Ryshkewitch-Duckworth Porosity-Strength Relation

Eugen Ryshkewitch measured the strength of alumina and zirconia over a range of densities and published the result in 1953 in the Journal of the American Ceramic Society; Winston Duckworth gave it the exponential form everyone now uses, in the same volume. Sixty years and a great many sintered materials later, $\sigma = \sigma_0 e^{-bP}$ still fits.

The number that makes the point: at 10% porosity with a typical $b$ of 7, the strength is 49.7% of the fully dense value. Ten percent of the material is missing and half the strength is gone. A simple load-bearing-area argument — the section is 10% smaller, so it should carry 10% less — would have predicted 90%. The gap between 90% and 50% is the whole content of the equation, and the whole reason the last few percent of sintered density are worth so much trouble.

Pores are not merely absent material. They are stress concentrators, they are crack starters, and in a brittle body governed by weakest-link statistics the largest pore sets the strength all by itself. The Griffith relation gives the mechanism: strength goes as the inverse square root of flaw size, so a pore that is twice the size of its neighbours costs 30% of the strength no matter how few of them there are. This is why hot isostatic pressing exists, why sintering aids and dopants that promote densification are worth their cost, and why a datasheet strength quoted without a density is close to useless — the two numbers only mean something together.

The constant $b$ carries the pore shape, and it is not universal. Around 4 for well-rounded, isolated, evenly distributed pores; 7 or more for the angular, interconnected porosity that a partly sintered body has. Two materials at identical measured porosity can therefore differ substantially in strength, which is a useful thing to know when a supplier's density figure looks reassuring. Fit $b$ to your own process, from at least three or four densities across the range you care about, and plot $\ln\sigma$ against $P$: the relation says that plot is a straight line of slope $-b$. If it curves, the pore population is changing character as it changes in amount, and that curvature is real information about the sintering rather than an inconvenience.

The relation is empirical. It has no derivation. It fits well from a few percent porosity up to perhaps 40%, and beyond that the body is a foam and obeys cellular-solid scaling — Gibson and Ashby's power laws — instead. It also has a companion for stiffness, $E = E_0 e^{-b_E P}$, of the same form but with a different constant. That difference matters more than it looks: porosity does not reduce strength and modulus by the same factor, and their ratio is precisely what governs thermal shock. A porous refractory can end up with better thermal shock resistance than a dense one because $E$ fell faster than $\sigma_f$, which is the quantitative version of why furnace linings are not made of dense alumina.

$\sigma_0$ here is an extrapolation to zero porosity, not a measured property — most real ceramics are never made fully dense. Treat it as a fitted intercept, quote the porosity range it was fitted over, and use it for what it is genuinely good at: comparing processes. Two routes to the same material should extrapolate to the same $\sigma_0$, and when they do not, something other than density differs between them. Grain size is the usual culprit, since strength falls with grain size on a Hall-Petch-like relation of its own, and a coarse microstructure can look identical on a density measurement while being substantially weaker.