, mix 1:1, vacuum degas, rigid jacket. Ten demolds on a single-cavity coupon of the chin: no white stress mark, no nick. Then we copied to 12. A parallel LSR quote for a silicone toy of the same outline does not dump pectin.

Draft versus stretch: pick one primary trick

Self-demold is not “zero draft everywhere plus vibes.” A clean self-demold feature usually has:

• Draft on the long walls of the *master*
• Stretch reserved for one named lock
• Radii on that lock
• A peel start
• A rubber thickness that can take the strain
• Hardness chosen for that strain (10A skin vs 30A block)

Two named locks in two axes is a split. One named lock with a knife corner is a tear. Zero named locks and zero draft on a 40 mm-deep RTV block is a stuck master.

The test: print or machine a coupon of the worst lock at real size, real fillet, real hardness, real skin thickness. Ten demolds. If the tenth still has no white stress mark and no nick, you may skip the extra mold piece. If the first demold needs two people and a plastic spatula, you don't have self-demolding geometry. You have a fight.

When this article is the wrong quote

If you actually need vulcanized silicone parts that self-strip off steel, this RTV peel path is the wrong asset. Quote HCR compression or LSR injection with a cold-runner on a parts OEM. A tiny, shallow tray can skip the stretch math entirely. Geometry can override quantity: a full ID bulb at 80 dumps is still a core, not self-demold.

Related reading: custom RTV molds, keying two-part molds, radii for mold longevity, gummy trays.

Coupon and RFQ lines to copy

Copy this. A “flexible mold” price without these lines is not comparable.

• Self-demold on [feature] only; peel direction [axis]
• Mold Shore and skin thickness at the lock
• All other walls drafted on the master
• One-piece vs keyed split; jacket yes/no
• Coupon of the worst lock before deleting the split
• Platinum or tin; recast of the skin if the lock tears
• Casting material and monthly dumps

Send the STEP of the master, a section of the lock, and the material you will cast. We will mark whether the peel is honest or whether you need a two-part / plug / core.

We would not recommend a one-piece glove for this geometry

We would not recommend a one-piece open-back mold if the part has a bail, a letter with negative draft, or a two-sided logo. Raising Shore A first is usually the wrong fix for a stuck pull; fillet the lock or split. We would not copy a 10A glove undercut into LSR steel “so we can scale later.” We would not pick tin so the mold “releases better.”

RUUIPON Mold Structure Decision: Flat back? open-back. Else deep undercuts + peel? If the silicone cannot peel safely → two-part / removable plug / inner core. Large wet casting or soft Shore? mother mold. Tiny shallow tray? skip the jacket.

flowchart TD
  F[Flat back?] -->|yes| O[Open-back RTV]
  F -->|no| U[Deep undercuts?]
  U -->|silicone can peel safely| O
  U -->|cannot peel| S[Two-part / plug / inner core]
  W[Large wet casting or soft Shore?] -->|yes| M[Mother mold required]
  W -->|tiny shallow tray| O
  P[Named lock with knife corner?] -->|yes| R[Fillet or split; do not just drop Shore]

Typical shop values

Caption these as typical shop values, not a named lot.

| Item | Typical shop value | | --- | --- | | Cavity Shore | 10A–20A platinum RTV, jacket behind it | | Mix | Platinum 1:1 or 10:1 by grade; vacuum degas until the foam collapses | | Linear shrink | ~0.1–0.3% (platinum); tin keeps moving in storage | | Loaded-tray edge sag | About 0.5–1.5 mm on an unsupported large food tray | | Coupon | Ten demolds of the worst lock at real size and skin thickness | | 12-cavity 18 mm bear | ~2,000 dumps/month is a common first production band |

FAQ

Can I copy an RTV glove undercut into an LSR steel cavity? Usually not. In an RTV mold the rubber stretches off the master; in LSR the part stretches off steel. A lock that peels from Shore 10A–20A skin will shear a 60A–80A part. Sketch the peel path, then coupon the worst lock at production hardness before you delete a split. That LSR job is a different quote.

Does tin-cure RTV self-release better than platinum? Tin versus platinum does not change the geometry rule. It changes shrink and inhibition. Don't pick tin so the mold "releases better." Fillet the lock; a small radius on a bead cuts strain more than dropping a few points of hardness.

How do I know a return flange will actually strip? Treat it as calculated stretch: extra width versus the opening, skin thickness, and hardness. A 10A glove with honest wall may roll off a modest flange; a 30A block of the same shape will take the flange with it or tear. Shell the skin so you roll rubber, not lift a floppy tub.

Will a 30A production block invert like a 10A glove? No. Harder RTV strips like a stiff boot. Keep undercut height small relative to local wall and opening, round the master so it does not cut on strip, and put draft on everything that is not the named lock. A full ID bulb is cores and splits, not self-demold.

What coupon lets me skip an extra mold piece? Machine or print the worst lock at real size, fillet, hardness, and skin thickness, then do ten demolds. If the tenth still has no white stress mark and no nick, you may skip the split. If the first pull needs two people and a spatula, you do not have self-demolding geometry.