End of Range Defects
# End-of-Range Dislocation Loops
An ion implant leaves behind more disorder than just the dopant atoms it placed. Every nuclear-stopping collision knocks a silicon atom out of its lattice site, creating a vacancy and sending that atom off as an interstitial. Most of these recombine harmlessly during anneal. But past a certain implant dose, a population of excess interstitials survives, collects in a thin band just past the original amorphous/crystalline interface, and condenses into extended defects — rod-like {311} defects that ripen into dislocation loops — sitting at a specific depth the implant profile itself defines.
The excess interstitials come from damage that has nowhere to recombine. A high-dose or preamorphizing implant drives the near-surface region fully amorphous; deeper than that, the silicon stays crystalline but takes a beating from nuclear stopping on the way in, and every displaced atom there is an interstitial with no vacancy nearby to pair off with, since the vacancy-rich damage is concentrated closer to the surface. The result is a thin, buried layer of excess interstitials sitting right at the amorphous/crystalline boundary — the end-of-range band. During anneal those interstitials diffuse laterally within the band and agglomerate into {311} defects, which are metastable; hold the anneal long enough and the {311}s dissolve and re-condense into dislocation loops, extended line defects that are far more thermally stable and can persist through the rest of the flow.
Loops don't self-anneal the way ordinary point-defect damage does, and that's what makes them dangerous. Scattered vacancies and interstitials elsewhere in the implanted region mostly recombine with each other during the thermal budget and the lattice reorders; EOR loops break that rule twice over. They're thermally stable — once interstitials condense into a loop, dissolving it back into point defects costs real activation energy, so a loop that survives the main anneal can persist through the rest of the process flow as a permanent defect sitting at a fixed depth. And they act as a slow-release interstitial reservoir — as a loop ripens or partially dissolves, it keeps emitting interstitials back into the lattice over an extended time at anneal temperature, and silicon self-interstitials are exactly the species that drives enhanced diffusion of common dopants. That second point has a direct device consequence known as transient enhanced diffusion (TED): a dopant profile placed near an EOR band diffuses faster and farther than the same dopant would in undamaged silicon at the same anneal, which is a major reason shallow junctions are hard to hold at advanced nodes — the shallower the junction, the closer it sits to a preamorphizing implant's own EOR band.
Intel, TSMC, and Samsung process-integration teams fight this with three levers: preamorphization depth control, choosing the preamorphizing species and energy so the EOR band sits deeper than the target dopant profile; carbon or fluorine co-implantation, since both species are known interstitial traps that capture excess interstitials before they reach the active junction; and spike or flash/laser anneal schedules that reach peak temperature and ramp back down fast enough to activate dopants before interstitial-driven diffusion has time to act, trading a slower conventional furnace anneal for one that outruns the defect's own release kinetics.
Stopping power explains why implant damage exists and where along the ion's path it concentrates. End-of-range dislocation loops are what that damage becomes once a high-dose implant and a thermal budget interact — the direct physical reason a junction-depth target and an anneal recipe can never be chosen independently of each other.