ICF10B

the material (grain boundaries, surface heterogeneities, pre-existing microcracks, dislocation tangles). These heterogeneities usually evolve in time (fatigue, work hardening, electromigration, void formation). This makes the failure rate history dependent, and too complicated a problem to start with. So, we’ll focus on brittle fracture, where we perhaps can assume that the material far from the crack tip does not irreversibly transform. With this approximation, we have two cases. (1) The important case is when the failure is due to the inhomogeneities: distributions of microcracks[4], distributions of bond strengths[5], or other randomness[6]. In all of these cases, the rate is number of failures for a given increment of strain, not per unit time. Either the stress is large enough to trigger failure or not: a given sample, if it survives the initial loading, will survive forever. This leads to some elegant theories and simulations, but isn’t what I’m focusing on today. (2) The case we study is when the material is homogeneous, and the failure is triggered by a thermal fluctuation. This basically never happens, but it’s an interesting theoretical problem. So, fundamentally, what is the fracture nucleation rate for our brittle, homogeneous material? This is actually a deep theoretical physics question. The basic problem is defining what is meant at finite temperature by a material under tension. The lowest energy state (free energy, to be pedantic) for a large cube of brittle material under external strain can easily be seen to be broken in two. The elastic energy released is proportional to the volume; the surface energy cost is proportional to the cross-sectional area; for fixed external strain there will be a length at which breaking wins. One should note that this breaking strain in real materials isn’t very large: a cubic meter of iron will prefer to break in two when stretched by 10 m(140 m if one includes the energy lost due to plastic deformation during fracture). Crack nucleation is precisely the mechanism by which a material goes from the metastable stretched state to the state which is broken in two. There is a close analogy to the nucleation of crystals in supercooled liquids. In supercooled liquids[7] the liquid is a metastable state with a decay rate given by the thermal probability of generating a critical nucleus. For fracture, the stretched state is a metastable state with a decay rate given by the thermal probability of generating a crack of the size given by the Griffiths threshold. (The analogy even extends to the practicality of the thermal nucleation mechanisms. Homogeneous nucleation of crystals in supercooled liquids almost never happens: the crystal almost always nucleates on a dust particle, or on the surface...) Let’s do this calculation explicitly, for a two-dimensional straight crack of length ` , in a medium under external hydrostatic tension (a pressure P <0 ). One can solve the elastic equations to find the energy of this crack. Let be the surface energy (energy per unit length, in two dimensions, of the crack surface), Y be the Youngs modulus, and be Poisson’s ratio. If we define a critical crack length `c (the Griffiths[8] threshold) by `c =4Y = P 2 (1 2 ) (1) then the energy of the crack is E=2 ` ` 2 =`c: (2) It follows that cracks with ` > `c will grow to lower their energy, and that cracks with ` < `c will heal (at least in our model). The energy at the top of the barrier is E(`c) = `c = 4Y 2 =( P 2 (1 2 )) , and the probability for this critical crack to form by a thermal fluctuation is suppressed by the Boltzmann factor exp( E(`c)=kBT) . The fracture nucleation rate failure per unit volume per unit time will be given by some prefactor times this Boltzmann factor: failure = 0 exp( E(`c)=kBT): (3) (Calculating this prefactor 0 is quite a subtle problem[3].) This calculation is precisely analogous to the corresponding one for a critical nucleus of solid forming in a supercooled liquid[7]. Indeed, the crack as it nucleates will be filled by a gas of vapor atoms (think of fracture in a piece of dry ice, where the vapor pressure isn’t vanishingly small). The spontaneous fracture rate is precisely the nucleation rate of a gas forming in a super-stretched crystal. Remember, though, that we claimed from a fundamental point of view the difficulty is defining the stretched material. A stretched piece of material is fundamentally a transient: how do we use equilibrium statistical mechanics to study what is only a metastable equilibrium? The precise definition of the stretched, metastable state is as an analytical continuation from the compressed state[7, 2]. In particular, suppose we study the

RkJQdWJsaXNoZXIy MjM0NDE=