a defect cluster
Point defects do not always stay lonely and scattered. When there are many of them, or when they carry opposite charges, they attract one another and clump together into small groups — a defect cluster. Think of magnets sprinkled on a table: a few isolated ones stay put, but pack in enough and they snap together into clusters. In crystals the driving force is the same idea of lowering energy: two defects whose strain fields or charges partly cancel are more comfortable side by side than far apart, so they associate.
The simplest cluster is a pair: a positively charged defect and a negatively charged one binding together (like a vacancy and a nearby dopant ion of opposite effective charge), or two vacancies joining into a divacancy. But clusters can grow far more elaborate. The most famous is the Koch-Cohen cluster in nonstoichiometric wustite, Fe(1-x)O, where iron vacancies and interstitial iron ions organise into a repeating little unit (roughly four cation vacancies grouped around one interstitial Fe3+) rather than scattering at random — the crystal's way of accommodating a large defect population efficiently. At high enough concentrations such clusters can even order into a superstructure of their own.
Clustering matters because it changes everything the individual defects would have done. Bound-up defects are less mobile, so clustering slows diffusion and ionic conduction; the simple dilute-solution mass-action equations of defect chemistry break down once defects associate, which is exactly why heavily nonstoichiometric oxides need cluster models rather than isolated-vacancy models. Under irradiation, clustering of self-interstitials and vacancies into loops and voids is what causes metals to swell and embrittle. So the tendency of point defects to gather is not a footnote — it is often the main event in real, defect-rich materials.
In wustite the iron vacancies do not sit alone: they gather with interstitial Fe3+ ions into Koch-Cohen clusters, a repeating grouping of about four vacancies around one interstitial. This clustering is why real Fe(1-x)O deviates from what a simple isolated-vacancy model predicts, and why its defect population had to be worked out by diffraction rather than assumed.
Enough oppositely charged or strained defects attract and clump; in wustite they form ordered Koch-Cohen clusters.
Once defects cluster, the dilute-solution assumption behind simple defect chemistry fails — you can no longer treat each defect as independent. Highly nonstoichiometric and heavily irradiated crystals almost always demand cluster models, not isolated-point-defect models.