Why a Glass Pipe Cracks Hours After Cooling

A piece cracks hours after it has cooled because the kiln brought it down through the strain point too fast, or held it too cool at the wrong moment, so stress stayed locked in the wall. The crack is not the cooling that did it. The crack is the schedule that did it, and the glass only reports the damage once the bench has stopped watching. That habit is the same one a painter keeps when tracking a wash on wet paper, and it is the reason a working practice journal is worth keeping on the bench next to the kiln log.

Why does a piece crack hours after it has cooled?

Borosilicate does not freeze into a rigid solid at one temperature. It passes through a range. Above roughly 510 C the wall is still soft enough to relieve its own stress. Below that, the wall is stiff, and any stress built into it stays built in. The strain point sits near 510 C for a standard borosilicate tube, and the annealing point sits near 565 C. Between those two numbers the glass is doing the only work that matters: relaxing.

A bench that pulls a piece out of the kiln at 200 C and sets it on the bench has skipped the range that counts. The piece feels cool in the hand. Nothing looks wrong. Then the wall equalizes with the room over the next two to eight hours, the last of the internal stress finds the weakest point, and the crack opens. That is why the failure shows up at hour three and not at minute three.

Thick sections make it worse. A 7 mm wall on a hammer bubbler holds heat longer than a 3 mm spoon wall, so the two sections cool at different rates inside the same piece. The thick section lags, the thin section leads, and the joint between them carries the difference. A crack at the neck of a heavy piece is usually a schedule problem, not a torch problem.

What does the annealing schedule change in the glass?

The schedule changes one thing: how much stress is left in the wall when the piece reaches room temperature. Nothing else. It does not change color, shape, or the ground joint. It changes whether the piece survives the trip.

A schedule has four parts. The ramp up, the hold, the ramp down, and the final drop. The hold is the part most benches get wrong. A piece needs to sit at the annealing point long enough for the whole wall, thick and thin, to reach the same temperature. A common shop rule is one hour of hold per 6 mm of wall thickness at the thickest point, plus a soak for the kiln itself. A 12 mm marble gets two hours. A 3 mm spoon wall gets thirty minutes if the kiln is already warm.

The ramp down is where the discipline lives. A safe rate for borosilicate is roughly 60 C per hour down to the strain point, then slower through it, then a free fall to room temperature once the piece is below about 370 C. Benches that run a fast ramp down to save kiln time are the benches that file breakage claims.

The record matters as much as the schedule. A bench that writes down the ramp, the hold, the piece, and the result can compare one firing to the next. The kiln log answers what the schedule was. The practice journal answers what the bench did with it.

How does a bench know its schedule is right?

It reads the result before it changes a setting. That order is the whole method.

A bench reads three things. First, the break itself. A crack that runs clean and straight through a wall is a stress crack. A crack that follows a scratch or a cold seal is a workmanship crack. The two look different under a loupe, and they call for different fixes. Second, the location. A crack at the base of a can, where the wall is thickest, points at the ramp down. A crack at a thin neck points at the hold. Third, the timing. A piece that cracks in the first hour out of the kiln was pulled too hot. A piece that cracks overnight was ramped down too fast.

Only after reading those three does a bench touch the controller. Changing the hold and the ramp at the same time tells the bench nothing, because the next failure cannot be traced to either one. One variable per firing. That is the rule.

A bench can also test a schedule with a polariscope. Two pieces of polarized film and a light source will show stress in a finished piece as bright bands. A piece that reads clean under the film is a piece the schedule worked on. A piece that shows a bright halo at the joint is a piece the schedule missed. The test costs a few dollars and takes a minute, and it turns a guess into a reading.

The kiln log and what goes in it

A kiln log is a sheet per firing. It carries the date, the kiln, the pieces in the load, the ramp up, the hold, the ramp down, and the room temperature at the time. It carries the result, checked at one hour, at eight hours, and at twenty four hours. It carries the polariscope reading if the bench took one.

A log that only records the schedule is half a log. The result column is the part that teaches. After fifty firings a bench can look down the result column and see the pattern: every failure in the load sat on the top shelf, or every failure came from the same tube batch, or every failure followed a ramp down faster than 60 C per hour. That is a schedule the bench can trust, because the bench has read it against its own results.

Case pack, freight, and the breakage claim

A schedule that holds up in the kiln still has to survive the crate. A piece that leaves the bench with residual stress will crack in transit, and a crack in transit is a breakage claim, not a kiln problem. The two get confused on the claim form.

A bench that ships should pack to the same standard it anneals to. Double box, 5 cm of cushion on every face, no glass touching glass, and a piece count on the outside of the carton. A case pack of 24 spoons in a single wall carton with paper wrap is a case pack that will produce claims. A case pack of 24 in a double wall carton with foam and a piece count will not.

When a claim does come in, the bench reads the break the same way it reads a kiln failure. A crack that runs from the rim down is usually a drop. A crack that runs across the base is usually a stack load. A crack that appears in a piece that was fine on arrival and cracked on the shelf is a schedule problem that shipped. The claim form should say which one, because the fix is different in each case.

Reading a result before changing a setting

The discipline is simple to state and hard to keep. Fire the schedule. Wait the full twenty four hours. Read the break. Write the result. Then change one thing.

A bench that changes the hold, the ramp, and the tube supplier in the same week has no idea what fixed the crack, and no idea what will bring it back. A bench that changes one variable per firing builds a schedule it can defend, because every number in that schedule has a result behind it.

The piece that cracks hours after cooling is not a mystery. It is a reading the bench did not take. Take the reading, keep the log, and the next piece comes out of the kiln whole.