Thermal Expansion, Anisotropy, and Microcracking

The asymmetric bond well: why heat swells a solid

In guide 1 you saw that heating a ceramic sets its atoms vibrating harder — that stored jiggle is its heat capacity. Here we follow one more consequence of the jiggle: almost every solid grows a little when it is heated, and shrinks when cooled. The number that says how much is the coefficient of thermal expansion, alpha (often called the CTE) — the fractional change in length per degree of temperature. It sounds like a dull materials constant, yet it is the single most important thermal property a ceramic engineer worries about, because it is the seed of thermal stress and, a few guides from now, of thermal shock.

Why should warming a solid push its atoms apart? Picture two bonded atoms as a marble resting at the bottom of a valley — the bond-energy well. If that valley were a perfectly symmetric bowl, heating would make the marble swing wider from side to side, but its average position would stay dead centre, and the solid would not expand at all. The real well is lopsided: the wall is steep on the squeeze-together side — atoms fiercely resist being pushed closer once their electron clouds touch — and gentle on the pull-apart side. So as the swings grow with temperature, the marble spends more of its time out on the shallow side, and its average spacing drifts outward. That outward drift, summed over billions of bonds, is thermal expansion. The stiffer and deeper the ionic-covalent bond, the more symmetric and steep the well, and the smaller alpha — which is exactly why most ceramics expand less than metals.

Reading alpha, and why low expansion is prized

Put a number on it. The coefficient is defined as alpha = (1/L) times (dL/dT), so the length change of a part is simply delta L = L times alpha times delta T. Work an example: an alumina rod 100 mm long, with alpha near 8 x 10^-6 per degrees C, taken from room temperature up to about 1000 degrees C (delta T = 1000). Then delta L = 100 mm times 8 x 10^-6 times 1000 = 0.8 mm — not much, but enough that a kiln shelf, a furnace tube, or a tight metal fitting must be given room to breathe. Multiply that swing by a stiff modulus and clamp it, and you get the stress that guide 4 will chase.

  MATERIAL               alpha (10^-6 per degrees C)   NOTE
  -----------------------------------------------------------------------
  fused silica           ~0.5      pure-SiO2 network too rigid to swell
  borosilicate glass     ~3.3      the labware that shrugs off a splash
  Si3N4                  ~3        low-alpha engineering ceramic
  SiC                    ~4.5      low-alpha engineering ceramic
  alumina  (Al2O3)       ~8        the workhorse oxide
  soda-lime glass        ~9        ordinary window / jar glass
  zirconia (YSZ)         ~10       high, for a ceramic
  MgO                    ~13       ionic rock-salt, loosely bound
  ---- for comparison ----
  steel                  ~12       a "low" metal
  aluminium metal        ~23       a "high" metal
  aluminium titanate     ~1        single crystal wildly anisotropic;
                                   microcracks to a near-zero bulk value

Now the payoff of a small alpha. Pour boiling water into an ordinary soda-lime tumbler (alpha near 9 x 10^-6) and it may crack; pour it into borosilicate labware (alpha near 3.3 x 10^-6) and it shrugs, because a smaller alpha means smaller strains and smaller stresses for the same temperature jolt. Push further to fused silica (alpha near 0.5 x 10^-6), whose stiff pure-SiO2 network hardly swells at all, and you get a glass you can heat white-hot and plunge into water. Engineers even chase near-zero on purpose: a glass-ceramic cooktop or telescope-mirror blank is grown with a crystalline phase that shrinks on heating, so it cancels the glass's expansion and the finished body barely changes size across hundreds of degrees. Low, and matched, expansion is the quiet superpower behind everything that survives the heat.

Anisotropy: expansion that has a direction

So far we have treated alpha as one number, and for a cubic crystal that is true: rock-salt MgO or a cubic spinel expands the same amount along every axis, because the bonds look identical whichever way you turn — the expansion is isotropic. But most ceramic crystals are not cubic. In a hexagonal, tetragonal, or lower-symmetry lattice the bonds are packed more tightly along some directions than others, so the well is a different shape along each axis, and the crystal expands by a different alpha in each direction. This directionality is thermal expansion anisotropy, and it is the rule, not the exception, for the oxides and silicates that make up most ceramics.

The spread can be mild or wild. Corundum alumina, the workhorse oxide, is hexagonal and expands only about 10 percent more along its c-axis than across it — a gentle anisotropy. Quartz is stronger, and famously lurches when it flips crystal form near 573 degrees C. At the extreme sit graphite, which expands many times more perpendicular to its sheets than within them, and a handful of oddballs like beta-eucryptite and aluminium titanate that actually shrink along one axis while growing along another, giving a near-zero or even negative average. Anisotropy is not a nuisance to memorise; it is a design knob — but, as the next section shows, it can also quietly wreck a part.

Microcracking: anisotropy tearing a polycrystal as it cools

Here is the trap that anisotropy sets. A fired ceramic is not one crystal but a polycrystal — millions of grains, each frozen in at a random orientation. Cool that body down from its firing temperature and every grain tries to shrink, but by a different amount along its different crystal axes, while its neighbours are all pulling different ways. Because the grains are cemented rigidly at their grain boundaries, none is free to shrink as it pleases, and the mismatch locks in a field of internal stress — tension across some boundaries, compression across others. If that stress is large enough it does exactly what stress does to a brittle solid: it opens tiny cracks along the boundaries, all on its own, with no external load. This spontaneous cracking on cooling is thermal expansion microcracking.

1. Is the crystal non-cubic? A cubic phase expands the same in every direction, so it carries no anisotropy stress and cannot microcrack this way. Only lower-symmetry crystals are at risk.
2. How big is the mismatch? The larger the difference in alpha between the crystal's axes (delta-alpha), the more strain the boundaries must swallow. Aluminium titanate, with one axis shrinking as another grows, is a champion offender; mildly anisotropic alumina is fairly safe.
3. How coarse are the grains? The stored stress in a grain scales with its size, so there is a critical grain size: below it the stress can never reach the Griffith threshold to open a boundary crack, above it it can. Fine grains stay whole; coarse grains split.
4. Add it up. Non-cubic, large delta-alpha, and grains past the critical size together decide whether a body cools intact or riddles itself with microcracks — which is why the same material can be sound when fine-grained and self-destruct when over-fired to coarse grains.

Microcracking sounds like pure bad news, and often it is: a web of boundary cracks drops the strength and the stiffness of the part, just like any other population of flaws, and can let moisture and corrosion creep in. But — honestly — it is not always the enemy. A gently microcracked body has a low elastic modulus (the cracks make it springy) and a very low apparent expansion (the cracks open and close to absorb the swelling), and both of those, as guide 4 will show, are exactly what buys thermal-shock resistance. That is why heavily anisotropic aluminium titanate, riddled with microcracks, is deliberately used for engine exhaust-port liners and thermocouple protection tubes: it is weak, but it laughs off the ferocious thermal cycling that would shatter a stronger, stiffer ceramic.

Taming and matching expansion in real parts

Step back and the theme of the whole rung comes into focus. Constrain a body from expanding freely and the stress it feels is roughly sigma = E times alpha times delta T — stiffness times expansion times temperature swing. Alumina (E near 380 GPa, alpha 8 x 10^-6) chilled suddenly by just 200 degrees C can feel about 380 x 10^9 times 8 x 10^-6 times 200 = 600 MPa, comfortably above its tensile strength, so it can crack from temperature alone. That is thermal shock in one line, and it shows why a low alpha is worth so much: it scales the stress down directly. Guide 4 turns this sketch into the proper resistance parameters R and R'.

There is a second, sneakier way expansion bites: mismatch between materials that are joined. A glaze fired onto a pot must have an alpha close to the body's; if the glaze contracts more on cooling it crazes into a web of fine cracks, if it contracts less it can flake off. A glass-to-metal seal, a metal fitting shrunk onto a ceramic, and a thermal-barrier coating sprayed onto a turbine blade all live or die by how well their expansions are matched across every temperature they will see. Engineers tune alpha to fit — blending a second phase into a ceramic, or choosing a composition — precisely so that neighbouring parts expand in step rather than tearing at each other.