The Design Question, in One Inequality
The last four guides built the physics; this one spends it. You know the villain by now: heat or cool a ceramic faster than it can even out its own temperature, and the hot skin and cold core pull against each other. Because the material cannot yield, that fight becomes thermal stress, and when the stress reaches the strength, the part cracks. That is thermal shock in a sentence. The whole design problem reduces to a single inequality: keep the thermal stress a body generates below the strength it can muster. Everything in this guide is a way to widen that gap.
Guide four turned that inequality into two ranking numbers. The first, the thermal-shock parameter R = sigma_f times (1 - nu) / (E times alpha), is the temperature drop a body can survive in an instant quench — it has units of degrees C, and you read it as 'plunge me into water this many degrees colder and I just barely make it.' The second, R' = R times k, folds in the thermal conductivity k; it ranks parts under a steady heat flux, where a material that carries heat away fast never lets a big gradient build in the first place. Two numbers, because there are two different shocks.
Five Levers You Can Pull
Read R and R' as shopping lists and five design levers fall out. Lower the thermal expansion coefficient alpha — it sits in the denominator of R, so halving alpha doubles the survivable quench, and this is by far the strongest lever. Lower the elastic modulus E — a floppier material stores less stress for the same strain (yes, sometimes softer is safer). Raise the strength sigma_f, the numerator — a stronger body simply reaches its limit later. Raise the thermal conductivity k — it never appears in R but doubles R', flattening gradients before they bite. And shrink the part, or at least the thickness heat must cross: a small, thin body equalizes its temperature almost as fast as you change it, so the gradient — and the stress — never grows large.
There is a sixth, subtler lever the parameters hide: make the flaws forgiving. R assumes the part cracks the instant stress meets strength, but a tough, flaw-tolerant ceramic does not shatter — it grows a network of tiny cracks that blunt and share the strain, and it loses strength gradually rather than exploding. Recall from the mechanics rung that strength is flaw-controlled through the fracture toughness K_IC: a higher K_IC and a finer grain size mean a thermal microcrack has to work harder to run. So two ceramics with identical R can behave utterly differently — one dices into fragments on the first quench, the other quietly downgrades and keeps holding. Designing for graceful degradation, not just for a high R, is the mark of a real thermal-shock part.
- Pin down the shock: how large a temperature swing (delta-T), how fast, and across how thick a section? A fast quench of a thick part points you at R; a slow gradient in a thin one points at R'.
- Chase low expansion first. Because alpha dominates R, a near-zero-expansion material buys more margin than any amount of extra strength — this is why fused silica and cordierite beat much stronger aluminas.
- Then trade E, sigma_f, and k against cost and the rest of the spec, and thin the section wherever heat has to cross it.
- Choose a tough, fine-grained microstructure so the part degrades gracefully instead of dicing — buy thermal-shock-resistant design margin from toughness, not just from R.
- Prove it with a quench test: heat identical specimens to a rising ladder of temperatures, drop each into water, and measure the retained strength — the delta-T where strength falls off a cliff is your real, as-built limit.
Reading It Off Real Materials
Put the levers to work and the champions announce themselves. Fused silica has an almost unbelievable alpha of about 0.5 times 10^-6 per degree C — sixteen times lower than alumina — plus a modest modulus, and that combination gives an R near 1600 degrees C. You can heat a fused-silica tube until it glows and quench it in water without a crack; nothing else on the shelf comes close. Borosilicate glass (Pyrex) is the everyday version, its alpha of about 3.3 times 10^-6 giving an R around 250 degrees C — enough to go from oven to countertop, which soda-lime glass at three times the expansion cannot survive. And silicon carbide wins the other way: an unremarkable R near 180 degrees C but a huge k of about 120 W per metre-kelvin, so its R' dwarfs everything — under a steady heat flux, SiC is the material that refuses to build a gradient.
material alpha E sigma_f k R R'=R k
(10^-6/C) (GPa) (MPa) (W/m.K) (deg C) (W/m)
--------------------------------------------------------------------------
fused silica 0.5 73 70 1.4 ~1600 ~2200
borosilicate 3.3 64 70 1.1 ~ 250 ~ 280
cordierite g-c 1.5 140 120 2.5 ~ 400 ~1000
alumina 8.0 380 350 30 ~ 90 ~2700
silicon carbide 4.5 410 400 120 ~ 180 ~22000
--------------------------------------------------------------------------
R = sigma_f (1 - nu) / (E times alpha) -> deg C of a sudden quench survived
R' = R times k -> ranks parts under a steady flux
low alpha wins a QUENCH; high k wins a slow GRADIENT.Two honest lessons hide in that table. First, look at where alumina lands: strong (350 MPa) and stiff, yet an R of only about 90 degrees C, because its high modulus and eightfold-larger expansion swamp its strength. Alumina is a superb wear part and a poor thermal-shock part, and no amount of extra strength fixes that — the lever it needs is lower alpha, which its chemistry will not give. Second, the winner of the quench, fused silica, is one of the weakest materials in the table at about 70 MPa. Strength is not destiny here; the ratio sigma_f / (E times alpha) is. That is also why the glass-ceramic cordierite, with an alpha near 1.5 times 10^-6, quietly runs your ceramic cooktop and the honeycomb of a catalytic converter, shrugging off the flame-to-cold cycling that would shatter a stronger, higher-expansion body.
The Other Clock: High-Temperature Creep
Thermal shock is the fast death — a crack in milliseconds. But a ceramic held hot and loaded for months faces a slow one: high-temperature creep, the quiet, permanent sag of a part under a stress far below its fracture strength. A furnace roller bows between its supports; a turbine component stretches a hair each thousand hours until its clearances close. Creep is a rate, not an event — you design against a strain accumulated over a lifetime, so the right question is not 'will it break?' but 'how much will it move in ten thousand hours at temperature?'
Where does that motion come from? Two paths, and the microstructure rung already handed you both. In a clean ceramic, atoms diffuse — vacancies drift through the lattice and along grain boundaries (Nabarro-Herring and Coble creep), so grains slowly change shape. Here fine grains, the very thing that raised your strength, hurt you: more grain-boundary area means faster diffusional creep, an honest trade-off you cannot escape. The second, usually nastier path is the glassy grain-boundary film a liquid-phase sinter leaves behind: nanometres thick and solid at room temperature, it softens to syrup near its glass transition and lets neighbouring grains slide past one another. This is why the same silicon nitride that is superb at 1200 degrees C can be crippled by a percent of the wrong boundary glass — and why creep-critical parts are fired to crystallize or minimize that film. High-temperature strength, in the end, is a grain-boundary property.
Turning the Logic Inside Out: The Thermal Barrier
Everything so far said high conductivity is your friend. Now meet the design where it is the enemy. A jet-engine or gas-turbine blade is a superalloy that would soften and fail above roughly 1100 degrees C, yet the gas rushing past it is far hotter. The fix is a thermal barrier coating — a thin ceramic skin, a fraction of a millimetre thick, whose entire job is to be a bad conductor of heat. With cooling air flowing inside the hollow blade, that skin sustains a temperature drop of one to three hundred degrees across its thickness, keeping the metal below its limit while the gas roars over the outside. Here you want the thermal conductivity as low as physics allows — the exact opposite of the quench problem.
The material of choice is yttria-stabilized zirconia, and its microstructure is deliberately, gloriously imperfect. Its heavy zirconium atoms and the oxygen vacancies the yttria dopant creates already scatter phonons hard, giving zirconia one of the lowest conductivities of any oxide. Then the coating is plasma-sprayed as a stack of splats riddled with 10 to 15 percent porosity and a lace of microcracks — pores and interfaces that scatter phonons even harder, dropping k toward 1 W per metre-kelvin. And that engineered porosity does a second job for free: the same microcracks make the coating strain-tolerant, so when the blade heats and cools and the ceramic and metal expand by different amounts, the cracks open and close instead of the coating spalling off. Zirconia is even chosen partly because its expansion (about 10 times 10^-6) sits close to the superalloy's, softening the mismatch further. This is the whole ladder in one part: porosity, usually the enemy of strength, engineered in on purpose to buy low conductivity and strain tolerance at once.