Current (and future) microelectronics fabrication requirements place unprecedented demands on the fidelity of plasma etching. As device features shrink to atomic dimensions, the plasma etching processes used to define these devices must resolve these scales. By separating etching processes into cycles of multiple, self-limited steps, different physics processes which are closely coupled in traditional plasma etching can be largely decoupled and separately optimized. This technique, atomic layer etching (ALE), can ideally remove uniform layers of material with consistent thickness in each cycle. ALE holds the promise of improving uniformity, reducing damage, increasing selectivity, and minimizing aspect ratio dependent etching (ARDE) rates. The practical implementation of ALE depends on how close to ideal the system can be operated and the tolerance to nonideal conditions. In this paper, results are discussed from a computational investigation of the consequences of nonidealities in the ALE of silicon using Ar/Cl 2 plasmas for both two dimensional trenches and three dimensional features. The authors found that ideal ALE requires self-limited processes during all steps of the ALE cycle. Steps that include continuous (non-self-limited) etching reactions reduce the ability of ALE to decouple process parameters. In addition to an etch depth that depends on pulse length per cycle, non-self-limited processes can reintroduce ARDE and produce surface roughening. By controlling subcycle pulse times, these deleterious effects can be minimized, and many of the benefits of ALE can be restored. Even nonideal ALE processes, when properly optimized, still provide benefits over continuous etching with similar chemistries and ion energy distributions. Using fluxes generated by a conventional inductively coupled plasma reactor, an example ALE process is able to clear the corners in a three-dimensional fin based field effect transistor case study with significantly less over-etch than the continuous process.