Splitting Complex Asymmetrical Master Models for 3-Part Silicone Molds

A symmetric bottle wants a midplane. An asymmetrical master (a sculpted housing, a medical grip, a figurine, a duct with one twisted branch) does not have that plane. Forcing a 2-part split puts an undercut on the wrong half and then everyone argues about “silicone stretches.”

A 3-part (sometimes 3-piece, 3-plate) silicone mold is a parting *network*: two splits that let three rigid or semi-rigid pieces leave in three directions, or two jacket pieces plus a core. The job is to choose those directions from the undercut map, not from the pretty view in the render.

Two industries share this headache and must not share one recipe:

• RTV mold making (platinum or tin jackets poured on a master, often with a mother mold). Three jacket pieces, or two jackets plus a core/mandrel.
• Production LSR/HCR tools (steel) where a third plate, a side core, or a split cavity does the same geometric job at a different cost.

This article starts with the master split for RTV and steel-intent DFM, then notes what changes when the third piece becomes a LIM action.

Draw the undercut map before you draw the clay wall

On the CAD (or on a printed master), color faces by which way they can unhook:

• Direction A: main open (usually the largest projected area).
• Direction B: the second open (often 90° or 180° from A).
• Direction C: cores, lost plugs, or a third jacket for a local barb.

Asymmetry means A is not a plane of symmetry. The split can be a stepped, wrapped, or joggled surface. That is legal. A joggle that leaks is not.

Rules that keep you out of a 4-part accident:

• Every undercut needs a home. If a hook cannot unhook in A, B, or C, you need a fourth piece, a lost core, or a geometry change. “Silicone will stretch” is a local permission for shallow barbs in soft RTV, not a global waiver.
• Splits should sit on low-cosmetic, easy-to-flash-control faces. On a product, hide the witness in a radius or a texture break. On a master for RTV, the split prints on every casting.
• Avoid splits that run across a sealing lip or an optical window.
• Keep the three pieces loadable in one obvious order. If the only assembly sequence requires three hands and a prayer, the shop will mis-key it.

If you ignore a side hook on an otherwise 2-part body, the defect is a torn jacket on pull one, or a steel tool quote that suddenly includes a lifter you did not budget.

How RTV 3-part jackets are actually split

Shop practice, not sculpture class:

1. Keys. Cones, sockets, or irregular jogs so the three pieces cannot assemble 180° wrong. Asymmetrical masters still need *more* keys, because there is no visual symmetry to save you.
2. Stable walls. Each jacket needs thickness and a rigid mother (plaster, fiberglass, printed shell) if the piece is large. A floppy third piece is a flash machine.
3. Pour / vent path that still fills after the split. Air traps at the new corners. Last-fill vents belong at those corners.
4. Registration to the core if there is a hollow. A 3-part OD with a wandering mandrel is a wall-thickness lottery (core shift’s RTV cousin).
5. Tin versus platinum. Tin is more forgiving of inhibition from clays and some printed masters; platinum copies better and lives longer. Neither fixes a bad split. Sulfur clay against platinum is a sticky jacket, not a split problem.

Shims and clay walls that follow a spline on the master are how you build a non-planar split. Do it once on a signed CAD plane if this master will later become LSR steel. A romantic freehand split on the prototype is a change order on the production tool.

When the third part should be a core, not a jacket

Hollows and deep draws: two OD jackets plus a mandrel is a 3-part mold. Trying to split the OD into three jackets because the ID scared you is usually backwards. Pull the core along its axis; split the OD where the exterior undercuts live.

Lost cores (soluble, low-melt, 3D-printed breakaways) are a fourth strategy when no third rigid piece can leave. They are not 3-part molds. Do not write “3-part” on an RFQ that needs a lost core.

Steel / LSR analog

In LIM, “3-part” often means a third plate, a stripper, or a side action. Asymmetrical products still need the same undercut map. Differences:

• LSR will flash the extra splits (0.005 mm class). Every new split is a land to maintain.
• Heat: three pieces at 180 °C must still meet. Alignment at operating temperature.
• Automation: the robot must see a consistent open. A 3-piece RTV peel does not teach a 3-plate LSR sequence unless you designed the production splits first.

If the program will go to steel, split the master on the production parting network. Prototype jackets that peel in impossible directions train no one.

RFQ notes for 3-part splits

Process: 3-part RTV from master and/or steel with third action. Send CAD with proposed open directions, faces that cannot carry a witness, undercuts you refuse to change, and whether the RTV split must equal the LSR split. If you ignore the production split, you will love the prototype and hate the tool quote.

FAQ

When is a 3-part split better than stretching silicone off a 2-part undercut?

When the undercut is deep, the rubber is hard, the wall is thin, or the jacket (or steel land) will tear or flash. Soft platinum RTV can peel shallow barbs in two parts. A twisted asymmetrical hook usually wants a third direction. Stretch is a local permission, not a split strategy.

Can the three parting surfaces be non-planar on an asymmetrical master?

Yes, and they often must be. They still need keys, a pour path, and a way to clamp without a gap. A spline split that leaks is worse than an ugly planar jog that shuts. For later LSR steel, prefer splits a toolroom can grind.

Should I split the RTV prototype the same way as the future LSR tool?

If the program is going to steel, yes, as far as the product allows. A glove-mold peel that no press can copy trains the wrong DFM. Use the extra RTV freedom only for features you have already agreed will change before LIM.

Where do 3-part RTV molds usually flash?

At keys that do not seat, at joggles that were not rigid in the mother mold, and along a clay wall that moved. Asymmetry hides a 0.5 mm mismatch that a symmetric bottle would show. More keys, a rigid shell, and a written assembly order beat another clamp strap.

Is a two-jacket-plus-core mold counted as 3-part?

Yes in the sense of three pieces and three leave directions. Name it that way on the RFQ (OD split + mandrel) so the shop does not bid three OD jackets and no core. The undercut map should show the ID leaving along the core axis.