Improved Ion Beam Incident Angle Control for Varian
E220 and E500 Implanters
†
David Hendrix, Zhiyong Zhao*, Reuel Liebert, Ken Gifford and Pierre Mitchell¶
†
*Spansion LLC, Austin, TX, USA
Varian Semiconductor Equipment, 35 Dory Road, Gloucester, MA, USA.
¶
Therma-Wave, Fremont, CA. USA.
Abstract. This paper discusses the need and a solution for improved implant angle control required to support advanced device
production at 90nm on Varian E220/E500 series ion implanters. The paper characterizes the software and hardware
improvements made to the implanter in order to achieve improved control over implant angle. In-situ Therma-wave metrology
checks along with split-lot test methodology and device electrical performance results are also characterized.
Keywords: Ion Implantation, Angle Control, Semiconductors
PACS: 61.72, 29.27, 41.75Ak, 41.85Ja, 41.85Si
should improve from 2.5% to 1.25% at the extreme of
the tolerance.
INTRODUCTION
The challenges presented by ever-advancing nodes
of semiconductor production have led to rapid
obsolescence of older generation ion implantation
systems. Upgrades in wafer size and angle control
have been major drivers of this trend. While wafer
size increase is primarily a productivity and cost of
ownership issue, the degree of angle control affects the
basic capability of the implanter to control dopant
placement to the level needed for advanced device
fabrication. By the 130nm node, it became evident
that the accuracy of implant angle is critical to control
the lateral dopant profile of pocket implants and the
substrate channeling effect that can dominate the
PMOS device sensitivity [1]. Device makers had to
very carefully select their device layout and
architecture to minimize these effects.
Angle
requirements become even more stringent with more
advanced nodes [2].
This paper evaluates an upgrade to the E220/E500
implanter series that substantially improves its level of
implant angle control. First introduced in 1989, the
E220 effectively operated with an implant angle
accuracy of ±1.2º (combining in quadrature the effects
of tilt angle and beam entry angle uncertainty). The
upgrade brings the implant angle (“beam incident
angle” or BIA) accuracy from ±1.2º to a specification
of ±0.5º. The corresponding dose error arising solely
from the cosine projection at an implant angle of 45º
DESCRIPTION
The Beam Incident Angle (BIA) upgrade hardware
consists of an improved accuracy high-resolution dualencoded platen tilter assembly and control electronics
for positioning the wafer in the process chamber, a
new “focus cup” faraday (used conventionally during
scan setup), a new positive position optical sensor for
tilter initialization, and a new traveling faraday
centerline alignment sensor. The new focus faraday
and the traveling faraday assembly, which is normally
used for setting up the scan uniformity, are used with
new software packages to measure the beam entry
angle (BEA) with respect to the wafer plane at zero
tilter angle. The implant angle (BIA) is calculated
from the set tilter angle and the measured BEA.
The architecture of the E220/E500 implanter has
been described elsewhere [3].
It features an
electrostatic scanner and a dipole angle correction
magnet before final acceleration. The angle correction
magnet converts the angled scan that emerges from the
electrostatic deflection plates to a parallel-scanned
beam. The alignment of the implanter and the setting
of the field in this magnet will determine the BEA for
a given ion beam. The implanter alignment was
verified before the upgrade was installed. The system
computer sets the field in the magnet according to a
table lookup ordered by beam stiffness. The software
CREDIT LINE (BELOW) TO BE INSERTED ON THE FIRST PAGE OF EACH PAPER
CP866, Ion Implantation Technology,
edited by K. J. Kirkby, R. Gwilliam, A. Smith, and D. Chivers
© 2006 American Institute of Physics 978-0-7354-0365-9/06/$23.00
425
from the implant to load position. The specification
of the design is that the tilter be held to within ±0.35º
to assure the overall goal and the data in the figure
(with 22.76 counts/degree) show variation well within
that target.
0.2
35.1
35.0
T2 Encoder at 35º Tilt - degrees
35º Tilt
Tilter Control Specification = ± 0.35º
0.15
34.9
34.8
0.1
34.7
0.05
34.6
34.5
0
34.4
T2 Encoder at 0º Tilt - degrees
provided with the upgrade provides for automated
generation of table entries using a machine BEA
routine known as the “Refraction Method”. The
Refraction Method measures the difference in position
of the ion beam’s centerline location in the process
chamber when voltage is applied to the acceleration
tube. If the angle of attack is perfect, i.e. the BEA is
zero degrees, then there should be no shift when the
acceleration tube voltage is turned on. The traveling
faraday is moved through the beam in 0.1mm steps
and the software calculates the position of the beam
center from the resulting plot of beam current vs.
profiler position. An initialization routine verifies the
calibration of faraday position readout by calibrating
to the new zero-position sensor. The resolution of the
method is ~0.004º based on geometry and faraday
position accuracy. . By using this method, the magnet
constant table used in beam setup was filled with
constants that were adjusted using an automated
routine to iteratively minimize the measured shift.
Once the table was constructed, a second method
was used to measure the BEA at the time of each beam
setup. The second method (the Shadow Method) is
faster, does not require changes in the beam position,
and measures the angle of the beam at its final energy
in the process chamber. The tradeoff for using this
fast method is a small degradation in the resolution to
~0.015º. The Shadow Method employs an ion beam
profiler faraday (new focus cup) located along the
back wall of the implant chamber and the traveling
faraday
The Shadow Method uses the motion of the
traveling faraday across the process chamber to block
the ion beam from entering the narrow slit in the focus
faraday mounted in the back of the implant chamber.
By measuring the drop in beam and correlating it to
the traveling faraday encoded position, the software
can measure and calculate the BEA. During beam
setup, this method is used to assure the BEA is within
the ±0.2º interlock that guarantees the BIA accuracy to
within ±0.5º. If the measurement is outside the
allowed error, then the angle correction magnet current
is adjusted, the beam is repositioned in the focus
faraday cup, and the measurement is repeated. In this
manner, closed loop control of the BEA is achieved.
0º Tilt
-0.05
34.3
34.2
-0.1
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
Samples
Figure 1. Position Repeatability of the Upgraded Tilter
Measured with New Driven-Side Encoder
The Refraction and Shadow methods were
compared to verify that the angles measured by the
two methods are equivalent. Testing of the prototype
unit at Varian and the Beta unit at Spansion showed
the upgraded implanter could routinely achieve a BEA
specification of ±0.20º on a variety of implant recipes.
1100
1050
Implant: 150keV B+ 2E13 at/cm2
TWmin is at 35.23 +/- 0.15 degrees
Theoretical Channel Peak = 35.26 degrees
Data Taken Over 7 Days
1000
TW Units
Data from 12-14 Trials per Angle
Gaussian Fit
950
Gaussian Sigma = 2.35°
900
850
800
33.25
33.75
34.25
34.75
35.25
35.75
36.25
36.75
Recipe Implant Angle, degrees
Figure 2. Alignment of the BEA system and implanter
is shown using a <112> channeling test
SYSTEM TESTING
Alignment of the implanter and BEA system at the
Beta site was verified by a channeling test using the
<112> axis that should align at a tilt angle of 35.26º.
Fig. 2 shows the results of a seven day run in which
five angles spanning the expected channeling
minimum were measured by Thermawave. The data
for each angle (12-14 trials) were used to determine an
average and an error-weighted Gaussian fit was used
The upgraded tilter assembly uses encoders for the
motor drive and also for the driven shaft of the tilter.
The software requires that the two encoders agree
when the tilt angle is set. Figure 1 shows the position
variation of the driven shaft encoder during a test that
cycled the system 7669 times through various angles
426
weighted average of) the standard deviations of
Thermawave data from A and B to the standard
deviation of C suggests that the upgrade reduced this
variability by 71%. .
to determine that the alignment was good to 0.03º
within the fit error of 0.15º.
.
90NM PROCESS QUALIFICATION
Split-Lot Tests
A Flash Memory device lot was run to test the BIA
upgrade. The splits were arranged to have three
wafers run on each of three different implanters for
several consecutive days. The splits are: six wafers for
two days on an E500HP without the BIA upgrade, six
wafers for two days on a VIISta810EHP, and thirteen
wafers for four days on the E500EHP with the BIA
upgrade. VIISta810EHP was used as a preferred
performance standard since it has an implant angle
correction package that has been proven to perform
better than the standard E500 implanters at Spansion
fab 25. The E500HP without the BIA upgrade is
utilized as a reference of a nominal E500 implanter.
The 90nm feature size device was utilized for this test.
A high tilt angle channel engineering implant into the
memory cell that is angle-sensitive was used to
enhance the response.
Two more device lots of 110nm feature size were
also tested. The results were similar to the 90nm lot.
This paper will present the data of the 90nm device lot
only for the purpose of clarity.
The following three figures represent the electrical
evaluation of the split lot. The data has been
normalized for proprietary reasons. The parameters
related to this implant did show responses in favor of
the BIA upgrade.
39
38.5
38
37.5
37
36.5
36
35.5
35
34.5
34
A
B
C
Number
13
9
18
Std.Dev.
0.96
0.33
0.20
10
Nominal Cell VTL
Programmed tilt for maximum channeling by ThermaWave
To monitor beam incident angle at high tilt, the
production B+ beam is implanted into test wafers at
34, 35 and 36 degrees over the period of months.
Because of axial channeling, a minimum Thermawave
measurement is expected at near 35 degrees. A
Thermawave Model TP-500 was used for all the
testing. A parabolic fit through the three Thermawave
measurements serves to estimate the programmed tilt
that would give in the lowest measurements. At this
tilt, the channeling is greatest. Typically, good
agreement between the values calculated from
Thermawave and sheet resistance measurements
demonstrated the accuracy of this method.
Implanter
Figure 3. The best achieved channeling angle for the <112>
axis based on Thermawave measurement. Three groups are
given: an E500EHP implanter before and after the BIA
upgrade as well as an E500EHP without the BIA upgrade.
(B,C and A). The theoretical tilt angle at the channel is
35.26º
5
0
-5
E5 no BIA
These calculated values are plotted in Figure 3.
Implanter A is a virtually identical E500HP without
the upgrade, B is the E500EHP before the upgrade,
and C is the same E500EHP after the upgrade The
programmed implant tilt angle determined to give the
minimum measurements is plotted for the E500EHP
for making comparisons before and after the BIA
upgrade. Comparing the data from E500HP without
the upgrade to data from the E500EHP with the
upgrade illustrates that the variability of beam incident
angle is reduced by at least 50%. Comparison of (a
E5 w/ BIA
VIISta 810
Implanter
Figure 4. The linear Vt variation in percentage as measured
on the tested device wafers.
Figure 4 shows the percent variation of the nominal
memory cell’s linear Vt. From the left, the first data
group is from an E500HP implanter without the BIA
upgrade. The second data group is from an E500EHP
with the BIA upgrade. Finally, the third data group is
427
from a VIISta 810 implanter serving as a preferred
performance standard. The data shows that the
E500EHP with BIA achieves Vt variation 56% smaller
than the E500HP without BIA. Further, the BIA
E500EHP performance is closer to that of a VIISta
810EHP, which has a reduced Vt variation of 66%
over the non-BIA E500HP.
a smaller implant angle variation. As evidence,
Figures 6 presents DIBL (drain induced barrier
lowering) variation of the same memory cell. With a
better defined drain or equivalently, a better defined
channel, variation in DIBL is reduced by 73% for the
E500EHP with the BIA and 83% for the VIISta
810EHP over the non-BIA E500HP. If this channel
engineering implant (pocket implant) is the only
contributor to the difference, then this would imply
that the smaller Vt and delta Vt variations are due to a
better controlled implant angle during the implantation
process.
Nominal Cell Delta Vt1
5
2.5
0
-2.5
CONCLUSIONS
-5
-7.5
The BIA Upgrade on E500EHP, based on the
limited data on test wafers and device wafers, has a
measurable improvement over the angle control
capability of standard E500 series implanters. The
design goal for the BIA on E500-series implanters is to
improve BIA from a capability of ±1.2º to a
specification of ±0.5º. This is an improvement of 58%
over its current performance. The data on the Vt and
delta Vt suggests that the E500 with BIA can control
the beam incident angle to its design specification of
±0.5o and is compatible with 90nm processes. In fact,
the data shows it behaves closer to the VIISta 810
implanter that has an angle controllability of ±0.2o.
-10
E5 no BIA
E5 w/ BIA
VIISta 810
Implanter
Figure 5. The Delta Vt variation in percentage as measured
on the tested device wafers.
Figure 5 shows the corresponding delta Vt of
Figure 4. The E500EHP with the BIA upgrade has an
improved delta Vt over the non-BIA E500HP and it
behaves closer to a VIISta 810EHP. The magnitude of
improvement is 58% for the E500 with the BIA and
63% for the VIISta 810EHP. This correlates to the Vt
variation improvement observed in Figure 4. Figure 4
illustrates that the Delta Vt for the BIA E500EHP
tends to have higher values than either of the other
implanters while its data distribution is the tightest.
This phenomenon is not understood at this time.
ACKNOWLEDGEMENTS
The Spansion authors would like to thank Linda
Wang of Fab25, Spansion LLC, for her helpful
discussion on device parameters. The Varian authors
would like to thank Dave McClellan and Dave Olson
for their help with the test data and Moussa Haddad
for his help and guidance with respect to all the
software tasks.
80
Nominal Cell DIBL
60
40
20
REFERENCES
0
1. Kenn S.Y. Yeh, M.C. Chiang, C.J. Tsai, Y.L. Wang and
J.K. Wang, “Optimization of High Tilt Pocket Implant
Process for Improving Deep Sub-micro PMOS Device
Sensitivity” in Proc. of 14th Int. Conf. on Ion
Implantation Technology, 2002, p.13-16
2. U. Jeong, Z. Zhao, B.N. Guo, G. Li, and S. Mehta,
“Requirements and Challenges in Ion Implanters for
Sub-100nm CMOS Device Fabrication,” CP680,
Application of Accelerators in Research and Industry:
17th Int'l. Conference, edited by J.L.Duggan and
I.L.Morgan, AIP press, New York, (2003), 697
3. D.W. Berrian, R.E. Kaim, J.W. Vanderpot, and J.F.M.
Westendorp, Nucl. Instr. And Meth. B37/38 (1989) 500
-20
-40
E5 no BIA
E5 w/ BIA
VIISta 810
Implanter
Figure 6. The Drain induced barrier lowering variation in
percentage as measured on the tested device wafers.
The smaller variations of Vt and delta Vt as
observed in Figure 4 and Figure 5 can be attributed to
428