Silicone Molding Defects: A Field Guide to Bubbles, Shorts, Sticking and Under-Cure

Of the four defect families, bubbles and under-cure come mainly from cure parameters, short shots come mainly from charge and tooling, and sticking comes mainly from mold condition and release agent. Classify before you adjust — get the order wrong and you make it worse: treating a short shot as under-cure and adding time only embrittles the part; treating under-cure as a short shot and adding material only thickens the flash.

This guide is for the moment you open a first article and find a problem. It answers “what do I check first,” not “what number goes in the process card.”

The 30-Second Triage Table

Bubbles and Porosity

Three kinds of bubbles, three different causes

Treating all three as one problem is the main reason bubble issues never get resolved.

Trapped gas in the core of a thick section has a different cause. Gas entrapment from abrupt wall-thickness change is a design problem, not covered here — see why the thick-thin junction always traps gas.

The mechanism: why the gas cannot escape

Gas in compression molding has three sources: air in the cavity, moisture and low-molecular volatiles carried by the compound, and byproducts generated by the crosslinking reaction. All three must leave while the compound still flows. Once the skin crosslinks, the escape route closes.

The third source deserves separate mention. Peroxide systems generate acidic and small-molecule byproducts during cure; Fan and co-authors describe them as “strongly acidic byproducts produced during vulcanization” (Acta Materiae Compositae Sinica, 2024, 41(3): 1259–1269, DOI: 10.13801/j.cnki.fhclxb.20230814.004). That gas is inherent to the chemistry — no amount of charging accuracy avoids it. Only venting removes it.

Lei, Shi, Luo and co-authors studied the effects of cure time, temperature-ramp interval, demolding temperature and venting amount on part quality in Preparation of Thick Functional Products of Silicone Rubber, Silicone Material (有机硅材料), 2012, 26(3): 164–169. Their conclusion: a staged temperature ramp, extended cure time, and pressure-holding cooldown produce dense, defect-free thick sections. Reported process windows include staging increments of 15 °C, ramp intervals of 10–20 min, average heating rate 0.75–1 °C/min, demolding temperature selected by section thickness from room temperature up to 80 °C, and a mold design using large clearances with small flash grooves.

The same work contains a counter-intuitive finding that is easy to miss: as the venting amount increases, large parts show a tendency toward core cracking (the reported preferred venting window is 0.20%–0.27%). More venting is not better — too little traps gas, too much can split the core.

The skin curing first is the number one reason bubbles get locked in

Heat travels from the mold surface inward, so the skin always reaches crosslinking temperature before the core. Ramp too fast and the skin completes crosslinking before core gas has migrated out, forming a dense shell — and the gas has nowhere left to go.

That is why the literature prescribes staged ramping rather than a fast push: splitting the rise into steps with dwell time in between lets the whole mass approach the crosslinking range more synchronously, giving gas a window to escape.

Diagnosing it is straightforward: cut the blistered part open. Voids near the skin → skin-lock. Voids concentrated in the core → heat transfer or venting-amount problem.

Three levers you can actually pull

• Vent count and timing. The first vent must precede skinning. Count depends on section thickness and compound flow; it is not a fixed number.
• Compound moisture and storage. Compound that has absorbed moisture adds a gas source; storage after opening must be controlled.
• Mold venting design. Vent groove location and depth determine where gas goes. This belongs to the tool, not the parameters — tuning parameters only masks it.

Short Shots

Three variables: charge weight, placement, flow

A short shot looks like “not enough material,” but the real cause is usually a mis-ranking of these three:

• Charge weight. The most obvious, and the easiest to overdo — overshoot and you get flash.
• Placement. Charge placed to one side fills the far end last, so corners go short. Multi-cavity tools show this clearly.
• Flow. Tied to compound grade, storage time, and whether it has been re-milled. The same batch flows differently after sitting.

The correct order is placement first, then charge weight, and temperature last — raising temperature improves flow but accelerates scorch at the same time.

Scorch: the false short shot

There is a short shot that adding material does not fix and actually worsens: the compound begins crosslinking before it has filled the cavity. Viscosity rises and flow stops. It looks like a short shot; the mechanism is scorch.

Zhang, Zhen, and Zhao examined formulation effects on cure kinetics by DSC in Effects of formulations of silicon rubber on its non-isothermal vulcanization kinetics, Journal of Chemical Engineering of Chinese Universities, 2020, 34(1): 222–229, DOI: 10.3969/j.issn.1003-9015.2020.01.028. Two findings bear directly on scorch tendency: increasing curative loading markedly lowers the crosslinking onset temperature, driving crosslink density too high early and suppressing later conversion; and fumed silica forms bound rubber with the matrix, so its loading affects the cure process non-monotonically — neither more nor less is automatically better.

In shop-floor terms: curative loading and silica loading jointly set where the “crosslinking starts” threshold falls, which determines how much flow time the compound still has. When chasing a false short shot, those two formulation variables matter as much as temperature.

Short shot vs flash: opposite directions on the same variables

Sticking and Demolding Difficulty

Three independent causes

Sticking is the most frequently misdiagnosed of the four, because it has three independent causes:

• Mold surface condition. Rough cavity, cured residue, or scoring mechanically resists release.
• Release agent problem. Wrong type, uneven application, or insufficient dosage.
• Under-cure. The part simply is not cured through and the surface is tacky. In this case adding more release agent only hides the problem.

So the sequence must be state of cure first, then the tool. Reverse it and you will treat under-cure as insufficient release, spray more, cover the defect for a while — and then have it reappear at the customer as “tacky.”

Internal vs external release agent

The difference is not which works better. It is where the side effects land.

“Tacky surface” does not mean “sticking”

Investigate a tacky surface as a state-of-cure problem first, and as a release problem second.

• Uniform tackiness across the batch, with the part soft → most likely under-cure.
• Tack only on the mold-contact face, rest of the part normal → likely release residue or mold surface contamination.
• Tack only in certain cavities → check whether that cavity runs cool.

For platinum-cured parts, catalyst poisoning is another route to surface tack — a separate failure chain, covered in Platinum Cure vs Peroxide Cure Silicone: How to Choose (With a Selection Matrix).

Under-Cure and Over-Cure

What under-cure looks like on the floor

Under-cure does not announce itself loudly. It usually shows up as:

• Soft and slow to rebound — crosslink density never came up.
• Tacky surface — unreacted species still present.
• High compression set — the network cannot hold rebound.
• Unstable dimensions — continued shrinkage after demold, drifting lot to lot.

Two or more of these together is enough to lock onto under-cure without suspecting the formulation first.

The cure-rate window is set by the formulation

Zhang et al. (2020, as above) measured the apparent activation energy of the DCP-containing system at 138.77–149.65 kJ·mol⁻¹ (Kissinger method) and noted the cure shows classic autocatalytic behavior. That number describes how readily the reaction starts — a different activation energy means a different temperature and time window for the same part.

The sourcing implication: do not carry a previous supplier’s cure conditions straight onto a new compound. Curative type, molecular weight, and filler loading all shift this window. Changing compound means re-confirming t90.

What post-curing fixes — and what it does not

Post-curing removes residual byproducts and volatiles, stabilizes the network, and improves compression set. What it cannot do: eliminate gas already locked in during first-stage cure, or finish curing an under-cured part. Pushing under-cure onto post-cure is one of the most common misuses on the floor. See Why Silicone Needs Post-Curing: Byproducts, Odor, and the Four Limits for the full boundary.

Under-cure / correct cure / over-cure compared

Defects You Cannot Fix by Tuning Parameters

Five classes of problem where the tool and the compound have already decided the outcome. Tuning changes how often a defect shows up, not whether it can.

• Structural gas entrapment from abrupt wall-thickness change → only a design change fixes it (uniform wall, transition radii). However you tune, the thick-thin junction still traps gas. See the thermal gradient in thick sections.
• Poor mold venting design → vent location, depth, and count are tooling attributes. Parameters only relieve; fixing requires reworking the tool.
• Incoming compound over the limit on moisture or volatiles → that is incoming material control, not process. On the floor you can only reject or downgrade it.
• Geometry unsuited to compression molding (very thick plus very low volume) → change the process or the design; stop tuning parameters.
• Flash from a worn parting line → repair the tool, not the pressure setting.

We do not promise “zero defects,” and we do not publish defect-rate numbers. How far each defect class can be improved on your specific part has to be confirmed by trial molding — quoting a generic percentage would be dishonest and useless as an acceptance criterion.

How to Surface These Problems During Sampling

Five checks on the first article

• Cut a section. Void location (near skin vs centered) separates skin-lock from heat-transfer problems immediately.
• Measure hardness against target. If it is off by more than ±5 Shore A, check state of cure before touching the formulation.
• Inspect corner fill. Decide whether it is charge weight or placement.
• Check flash thickness. Flash is the most direct feedback on whether charge weight was calculated correctly.
• Record demold feel. When demold resistance is abnormal, judge state of cure before inspecting the tool.

Six data points to record on every trial

Trial records exist so the second round has something to stand on. Fix at least these six:

• Charge weight (including preform shape and placement)
• Mold temperature (measured, recorded separately for upper and lower halves)
• Cure time (including vent count and the timing of each vent)
• Clamp pressure
• First-article hardness and critical dimensions
• Defect description (with the section location noted)

For parts on a platinum cure system, we can sample both cure paths so you can compare defect behaviour on the same geometry before committing. The full sampling workflow and answers to common questions are in our sampling and trial-moulding workflow and the silicone FAQ centre.

Frequently Asked Questions

Is a bubble on a silicone part a quality problem?

It depends on the type and distribution — you cannot judge from the word “bubble” alone. A surface blister with a smooth inner wall is usually a vent-timing issue and is process-correctable; internal honeycomb porosity usually traces to compound moisture or an insufficient charge; bulging after post-cure means the problem sits in first-stage density. All three are traceable and improvable — but “there is a bubble” is not a verdict. Classify first.

Why do some parts in the same lot bubble and others not?

In-lot variation usually points to one of three: cavity-to-cavity mold temperature differences, inconsistent preform placement, or differing moisture pickup because compound sat open for different lengths of time. Mold temperature spread is the one most often missed — measured temperature can differ meaningfully across one tool, so the setpoint on the controller is not enough.

Can post-curing remove bubbles?

No. Post-curing removes residual byproducts and volatiles, stabilizes the network, and improves compression set — but it cannot remove gas already locked in during first-stage cure, and it cannot finish curing an under-cured part. A part that bulges after post-cure is telling you first-stage density was insufficient; the fix belongs in the first stage. See what post-cure removes and what it does not.

After changing suppliers the defect disappeared — formulation or process?

Usually it is process-window matching, not formulation quality. Curative type, molecular weight, and filler loading all shift the cure-rate window (Zhang et al., 2020 measured 138.77–149.65 kJ·mol⁻¹ for the DCP system), so the same time and temperature is no longer optimal on a different compound. The more reliable way to tell: compare the new supplier’s six trial data points with the previous one’s. Whichever item differs is where the problem sits.

Bottom Line

Bubbles and under-cure mean checking parameters; short shots mean checking charge and tooling; sticking means checking state of cure before the tool.

• Classify before treating. The three bubble types have entirely different causes; treating them as one is why they never get fixed.
• Short shot and flash are opposite directions on the same variables. Calculate the charge first, then talk about anything else.
• Treat a tacky surface as under-cure first. More release agent hides the problem until it resurfaces at your customer.
• Five defect classes cannot be fixed by tuning (abrupt wall-thickness entrapment, venting design, out-of-spec incoming volatiles, geometry unsuited to compression molding, worn parting line). Those need a design change or a tool change.

We do not publish generic defect rates and we do not promise zero defects — how much a given part can improve has to be confirmed by trial molding. Most defect photos can be classified at a glance. But classifying it and suppressing it by tuning are two different questions — the second one needs the trial record. Send the drawing, part photos, and trial records together: within 24 hours you get three answers — which class it falls in, whether tuning can suppress it, and if not, what to change. (If you already have trial records, include the six data points above.)

Shenzhen Lixinyuan Technology Co., Ltd. | Custom silicone OEM/ODM | MOQ 50 pcs | Sampling in 5–7 days | Contact us