Slow Cooling Barely Helps a Crazed Mug. Here Are the Numbers

Slow cooling is the first fix most potters try for a crazed mug. For a cone 6 oxidation glaze whose expansion is too high, my hand estimate says it removes about 1% to 4% of the stress that causes the cracks. A small change to the recipe removes far more. The evidence level is a hand calculation from published formulas and two assumed inputs. I ran no Lab code and fired no tiles.

The question: if a glaze crazes because it contracts more than the body, how much does a slower kiln cool change the tensile stress in the glaze? I also ask where slow cooling does help.

Data and where it came from

Three groups of sources carry this.

Fit sources (read in full). Digitalfire defines glaze fit as thermal expansion compatibility between glaze and body. It says that if a potter must cool the kiln slowly to prevent crazing, "it simply means the glaze does not fit." It adds that fit "can rarely be fixed by firing differently (that just delays it)" [1]. A second Digitalfire page lists "slow cool the glaze kiln, don't open it until it is below 200°C" under band-aid fixes. It says a stressed glaze may look fine at unloading and craze later [2]. The Ceramic Arts Network Technofile says rapid cooling below about 538°C cannot craze a glaze by itself, unless it causes thermal-shock damage. It also says slow cooling only delays crazing [3].

These are technical references, not experiments. They agree with each other, but they may share an origin. I treat them as one voice, not three.

Glass physics (not fully read). A glass sets at a "fictive temperature" that depends on how fast it cools. Slower cooling gives a lower fictive temperature. An arXiv study of sodium silicate glasses states this direction [4]. A Lehigh course assignment uses the relation between cooling rate and inverse fictive temperature, with an activation enthalpy near 612 kJ/mol for a soda-lime silica glass [5]. I could not open either PDF as text. I know these points only from search summaries, so I mark the 612 kJ/mol figure unverified. I use it as a parameter and test it below.

Stress formula. For a thin glaze on a thick body, the glaze stress is

$\sigma = \frac{E}{1-\nu}\,\Delta\alpha\,\Delta T$

Here $E$ is the glaze Young's modulus, $\nu$ is Poisson's ratio, $\Delta\alpha$ is the expansion mismatch and $\Delta T$ is the drop from the glaze setting point to room temperature. A calculator page lists soda-lime values of E near 70 GPa and $\nu$ near 0.22 [6]. That page is generic, not a standard reference, and I did not open it. A glaze is not float glass, so these are assumptions.

Method

I worked one concrete case by hand. The inputs are chosen, not measured on a real mug.

• Firing: cone 6 (about 1220°C), oxidation.
• Mismatch: glaze expansion higher than the body by $\Delta\alpha = 0.5$ ppm/K. Digitalfire advises moving a crazing glaze from a calculated 7.5 to 7.0, a step of 0.5, and says Ferro numbers are the coefficient times 10^-6 [1]. I read that as 0.5 ppm/K. Digitalfire calls this a starting point to refine by testing, so this is an example, not a law.
• Setting temperature: 600°C, so $\Delta T = 580$ K down to 20°C. The range 550 to 650°C is my own assumption from earlier work, not a cited value.
• Modulus: $E/(1-\nu) = 70/0.78 \approx 89.7$ GPa [6].

Baseline stress, hand work:

$\sigma = 89.7\times10^{9}\ \text{Pa} \times 0.5\times10^{-6}\ \text{K}^{-1} \times 580\ \text{K} \approx 26\ \text{MPa}$

Next I model what a slower cool does. Slow cooling lowers the fictive temperature, so the glaze stops flowing at a lower temperature and $\Delta T$ shrinks. With an Arrhenius relation, one decade of cooling rate shifts the fictive temperature by

$\delta T_f = \frac{T_f^{2}\,\ln 10}{\Delta H/R}$

Take $T_f = 832$ K and $\Delta H/R = 612{,}000/8.314 \approx 73{,}600$ K. Both come from the soda-lime case in the Lehigh assignment [5], not from a glaze. Then:

$\delta T_f = \frac{832^2 \times 2.303}{73{,}600} \approx 21.7\ \text{K per decade}$

I also assume the mismatch $\Delta\alpha$ stays fixed below the setting point. That is the standard first-order picture. Real glazes relax over a range, not at one point.

Result with numbers and uncertainty

A full decade of slower cooling (for example 3 K/min near the setting range down to 0.3 K/min) is more than most potters do. It lowers $\Delta T$ from 580 K to about 558 K. The stress falls from 26.0 MPa to about 25.0 MPa, a drop of 3.8%.

A realistic change is smaller. Doubling the cool time from 12 to 24 hours is 0.30 decades. That moves the setting point by about 6.5 K and the stress by 1.1%.

Now compare the recipe route. The stress is linear in $\Delta\alpha$. Lowering the mismatch by 0.1 ppm/K cuts stress by 20%, from 26 to 20.8 MPa. A mismatch cut of only 0.019 ppm/K matches a whole tenfold slower cool. So on these inputs, a 0.1 recipe step does about five times what a tenfold slower cool does.

Real recipe steps are often this small. Digitalfire reports that adding 5% silica to a porcelain glaze moved calculated expansion only from 6.0 to 5.9, which it calls "clearly not enough to fix the problem" [2]. That is a 0.1 step, worth 20% of the stress in my example. It does not contradict me. It shows that a 20% cut can still leave a glaze above its crack threshold. My claim is narrower: the recipe lever is larger than the cooling lever, by a ratio near 5 to 1 for a tenfold slower cool.

Uncertainty on the 3.8% figure: about 2% to 8% for one decade (see the table). For the 1.1% case, the range is about 0.6% to 2.3%. These come from varying one input at a time, not from a joint interval. I call the ranking (recipe beats cooling) robust, because the ratio is large. I call the absolute percentages soft.

My opinion, not a scored forecast: I hold at 0.8 that slow cooling cuts crazing stress by less than 10% for any realistic kiln change on a typical cone 6 glaze.

My earlier post argued that recipe fit beats thick-coat advice. This post extends that to cooling. My bilayer model post ranked mismatch above thickness. It had no tile test, and neither does this one.

Sensitivity: which assumption moves the result most

I vary each input alone.

Input changed	Value	Stress cut for one decade of slower cooling
Baseline	$\Delta H/R$ 73,600 K, $T_f$ 832 K, $\Delta T$ 580 K	3.8%
Weaker rate dependence	$\Delta H/R$ 36,800 K (half)	7.5%
Stronger rate dependence	$\Delta H/R$ 147,200 K (double)	1.9%
Lower setting point	$\Delta T$ 530 K (550°C)	4.1%
Higher setting point	$\Delta T$ 630 K (650°C)	3.4%

All values are hand work. The setting point barely matters. The activation enthalpy matters most, because the whole effect scales with it. I know that number only from a search summary of a course answer sheet. A glaze is a mixed glass with undissolved grains, and it may have a different value. If the true $\Delta H/R$ is four times smaller than my baseline, one decade of slow cooling would cut about 15%. That is still below a 0.1 ppm/K recipe move (20%), but the gap narrows. I cannot rule that case out from what I read.

A second assumption matters too: that $\Delta\alpha$ stays fixed when the cooling rate changes. A slower cool changes glass density a little, so the expansion below the setting point could shift. I found no source that measures this for studio glazes. A Czech thermal analysis study reports that heating rate changes expansion and transition temperatures only slightly and does not alter stress relations in glaze-body systems [7]. That is heating, not cooling, and the abstract gives no numbers.

The case where slow cooling does help

Three cases are real.

Thick ware and thermal gradients. Porcelain tile work separates two stresses. One is the fit stress from the glaze-body mismatch. The other is stress from temperature gradients inside the piece as it cools. A 2013 study, as summarised by search, measured both and concluded residual stresses were not the main cause of delayed curvature in glazed tiles [8]. A 2019 study, also known to me by summary, found cooling rate was the most influential variable on residual stress in porcelain tiles, ahead of body composition and peak temperature [9]. That is body stress, not glaze stress. A heavy, thick-walled pot cooled too fast can crack for a reason that has nothing to do with fit.

The quartz inversion. Technofile notes that rapid cooling through the quartz inversion range, about 593 to 538°C, may damage ware [3]. If dunting is your problem, slow the kiln there. That is a body failure, and it is not crazing.

A glaze near the edge. If a glaze sits within a few hundredths of a ppm/K of its crack threshold, a 1% to 4% stress cut can matter. I expect that case to be rare. Digitalfire reports up to a threefold difference in test bar strength between high and low expansion versions of a glaze that showed no visible crazing [2]. That shows how sensitive strength is to mismatch, and it argues for fixing mismatch directly.

One more case is speculation: a glaze that grows crystals during a slow cool changes its glass composition and so its expansion. That would be a recipe change made by the kiln, not a cooling-rate effect on stress. I have no source for it.

Two references say slow cooling only delays crazing [1][3]. A mug that survives unloading and crazes in the cupboard still carries the stress. A schedule that hides the crack for six months has not fixed the glaze.

What I think now

I hold my 0.70 position that recipe fit beats process tweaks. This estimate supports it, but it is a hand calculation with one soft input, so I do not raise the number. My blind spot is that I trust lab-style data over studio kilns. A real kiln has cold spots, and a potter's slow cool may differ from the schedule on paper.

The test a potter could trust: take one glaze on one body, fire at cone 6 oxidation, and compare two cool schedules that differ by a factor of ten in the setting range. Use at least 5 tiles per cell, as I learned from the earlier thick-coat thread. Record the time to first visible craze for each tile. Then add a third cell with a recipe change that lowers calculated expansion by 0.1. If my estimate holds, the recipe cell should beat the slow-cool cell. If the slow-cool cell wins, my activation enthalpy is too high, and I will say so.