Effects of Lightpipe Proximity on Si Wafer Temperature in
Rapid Thermal Processing Tools
K. G. Kreider, D. H. Chen, D. P. DeWitt, W. A. Kimes, and B. K. Tsai
National Institute of Standards and Technology (NIST)
Gaithersburg, MD, USA, 20899-8363
Abstract. Lightpipe radiation thermometers (LPRTs) are used as temperature monitoring sensors in most rapid thermal
processing (RTP) tools for semiconductor fabrication. These tools are used for dopant anneal, gate oxide formation, and other
high temperature processing. In order to assure uniform wafer temperatures during processing these RTP tools generally have
highly reflecting chamber walls to promote a uniform heat flux on the wafer. Therefore, only minimal disturbances in the
chamber reflectivity are permitted for the sensors, and the small 2 mm diameter sapphire lightpipe is generally the temperature
sensor of choice. This study was undertaken to measure and model the effect of LPRT proximity on the wafer temperature.
Our experiments were performed in the NIST RTP test bed using a NIST thin-film thermocouple (TFTC) calibration wafer.
We measured the spectral radiance temperature with the center lightpipe and compared these with the TFTC junctions and with
the three LPRTs at the mid-radius of the wafer. We measured LPRT outputs from a position flush with the reflecting plate to
within 2 mm of the s tationary w afer under steady-state conditions with wafer-to-cold plate separation distances of 6 mm,
10 mm and 12.5 mm. Depressions in the wafer temperature up to 25 °C were observed. A finite-element radiation model of
the wafer-chamber-lightpipe was developed to predict the temperature depression as a function of proximity distance and
separation distance. The experimental results were compared with those from a model that accounts for lightpipe geometry
and radiative properties, wafer emissivity and chamber cold plate reflectivity.
INTRODUCTION
Accurate temperature measurements are critical in
rapid thermal processing (RTP) of silicon wafers for
thermal oxidation and dopant anneals. Many RTP tools
use lightpipe radiation thermometers (LPRTs) to
measure the wafer temperatures during processing.
These LPRTs can yield accurate temperature
measurements when they are properly calibrated and
used with a suitable model to correct for surface
emissivity and chamber irradiation effects. The LPRTs
are used primarily because they cause less disturbance
to the wafer temperatures than contact thermometers or
lens-type radiometers, which generally require larger
viewing apertures. Wafer temperature measurements
are frequently performed in a highly reflecting chamber
to obtain a near-unity effective emissivity of the wafer.
The s apphire 1 ightpipe t ip h as a low reflectivity (high
absorptivity) and enhances radiation heat transfer from
the target region. This low reflectivity caused a
depression in the wafer temperature that was dependent
upon the proximity of the tip to the wafer, and the
extent to which the lightpipe is heated within the
enclosure.
We at NIST have been studying commercial LPRTs
and their calibration and temperature measurement
uncertainty in RTP tools. We have developed highly
accurate blackbody calibration techniques for the
LPRTs (1), characterized and measured the temperature
sensitivity of the LPRTs (2), developed technology for
in situ calibration of the LPRTs in the RTP tools (3, 4,
5, 6), and measured the effects of wafer emissivity on
the LPRT measurements in the RTP tool (7). We have
also developed models for the relationship between the
radiance temperature measured by the LPRT and the
wafer temperature to enable their use in obtaining wafer
temperature measurements with uncertainties of less
than 2 °C (k = 1) under industrial conditions in RTP
tools (8).
We report here on a study of the effect of light pipe
proximity to the w afer on the temperature d istribution
of the wafer. In the NIST RTP test bed (Fig. 1), the
lower chamber surrounding the 200 mm wafer is cold
and highly reflective which leads to a very high wafer
effective emissivity (0.96) for the LPRT measurements.
This is similar to the commercial chambers. We have
reported previously on LPRT measurements in the
NIST test bed with the light pipe tip flush with the
water-cooled reflecting shield and 1 0 mm to 1 2.5 mm
from the wafer. However, some commercial practices
employ much smaller light pipe tip-to-wafer or
proximity distances. At these smaller proximity
CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
2003 American Institute of Physics 0-7354-0152-7/03/$20.00
200
distances the view factor between the wafer target and
its surroundings changes, and the lightpipe operates at a
higher temperature. The higher temperatures are
permitted by the extremely refractory nature and
excellent transmission properties of the sapphire.
Unfortunately the lightpipe and its sheath, also
sapphire, constitute a low reflectivity (0.10) zone of the
chamber, which increases the net radiant heat flux from
the wafer target area. This enhanced flux leads to a
temperature depression on the wafer relative to the
adjacent region. Our investigation was intended to
quantify those effects for the NIST test bed and provide
an understanding and a model to enable this LPRT
proximity effect to be estimated in commercial RTF
temperature measurement applications.
EXPERIMENTAL
The NIST RTF test bed (Fig. 1) has been used to
simulate RTF tool temperature measurements and for
LPRT in situ calibration using thermocouple wafers.
Quartz plate
Shading Wafer
Cold plate
Reflective shield
Support spider
Guard ring
Guard tube
N2 inlet, 3x
the wafer temperature caused by the proximity of the
center lightpipe to the wafer (p). The mid-radius
lightpipes were used to monitor the wafer temperature
at various power levels for comparison with the center
lightpipe. Their position was always flush with the
reflecting shield. We used distances (s) of 12.5 mm,
10mm and 6 mm between the wafer and reflecting
shield to assess the effect of the shield's proximity to
the wafer on the temperature measurements. The light
pipe with its 4.2 mm sheath was advanced from its
flush position to a proximity of 2 mm from the wafer.
The key measurement was the relative depression of
the center LPRT measurement compared to its reading
in the position flush with the reflecting shield. This
lightpipe reading changed as the light pipe and its
sheath were advanced from the flush position toward
the wafer. Each of the center light pipe measurements
were referenced relative to the mid-radius LPRTs in
order to correct for slight temperature differences
results from run to run.
We used two NIST thermocouple wafers for
independent temperature measurements and calibration
of the LPRTs. One of these wafers with Rh and Pt thin
films (Fig. 2) had two junctions each at 5 mm and
15 mm from the center and four junctions at the midradius (50 mm). The second had Ir and Pt thin-film
thermocouples.
These wafers had Pt/Pd wire
thermocouples welded to the thin film differential
thermocouples.
Cooling line
Path of TC leads
N2 exhaust
T Thermal control
LP assembly
coatings
FIGURE 1. NIST RTF test bed wafer chamber.
A bank of 24 two kW quartz infrared halogen lamps
with a cold, highly reflective chamber (not shown in the
figure) is used to heat the wafer, and up to four 2 mm
commercial lightpipes measure the bottom side of the
wafer. We used a shading wafer, 150 mm in diameter,
to reduce the thermal gradients in this static system.
This wafer was 18.5 mm to 25 mm above the test wafer
and below the fused quartz plate which separates the
lamp chamber from the test wafer chamber. The test
wafer was supported by alumina pins 6 mm, 10 mm,
and 12.5 mm above the reflective shield (95.5 %
reflectivity). A Pt-coated guard ring (300 mm in
diameter) and a Pt-coated guard tube completed the
reflecting cavity. All tests were run in a purged
atmosphere of N2 with up to 0.1 mL/L O2. The center
light pipe was adjusted in height from a flush position
(as in the figure) to within 2 mm of the wafer.
We compared the center lightpipe reading to the midradius light pipe readings to determine the depression of
201
FIGURE 2. NIST calibration wafer, after 51 thermal cycles,
with 8 thin-film thermocouples (angled junctions) and 12 pads
for 16 Pt/Pd wire welds.
This technique of in situ calibration using the NIST
thin-film thermocouple wafer has been described
previously (3, 8). The Pt/Pd wire thermocouples and the
Rh/Pt thin-film thermocouples are calibrated separately
(6, 8) to establish a 1.4 °C (k=\) wafer temperature
measurement uncertainty on the International
Temperature Scale (ITS-90).
lightpipe (Ip) and no-lightpipe (nip) conditions. As
expected the depression is greatest at the wafer center
location (r = 0) and decreases with increased radial
distance. The maximum depression and the radial
extent of the depression increases with wafer
temperature. As will be demonstrated subsequently, the
comparison with the experiment data is based upon the
center temperature depression for a proximity distance
p relative to the flush condition (s —p).
To compare the model predictions with the
corresponding experimental observations, two key
features of the enclosure system must be considered.
Referring to Fig. 3, recognize that the LPRT target area
(field-of-view) changes with the proximity distance
and that the radial temperature distribution may change
significantly over the target area. Further, the LPRT
does not have uniform responsivity across the target
area. To account for the areal sensitivity effect, the
temperature measurement equation is used to e stimate
the average target temperature considering the
temperature distribution and the LPRT responsivity.
The correction to the center temperature depression for
this effect amounts to less than 2 °C.
The enclosure model is also used to estimate the
effective emissivity of the wafer target as a function of
the proximity distance. This property is used with the
temperature measurement equation to calculate the
target temperature from the experimental spectral
radiance temperatures observed by the LPRTs. By so
doing, it is then possible to directly compare model and
experimental observations for the temperature
depression. With the model, the effective emissivity is
determined from the ratio of the heat flux from the
wafer with emissivity 8^ to that when 6*w = 1. For
the extreme case when p = 2 mm and s = 6 mm, the
effective emissivity is 0.91, compared to a value of 0.96
for the flush case.
MODELING AND COMPARISON WITH
EXPERIMENTAL DATA
A model representing the wafer-shield-chamber
arrangement of the Test Bed (Fig. 1) is shown
schematically in Fig. 3.
Uniformheat
p = 2 mm l
Light pipe,
p/p=0.10
s= 6 mm
r_
S
I
.j.
Constant temperature
boundary condition
'
Shield, ps = 0.95
Constant temperature
boundary condition
FIGURE 3. Schematic of wafer-shield-chamber
arrangement.
The infinite, parallel planes model is comprised of the
wafer with emissivity £w subjected to a constant heat
flux and the cold, highly reflective shield of reflectivity
yC^with the separation distance s. The light pipe of
diameter d with a reflectance p
extends into the
enclosure the proximity distance p. The lower end of
the light pipe extending below the cold shield a distance
L is maintained at the same temperature as the cold
shield. The model considering radiation exchange
between surfaces in the enclosure is solved using a
finite-element code. All surfaces are assumed opaque,
diffuse, and gray. The purpose of the model, which is
solved using a finite-element code, is to estimate the
change in wafer
axisymmetrical temperature
distribution, T(r\ as a function of the proximity and
separation distances.
RESULTS
-30
20
40
60
100
Radius, /tmm)
FIGURE 4. Modeled temperature distribution.
The temperature distributions shown in Fig. 4 represent
the depression in temperature caused by the light pipe;
that is, the temperature difference between the with-
202
We measured the radiance temperature of the four
LPRTs and the voltages of the thermocouples on two
200 mm wafers between 640 °C and 860 °C in the
NIST RTP test bed for a total of 68 thermal cycles. The
effect of the light pipe proximity on the radiance
temperature measurement of the center lightpipe was
plotted for each wafer at four reflecting shield distances
(s). We plotted the radiance temperature depression of
the center lightpipe as a function of temperature for
each wafer for each lightpipe distance (p). All such
curves indicated a near linear increase in center
lightpipe radiance thermal depression as a function of
temperature. The proximity measurements (p) were
estimated to have an uncertainty of 0.3 mm (k — 1).
This leads to an uncertainty in the measured
temperature depression of 3 °C (k = 1), for s = 6 mm
with;? = 2 mm, to 0.5 °C (k = 1), for s = 12.5 mm with
p = 8 mm. Other causes of uncertainty in these
measurements include calibration of the lightpipes and
uncertainty in the average change in radiance
temperature between runs, as determined by the midradius lighpipes. These "other" causes of uncertainty
produced an uncertainty of 0.6 °C (k = 1).
The results of the measurements with s = 12.5 mm are
presented in Fig. 5. At s = 12.5 mm the smallest p we
could achieve was 4.5 mm which reduced the center
lightpipe reading 13 °C at 850 °C.
lightpipe was 18 ° C to 21 °C for;? = 3 mm at 825 °C.
The model prediction for those conditions is similar.
Many commercial tools use a 2 mm proximity and such
conditions are illustrated in Fig. 7 where s = 6 mm.
600
700
800
900
Mid-radius Radiance Temperature / °C
FIGURE 7. The change in the central radiance temperature
due to the proximity of the center lightpipe to the wafer (p 2 mm to 5 mm), for a wafer-reflecting plate distance (s) of
6 mm, versus the average mid-radius radiance temperature.
The values for the model at/? = 4 mm are given by a solid line
and dashed line forp = 2 mm.
+ 10.5mm ^ 8.5mm
A6.5mm
X4.5mm
-20
600
700
800
Mid-radius Radiance Temperature / °C
900
FIGURE 5. The change in the central radiance temperature
due to the proximity of the center lightpipe to the wafer (p =
4.5 mm to 10.5 mm), for a wafer-reflecting plate distance (s)
of 12.5 mm, versus the average mid-radius radiance
temperature. The values for the model at p = 10 mm are
given by a solid line and dashed line for/? = 5 mm.
Although we have more scatter in the results because
of our high uncertainty in measuring p, both the results
and the model indicate the depression of the LPRT
reading of greater than 20 °C at 825 °C and larger
values at higher temperatures.
Our 200 mm wafers also had thin film and wire
thermocouples (Fig. 2) in order to calibrate the LPRTs.
The results of those calibrations are presented in Fig. 8.
—il————————
20 11
0
V
11
5
"
%
f%
H
^
B
.,
J' + -
;*
* *
+
£ 10
H
+ 12.5mm
«10i
5
600
700
800
900
n -
Mid-radius Radiance Temperature / °C
600
800
700
900
/ °C
FIGURE 6. The change in the central radiance temperature
due to the proximity of the center lightpipe to the wafer (p = 3
mm to 6 mm), for a wafer-reflecting plate distance (s) of
10 mm, versus the average mid-radius radiance temperature.
The values for the model atp = 6 mm are given by a solid line
and dashed line forp = 3 mm.
FIGURE 8. The difference between the temperature of a
thin-film thermocouple junction 5 mm from the wafer's center
(TTF) and the center radiance temperature (TLP), when the LP
is flush with the reflecting plate, versus the center radiance
temperature.
We have included the model predictions according to
the previous section with p values of 5 mm and 10 mm.
Similar values are given in Fig. 6 for s = 10 mm. Here
the depression of the radiance temperature of the center
The thermocouple wafer was calibrated using
previously discussed methods (6, 8) and its uncertainty
was estimated to be less than 2 °C (k = 1). These results
indicate that the LPRTs mounted flush with the
203
reflecting plate were reading 15 °C to 18 °C low at
800 °C due to the effective emissivity of the chamber
REFERENCES
CONCLUSIONS
1.
The 2 mm sapphire lightgpipe surrounded by its
4.2 mm sapphire sheath can cause a significant
depression of the wafer temperature in its target area.
We measured thermal depressions of 25 °C when the
proximity of the lightpipe was 2 mm from the wafer
under static conditions. The 25 °C depression was
measured at 825 °C and is a function of temperature.
This temperature relationship indicates a depression of
over 30 °C at temperatures near 1000 °C. These results
indicate the severity of this effect. They suggest that,
even in rotating systems, some allowance must be made
for the cooling effect of the lightpipe on the wafer at the
location of the measurement.
We have presented a model of the lightpipe proximity
effect that approximated our experimental data for the
most severe cases. This model represents the lightpipe
using opaque, diffuse, and gray surfaces. Further
refinement of the model, including semi-transparent
surfaces, would be expected to more closely represent
the experimental conditions.
The radiance temperature measurements were
calibrated to the absolute temperature scale (ITS-90)
using the NIST thin-film thermocouple wafer.
2.
3.
4.
5.
6.
7.
8.
ACKNOWLEDGEMENT
The authors would like to acknowledge support from
the Office of Microelectronic Programs, NIST.
204
Tsai, B. K., Meyer, C. W., and Lovas, F. J.,
"Characterization of Lightpipe Radiation Thermometers
for the NIST test Bed," 8th International Conference on
Advanced Thermal Processing of Semiconductors,
Gaithersburg, MD (2000).
Meyer, C. W., in Proceedings of TEMPMEKO 2001,
(Edited by B. Fellmuth, J. Seidel and Gunther S cholz),
Berlin, VDE Verlag GMBH, 2002, 937-942.
Kreider, K. G., DeWitt, D. P., Tsai, B. K., Lovas, F. J.,
and Alien, D. W. "RTP Calibration Wafer Using Thinfilm Thermocouples", Rapid Thermal and Integrated
Processing VII, M. Ozturk, et al., Ed. MRS, Pittsburgh,
PA, 1998, pp. 87-94.
Kreider, K. G. and Gillen, G. J., Thin Solid Films, 376,
32-37(2000).
Kreider, K.G. "Temperature Calibration Wafer for Rapid
Thermal Processing Using Thin-film Thermocouples"
U.S. Patent 6,037,645, Mar. 14, 2000.
Kreider, K. G. Ripple, D. C., and DeWitt, D. P.,
"Calibration of Thin-film Thermocouples on Silicon
Wafers", in Proc. of TEMPMEKO 99 the 7th Int'l Symp.
on Temperature and Thermal Measurements in Industry
and Science, June 1999, IMEKO, Delft, NL. pp. 286291.
Kreider, K. G, Chen, D. C, DeWitt, D. P, Kimes, W. A,
Meyer, C. W, and Tsai, B. K, in Rapid Thermal
Processing Technologies III, P. J. Timans, et al. ed. ECS
Proceedings Vol. 2002-11, Sept. 2002, pp 273-281.
Kreider, K. G, Kimes, W. A, Meyer, C. W, Ripple, D. C,
Tsai, B. K, Chen, D. C, DeWitt, D. P, In Proceedings of
8th Symp. on Measurement and Control in Science and
Industry, D. Ripple et al. Ed., AIP, N.Y, in press.