Stress Concentration and the Griffith Criterion

The Strength That Went Missing

In the last guide you met a ceramic as it really is: fantastically stiff, held together by strong ionic-covalent bonds locked into a rigid cage, and brittle — it barely yields, it just breaks. Those bonds are ferocious. Add up the pull it would take to rip one plane of atoms clean away from the next, and you predict a theoretical strength of roughly one-tenth of the Young's modulus — for alumina, with a modulus near 400 GPa, that is around 40 GPa. Yet take a real alumina rod, bend it, and it snaps at a few hundred MPa — around a hundred times weaker than its own bonds. Ninety-nine percent of the strength has gone missing. Where?

The answer is the single most important idea in this whole rung: a ceramic does not break where it is — it breaks where it is weakest. The strength you measure is never the strength of the bonds; it is the strength of the worst flaw the part happens to carry. This guide tells the story of that flaw in two acts. First, how a tiny crack magnifies an ordinary stress into a monstrous one at its tip — stress concentration. Then, how A. A. Griffith worked out, with nothing but energy bookkeeping, exactly how big a flaw a given stress can tolerate before the crack tears loose — the Griffith criterion.

Stress Concentration: Why a Sharp Crack Is Deadly

Picture a wide sheet pulled evenly in tension. If it is flawless, the load streams straight through in parallel lines, every bond carrying its fair share. Now cut a small hole. The load lines cannot pass through empty space, so they crowd around the ends of the hole, bunching together like four lanes of traffic squeezing past a lane closure. Where the lines pack tightest, the local stress soars far above the gentle average you applied. That local pile-up is stress concentration, and it is the reason a ceramic gives up so far below its bonds.

In 1913 C. E. Inglis put a number on it. For an elliptical hole of half-length c with a tip radius rho, the stress right at the sharp end is sigma_tip = sigma_applied x (1 + 2 x sqrt(c/rho)). The whole story lives in that 2 x sqrt(c/rho): the longer the flaw (big c) and, above all, the sharper its tip (tiny rho), the bigger the magnification. A round hole (rho = c) merely triples the stress — a factor of 3, survivable. But a crack is not round. Its tip can be as sharp as a single broken bond, rho of a fraction of a nanometre, and then the factor explodes.

  LOAD LINES CROWDING AROUND A SURFACE CRACK
  (uniform pull from top and bottom; flaw of depth c, tip radius rho)

        sigma_applied  (spread evenly, far away)
     |    |    |    |    |    |    |    |
     v    v    v    v    v    v    v    v
  ---+----+----+----+----+----+----+----+---
                    ||  <- lines can't cross the crack,
                   /||\    so they detour around it and
                  / || \   PILE UP at the tip
     ===========<   ||   >===========  <- crack, depth c
                  \ || /
                   \||/
                    ##  <- TIP (radius rho): here the local
                        stress = sigma x (1 + 2*sqrt(c/rho))

   round hole (rho = c):  factor ~ 3      (harmless)
   sharp crack (rho << c): factor = HUGE  (deadly)

Put numbers on it. A surface crack just c = 10 micron deep (10^-6 m x 10 = 10^-5 m) with an atomically sharp tip, rho about 0.5 nm (5 x 10^-10 m), gives sqrt(c/rho) = sqrt(10^-5 / 5x10^-10) = sqrt(2 x 10^4), about 140. So sigma_tip is roughly 2 x 140 = 280 times the applied stress. Apply a mild 100 MPa far away, and the bonds at the tip feel about 28 GPa — right up at the theoretical bond strength. They snap, the crack steps forward one notch, its tip stays just as sharp, and the concentration on the new tip is just as savage. That self-feeding sharpness is why brittle fracture, once it starts, does not stop.

Griffith's Energy Bargain

Inglis's formula has a paradox baked into it. Let the tip get perfectly sharp (rho going to 0) and the predicted tip stress goes to infinity — which, taken at face value, says any flaw at all should make any material fail under any load. That is plainly false; cracked things sit around under load all the time. In 1920 A. A. Griffith cut the knot in a beautiful way: stop asking about the stress at the tip, which misbehaves, and ask instead about energy — over the whole crack, does letting it grow one step forward pay for itself?

The bargain has two sides. Growing a crack a sliver forward creates fresh surface, and every square metre of new surface costs energy — two new faces at surface energy gamma_s each, so 2 x gamma_s per unit area — because you are breaking bonds and leaving them dangling. That is the bill. But a longer crack also lets the stretched material on either side of it relax, releasing elastic strain energy that had been stored in those loaded bonds, like letting a bank of compressed springs spring back. That is the income. Griffith's rule is simply this: the crack advances only when the income covers the bill.

1. Store the energy. Pull the plate to a stress sigma. Every stretched bond now holds elastic strain energy — think of a whole room packed with loaded springs, primed to let go.
2. Add up the bill. A crack of half-length c has two new faces whose surface-energy cost grows in simple proportion to c — a straight line climbing with slope 4 x gamma_s. Small crack, small bill.
3. Add up the income. The crack relaxes a roughly circular patch of material around itself, so the strain energy it releases grows as c^2 — a downward parabola that starts flat but steepens fast as c grows.
4. Combine them. The total energy first RISES (the linear bill wins for a short crack, so it is stable and will not grow on its own), reaches a peak, then FALLS (the c^2 income wins for a long crack, so it runs away). That peak is the critical point.
5. Read off the answer. At the peak the two slopes are equal (dU/dc = 0). Solve it and out drops the Griffith fracture stress: sigma_f = sqrt(2 x E x gamma_s / (pi x c)).

Read that equation slowly, because it is the master key of the rung. Strength is not a fixed material constant — it depends on the flaw size c, and it scales as 1 / sqrt(c). Stiffen the material or raise its surface energy and strength climbs only as a square root; but a flaw four times bigger halves the strength. Put numbers on it for alumina: E = 400 GPa = 4x10^11 Pa, gamma_s about 1 J/m^2, worst flaw c = 10 micron = 10^-5 m. Then sigma_f = sqrt(2 x 4x10^11 x 1 / (3.14 x 10^-5)) = sqrt(2.5 x 10^16), about 160 MPa. One 10-micron flaw, and there goes your strength — squarely in the range a real alumina actually shows.

A Ceramic Is Only as Strong as Its Worst Flaw

Here is the whole idea in a picture from your kitchen drawer. A clean sheet of paper is surprisingly hard to tear — pull it straight and it just holds. But nick the edge with your nail and it tears effortlessly, straight from the nick, every time. You did not weaken the paper's fibres one bit; you simply gave the stress a place to concentrate and the tear a place to start. A ceramic is exactly the same: its strength is decided at its worst flaw, wherever the largest, sharpest defect happens to sit. Change nothing but the size of the biggest nick and you change the strength.

And flaws are everywhere in a real body — you met most of them climbing earlier rungs. A pore left behind by incomplete densification; an over-grown grain from abnormal grain growth; a hard agglomerate that never broke up in milling; a foreign inclusion; a groove ploughed by a machining diamond; even the rough valleys of a fired surface. The Griffith equation is magnificently indifferent to what the flaw is made of — it cares only how big it is and how sharp. As a rule the strength-limiting flaw is about the size of the largest pore or the largest grain, which is precisely why the whole microstructure rung fought so hard for fine grains and full density: shrink the worst flaw and you buy strength as 1 / sqrt(c).

Two honest riders before we leave Griffith. First, his ideal bill of 2 x gamma_s is too cheap for a real ceramic: a moving crack also spends energy on messy little processes near its tip, so the true cost — the fracture energy — is larger, and engineers bundle it into one measured property, the fracture toughness K_IC, which the next guide builds and which turns the same law into sigma_f = K_IC / sqrt(pi x c) — the very same 1 / sqrt(c), now with a number you can actually measure and a matching critical flaw size. Second, all of this assumes the crack is pulled open. Squeeze a ceramic instead and its cracks clamp shut, the concentration vanishes, and the same material can be roughly ten times stronger — which is exactly why ceramics live happily in compression (arches, bricks, dies, bearings) and are feared in tension.

What Griffith Starts, the Rest of the Rung Finishes

Griffith hands you the single most useful sentence in ceramic mechanics — strength equals toughness divided by the square root of the worst flaw — but it is an idealisation, and real ceramics bend its rules in four ways that the rest of this rung is built to handle. Naming them now is a map of where you are headed next.

First, the worst flaw is not the same size in every part. Nominally identical rods hide different largest defects, so their strengths scatter — a ceramic has no single strength but a whole distribution, described by Weibull statistics, the weakest-link idea that a chain breaks at its feeblest link. And because a bigger part packs in more flaws, it is more likely to hide a nasty one, so on average it is weaker — the size effect. That is guide 4. Second, some ceramics fight a growing crack: zirconia's transformation toughening throws an airbag in front of the tip, fibres bridge and pull out of a composite, tiny microcracks blunt the advance — so the toughness itself rises as the crack extends (R-curve behaviour). That is guide 5.

Third, a crack need not politely wait for the full Griffith stress. Hold a ceramic below it in a damp atmosphere and water molecules quietly attack the strained bonds at the tip; the crack creeps forward for hours, days, even years, until it finally reaches critical size and the part fails long after it was loaded — subcritical (slow) crack growth, also called static fatigue, a genuinely dangerous surprise. And fourth, the engineer's answer to all this scatter and slow creep: proof testing — deliberately overload every single part so the ones with the worst flaws break on the bench, in your hands, not later in service — and then design not to one heroic strength but to an accepted survival probability. Guide 5 again.