Left - Fig.4, Sheet resistance map of wafer implanted with Si ion dose of 3.5 (10¹²/cm² at 140keV.)

Right - Fig.3, Sheet resistance map of wafer implanted with Si ion dose of 8.0 (10¹²/cm² at 135keV.)

This 'standard implant' is always performed on the same implanter, and becomes the reference implant. Twice per week on all implanters and whenever a source change is required, the other side of one of these wafers is implanted with the same Be and Si dose as the standard side. Subsequently, the wafer is annealed, contactless sheet resistance measurements are made, and the delta in sheet resistance between the standard and test side of the wafer is monitored. Our gage sensitivity tests of this monitoring technique have shown that it can accurately detect activated dose errors less than 0.5% which is well within the specifications of typical production ion implantation equipment. This technique has greatly enhanced our ability to detect machine problems that cause minor dose perturbations, and which otherwise would have gone undetected. We have devised similar techniques for monitoring absolute temperature, and matching for our rapid thermal annealing equipment. A dedicated boule is implanted with a Si implant dose that ensures dose saturation in the GaAs substrate. This results in the activated sheet resistance being a function of the anneal temperature, and not small variations in the implant itself. We have determined that this monitoring technique gives us a sensitivity of 1 ohm/^oC. In general, this sensitivity provides capability to ensure temperature repeatability of ±3^oC."

The implant activation was first studied for a number of commercially available GaAs wafers from different boules grown by both VGF and LEC technique. The substrates were blanket implanted at Raytheon and TriQuint, in both cases through a dielectric cap. The resulting sheet resistance values and uniformity were measured with Lehighton contactless conductivity probes at Raytheon and TriQuint.

The main focus of the study is on the evaluation of performance of E-FETs and D-FETs fabricated on VGF and LEC substrates. The FETs were produced in routine fabrication runs of TriQuint's QED/A process. That process is TriQuint's most popular process and features 1 µm gate length recessed-gate D- and E-FETs.

Sheet resistance mappings were made of typical 100 mm VGF substrates after Si implantation and rapid thermal annealing performed at Raytheon and TriQuint. The sheet resistance measured by the Lehighton contactless conductivity probe is (332 ohm/sq. [Fig. 3] and 840 ohm/sq. [Fig. 4] for 29Si-ion doses and energies

of 8.0 (1012/cm2) at 135 keV and 3.5 (1012/cm2) at 140 keV at Raytheon and TriQuint, respectively. These sheet resistance values are comparable to those that can be achieved by LEC substrates, and within the specifications of these two companies. The uniformity is observed to be excellent. Typical standard deviation of sheet resistance after ion implantation and rapid thermal annealing is only on the order of 0.5-1%. This is also comparable to, or sometimes better than LEC substrates.

A low energy implant (85 keV) at Raytheon showed that the implant activation on VGF substrates can be 5-10% lower than on LEC wafers. However, further investigations show that the lower activation can be readily offset by slightly adjusting the implant dosage. This yields consistent results both for sheet resistance and uniformity.

Slightly lower implant activation and consequently higher sheet resistance was observed at TriQuint on some blanket implants of VGF substrates. However, E- and D-FET devices were all fabricated without any adjustment of implantation conditions. In one case where sheet resistance was higher (opposite to the trend just mentioned) for the LEC substrates, a recess etch adjustment was used to achieve comparable FETs. Overall, no inferior device performance related to the slightly lower implant activation for VGF substrates in some blanket implants has been observed. It can even be seen that VGF substrates yield better device properties and tighter parameter distributions in some test runs.

Brophy says [4], "For our HEMT production, one instrument is used for incoming inspection of GaAs epi wafers with sheet resistances of 30k ohm/square or higher. In that case we use the Lehighton system to clear wafers for use." Work is continuing to enable better repeatability at these high sheet resistances.

Contactless sheet resistance measurement has been used to improve threshold voltage control. Lin et al., of Hewlett-Packard Labs, Palo Alto, CA state [7] that: Reproducible MBE growth of epi-layers and a uniform and reproducible RIE process have resulted in excellent threshold voltage (VTH) control. This was achieved using three techniques: daily flux measurements of the group III sources, calibration wafer growths before MODFET IC wafer lots, and individual screening of wafers. The wafer screening is accomplished with a non-contact resistivity monitor which allows the overall sheet resistance of the as-grown material to be measured with a repeatability of ±2% over approx. 270 to 330 ohms per square. ( Note that the LEI 1310 1 sigma repeatability over this range is ±0.05%.)