The Trap Inside a Temperature Gradient
Everything in the last three guides now converges into one dangerous moment. You met heat capacity (how much energy a body soaks up), the thermal expansion coefficient alpha (how much it grows per degree), and thermal conductivity k (how fast heat spreads). Thermal shock is what happens when these three fight each other. Heat or cool a ceramic faster than k can even out the temperature, and different parts of the same body end up at different temperatures — which means, thanks to alpha, they all want to be different sizes at the same instant.
But they cannot: every region is bonded to its neighbours in that rigid ionic-covalent cage. A part that wants to shrink but is held stretched by a hotter neighbour is under tension; a part that wants to grow but is squeezed by a cooler one is under compression. This self-inflicted thermal stress needs no external load at all — the temperature difference alone generates it. Picture pouring boiling water into a cold glass tumbler: the inner surface leaps to expand while the still-cold outer wall holds it back, and the glass tears itself apart from the inside.
How Big Is the Stress?
Put numbers on it. If a piece of material wants to shrink by a strain alpha times ΔT but is completely prevented from doing so, the stress it develops is stiffness times prevented strain. Because a quenched surface is held in two directions at once (biaxial constraint), a factor 1/(1 - nu) creeps in, where nu is Poisson's ratio. The result is the master equation of thermal shock: sigma = E alpha ΔT / (1 - nu). Here E is the elastic modulus and ΔT the temperature difference the body is trying to hold across itself.
Read the equation like a warning label. Every term on top is a liability: a high modulus E (a very stiff cage transmits the mismatch as huge force), a high expansion alpha (bigger size mismatch per degree), and a large ΔT (a fiercer gradient) all pile on stress. Work a case: alumina, with E = 380 GPa, alpha = 8 x 10^-6 per degree C, nu = 0.2, quenched through ΔT = 200 degrees C. Then sigma = (380 x 10^9)(8 x 10^-6)(200) / 0.8 = 760 MPa. That dwarfs alumina's strength of roughly 300 MPa, so a 200-degree quench cracks it outright.
But this equation quietly assumes the worst case — that the surface is held at the full ΔT the instant it is quenched, as if heat never flowed at all. That would only be true for an infinitely fast quench into a perfect heat sink. In reality the surface stress builds only as fast as heat can pour out, and a material that conducts heat well (high k) never lets a big gradient form in the first place. That missing role of conductivity is exactly why one thermal-shock number is not enough — and why the next section needs two.
The R and R' Parameters
Failure strikes when the thermal stress reaches the material's fracture strength sigma_f. Set sigma = sigma_f in the master equation and solve for the temperature drop the body can just barely survive. That critical ΔT is the first thermal-shock resistance parameter: R = sigma_f (1 - nu) / (E alpha), measured in degrees. R answers 'how big a sudden quench can this ceramic take before it cracks?' — and, being a critical temperature difference, higher is better. It rewards exactly the opposite of what the stress equation punished: strong, floppy (low E), low-expansion materials win.
R describes the most brutal shock, where heat flow does not help you. For gentler, more realistic shocks the conductivity finally earns its place: multiply R by k to get the second parameter, R' = k sigma_f (1 - nu) / (E alpha), in watts per metre. R' rewards a material that whisks heat away before a steep gradient can build. The practical rule of thumb is the Biot number Bi = h L / k, comparing surface heat-transfer h and size L against conductivity: a violent water quench of a big low-k part (high Bi) is ranked by R, while a milder shock of a small conductive part (low Bi) is ranked by R'.
Thermal-shock ranking of some ceramics (round numbers, nu ~ 0.2)
material alpha E sigma_f k R R'
(1e-6/K) (GPa) (MPa) (W/m.K) (K) (W/m)
--------------------------------------------------------------------
fused silica 0.5 73 100 1.4 ~2200 ~3100
borosilicate 3.3 64 70 1.1 ~265 ~290
soda-lime glass 9.0 70 50 1.0 ~64 ~64
alumina 8.0 380 300 30 ~79 ~2400
silicon carbide 4.5 410 400 110 ~175 ~19000
R = sigma_f (1 - nu) / (E alpha) <- severe quench, high Bi
R' = k times R <- milder shock, low Bi
note how the winner CHANGES with the column:
* severe shock -> fused silica runs away (tiny alpha)
* milder shock -> SiC leaps ahead (k does the work)The table tells a story worth pausing on. Fused silica has an almost unbelievable R of ~2200 K purely because its alpha is near 0.5 x 10^-6 — this is the material you can heat white-hot and plunge into cold water without a crack, the classic lecture demonstration. Borosilicate (Pyrex) has a modest R of ~265 K but that is enough to survive oven-to-counter, while ordinary soda-lime glass at ~64 K shatters when you pour boiling water in. Alumina looks poor on R (~79) yet decent on R' (~2400) because it conducts. Silicon carbide is only middling on R but stellar on R' — its conductivity, not its strength, is the secret.
Measuring It: The Quench Test
R and R' are predictions; the quench test is how you actually pin a material's thermal-shock resistance down in the lab. The idea is disarmingly direct: heat samples to a known temperature, drop them into water, and then ask not whether they look cracked but how much strength they have secretly lost. You look for the temperature drop at which the retained strength falls off a cliff.
- Prepare many identical bars and measure the baseline bend strength of an unshocked group.
- Heat separate groups to a range of furnace temperatures, giving a series of quench temperature differences ΔT above the water-bath temperature.
- Drop each hot group into the water bath, so it cools almost instantly through its ΔT.
- Dry the quenched bars and measure the retained bend strength of each group.
- Plot retained strength against ΔT: the ΔT where it suddenly drops is the critical quench, ΔT_c, and it should track R.
The shape of that curve carries a second lesson. A coarse-grained, dense ceramic like well-made alumina shows a sharp cliff: below ΔT_c strength is untouched, and at ΔT_c a single crack runs catastrophically and strength collapses to a fraction of its former value. But a material laced with tiny microcracks — many refractories, and coarse zirconia bodies — shows a gentle, gradual decline instead, because incoming cracks are short and get arrested by pores and grain boundaries. That difference between crack initiation and crack survival is the doorway to the last guide in this rung.
Reading the Rankings Honestly
R and R' are powerful for ranking, but be honest about what they do and do not say. They predict when a crack will start, not what happens next. A material can have a high R yet, once a crack does begin, shatter it in one clean fracture; another can start cracking early (low R) yet stop every crack in its tracks and merely degrade gracefully. Full thermal-shock engineering therefore needs a second family of fracture-mechanics parameters built from strength and toughness that rank crack propagation and damage tolerance — which is precisely where the design guide picks up.
One last honest caveat. R and R' assume you shocked the body once, quickly, and that its strength was fixed. Real service is messier: repeated cycles slowly grow microcracks (thermal fatigue), and at very high temperatures the material may creep and relax some stress rather than store it elastically — so a hot part can shrug off a gradient that would crack the same material cold. Treat R and R' as an honest first sort of candidate materials, not as a guarantee. They tell you where to start looking, and that is exactly what a good thermal-shock number should do.