More on high-current ion implanters
02/01/2003
Regarding SST's October feature article, "Angle control in high-current ion implanters" (p. 39, by Leonard Rubin of Axcelis Technologies Inc.), the author's assertions can be accepted for legacy technologies. Today, however, it is much more relevant to assess angle control requirements for 300mm wafers at 130nm and below than to assess 200mm wafers at 350nm.
For example, the article understates the importance of beam incident angle control in the critical self-aligned source-drain extension (SDE) implant application, as compared to dose and energy control. At 130nm technology, incident angle variations in the extension implant are a more important effect than dose variations; angle variations of only ~1.0° produce threshold voltage shifts equivalent to more than 12% dose variation. Device performance is more sensitive to angle control than to dose control in the self-aligned extension structure implant process at and below the 130nm node.
The article also states that on single-wafer, high-current tools, "Incident angle errors may therefore be different with each beam setup, leading to lot-to-lot variations in incident angle control." In reality, the incident angle on all implanters will vary unless it is measured and action taken to compensate. Only single-wafer, high-current implanters do this in production today. Batch tools use beam current maximization routines without measurement and control of incident angles.
Single-wafer, high-current implanters are in production that use the control of incident angles in the ribbon beam and apply two-axis adjustment of the wafer orientation to the beam to deliver repeatable incident beam angle. This technology monitors, controls, and interlocks total incident angle. Batch high-current tools do suffer from significant beam angle variations due to geometric effects. Therefore, they cannot provide a true 0° implant process with <0.5° beam angle variations (cross-wafer, lot-to-lot, tool-to-tool) for the critical extension implant.
Batch tools compensate for this 0° tilt deficiency by performing quad-repositioning implants with low tilt angles, which attempt to "average out" the unacceptable cross-wafer angle variations at 0° tilt. This approach was adopted when undesirable device variations began to appear in the 0° SDE process on batch tools around the 250–180nm technology nodes.
The statement, "Serial high-current implanters using asymmetric magnetic angle correction are forced to trade off beam parallelism for dose uniformity, and the situation changes with each beam set up," is incorrect for ribbon beam, single-wafer, high-current tools. Before optimization, uniformity and incident angles have some characteristic value. After optimization, both the uniformity and angles of the ribbon beam have lower deviation values. There is no inherent trade-off between the two parameters.
The reference employed to substantiate the key claim of this paper is misused. Jasper, et al. of Motorola compared serial and multiwafer implanters for extension and threshold adjust implants, and concluded, "there was no difference in the devices between the two implanter architectures, and that the use of quad-repositioning implants provided the best device control." This was a well-executed study by a credible source; however, the work in the reference cited used an NMOS, RF device at 350nm technology fabricated on 200mm wafers.
For the older technology cited by the author, batch architecture is sufficient. Implant angle control becomes fundamentally important for CMOS technology at and below 130nm, where angle sensitivity in the extension implant becomes a dominant effect and is a primary implant process control, as important as gate length control, at the 90–100nm and 65–70nm nodes.
For ≤130nm devices, however, the sensitivity to ion beam incident angle variations in the self-aligned extension implant application is a major effect, producing significant and undesirable device performance variations that become more severe as scaling proceeds and 300mm wafers are adopted. These variations can only be avoided by using a low-current density ribbon beam with in situ measurement, closed loop control, and setup-to-setup, beam-to-wafer incident angles interlocking.
Robert Simonton, Sandeep Mehta,
Michael Chase
Varian Semiconductor Equipment
Associates Inc.
Newburyport, MA
Axcelis' response
Axcelis stands by all the statements made in our October article. We agree that the need for precise ion beam angle control increases with shrinking device geometries. The point of the article is that when all sources of beam-to-wafer angle errors are considered, single-wafer and multiwafer implanters provide nearly identical performance.
Since the SST article was written, a presenter from Varian admitted across-wafer (beam parallelism) errors of ±1.15° on a single-wafer high-current implanter at the IEEE XIV Conference on Ion Implantation Technology (Sept. 23–27, 2002, Taos, NM). Two-axis adjustment of the wafer can only compensate for errors in the beam centroid angle; it is impossible to correct beam parallelism errors with this technique. Even with wafer angle adjustment, the same presentation admitted residual wafer-to-wafer variations of ±0.5°. Multiwafer implanters routinely achieve this level of performance.
Furthermore, there is an inherent trade-off between uniformity and angle control. If a ribbon beam is formed with any across-beam variations in dose, the only way to compensate is to steer ions from the denser regions to the other ones by altering their trajectory, and hence their angle. The writers admit this, saying that optimization of the ribbon beam leads to "lower deviation values."
It is misleading to infer that the results in the Jasper, et al. reference are irrelevant to future geometries. CMOS logic was not investigated, but rather an advanced RF circuit that requires minimal threshold voltage variations between matched transistor pairs for proper operation. This results in an extreme sensitivity to shifts in beam angle, which is precisely why these advanced devices were chosen for investigation.
Because single-wafer and multiwafer implanters have essentially identical angle control performance, tool-purchasing decisions are based on other factors such as productivity, where multiwafer enjoys a clear advantage.
Leonard Rubin
Axcelis Technologies Inc.
Beverly, MA
Clarification
In our January issue, in "Accurate measurements in the DUV PEB process" on p. 56, the equation should read:The time constant is expressed as
 |
null