Overmolding Silicone on Glass: Primer Requirements and Thermal Shock Control

Glass is not PC. It does not melt, it does not take a knurl the same way, and it will crack in a hot LSR tool if you treat it like a plastic first shot.

Overmolding silicone onto glass (lenses, sight glasses, lighting covers, cookware lids, diagnostic windows, some appliance knobs) is a bonding-and-survival problem. The rubber can be LSR injection or HCR transfer/compression. The glass is an insert. Self-bonding grades that hang onto polycarbonate are not a substitute for a glass primer, and a primer that passed a room-temperature peel will still fail if the insert starred on the first heat-up. Design the chemistry and the thermal path as one system.

Room-temperature platinum RTV painted onto a bottle is a different, weaker process. Do not write bench adhesion into an LSR overmold RFQ.

Why glass almost always wants a primer

Cured silicone sits at low surface energy. Clean glass is silica with hydroxyls. That sounds like a perfect silane job, and it is, if you actually run the silane job.

A typical primer for this stack is a silane coupling system (alkoxy silanes that hydrolyze, bond to Si-OH on the glass, and present groups the silicone network can react with during vulcanization). Some suppliers sell glass-specific or multi-substrate primers; some self-bonding LSR grades list glass. Treat the datasheet as a starting claim. Qualify on *your* glass (soda-lime, borosilicate, chemically strengthened, coated, printed) with *your* clean, *your* primer lot, and *your* cure.

What kills the bond before thermal shock ever shows up:

• Fingerprints, cutting fluid, mold release, and detergent film. Glass looks clean and is not.
• Primer applied too thick, or flashed off too little, so solvent or uncured primer foams at the interface.
• Primer applied too early and then parked in a dusty rack. The coupling surface is not a warehouse coating.
• Printing, metallization, AR coatings, and oleophobic topcoats on the bond land. You are bonding to the coating, not to glass.
• Moisture on the glass going into a hot tool, which can steam the interface.

Plasma or flame can raise surface energy as a prep. On glass, cleanliness plus the right silane still does the chemical work. Mechanical locks still help: a wrap around the edge, a groove the rubber can freeze into, a through-feature if the design allows. A 0.4 mm LSR film on a glossy plateau with a square peel edge is asking the primer to be a structural adhesive. It is not.

If you ignore the coating on a “glass” insert, the defect is a clean peel at the ink, not at the silica.

Thermal shock is a CTE problem, not a mystery crack

Approximate orders of magnitude, not a lab certificate: many silicones sit around 200-300 × 10⁻⁶ /K linear expansion. Soda-lime glass is around 9 × 10⁻⁶ /K. Borosilicate is lower still. The rubber wants to grow and shrink. The glass barely moves.

Two thermal events matter:

1. The molding cycle. LSR cavities run hot. Dropping a room-temperature glass disk onto a 180 °C land, then clamping, then injecting, puts a gradient through the glass before the rubber is even there. Edge chips and residual tempering stress concentrate that gradient into a crack. HCR compression is slower but still hot.

2. Service. Dishwasher, oven door, outdoor night-to-day, autoclave, or a lighting fixture that goes from cold soak to lamp-on. The silicone is now a bonded ring. Every temperature swing shears the interface. If the bond is only a face film, it peels. If the glass was already micro-cracked from molding, service finishes it.

Control is boring and specific:

• Choose the glass for the thermal duty. Tempered soda-lime, annealed, borosilicate, and chemically strengthened glass do not fail the same way. Name the type on the RFQ. Edge quality (ground, seamed, fire-polished) matters more than people who design in CAD think.
• Preheat the insert so it does not see the full ΔT in one slap. Preheat is also good for bond: a cold insert steals cure from the interface. Preheat is not a license to throw cold glass on a hot core.
• Ramp and support. Nest the glass so clamp is on a designed land, not a point contact. Do not let LSR pack under an unsupported pane and bend it.
• Bond geometry that can flex. A fillet and a wrap beat a knife-edge flash of rubber on a sharp glass corner. The corner is a crack starter in the glass and a peel starter in the rubber.
• Qualify the stack with the actual thermal cycle (and humidity, if the product will see it). A dry 90° peel on a plaque is not a dishwasher or autoclave result.

ISO 10993 and FDA 21 CFR 177.2600 speak to the *silicone grade* and intended contact, not to the glass bond. USP Class VI on the rubber does not mean the primer is approved for the same contact. Name the primer in the biocompatibility file if this is a medical window.

Process knobs that are not folklore

• No release on the bond land. Release on the cavity face if you must.
• Cure at the interface. The sprue looking done is not the glass looking bonded. Cold glass, short cycle, weak ring.
• Post-cure if the grade and the primer system expect it. Post-cure can improve some bonds and can also stress a poorly supported insert. Try it on the real geometry.
• Shut-offs. LSR will flash under a glass edge through a gap on the order of 0.005 mm. Flash under the pane is a peel-start and a crack-start. Steel must pinch a designed land, not smash the glass.
• HCR versus LSR. HCR overmold onto glass exists (transfer or compression). Cycles are longer; primer discipline is the same. Do not assume an LSR self-bonding grade applies to a millable HCR.

RTV overmold (pour or brush platinum onto glass at room temperature) avoids the hot-tool shock. Bond strength and heat aging will not match a properly primed, heat-cured production joint. Prototype the *look* in RTV if you must. Prototype the *bond and the crack* in the production chemistry and the production heat.

RFQ notes for glass overmolds

Send the glass type and coating stack, the exact bond land, whether a wrap or groove is allowed, the primer policy (supplier system versus “shop standard”), insert preheat, and the thermal-shock and peel tests that match the product. If the window is printed, send the print that sits on the bond land. If you ignore that print, the shop will bond to ink and you will learn about it in the field.

FAQ

Can self-bonding LSR skip primer on glass the way it can on some PC grades?

Sometimes a grade lists glass. Treat that as a claim to verify on your glass, coating, and clean. Most production glass overmolds still run a silane primer because coatings, fingerprints, and peel edges eat primerless windows. Qualify peel after the real thermal cycle, not only a dry tug on a plaque.

Will preheating the glass stop thermal shock cracking?

It reduces the slap from room temperature onto a 160-200 °C tool, which is the usual molding crack. It does not fix a chipped edge, a point-loaded nest, or service shock later. Preheat and a designed support land belong together. Preheated glass with a smashed corner still stars.

Does borosilicate always survive LSR overmolding better than soda-lime?

Borosilicate’s lower expansion helps some thermal gradients, but thickness, edge quality, residual stress, and how the rubber wraps the pane decide the crack. Name the glass and the duty. Do not swap types in production after you qualified one.

How do I test the bond if a peel tab is not in the product?

Mold a witness tab on a process coupon with the same glass, primer, and cycle, and still run thermal shock on the real part. A tab that peels cohesively in the rubber is the usual goal. Adhesive peel at the glass after a dishwasher or autoclave cycle is a fail even if the day-one tug felt strong.

Is the primer part of a medical biocompatibility file?

If the primer can contact the patient or a fluid path, it belongs in the material list next to the LSR grade (ISO 10993, and any USP Class VI claims you actually need). A biocompatible rubber with an unlisted solvent-borne primer is not a complete stack. Ask the primer supplier what they will support.