Kirkendall Effect: Void Formation from Unequal Interdiffusion

The original experiment: brass, copper, and molybdenum markers

In Kirkendall's classic 1947 experiment, a block of alpha-brass (roughly 30% zinc, 70% copper) was wrapped with fine molybdenum wires and then electroplated with a thick layer of pure copper, creating a diffusion couple with inert markers positioned precisely at the original brass-copper interface. Molybdenum was chosen specifically because it is essentially insoluble in and does not diffuse appreciably into either copper or brass at the annealing temperatures used, making it an ideal passive tracer for the local lattice position. After annealing at temperatures around 785 degrees Celsius for extended periods, careful metallographic sectioning revealed that the molybdenum markers had moved inward, toward the center of the brass, by an amount that increased with the square root of annealing time, the classic signature of a diffusion-controlled process. Because the markers demonstrably moved relative to the ends of the sample, this proved that the two atomic species were not swapping places at equal rates through simple direct exchange or ring mechanisms as some earlier diffusion theories had assumed; some other conserved carrier of atomic motion had to be flowing asymmetrically as well. This single, carefully controlled experiment overturned decades of implicit assumptions about diffusion mechanisms and is now considered one of the foundational experiments of physical metallurgy. Smigelskas and Kirkendall's original 1947 paper was initially met with considerable skepticism, since it directly contradicted the widely accepted Frenkel and Wagner assumption that solid-state diffusion in substitutional alloys proceeded through a direct interchange mechanism in which two neighboring atoms simply swap lattice sites without needing any vacant sites at all; the observed marker motion was incompatible with such a direct-exchange picture, since direct exchange conserves the local lattice site count exactly and predicts no marker drift whatsoever, and it took several years of follow-up work before the vacancy-mediated explanation became the accepted consensus.

Vacancy-mediated diffusion and Darken's equations

The theoretical resolution, developed by Lawrence Darken shortly after Kirkendall's results were published, rests on recognizing that atomic diffusion in most crystalline solids occurs via a vacancy mechanism: an atom can only hop to a neighboring lattice site if that site happens to be vacant. If zinc atoms in brass have a higher intrinsic diffusion coefficient than copper atoms, meaning zinc atoms exchange with adjacent vacancies more readily, then the net atomic flux of zinc out of the brass region will exceed the net atomic flux of copper into it. Since each atomic jump into a vacancy is simultaneously a jump of that vacancy in the opposite direction, this imbalance necessarily corresponds to a net flux of vacancies flowing from the copper-rich side toward the brass-rich side, precisely canceling out the imbalance in atomic fluxes so that the total number of atomic sites is locally conserved. Darken formalized this by distinguishing the intrinsic diffusion coefficients of each species, D_Zn and D_Cu, from a single interdiffusion coefficient D̃ that describes the observable, marker-frame-independent evolution of the composition profile, related by D̃ = X_Cu D_Zn + X_Zn D_Cu, where X denotes local mole fraction. Darken further showed that the marker velocity itself is directly proportional to the difference between the two intrinsic diffusion coefficients and the local composition gradient, providing the quantitative link between the observed marker-plane drift and the underlying asymmetry in atomic mobilities, a relation now known as the Darken equation. One important subtlety Darken's analysis clarified is that the interdiffusion coefficient D̃, unlike the two intrinsic diffusion coefficients, is composition-dependent even when the intrinsic coefficients themselves are treated as constants, because it is a concentration-weighted combination of the two; extracting the individual intrinsic diffusivities of each species from a measured interdiffusion profile therefore requires combining the classical Boltzmann-Matano analysis of the concentration profile with an independent measurement of the marker velocity itself, a technique that remains the standard method for determining intrinsic diffusion coefficients in binary alloy systems today.

From vacancy supersaturation to void nucleation

The vacancy flux required to compensate unequal atomic diffusion does not simply disappear once it arrives at its destination; it must be absorbed somewhere, typically at dislocations, grain boundaries, or free surfaces that act as vacancy sinks capable of annihilating excess vacancies by allowing atomic planes to locally contract. If the rate at which vacancies arrive at a given region exceeds the rate at which nearby sinks can absorb them, which is common when diffusion couples are annealed at temperatures where sink density and sink efficiency are limited, the local vacancy concentration rises well above its thermal equilibrium value, a state called vacancy supersaturation. Beyond a critical supersaturation, it becomes energetically favorable for vacancies to aggregate directly into three-dimensional clusters rather than being individually absorbed at distant sinks, nucleating small voids that then grow by continued vacancy condensation, exactly analogous to how supersaturated vapor condenses into liquid droplets. These Kirkendall voids characteristically form in a band located on the side of the diffusion couple from which the faster-diffusing species departed, since that is the region experiencing the net inward vacancy flux, and they are readily observed by cross-sectional electron or optical microscopy as porous bands running roughly parallel to the original interface. The size, density, and distribution of Kirkendall voids depend sensitively on annealing temperature and time, the diffusivity mismatch between the two species, and the density of dislocations and grain boundaries available to act as competing vacancy sinks. Because sink efficiency generally improves with increasing dislocation density, cold-worked or fine-grained metals tend to develop less severe Kirkendall porosity than well-annealed, low-dislocation-density samples subjected to the same interdiffusion conditions, an observation that has motivated deliberate microstructural engineering, including pre-straining or controlled grain refinement, as a practical strategy for suppressing unwanted void formation in diffusion-bonded and coated components.

Kirkendall porosity as an engineering reliability concern

While scientifically illuminating, Kirkendall voiding is a genuine practical liability in modern electronics and materials engineering, most prominently in solder joint reliability. In lead-free solder interconnects, particularly the widely used tin-based solders bonded to copper pads, intermetallic compound layers form at the solder-copper interface, and the interdiffusion of copper and tin through these intermetallic layers is markedly asymmetric, leading to Kirkendall void formation concentrated near the copper-to-intermetallic interface. Because these voids weaken the joint mechanically and can coalesce into larger cracks under thermal cycling or mechanical stress, Kirkendall voiding has become a significant reliability concern in ball-grid-array packages and other solder interconnects used throughout the semiconductor packaging industry, driving substantial process-engineering effort toward controlling intermetallic growth kinetics, using diffusion-barrier layers such as nickel, and optimizing reflow and aging thermal profiles to minimize void density. Kirkendall-type porosity is likewise an important consideration in diffusion bonding and cladding processes used to join dissimilar metals in aerospace and nuclear engineering applications, where unexpected void bands can compromise the mechanical integrity of a joint that otherwise appears metallurgically sound under casual inspection. Nondestructive and cross-sectional characterization methods, including scanning acoustic microscopy and focused-ion-beam cross-sectioning combined with electron microscopy, are routinely used in failure analysis laboratories to detect and quantify Kirkendall void bands before a component is placed into service, and predictive interdiffusion models based on measured intrinsic diffusion coefficients are now integrated into solder joint reliability simulations used during electronic package qualification. Beyond electronics, Kirkendall porosity is also a recognized concern in the aluminizing and chromizing diffusion coatings applied to nickel-based superalloy turbine blades for high-temperature oxidation protection, where void formation at the coating-substrate interface can locally reduce load-bearing cross-section and accelerate crack initiation under the combined thermal and mechanical stresses experienced in a jet engine, making Kirkendall-aware coating process design an important part of aerospace component qualification.

The Kirkendall effect as a synthesis tool: hollow nanostructures

In a striking reversal of its role as an engineering defect, the Kirkendall effect has become a deliberate, powerful synthetic strategy in nanomaterials chemistry for producing hollow nanoparticles and nanoshells with precisely controlled wall thickness. When a nanoscale metal particle, such as cobalt or nickel, is exposed to a reactive species like oxygen or sulfur, a shell of the corresponding compound (oxide or sulfide) forms at the particle surface, and this shell effectively acts as a nanoscale diffusion couple, since the reaction proceeds through the outward diffusion of metal cations through the growing shell far faster than the inward diffusion of oxygen or sulfur anions. The resulting flux imbalance drives a net inward flow of vacancies into the shrinking metallic core, and because a nanoparticle core has no distant interior sink comparable to the grain boundaries and dislocations available in a bulk polycrystalline sample, the vacancies condense directly at or near the particle center, nucleating a single central void that grows as the reaction proceeds, ultimately consuming the entire metallic core and leaving a hollow, chemically transformed shell, a phenomenon named the nanoscale Kirkendall effect after its discovery and popularization by Alivisatos and coworkers in the mid-2000s. This mechanism has since been generalized to produce hollow and multi-shelled nanostructures from a wide range of material combinations, with applications spanning catalysis, battery electrode materials, and drug delivery, all exploiting the same fundamental diffusional asymmetry that Kirkendall first identified as an engineering curiosity in bulk brass-copper diffusion couples. Researchers have further learned to control not just whether a hollow structure forms but its detailed morphology, including multi-chambered or yolk-shell architectures, by tuning reaction temperature, particle size, and the specific cation-anion diffusivity ratio of the chosen material system, effectively turning what began as an unwanted defect mechanism in bulk metallurgy into a precision nanoscale fabrication toolkit whose design rules trace directly back to the same vacancy-flux physics Kirkendall and Darken worked out for brass and copper more than seventy years ago.

Frequently asked questions

Try it live

Everything above runs in your browser — open Kirkendall Effect: Void Formation from Unequal Interdiffusion and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.