Silicon-on-insulator (SOI) sub-strates are considered among the most promising materials for the continued production of faster ICs with low power consumption. From the device-scaling road map, the use of SOI substrates is predicted to increase in the near future.
The buried-oxide (BOX) layer present in SOI substrates is effective in controlling device leakage. Another advantage of the BOX layer is that the interface formed by the SOI/BOX interface seems to inhibit diffusion of implanted species past the interface. As channel lengths are reduced with each new device node, the transistor junction depth must also be decreased to control leakage and reduce resistance.
One of the critical challenges to forming 65-nanometer transistors is controlling the diffusion of dopants used to form the source and drain extensions. During the post-implant anneal step, the arsenic (AS)-implanted dopant profile is driven farther into the substrate, resulting in junctions that are deeper than desired. In a joint study into the effects of SOI substrates on dopant profiles conducted by researchers at Applied Materials Inc. and Soitec SA, it was found that the BOX interface provides a very sharp junction that is much shallower and more abrupt than junctions obtained with bulk silicon substrates. This study indicates that the use of SOI substrates can extend today's ion implant and anneal technologies to the 65-nm node and possibly beyond.
Three different thicknesses of SOI substrate were used in this study, all of them produced by the Soitec Smart Cut process. These had top SOI layers of p-type Si with approximate thicknesses of 550 angstrom , 700 angstrom and 1,100 angstrom . To provide a direct comparison of dopant behavior in SOI relative to crystalline silicon, these were implanted using a Quantum LEAP system and subsequently annealed using an RTP Centura system, in the same batch as prime p-type bulk-Si wafers. Additionally, some prime n-type wafers were implanted with boron (B) to enable sheet resistance comparisons.
Dopant profiles of low-energy B, BF2 and arsenic implants at doses and energies typical for source/ drain extension formation for 130-, 90- and 65-nm devices were investigated. The B and BF2 implants were annealed under N2 ambient whereas the arsenic implants were annealed under 10 percent O2 in N2.
To provide a comparison of diffusion, activation and uniformity between SOI and bulk substrates, three different thermal budgets were used: 10 seconds at 900 degrees C, 10 seconds at 1,000 degrees C and a spike anneal at 1,050 degrees C. In general, it was found that the AS-implanted dopant profiles are similar within the top Si layer of SOI and bulk Si. However, at the Si/BOX interface of the SOI substrates, the dopant concentration increases sharply, then falls off rapidly, forming a peak at the interface. Past the interface, the concentrations are less than those at equivalent depths in bulk Si. If the junction is at a concentration of 1 x 1,018 atoms/cm3, it can be seen that for the 5-keV boron implants, the SOI substrate yields a junction roughly 200 angstrom shallower than bulk Si. In the 200-eV case, where the projected range of the implant is much less than the depth of the BOX, there is no effect on junction depth. But if SOI substrates with shallower BOX layers are used, then the junction depth can be reduced. It was observed that the implants into shallower SOI substrates gave correspondingly shallower junction levels. In SOI substrates, the anneal tends to diffuse the boron up to the interface of the BOX. Therefore, a shallower BOX can produce a shallower junction.
Our study's conclusion: The comparison of thin-bonded SOI substrates with conventional bulk-Si wafers has shown a good overlap, using SIMS analysis of AS-implanted and annealed samples, and four-point-probe measurements of annealed samples. For boron implants into SOI, where the junction of the implant is remote from the top BOX interface, the implant profile and sheet resistance values are equivalent to those in bulk Si. For thin SOI, the junction depth is defined by the interface position. In the case of arsenic implants, equivalent activation is achieved with shallower junction depths, even when the BOX is remote from the junction. But more work needs to be done, including optimization of annealing and metrology techniques. We believe this study indicates that SOI substrates can be instrumental in inhibiting dopant diffusion during anneal and can contribute to junction scaling.