“In our derivation, individual atoms have no idea which way the concentration gradient points—every atom jumps blindly at random, and the net flux arises purely from the statistical difference in population between adjacent crystal planes.”Fundamentals of Materials Science · Ch. 7 Macroscopic PDE Meets Microscopic Random Walk
Why does high-pressure \(\mathrm{H_2}\) easily permeate steel walls at room temperature while \(\mathrm{N_2}\) barely leaks at all? And why does doubling \(\mathrm{H_2}\) pressure on both sides of a \(\mathrm{Pd}\) membrane boost permeation flux by only \(\sqrt{2}\) rather than \(2\times\)?
Why must furnace holding time be multiplied by \(4\) to double a gear's carburized case depth from \(0.5\text{ mm}\) to \(1.0\text{ mm}\)? And why does dendritic microsegregation in cast steel vanish in \(\sim 20\text{ h}\) at \(1100\,^\circ\mathrm{C}\), yet persist for trillions of years at room temperature?
Near its melting point, a \(\mathrm{Cu}\) atom makes \(10^8\) jumps per second—traveling \(2.5\text{ cm}\) in total path—yet achieves a net displacement of only \(2.4\;\mu\text{m}\). Why? And why does a compositionally uniform \(\mathrm{Cu}\) chip interconnect spontaneously grow voids and break open under high current density?
Tiny interstitial \(\mathrm{H}\) diffuses \(11\) orders faster than \(\mathrm{N}\) in \(300\text{ K}\) \(\alpha\text{-Fe}\), while \(\mathrm{H_2\rightleftharpoons 2H}\) dissociation yields Sieverts' \(\sqrt{p}\) scaling; scale-free PDEs depend on \(x/\sqrt{Dt}\) so \(2\times\) case depth takes \(4\times\) time; cast-steel dendritic homogenization \(\tau=l^2/(\pi^2 D)\) takes \(\sim 20\text{ h}\) at \(1100\,^\circ\mathrm{C}\) vs. \(\gt 10^{12}\text{ yr}\) at \(300\text{ K}\).
Uncorrelated jump directions cancel \(10^8\) steps down to \(\sqrt{n}\lambda\approx 2.4\;\mu\text{m}\) RMS drift; under \(10^6\text{ A/cm}^2\) electron wind, \(\mathrm{Cu}\) atoms drift along grain boundaries at \(J=(DC/k_{\mathrm{B}}T)Z^*e\rho j\), nucleating open-circuit voids at triple-junction flux divergences.
Next Lecture — Point Defects & Atomic Diffusion III: Solid-State Li-Battery Ionics, Kirkendall Marker Drift & Darken Equations, Parabolic Oxidation & Diffusional Creep.
1. At \(1000\,^\circ\mathrm{C}\), carbon in \(\gamma\text{-Fe}\) has \(D=3.5\times 10^{-11}\text{ m}^2/\text{s}\). After \(t=2\text{ h}\) (\(7200\text{ s}\)), what is the diffusion length \(L_{\mathrm{d}}=2\sqrt{Dt}\), and what total time doubles that depth?
\(\sqrt{Dt}=\sqrt{3.5\times 10^{-11}\times 7200}\approx 0.50\text{ mm}\), so \(L_{\mathrm{d}}=2\sqrt{Dt}\approx 1.00\text{ mm}\); since \(L_{\mathrm{d}}\propto\sqrt{t}\), doubling depth requires \(4\times 2\text{ h}=8\text{ h}\).
2. In a cubic crystal with jump distance \(\lambda=0.3\text{ nm}\) and jump rate \(\Gamma=10^8\text{ s}^{-1}\), what is \(D=\frac{1}{6}\lambda^2\Gamma\)? Why is \(Q\) for \(\mathrm{C}\) in \(\gamma\text{-Fe}\) far lower than \(\mathrm{Fe}\) self-diffusion?
\(D=\frac{1}{6}(3\times 10^{-10})^2\times 10^8=1.5\times 10^{-12}\text{ m}^2/\text{s}\); vacancy self-diffusion requires \(Q=\Delta H_{\mathrm{v}}+Q_{\mathrm{m}}\), whereas interstitial sites are mostly empty so \(Q=Q_{\mathrm{m}}\).
3. Combining Sieverts' permeation law, homogenization kinetics, and Einstein's relation, which statement is completely accurate?
\(\mathrm{H_2\rightleftharpoons 2H}\) gives \(C=S\sqrt{p}\), so doubling \(p\) raises \(J\) by \(\sqrt{2}\); negative thermodynamic factor \(\Phi\lt 0\) drives uphill diffusion; homogenization obeys \(\tau=l^2/(\pi^2 D)\propto l^2\).