Advancements in Angle Control and Correction to Meet Today's Critical Device Requirements
By Dennis Kamenitsa, Applications Scientist
The implant beam angle, for current and future device scaling, has become critical with respect to device parametric control.
Historically, control of the basic implant parameters of dose, energy, and species were sufficient to allow acceptable implant results. When issues such as wafer positional control became a factor, due to device shadowing or channeling (and its profile-related effects) they were often addressed simply by using “non-channeled” wafer orientations of tilt and twist (e.g. 7 degrees of tilt and 27 degrees of twist), and often quad implants. As device geometries have decreased, the necessity of implant beam angle control both for the control of shadowing and channeling have increased substantially. One of the most sensitive implant angle conditions is at an implant angle of 0 degrees of tilt, where small variations in the implant angle can have major effects on geometric shadowing of the devices along with major channeling changes due to interaction of the incoming ions with the <001> crystallographic axis. Therefore control of the incident beam angle has become a primary implanter consideration.
Implanter Beam Angle Measurement and Correction Hardware
On the Axcelis Purion MTM the need for implant tilt angle control has been addressed by the implementation of a sophisticated measurement and control system. It consists of a precisely controlled scanning fixture (the Wafer Head Assembly) that supports the Electrostatic Wafer Chuck on one side and a collimated Vertical Beam Angle (VBA) Faraday on the opposite side. As part of the Autotune sequence, a measurement is made of the exact beam angle, using the VBA fixture, and then the Wafer Head Assembly is rotated to have the Electrostatic Wafer Chuck (with the wafer) face the Beam during implant. The angle of the incoming beam to the plane of the Electrostatic Chuck (and therefore the wafer) is automatically corrected by the amount indicated by this Vertical Beam Angle measurement.
In an attempt to verify the Vertical Beam Angle system’s ability to accurately and precisely correct the tilt angle, the widely-used approach of implanting a “V-curve” around a known channeling feature was employed. This consists of implanting a series of wafers (usually 3-5) at implant tilt angles offset around a know channeling feature, in this case the <001> axis at 0 degrees of tilt and then fitting a second-order polynomial curve to the resultant wafer ThermaWave (or Rs) results. The minimum of the curve will then give the exact crystallographic alignment of the system. An example of this technique is shown for the three wafers which were implanted as the first V-curve set in this study (Figure 1).
Figure 1 A typical V-curve implanted with three wafers at tilt angles of -1, 0, and +1 degrees of tilt. Solving for the minimum of the polynomial equation indicates a crystallographic alignment of + 0.0119 degrees.
However, to do this testing accurately, the crystal-cut error of the wafers (the error in the crystal structure of the wafers, as manufactured) must be known and controlled. A group of wafers (20) with a known high alignment accuracy were employed (earlier testing had shown these wafers to have a crystal cut error of +0.015 degrees). These wafers were then used to calibrate the VBA fixture to the plane of the Electrostatic Chuck, again using the V-curve technique. Subsequently, these wafers were used to create a series of V-curves, some consisting of three wafers and some using five, to determine the absolute accuracy and the repeatability of the VBA correction over time. An implant condition of phosphorus, 500 keV, 5x1013 atoms/cm2, was chosen for the implants due to the high angular (i.e. channeling) sensitivity and good ThermaWave signal/noise ratio. In order to prevent any variation due to differences in the crystal cut from different groups (boules) of wafers, the same set of wafers was repeatedly used. To do this an anneal step after each implant/measurement sequence was employed to return the wafers TW values to an “un-implanted” signal level (usually around 50 TW units as measured on a TP630). As a result, most of the wafers in the original group of 20 have been used at least 3 times over a period of more than four months and crystallographic variations have been eliminated. During the extended measurement interval, significant hardware changes occurred, including source PMs, Electrostatic Chuck replacements, VBA Faraday removals, and even ThermaWave laser replacements, with no re-calibration of the VBA measurement system. The resulting V-curve plots, with their second order polynomial fits, can be seen in Figure 2.
Figure 2 ThermaWave Tilt Repeatability V-curves for phosphorus, 500 keV, 2x1013 atoms/cm2 (data from 12/7/2011 to 4/20/2012)
While the absolute values of the sets of V-curve implants do vary somewhat (due to repeated implant/measurement/anneal cycles and metrology issues (e.g. ThermaWave laser replacement), etc., the minima, which indicate the angular alignment, stay remarkably consistent. In fact, for the multi-month measurement period the values are:
average minimum value = +0.023 degrees and the standard deviation = 0.019 degrees.
These V-curve minima values for each V-curve are plotted in Figure 3.
Figure 3 Wafer implant tilt angles as established by multiple V-curve minima. Average = +0.023 degrees, sigma = +0.019 degrees
The basic V-curve technique, when employed with appropriate implant conditions and experimental control (e.g. appropriate wafer crystal-cut quality), is able to demonstrate implanter angle control and repeatability to significantly better than 0.1 degrees. This degree of control has been maintained even with substantial end station and other machine maintenance, although proper procedure would dictate that re-calibration be done whenever major end station, especially chuck-related, changes are made. For more information on this topic, please download this whitepaper.