20 Ion Implantation

Dr. Ayushi Paliwal and Dr. Monika Tomar

7. Introduction

Ion implantation, as applied to semiconductor technology, is a process by which energetic impurity atoms can be introduced into a single-crystal substrate in order to change its electronic properties. Implantation is ordinarily carried out with ion energies in the 50- to 500-keV range. Basic requirements for implantation systems are ion sources and means for their extraction, acceleration, and purification. This is followed by beam deflection and scanning prior to impingement on the substrate. Ion implantation provides an alternative to diffusion as a means for junction fabrication in semiconductor technology. There are some disadvantages to ion implantation as well. The equipment is highly sophisticated and expensive, so that the technology is at an economic disadvantage when compared to diffusion (in those areas where diffusion can be used). The competitive disadvantage is worse for GaAs (where energy require• ments are in the 200- to 500-keV range) than for silicon (50- to 150-keV range). This, however, is offset by the fact that this technique lends itself to a high degree of automation, including in-line process monitoring as well as end-point determination. As a result, ion implantation is extensively used in both technologies today. A second disadvantage is that ion implantation results in damage to the semiconductor. Annealing at elevated temperatures is necessary to heal some or all of this damage. This is not a problem with silicon since its vapor pressure is extremely low at annealing temperatures. Furthermore, silicon is often subjected to later high-temperature processes, where this damage can be completely annealed. With GaAs, however, it is necessary to use a cap, or to conduct the anneal in an arsenic overpressure, in order to avoid dissociation and loss of stoichiometry. Neither approach is completely satisfactory, and this problem has not been solved at the present time.

8. Characteristics of ion implantation

The technique, however, has many unique characteristics which have led to its rapid development from a research tool to an extremely flexible, competitive technology. Some of these unique characteristics are now considered.

• Mass separation techniques can be used to obtain a monoenergetic, highly pure beam of impurity atoms, free from contamination. Thus a single machine can be used for a wide variety of impurities. Furthermore, the process of implantation is carried out under high-vacuum conditions-that is, in an inherently clean environment.
• A wide range of doses, from 1011 to 1017 ions cm-2, can be delivered to the target, and controlled to within ±1 % over this range. In contrast, control of impurity concentration in diffusion systems is at best 5-10% at high concentrations, and becomes worse at low concentrations. Furthermore, dopant incorporation during diffusion is sensitive to variations in the electronic character of the surface, whereas this is not the case with ion implantation. Thus ion implantation provides inherently more uniform surface coverage than does diffusion-particularly when low surface concentrations are required, and when the surface has a pattern of doped regions.
• Ion implantation is usually carried out at room temperature. As a result, a wide variety of masks (such as silica, silicon nitride, aluminum, and photoresist) can be used for selective doping. This gives great freedom in the design of unique self-aligned mask techniques for device fabrication, which are not possible with diffusion technology.
• Ion implantation provides independent control of dose and penetration depth. Many types of dopant profiles can be obtained by controlling the energy and dose of multiple implants of the same or different impurities. Both hyper abrupt and retrograde doping profiles can be made with relative ease in this manner.
• Ion implantation is a nonequilibrium process, so that the resulting carrier concentration is not limited by thermodynamic considerations, but rather by the ability of the dopant to become electronically active in the host lattice. Thus it is possible to introduce dopants into a semiconductor at concentrations in excess of their equilibrium solid solubility.
• Ion implantation can also be used for depositing a controlled amount of a charge species in a specific region of a semiconductor. Thus it has important applications in the threshold control of MOS devices.9.

PENETRATION RANGE

There are two basic stopping mechanisms by which energetic ions, upon entering a semiconductor, can be brought to rest. The first of these is by energy transfer to the target nuclei. This causes deflection of the projectile ions, and also a dislodging of the target nuclei from their original sites. If E is the energy of the ion at any point x along its path, we can define a nuclear stopping power Sn = (1/N)(dE/dx)n to characterize this process. Nuclear stopping results in physical damage to the semiconductor, which takes the form of point as well as line defects. Often the semiconductor can become amorphous and/or semi-insulating as a result of this process.

• A second stopping process is by the interaction of the ion with both bound and free electrons in the target. This gives rise to the transient generation of hole-electron pairs as energy is lost by the moving ion. We can define an electronic stopping power Se= (1/N)(dE/dx)e to characterize this process.

• The average rate of energy loss with distance is then given by

where E0 is the initial ion energy. The quantity R is known as the range. A more significant parameter, of interest in semiconductor technology, is the projection of this range along the direction of the incident ion, as shown in Figure 1 (a).

Because of the statistical nature of this process, this projected range is characterized by its mean value Rp, as well as by a standard deviation DRp along the direction of the incident ion. Thus latter term is referred to as the straggle. In practice, the ion beam also has a spread at right angles to its incidence, as shown in Figure l (b). This transverse straggle is denoted by DRt and is of importance in determining the doping distribution near the edge of a window which is cut in a mask.

Tansverse straggle can be ignored if the width of this window is large compared to the implant depth. For this case, the statistical distribution for an amorphous target can be described by a one-dimensional gaussian distribution function of the form

This function has a maximum at Rp, and falls off rapidly on either side of this mean value. In a single crystal semiconductors, if the incident beam is aligned to a major crystallographic axis, the ions can be steered for a considerable distance through the lattice with little energy loss, by a process known as channeling. Such channeled beams results in greatly increased penetration depth, as well as in reduced lattice order. Unfortunately, the range is now critically dependent on the degree of alignment (or misalignment) of the beam and the crystallographic axis, and also on the ion dose.

The formation of a tail in the doping profile can also be caused by a rapid interstitial diffusion process undergone by the particles once they have lost their incident energy. Such particles continue diffusing, even at room temperature, until they encounter suitable trapping centers such as vacancies or a surface. If t is an average trapping time and D is the diffusion coefficient associated with this process, it can be shown that the impurity tail will be of the form given by

In some situations, this tail region is enhanced treatment) after the implantation. However, t can suppressing its formation.when the semiconductor is annealed (by heat be very short in heavily damaged material, thus

10. Nuclear Stopping

Some of the ions impinging on a semiconductor conductor surface are reflected by collision with the outermost layers of atoms. Some impart sufficient energy to target atoms which are then ejected from the substrate by a process known as sputtering. In this section, however, our interest focuses on the nuclear stopping process, by which the majority of the incident ions penetrate into the semiconductor and transfer their energy to the lattice atoms by elastic collisions. Detailed calculations for nuclear stopping were first advanced by Lin hard, Scharff, and Schiett (LSS) and are available in the literature. Here the approach used in developing them is outlined in order to gain some insight into the implantation process.

Consider first the elastic collision of two hard spheres, each of radius R0, as shown in Fig. 2. Let 0 and E0 be the velocity and kinetic energy of the moving sphere (projectile), of mass M1. The mass of the stationary sphere (target) is M2; after collision, its velocity and kinetic energy are 2 and E2, respectively. In like manner, let 1 and E1 be the velocity and kinetic energy of the projectile after impact. The distance between spheres is given by an impact parameter p. For this model, energy transfer only occurs if p 2R0. A head-on collision is represented by p = 0.

Upon collision, momentum is transferred along the line of centers of the spheres. In addition, the kinetic energy is conserved. Solving for these conditions, the projectile deflects from its original trajectory by an angle q, such that

This equation shows that the energy transferred to the target particle is related to the scattering angle q. In addition, the velocity n1 of the projectile after impact is given by

This is also the energy lost by the projectile M1.

The situation is more complex when the projectile and the target have an attractive (or repulsive) force between them. Associated with this force is a potential V(r), which usually extends out to infinity. If this is the case, both particles will be continually moving (without physical encounter) as they approach each other, as shown in Figure 3(a), so that the impact parameter p also extends out to infinity. The energy transferred in this process is thus somewhat less than Tm, and depends on the impact parameter. The problem now reduces  to a system of point masses  moving under  the influence of this potential, and has been classically  treated by two-particle elastic scattering theory.  Here  the approach consists of transforming from laboratory coordinates to a set of moving coordinates,  whose center is located at the center of mass of the system. This renders the problem symmetric, as shown in Fig. 3b, with a deflection angle between the particles,where

where p is the impact parameter, and V(r) the interaction potential system, and is given by RM the mirumum distance of separation between the particles, function. Er is the energy of the ion in the center-of-mass

The integral of Eq. (11) can be evaluated for any given interaction potential function. The magnitude of q, in laboratory coordinates, can be shown to be given by

and the energy transferred by this process is given by

The choice of V(r) is based on physical consideration of what happens if two atoms, with atomic numbers Z1 and Z2, approach each other at a distance r. The force between these atoms is coulombic in nature, and is given by

This situation holds only as long as the electrons of each atom are excluded. Their inclusion, however, results in a screening influence on the nuclear repulsion, which causes a modification of the potential function. One general form of such a potential function is

where a is  a screening parameter,  and f(r/a)  is known as a screening function. One such function,shown in Figure 4, is known  as the Thomas-Fermi screening function, and  has been  found  to  be useful in calculations of nuclear stopping power.  Often it is approximated by a/r, where

and a0 is the Bohr radius (0.53 Å). For this approximation, the potential is inversely proportional to the square of the distance.

Using either the Thomas-Fermi screening function or its approximation, the deflection angle (and hence ) can be computed for any impact parameter p. This, in turn, allows computation of Tn(E,p), the energy lost by the incident particle in an elastic encounter with a single stationary particle.

In the practical situation of an amorphous target, energy is transferred to all particles. The total possible values of impact parameter, so that of thickness x and having an atom density of N, energy transferred is obtained by integrating over all

Using this approach and the Thomas-Fermi screening function, the relationship for nuclear stopping takes the form shown in Fig. 5. Here, we note that as an ion enters a semiconductor (i.e., for large values of E), it first transfers energy to the lattice at a relatively slow rate. As the ion slows down, this rate of energy transfer increases, and then eventually decreases to zero as the ion finally comes to rest.

Use of the approximate form of the screening function, where the potential is inversely proportional to the square of the distance, leads to a constant rate of energy loss. This is shown by the dashed line Figure 5. The approximate expression for this rate of energy loss turns out to be

11. Electronic Stopping

A comprehensive treatment of inelastic energy exchange processes, which accompany the excitation and ionization of electrons by collisions with incident ions, can only be provided by a quantum mechanical approach. However, semiclassical approaches can give reasonable estimates for the energy loss rate due to these processes. The fact that electron excitation and ionization do indeed occur is readily seen by considering a head-on elastic collision between an ion of mass M1 and energy E0, and an electron of mass m0. Since M1>>m0, the maximum energy transfer due to this process is given by

Excitation energies are a few electron volts, whereas M1/m0 @ 1000-2000. Consequently, both electron excitation and ionization are possible since the energy of incident ions during implantation is normally in the 50- to 500-keV range.

Approximate computations of the inelastic energy loss rate associated with this process have been made [10] by considering the projectile and target atoms as forming a quasi-molecule, with energy transferred due to electronic interaction as the particles undergo an encounter. As before, the total energy transferred to a target of thickness Dx can be obtained by integrating over all values of the impact parameter so that

Thus, the rate of energy loss due to electronic stopping is several (15-50) electron volts per angstrom, which is somewhat less than the energy loss rate for nuclear stopping.

Summary

• Characteristics of ion implantation
• Expression for penetration range was discussed
• Nuclear and electronic stopping were discussed