Implant Damage Engineering with Reduced Temperature: How cold is enough?

By Michael Ameen, Ph.D., Chief Process Scientist

Axcelis introduced wafer temperature control as a process variable for matching implanter platforms over 10 years ago.  Now, chip manufacturers are investigating reduced temperature implants to control leakage and improve activation.  This article investigates the limits of cold implants as far as lowest temperatures required.

Reducing implant temperature has the effect of minimizing the “self-annealing” component of the implant; the relaxation of damage that occurs in the very short time after the ion penetrates into the silicon lattice, a diffusion process sensitive to substrate temperature.  Reducing or eliminating self-annealing results in a larger net displacement of Si atoms and altered damage profile leading to, for example, formation of a thicker amorphous layer, and reduced end-of-range (EOR) damage that can be responsible for device leakage. Other advantages are improved epitaxial regrowth during anneal, a sharper amorphous/crystalline boundary, and reduced dopant movement from transient enhanced diffusion (TED).

A natural question is “how cold does the wafer temperature need to be in order to achieve the desired improvements?”  Practically speaking, at what temperature does “damage saturation” occur, the point at which no further improvements are observed?  The answer to this simple question is actually quite complex, and depends on many variables, including primary factors species, energy, dose, and beam current, as well as equipment specific factors such as effective dose rate and wafer cooling efficiency. The final damage profile in an implant is determined through a complex interaction between the rate of damage introduction (dose rate) and elimination of damage relaxation (temperature).   A platform featuring a high dose rate, such as the Optima- HDx spot beam implanter, can achieve damage saturation at higher temperatures compared to other platforms.  This allows manufacturers to implement a cost-effective damage engineering solution without the need for cryogenic temperatures.

To illustrate these effects, Axcelis has studied the damage saturation for two species under active investigation for use in 20-nm devices, high mass Ge, and low mass C.    Ge is heavier and larger than Si (72 amu compared to 28 amu), and damages the lattice at a relatively high rate, leading to formation of an amorphous layer at low doses, typically 1-5 x 1014 ions/cm2.  For this type of implant, damage is saturated at about -20oC wafer temperature, and lower temperatures have no impact on the therma-wave signal, as seen in Fig.1.  For this implant, further reduction of temperature would have no impact on the materials characteristics of the silicon.

Figure 1
Figure 1: Ge implant indicating saturation of damage at -20oC.

In the case of carbon, a different behavior is observed.  Since carbon, 12 amu, is lighter than silicon, the damage rate is reduced, and temperatures between -50 and -60oC are required before damage is saturated.  Figure 2A shows the roll-off of TW signal as damage begins to saturate as a function of temperature.  An analysis of amorphous layer thickness by Transmission Electron Microscopy (TEM) (Fig 2b) indicates that no change in thickness at lower temperatures, a reliable indicator that damage saturation has been reached for these conditions.

Figure 2A
Figure 2A:  Carbon implant indicating damage saturation at -50-60oC.

Figure 2B
Figure 2B: Transmission Electron Microscopy (TEM) analysis of the amorphous layer formed by a carbon implant at reduced temperatures.

One additional benefit of damage engineering is the ability to form an amorphous layer at lower doses than possible using standard room temperature implants.  This is especially desirable when introducing carbon as a co-implant, as a high dose of carbon is known to cause leakage in devices.  For low mass, low dose implants, the importance of high dose rate in combination with reduced temperatures becomes critical.  Figure 3 shows the amorphous layer formed from a carbon 3 keV 4x1014 ions/cm2 implant.  In this case, the dose rate of carbon was increased through the use of a molecular implant.

Figure 3
Figure 3: Carbon implant (3 keV 4x1014 ions/cm2) showing amorphization at minimal carbon dose. A) Molecular Carbon showing amorphization B) Cold (-70oC) Monomer carbon shows incomplete amorphization

Control of implant temperature, both warm and cold, is becoming an important part of process control in modern implanters.  Lower implant temperatures offer several potential advantages, with each application requiring optimization of dose and temperature. 

To learn more about Axcelis’ product offerings and expertise in damage engineering, please download a poster on this topic presented at the recent IIT2012, or visit www.axcelis.com.