The fabrication of a state-of-the-art silicon chip consists of a carefully composed sequence of “steps.” The length, complexity, and place of each step within this process is analogous to that of a musical phrase within a long sonata. Just as a sonata consists of a series of recurring but evolving themes, a “process” utilizes a number of basic techniques that recur in a variety of contexts. A typical modern process consists of about 350 “steps.” Roughly a fifth of them belong to the class of ionassisted processing. These range from a variety of ion implantation modules, through sputter deposition and reactive ion etching of thin films, to photoresist etching and ashing. Ion-assisted processing thus has a role of major technological importance. With the range of new processes currently under active research, this trend seems likely to continue.

The technological applications of ionassisted processing rest on the basic concepts developed in research over the past 30 years. This special issue of the MRS Bulletin reviews the fundamental concepts that underpin this processing today, and highlights areas of current research which may have future technological potential. The emphasis is placed heavily on silicon, both for illustrative purposes, and because Si processing is the major customer for ion-assisted technology. The articles pinpoint areas where materials research can play a critical role in establishing the fundamental understanding needed for a realistic assessment of technological potential.

An energetic ion undergoes two major mechanisms of energy loss: (1) screened Coulomb collisions with target atoms (nuclear stopping), and (2) interactions with bound or free electrons in the solid (electronic stopping). This article discusses how these two energy-loss processes—and related quantities such as the ion range, collision cascade behavior, sputtering, and spike effects—depend on the experimental parameters of beam energy, atomic number, and target density. We restrict ourselves to an individual incident ion and a time scale (~10−12 s) sufficient for the cascade to fully develop (Figure 1), but short enough for subsequent quenching or diffusion processes to be ignored. What happens at longer times and higher ion f luences, due to diffusion and cascade overlap phenomena, is treated in the article by Brown and Ourmazd in this issue of the MRS Bulletin.

Energy-Loss Concepts

Figure 2 summarizes the velocity (ε1/2) dependence of the two energy-loss processes, nuclear stopping (dε/dρ)n and electronic stopping (dε/dρ)e, in terms of the dimensionless Thomas-Fermi (TF) energy, ε, and length, ρ, units derived by Lindhard. The energy unit ε is defined as the ratio a/b where a is the TF screening length (typically, ~0.01 nm) and b is the distance of closest approach in an unscreened head on collision. The following simple relationship exists between TF and lab energies.

where M1, Z1, M2, and Z2 are the masses and atomic numbers of projectile and target atoms.

The preceding article by Davies treated the behavior of single ions as they slow down in a solid, colliding with atoms of the solid and often transferring enough energy to initiate atomic displacement cascades. For a 200 keV As ion with a speed ~108 cm/s and a range in Si of ~ 2 × 10−5 cm, the cascade is completed in less than 10−12 s. The small volume excited in the cascade relaxes over 15 decades or more in time in a series of increasingly deliberate steps. The experimental evidence for most of these decades is indirect. No probes with sufficient spatial and time resolution exist to examine the spatial and temporal evolution of collision cascades created by individual incident ions. The first few decades in time have been followed in metals by molecular dynamics (MD) calculations and also approximated by continuum theory. The results of MD calculations for 5 keV cascades in Ni are shown in Figure 1. The behavior is in-terpreted as indicative of local “melting,” consistent with the concept of a “thermal spike” in the material. Extension of MD calculations to times of nanoseconds and longer is still prohibitively expensive in computer time, and application to covalently bonded systems, where the interatomic forces are much more complex than in metals, has not yet been reported.

As Davies has noted, the atoms of the cascade rapidly lose their high vibrational excitation to atoms in the surrounding material, and in only a few picoseconds the system cools. This constitutes a very rapid thermal quench—slower only than the quench associated with deposition of individual atoms (or small atomic clusters) on a cold substrate. The quench may produce a small volume of material with a different phase than its surrounding. For example, a small volume of “hot” Si atoms may be highly disordered and may quench to produce a small amorphous Si region in a crystalline Si matrix.

This article describes the current status of implantation in silicon for ion doses up to but not including the onset of the amorphous state. Physical phenomena, applications to VLSI, and equipment issues are discussed and linked from the viewpoint of vertically integrated technologies (see Figure 1). In such a view, engineering applications are the customer of research on phenomena, and the equipment industry is the customer of both the research and applications community. The article highlights areas in phenomena, applications, and equipment requiring particular attention.

Studies in physical phenomena associated with ion implantation are motivated by the need for very shallow junctions. The shallowness is limited by ion channeling effects in crystalline silicon (which directly affect the as-implanted impurity profile) and by diffusion (which occurs during implant activation and annealing of damage). The inventory of ion-damage induced defects includes point defects, cluster defects, and dislocations arising from the combination of damage and annealing. A recent survey by Schreutelkamp et al. of low-dose implantation describes advances in characterization of channeling effects, and the formation of dislocation damage which, in turn, impacts thermally processed profiles. These effects need to be understood in terms of modern MOS (metal-oxide-semiconductor) processes.

Experiments in ion implantation were first performed almost 40 years ago by nuclear physicists. More recently, ion implanters have become permanent fixtures in integrated circuit processing lines. Manufacture of the more complex integrated circuits may involve as many as 10 different ion implantation steps. Implantation is used primarily at f luences of 1012–1015 ions/cm2 to tailor the electrical properties of a semiconductor substrate, but causing only a small perturbation in the composition of the target (see the article by Seidel and Larson in this issue of the MRS Bulletin). Applications of implantation had been limited by the small beam currents that were available, but recently a new generation of high-current implanters has been developed. This high-current capability allows implanting concentrations up to three orders of magnitude higher than those required for doping—enough to create a compound.

In the early years of doping of semiconductors by ion implantation, atomic displacements and residual lattice damage were considered undesirable byproducts of an otherwise controllable doping process. Steps were taken to minimize disorder during implantation and/or to remove it as completely as possible during a subsequent annealing process. In many cases, such as boron- or phosphorus-implanted silicon, annealing temperatures exceeding 900°C were necessary to achieve the desirable electrical properties. Indeed, removal of implantation damage remains a crucial issue, particularly as device dimensions shrink and the need has arisen for substantially lower processing temperatures. The advent of high-energy (MeV) implantation in specific processing steps and the increasing use of more complex (often multilayer) compound semiconductors has added further to the need to understand and control ion damage and its annealing in semiconductors.

Over the past decade, there has been a growing realization that implantation induced atomic displacements and defects can have significant advantages in processing. For example, it was realized early that ion damage, and resultant defect fluxes to and from lattice disruptions, can “getter” and trap undesirable impurities that would otherwise interfere with device operation. More recently, it has been possible to use ion beams to tailor damage structures and form amorphous-crystalline superlattices, to remove pre-existing damage and induce crystallization of amorphous layers at very low temperatures, to form ultrapure amorphous silicon for studying thermodynamic properties of this phase, or to mix films with semiconductors and form stable compounds such as silicides. Indeed, ion damage has been used to electrically isolate devices, to form optical waveguides and cavities, and to improve the junction properties of deeply doped layers. These issues are briefly reviewed in this article.

Why are low-energy ions relevant to the surface processing of electronic materials? The answer lies in the overriding trend of miniaturization in microelectronics. The achievement of these feats in ultrasmall architecture has required surface processing capabilities that allow layer addition and removal with incredible precision. The resulting benefits of greater capacity and speed at a plummeting cost per function are near legendary.

The ability of low-energy ions to enhance the precision of surface etching, cleaning, and deposition/growth processes (Figure 1) provides one basis for the interest in ion-assisted processes. Low-energy ions are used, for example, to enhance the sharpness of side walls in plasma etching and to improve step coverage by metal layers in sputter deposition. Emerging optoelectronic applications such as forming ridges for wave-guides and ultrasmooth vertical surfaces for lasers further extend piesent requirements, and low-energy ions again provide one tool to help in this area of ultraprecise materials control. Trends associated with the decreased feature size include the movement from wet chemical processing to dry processing, the continuing need for reductions in defect densities, and the drive toward reduced temperatures and times in process steps.

How do the above trends focus interest on studies of low-energy ion-assisted processes? In current applications, these trends are driving the need for increased atomic-level understanding of the ion-enhancement mechanisms, for example, in reactive ion etching to minimize defect production and enhance surface chemical reactions.