• PDF / 1,960,163 Bytes
• 12 Pages / 415.8 x 637.2 pts Page_size
• 73 Downloads / 278 Views

DOWNLOAD

REPORT

---

miniaturisation have needed new regimes of thickness and performance [1,2]. As the feature size LF 0 77 has shrunk, the oxide thickness x has fallen as (x/20nm) - (LF I-tmn) [3]. Nor should one forget cost, since that implies consistency with existing methods, so far as that is possible. Further challenges, perhaps more relevant to insulators on silicon, include ultimate miniaturisation, the reliability of processes for producing microelectronic components, and the reliability of these components in operation. The scientific emphasis is different. Ideas include self-organisation, metastasbility, excitation controlled processes (including recombination enhanced processes), quantum effects, and a group of effects associated with small space regions: alloy composition fluctuations, large local electric fields from isolated charges, diffusion-imposed limits, and all the problems of feature control and lithography. These challenges are not trivial. Even for 0.25 micron technology, a 4-10 nm gate oxide should be accurate to 0.2 nm across a wafer, with negligible property variations; technology beyond the 0. 1 micron gate is still more difficult. As feature sizes fall, the trends are to thinner oxides and lower thermal budgets. Operation at perhaps 3.5 V should still have acceptable breakdown and wear-out behaviour. Whereas breakdown and wearout of ultrathin oxides may be acceptable today, the density of defects responsible for low field breakdown needs to be reduced significantly for 0.1 micron VLSI CMOS manufacture. Reduction particulate, metal and organic contamination will help. Noise and trapped charge must be kept within defined limits. Oxide quality is a major driving force in oxidation studies. One need is passivity: freedom from charge trapping and breakdown. 3 Mat. Res. Soc. Symp. Proc. Vol. 592 © 2000 Materials Research Society

Another need is dielectric constant SiO 2 constants is a compromise dielectric constants needed for storage and forcontrol, the lowsince dielectric wanted for for both speedthein high operation. The challenges lead to at least two major questions [4,5]. First, can silicon remain supreme? The roadmaps suggest yes. Even where new materials have promise (whether III-Vs, nitrides, II-VI nanodots, or organics), compatibility with silicon is crucial. A second broad question is "Can the silicon oxidation problem be solved?" (oxidation including here processes involving nitrogen or other variants). "Solved" means achieving three capabilities: to define what the best possible oxide would be like, to estimate the performance of such an oxide, and to define such process steps as would come acceptably close to that performance in practice. So one would need realistic prediction of growth, usually under conditions inconsistent with the well-known DealGrove mechanism [6]. One would need to understand realistically the failure mechanisms and their quantitative prediction. We believe that solutions, in this sense, are possible both for the oxide grown on silicon and for the substitutes now being propo