Heat-Soak Testing and Why Tempered Glass Breaks on Its Own

A tempered glass panel sits in a balcony guard for four years without incident. On an ordinary afternoon, with nobody near it, it shatters. No impact, no vandalism, no obvious cause. The building manager assumes someone hit it. Nobody did. What happened is a microscopic nickel sulphide inclusion inside the glass finally completed a phase change, expanded, and released enough energy in the tensile core of a tempered pane to detonate it. This is a well-documented failure mode, it is not rare, and it is the reason heat-soak testing exists.

For anyone specifying glass guards on a multi-storey building, this is not an academic topic. A spontaneously breaking panel at height is a life-safety and liability event, and remediating it across a tower after occupancy costs far more than preventing it at the factory.

Why tempered glass breaks with no apparent cause

Tempering works by rapidly cooling the glass surfaces after heating, leaving the outer layers in compression and the core in tension. That locked-in stress is what makes tempered glass four to five times stronger than annealed glass and what makes it break into small dice instead of shards. It also means the interior of every tempered pane is a reservoir of stored elastic energy waiting for a trigger.

Nickel sulphide inclusions are tiny particles — typically well under a millimetre — formed from trace nickel and sulphur contamination during float glass manufacture. Nickel sulphide exists in two crystal phases with slightly different volumes. The high-temperature phase gets frozen in when glass is rapidly cooled during tempering. Over years, at ambient temperature, it slowly converts to the low-temperature phase and expands by a few percent. If the inclusion happens to sit in the tensile core, that expansion is enough to initiate a crack, and the stored tempering energy does the rest instantly.

• The inclusion must sit in the tensile core to cause breakage. In fully tempered glass the surfaces are in compression and the central region is in tension, and that tensile core occupies most of the thickness — roughly the middle 60 percent. An inclusion within the compressed surface layers is harmless.
• Conversion is temperature-driven. Hot glass converts faster, which is why south and west elevations, and dark tinted or spandrel glass, statistically see more spontaneous breaks.
• There is no visual inspection that reliably catches it. The inclusion is too small and too deep to find by eye or by ordinary quality control.
• It cannot happen in annealed or heat-strengthened glass to the same degree, because there is not enough stored tensile energy in the core to propagate the crack catastrophically.

What heat-soak testing actually does

Heat-soak testing is a destructive screening process. After tempering, the glass is loaded into a dedicated oven, brought up to roughly 290 °C, held there for a defined dwell period, and cooled under controlled ramp rates. The elevated temperature accelerates the nickel sulphide phase conversion by orders of magnitude, so panes carrying a critical inclusion break in the oven instead of on a balcony in year four. Survivors ship. It is a brutally simple idea and it works.

Two things matter about how it is done. First, the process is defined by recognised standards — the European EN 14179 protocol is the common reference — and a proper heat-soak run means a controlled hold at temperature with monitored ramp rates, not simply parking glass in a warm oven for a while. Second, the process is not free of consequence: it puts thermal stress on the glass and a small percentage of the batch will break during the soak. That breakage is the point. It is also why heat-soak carries a real cost and a real schedule impact.

Heat-soak testing does not make glass safer. It removes the panels that were already going to fail — before they are 30 storeys above a sidewalk.

What heat-soak does not do

This is where specifications get sloppy. Heat-soak testing is a statistical risk reduction, not a guarantee. A properly executed soak to a recognised standard dramatically reduces the residual population of critical inclusions — the standards quantify this — but it does not drive the probability to zero. A small residual risk remains, and any consultant who tells you heat-soaked glass cannot spontaneously break is overselling.

• It does not address breakage from edge damage, impact, thermal shock, or hardware over-torque. Those are far more common failure causes than nickel sulphide and none of them are screened by a heat soak.
• It does not apply meaningfully to heat-strengthened or annealed glass, which do not have enough core tension to be at risk in the same way.
• It does not replace laminated construction. Even a heat-soaked panel that breaks is a hazard if there is no interlayer holding the fragments.
• It is only as good as the process control. An undocumented soak with no temperature logging and no certificate is worth roughly what you paid to skip it.

The most important point: heat-soak and laminated glass solve different problems and belong together, not as alternatives. Lamination handles the consequence of breakage — fragments stay bonded, the guard stays a barrier. Heat-soak reduces the frequency of breakage. On a tall building with frameless guards you want both.

When to specify heat-soaked glass

Heat-soak adds cost and lead time, so it is worth being deliberate about where it earns its place. Our general guidance on Canadian commercial projects:

• Specify it for any fully tempered glass in overhead or elevated guard applications on mid-rise and high-rise buildings, where a breakage event means falling glass over public space.
• Specify it where access for replacement is difficult or expensive — setback terraces, roof amenity decks, anywhere requiring a swing stage or crane to reach.
• Consider it strongly on dark tinted or heavily solar-loaded elevations, where operating temperatures accelerate conversion.
• It is generally less critical on ground-level and low-rise guards where a broken laminated panel is a maintenance call rather than an incident, though many owners specify it building-wide simply for consistency.
• Consider whether you need fully tempered glass at all. Heat-strengthened laminated glass sidesteps the nickel sulphide risk almost entirely and often has adequate capacity — it is a genuine design option rather than a downgrade, and it is worth asking the engineer to price both.

When you do specify it, write it properly: reference the standard, require the soak be performed after tempering and after any edgework, and require documentation — batch records with temperature logs and a certificate tied to the specific panels delivered. "Heat-soaked" as a bare word in a spec is a request that can be answered with a shrug.

The procurement and schedule reality

Heat-soak capacity is not universal, and this is where imported glass programs get uncomfortable. If the fabricator has an in-house soak oven, it is a step in the production sequence and adds days. If they do not, glass moves to a third-party facility, gets handled twice more, and adds weeks — plus the breakage from that extra handling, which is not trivial on large panels. When glass is coming from overseas the failed panels are discovered thousands of kilometres from the site, and the replacement runs on a container schedule rather than a truck schedule.

That supply chain question is worth resolving during design, not during submittals. It affects the railing package's position on the critical path, which we have written about separately in the context of railing lead times. Domestic fabrication and a short line between the engineer, the glass supplier and the install crew means a broken batch is a rescheduled week rather than a rescheduled quarter.