Stencil Mask Implanter Offers Flexibility, Process Simplification

Ion implantation using a stencil mask promises savings in manufacturing cost and process steps over traditional methods due to the elimination of resist coating, patterning, development and ashing steps. There also may be performance advantages because stencil mask implantation technology (SMIT) is performed on a chip-by-chip basis, potentially providing more uniform implantation conditions such as dose uniformity and, especially, angular distribution of the ion beam to the wafer. Researchers at Toshiba Corp. (Yokohama, Japan) and Ulvac Inc. (Tokyo) who developed this prototype equipment reported on these advantages in IEEE Transactions on Semiconductor Manufacturing (May 2002).

The SMIT tool consists of an ion source, analyzing magnet, ion scanner, ion deflector/collimator magnet, mask stage, 1:1 stencil mask, X-Y-Z-U stage and alignment optics. Electrostatic chucks hold the wafer and mask. Ions are implanted into selected regions of a chip through the mask, which can be positioned 10-100 µm above the wafer and controlled within 1 µm. The mask membrane is 5-10 µm thick. Alignment accuracy on the prototype system is 100 nm. The ion beam is scanned at 1 kHz, and parallelism of the incident beam is 0.2° compared with 0.5° on a traditional implanter. Typical ion beam current is 400 µA-1 mA with a shot size of 5 mm² .

This approach also reduces the charge build-up on the wafer by 90%, which usually is managed by a secondary electron flood or plasma electron flood hardware in traditional ion implanters. The researchers proposed that charge build-up in SMIT can be controlled through design and control of the mask bias, which were found to be linearly related.

Toshiba and Ulvac used the SMIT system to implant channel regions of a damascene metal gate (tungsten/titanium nitride) MOSFET. Control of threshold voltage values by implanted dose was as good as that obtained by conventional implantation. Different implantation conditions can be readily achieved on a chip-by-chip basis without adding process steps.

Challenges to SMIT include implanted pattern formation using a stencil mask and the need to harden the mask to prevent deformation with cumulative implanted dose. The standard silicon mask deformed after 1 × 10¹⁶cm^-2 doping, but a nitrogen-doped mask had higher Young's modulus and allowed cumulative dose >1 × 10¹⁸cm^-2 without deformation. This dose corresponds to 1000 wafers with 100 chips/wafer for a process requiring 10¹³ ions/cm^-2 doping. Studies are underway to further extend this lifetime.

Because it is impossible to make a stencil mask with isolated patterns, one solution comes from implanting twice or more, which allowed the formation of various patterns using a stencil mask with an array of 2 µm square holes with 4 µm pitch.

The SMIT approach is expected to deliver cost-of-ownership savings from the typical $9/wafer (¥ 1150/wafer) to less than half that value. This is due to a reduced raw process time of 1 hr/wafer lot (assuming 100 chips/wafer) relative to 3.8 hr/lot by conventional implantation processes. Ulvac expects to introduce the new technology implanter later this year.