Glass · Identification
Why did this glass break?
Find where the break started, then look at what is there. An impact leaves a mark and a cone. Thermal stress leaves one clean crack from an edge and nothing else. Edge damage leaves a nick you can feel with a fingernail. A nickel sulphide inclusion leaves a pair of wings around a speck in the middle of the glass. Four causes, four signatures — and the argument about who pays usually turns on which one it is.
Before anything else: which glass was it?
The answer narrows the list before you look at a single crack.
It fell in small blunt cubes. Toughened. Thermal stress is very unlikely — toughened glass tolerates around 200 °C of difference between centre and edge. That leaves impact, edge damage and nickel sulphide.
It is still in the frame in large pieces, with long curving cracks. Annealed, or heat-strengthened. Nickel sulphide is essentially off the list: the inclusion needs a big locked-in core tension to break the pane, and annealed glass has none. Thermal stress moves to the top, because annealed glass takes only about 30 to 40 °C of difference before it goes.
It cracked but held together in one sheet. Laminated. Read the break through the interlayer exactly as below; the plastic keeps the evidence in place, which makes this the easiest case of all to diagnose and the least urgent to clear up.
Not sure which it is? Five checks, in order — starting with the mark in the corner and the polarised-sunglasses trick, which works on a pane that is still in the wall.
Finding the origin
Everything below depends on this one step, and it takes about a minute.
Cracks branch away from the origin and never towards it. So follow the branches backwards. Where four cracks become two and two become one, you are close; where there is a single crack and no fork, you are there. That rule is the whole of it, and it holds on any glass, in any direction.
On a toughened pane that has diced, it is harder but not hopeless. The dice are not uniform: at the origin there is usually a pair of longer, thinner fragments lying together like two wings, and the fragments around them are noticeably larger than the rest. If the pane is still hanging in the frame or taped up, photograph it before anybody sweeps: once it is in a bin, the origin is gone and so is the answer.
The four causes, and what each one leaves behind
In the order they turn up in a real building.
-
Impact
By far the most common, and the easiest to confirm. At the origin there is a mark on the surface, and under it a cone: the crack starts at the point of contact and opens out through the thickness, so the chip is small on the struck face and wider on the far one. That is also how you tell which side was hit, which matters more often than people expect — a stone thrown from a mower, a ladder, a hail strike and a burglar all leave the cone opening the same way, away from the blow.
Radial cracks run out from the origin and concentric ones ring it. On thin annealed glass a hard, fast impact leaves very fine radial cracks; a slow heavy push leaves few and long.
Confirmed by: a mark you can feel at the origin and a cone that opens away from one face. Ruled out by: an origin at an edge with nothing on either surface.
-
Thermal stress
Sun heats part of the pane while the rest stays cold — typically the middle in the sun and the edge buried in the frame, still in shade and still cold from the night. The warm glass wants to expand and the cold edge will not let it, and the tension lands exactly where the glass is weakest: the cut edge.
The signature is unmistakable once you have seen one. A single crack, starting at an edge, leaving that edge at a right angle for the first centimetre or two, then wandering across the pane in smooth curves. No mark, no cone, no shattering, often no branching at all until much later in its run.
The published limit for annealed float is a difference of roughly 30 to 40 °C between the centre and the edge, around 100 for heat-strengthened and around 200 for toughened. Those are not material constants: edge quality decides them. A cleanly cut, well arrised edge takes far more than a nipped or shelled one, which is why two identical panes in identical frames behave differently.
Suspect it when: the break appeared in the morning after a cold night, in the first sunny spell; part of the pane is shaded by a reveal, a balcony or a tree; something dark sits close behind the glass — furniture, a blind pulled halfway, a stuck-on solar film, a sign; or the pane is annealed and nobody was near it.
-
Edge damage and restraint
A nick, a shell chip or a nipped corner from handling or from being levered into a frame sits precisely where the tension is highest. The pane may survive installation and break weeks later on the first cold night, when the frame moves and finds the flaw.
The same goes for a pane with nowhere to move: missing or wrong setting blocks, glass touching metal, a frame that has swelled, a building that has settled. The origin is at an edge or a corner, there is no impact mark, and running a fingernail along the edge finds the defect — carefully, because the edge of broken glass cuts.
Tells it apart from a thermal break: a thermal crack starts on an intact edge and leaves it square. This one starts at a visible defect, and often runs off at whatever angle the flaw dictates.
-
Nickel sulphide inclusion — toughened glass only
A microscopic nickel sulphide particle left in the melt sits in the tension core of a toughened pane. In service it slowly changes crystal phase, from α to β, and that change comes with an increase in volume. When the particle has grown enough, it cracks the pane from the inside — with nobody near it, on a hot day, typically months or years after the glass went in.
The evidence is a butterfly: two wing-shaped fragments meeting at a point in the middle of the thickness, with the tiny stone at the join, and larger polygonal fragments around them before the dicing becomes uniform. Origin inside the glass rather than at a surface or an edge is the thing that separates it from everything above.
It is nobody's workmanship. The accepted mitigation is heat soaking to EN 14179-1: the toughened panes are held hot for hours so that most of the affected ones break in the oven instead of on the building. It reduces the risk and does not abolish it. For overhead glazing, and anywhere a failure would drop glass on people, it is worth specifying — and when a pane does go, worth checking the paperwork to see whether it was.
Do not conclude this by elimination. "Toughened, no impact mark, therefore nickel sulphide" is the most common wrong answer in the trade. Edge damage does the same thing far more often, and it does leave a visible origin.
The short version
One row per cause. The origin column is the one that decides.
| Cause | Where it starts | What it looks like | Which glass |
|---|---|---|---|
| Impact | a point on a face, with a mark | cone opening away from the struck side, radial and concentric cracks | any |
| Thermal stress | an intact edge, at a right angle to it | one smooth wandering crack, no mark, no branching at first | mostly annealed |
| Edge damage | a visible nick or chip on an edge | crack from the defect, no cone, no square start | any, worst in toughened |
| Nickel sulphide | inside the glass, mid-thickness | butterfly pair around a speck, whole pane diced | toughened only |
What this page cannot do
Read this before you quote it at anybody.
This is how to make a sensible first reading in front of a broken pane. It is not fractographic analysis. A laboratory reads the fracture surface itself — the mirror around the origin, the mist and the hackle beyond it — and can put a number on the stress at failure; that is what settles a disputed claim, and it needs the actual fragments, kept and labelled, not a photograph.
Two honest cautions. A pane can carry more than one cause: edge damage that survived a year and then met a cold morning is a thermal break and an edge defect, and which one you call it decides who pays. And a toughened pane that has been swept up cannot be diagnosed at all — the answer went into the bin with the origin.