Ion Implantation Angle Variation to Device Performance and the Control in
Production
Z.Y. Zhao*, D. Hendrix*, L.Y. Wu**, B.K. Cusson***
*
Fab25, AMD, Mail Stop 608, 5204 E. Ben White Blvd., Austin, TX 78741
PCAL, Fab25, Mail Stop 613, 5204 E. Ben White Blvd., Austin, TX 78741
***
Advanced Process Control, Fab25, AMD, 5204 E. Ben White Blvd., Austin, TX 78741
**
As the device features get smaller and aspect ratios of photoresist openings get steeper, shadowing effect has more impact on the
performance of devices. Many of the traditional 7o tilt implants have progressed to 0o implants. But shadowing may still occur if the tilt
angle deviates from normal direction. Some implants, such as halo implants, demand even more stringent angle control to reduce device
performance variation. The demand for implant angle control and monitoring thus becomes more obvious and important. However,
statistical process control (SPC) cannot be done on shadowing effect without special test structures. Channeling, on the other hand,
provides good sensitivity in regard to implant angle changes. It is the authors’ intention to introduce channeling implant in different
channels to monitor the implant angle variation. The incoming <100> silicon wafers have a cut-angle spec of +/- 1.0o. This poses a
difficulty if one wants to control the implant angle’s accuracy within +/- 0.5o. Other measures have to be taken to ensure the consistency
of test wafers and to have prompt diagnosis feedback when needed. This paper will discuss the effect of implant tilt angle on device
parameters and how to control the angle variation in production. Correlations of implant tilt angle variation to ThermaWaveTM, sheet
resistance (Rs), Secondary Ion Mass Spectrometry (SIMS) and device parameters will be covered with certain implant conditions.
However, ion channeling is unavoidable when the implant
angle is set at 0o to <100> silicon wafers. In a perfect
channeling condition, i.e., the ion beam is well collimated,
ions penetrate the silicon crystal to a much deeper level
than the random ions. In commercial ion implanters,
however, fine ion beam collimation cannot be done since
the priority is given to high wafer throughput with large
beam current. The resulting dopant profile does not have a
single super-channeling peak as reported in literature [3].
Instead, the shape of implant profiles of wafers from
different ion implanters can vary, which places difficulties
in implanter matching and statistical process control. High
current ion implanters, for instance a batch implanter and a
serial implanter, have profile difference even if they are all
set with the same implant conditions with correct hardware
setup and calibration.
Historically, implant angle is controlled on medium
current implanters which are used for Vt and channel
implants where accurate angle is required. High current
implanters due to its traditional batch design do not have
strict angle control requirement. However, as the devices
are scaled downward, we start to see the need for implant
angle control even on high current implanters. With a
microprocessor, it is seen that when the implant angle is
off by 1o, the affected parameters of the device show
observable variations. When the angle is off by 2o, the
transistor parameters can be off the control limit. When
the implant angle is not controlled due to either hardware
failure or process setup, the device performance is affected
since the physical configuration of the device is altered, as
seen in Table I. In this table, one may see when the
implant angle is not controlled, the ratio of forward and
backward biased drive current of a microprocessor can be
INTRODUCTION
The ULSI circuits scaling is putting hundred of millions
of transistors on a single die. We are now making devices
in production with 0.13 µm feature sizes and it will drop to
sub 0.1 µm in the next couple of years. Depending on the
device, close to thirty ion implant operations may be
needed in the device manufacturing process.
In semiconductor device manufacturing, ion implanters
are used in the front end of line to define the transistors
and other electrical components in the integrated circuit.
Most implants in the past had 7o implant angle. This was
to avoid channeling in which the ions have minimum
collisions with the lattice sites and encounter minimal
electronic stopping power. Besides the requirements for
ultra-shallow junction formation and special species, such
as indium and antimony, the demand for implantation
angle accuracy and consistency is also very important.
Most implants have changed from a typical 7o implant
angle to 0o due to shadow effect since the aspect ratio of
photoresist opening is becoming greater. Besides the
shadowing effect, ion channeling at 0o implant angle
makes the dopant profile much deeper than a nonchanneled ion beam with larger lateral straggle. A small
variation in implant angle may make a big impact on
device parameters. Consequently implant angle control
becomes important and critical.
IMPLANTERS AND ION CHANNELING
When channeling is effectively avoided, the implanted
ions undergo random collision processes and the resulting
dopant profile can be simulated using mathematical
models, such as TRIM [1] or Monte Carlo process [2].
Such profiles usually follow a Pearson IV distribution.
CP680, Application of Accelerators in Research and Industry: 17th Int'l. Conference, edited by J. L. Duggan and I. L. Morgan
© 2003 American Institute of Physics 0-7354-0149-7/03/$20.00
666
TABLE II. Peak fitting result of the four SIMS profiles with
different implant angles shown in Figure 1. The centroids are in
Angstrom.
changed by more than 10% and the standard deviation is
up by seven times.
Tilt
TABLE I. The ratio of forward and backward drive current of a
microprocessor with implant angle variation. The involved
implant was a source/drain extension implant.
Angle Control
Good
Bad
Mean
1.005
1.140
0o
1o
2o
3o
Std. Dev.
0.006
0.042
0o IMPLANT ANGLE
With <100> Silicon wafers and 0o implant angle, it is
expected that most of the ions will under go a channeling
process. However, since the commercial ion implanters in
semiconductor industry do not have beam collimation as
those obtained in a physics laboratory, a super channeling
profile cannot be obtained on such implanters even with 0o
implant angle. This can be clearly seen in Figure 1.
Random
Peak %
40
52
77
82
Channeling
Peak %
60
48
23
18
Random
Centroid
3000
2980
2770
2590
Channeling
Centroid
4950
4630
4450
4380
In a mass production environment, one cannot rely on
analytical techniques such as SIMS for routine production
check due to the time and cost concerns. Sheet resistance
(Rs) and ThermaWaveTM are used due to their routine use
in semiconductor manufacturing environment. The same
implants as shown in Figure 1 were measured by Rs and
ThermaWaveTM. The data show ThermawaveTM is more
sensitive than Rs. However, it is prone to long term
drifting due to issues related to implant beam current
variation, laser aging and others. A control chart can be
set up with +/- 20 Twave units, which is equivalent to +/0.5o of implant angle change. Sheet resistance, on the
other hand, has demonstrated better long term stability
with good dose and tilt sensitivity. With true 0o implant
angle, the Rs value is 798 Ω/ . It goes up to 825 Ω/
when the implant angle is changed to +/- 1o. The sheet
resistance sensitivity to implant angle is then calculated to
be 27 Ω/ per degree. A control chart can be set up with
+/- 13 Ω/ control limit and it is equivalent to +/- 0.5
degs, or +/- 1.5% dose change. The above discussion is
based on Rs sensitivity to dose is 100%, which is defined
as the corresponding change of measurement to the change
in dose or angle.
FIGURE 1. The SIMS profiles of Boron implants into <100>
Silicon by a commercial ion implanter. The implant condition
was 70keV, 5E13/cm2. The implant angle was varied in 1o
increment to show the effect of ion channeling. The implants
were done on a commercial serial high current implanter.
The dopant profile for the 0o implant has a flat-top
distribution. When the implant angle is deterred from the
normal direction, the dopant profile tilts to the shallower
depth. Based on the understanding of ion channeling [3],
such profiles have two individual peaks, one channeling
and one random. An estimate of the percentage of
channeling peak and random peak can be obtained by
mathematical peak fitting. In Table II we list the result of
the fitting. The fitting was obtained by limiting the
variation of centroid position so that the relative positions
of the channeling peak and the random peak are locked.
The result seems to suggest that intrinsic ion beam
divergent angle is within 1o so that the four profiles are
separated into two groups.
FIGURE 2. The Rs and ThermaWaveTM response to the implant
angle vadiation. The implant conditions are the same as those in
Figure 1.
It is to be noticed that the strong channeling peak is not to
be observed when the implant energy is decreased. At
70keV, the ion beam may be parallel enough to produce a
plateau dopant profile. As the energy goes lower,
however, the same implanter with the same implant specie
667
Similarly the ThermaWwaveTM response may be used to
determine the wafer orientation which shows the greatest
sensitivity to changes in the beam incident angle. Figure 4
shows the ThermaWwaveTM sensitivity to a 5o change in
tilt for various twists. Clearly, a twist of 45o appears
optimal because as the twist is increased or decreased the
sensitivity of the measured signal to a change in tilt from
45 degrees to 40 degrees diminishes. Maximum <110>
channeling occurs with a programmed tilt and twist of 45o.
This largest channeling hole in the silicon crystal is most
open to the ion beam when the wafer twist is set to 45o but
begins to close as the wafer twist is changed from 45o.
cannot produce a profile with a similar shape due to factors
such as the beam divergence and channeling critical angle
change. With 0o angle implants at low energies, such as a
20keV B+ implant, the dopant profile showed a shallow
peak, presumably the random peak, with a prolonged
channeling tail. Such a profile makes it less desirable for
angle control monitoring purpose since the variation in the
tail does not have high influence in either sheet resistance
or ThermaWaveTM measurement.
45o IMPLANT ANGLE
Halo implants are common to modern ULSI devices.
These are implants with tilt angles ranging from 20o to 60o,
depending on the devices and processes. The device
performance can be affected if the implant angle varies
even by 0.5o. Also, at high implant angle, the wafer
orientation angle is critical since it determines which side
of the device gets the implant. It is then important to
monitor both the implant angle and orientation angle for
halo implants. Due to conventions in the semiconductor
industry, implant angle and tilt angle are used
interchangeably in this paper. So are the orientation angle
and the twist angle.
For example, if a 50 keV As+ beam is used in production
for a halo implant on a medium current implanter, the
optimal conditions for monitoring beam incident angle
may be determined similarly as in the earlier section. The
ThermaWwaveTM sensitivity to changes in the implant
angle is presented in Figure 3. The results are presented in
a 5o implant angle change for various doses when the
wafer twist is 45 degrees.
For lower doses, the
ThermaWwaveTM signal increases by less than 10% as the
tilt is reduced from 45 to 40 degrees and the crystal
damage comes closer to the surface. At higher dose the
ThermaWwaveTM readings exceed 3600, above which the
result becomes less reliable. Therefore, a dose of 4.5E13
atoms/cm2 of As+ appears optimal because the
ThermaWwaveTM signal increases by more than 25% as
the tilt is reduced from 45 to 40 degrees.
FIGURE 4. ThermaWwaveTM sensitivity to a 5 degree change in
tilt for various twists for a 4.5E13 As+/cm2 implant at 50keV.
SIMS analysis of these wafers confirms that the
minimum ThermaWwaveTM and Rs occurs at the point of
maximum <110> channeling when the channeling tail is
the greatest, as seen in Figure 5. Changing the tilt or twist
from 45o reduces this tail. The relatively low energy of
this 50keV As+ beam results in a dopant profile with less
longitudinal straggle than in Figure 1. This is similar to
the discussion at the end of the last section.
FIGURE 3. ThermaWwaveTM sensitivity to a 5 degree change in
implant angle for various doses of 50keV As+ with 45o
orientation angle.
FIGURE 5. The SIMS profiles of Arsenic implants into <110>
Silicon by a commercial medium current ion implanter. The
implant condition was 50keV, 4.5E13/cm2. The implant angle
and twist angle were varied to show the effect of ion channeling.
668
we believe this is the first time that such measure is
applied to high current implanters. The result from this
work shows a proper control of implant angles, both tilt
and twist, can be achieved by setting up the control limits
on SPC charts around the channeling direction where the
ThermaWaveTM and Rs values are minimal.
However, there are issues associated with this approach.
One of them is that the SEMI standard for wafer cutting
angle is +/- 0.5o. Consistency of our data indicates the
variation is much smaller. However, from time to time
one wafer may show up with a different cut angle.
Combined with the uncertainty of implant angle which is
to be qualified, extra analysis such as X-ray diffraction
maybe needed to confirm the data. In the meantime,
another test wafer can be used from a different source so
that the wafer’s cut angle uncertainty can be verified.
ACKNOWLEDGEMENT
The authors thank Advanced Micro Devices (AMD)
management for the support of this project and the
Diffusion module in Fab25, AMD for the data collection.
Using a 50keV As+ beam with the optimal dose and
twist, the beam incident angle may be found by
determining the wafer tilt which gives the minimum
ThermaWwaveTM and sheet resistance value. Figure 6.
shows the minimum ThermaWwaveTM and sheet resistance
values to be near the expected tilt of 45o. Assuming that
the minimum values should occur when the actual beam
incident angle is 45o, the error in the beam incident angle
may be found. Fitting a parabolic curve through each data
set, one may determine the tilt at which the curve reaches a
minimum. At this point of maximum channeling, the
actual beam incident angle is 45o or very close to it.
1.
2.
3.
FIGURE 6. ThermaWwaveTM and Sheet resistance responses to
change in tilt using production beam with optimal dose and twist.
4.
This determination of the tilt setting, which results in a
beam incident angle of 45o, may be repeated for various
implanters to confirm that the beam incident angles are
matched. Table III gives these results for five implanters
which are matched to within 1 degree. Correcting errors in
wafer platen position and any misalignment of the beam
through the accelerator column can eliminate this variation
of beam incident angle across implanters.
Table III. Comparison of beam incident angles across five
implanters.
Implanter
A
B
C
D
E
Implant Angle with
min. T-Wave
45.2
45.4
44.4
44.3
44.9
Implant Angle
with min. Rs
44.5
44.8
43.9
44.0
44.8
CONCLUSIONS AND DISCUSSIONS
This work demonstrates the necessity of implant angle
monitoring in semiconductor manufacturing. Although the
need existed for the medium current implanters in the past,
669
REFERENCES
J.F. Ziegler and J.P. Biersack, TRIM95: The Transport
of Ions in Matter, 1995.
A. Tasch, et al, UT-Marlowe 3.1 documentation,
Physical Department, University of Texas, Austin,
TX, 1998.
Z.Y. Zhao, A. M. Arrale, S.L. Li, D. K. Marble, D. L.
Weathers, S. Matteson, J.M. Anthony, B. Gnade and
F.D. McDaniel, Experimental evidence for a discrete
transition to channeling for 1.0-MeV protons in
Si<100>, Phys. Rev. A 57(4) 2742 (1998)
U. Jeong, S. Mehta, C. Campbell, R. Lindberg, Z.Y.
Zhao, B. Cusson, J. Buller, “Effects of beam incident
angle control on NMOS source/drain extension
applications”, Presented on Intern. Conf. On Ion
Implantation Tech., Taos,, NM, Sept. 22 –27, 2002.