Professional Documents
Culture Documents
Rinsing - A Critical Process in Contamination Removal
Rinsing - A Critical Process in Contamination Removal
Rinsing - A Critical Process in Contamination Removal
Removal
Kurt K. Christenson
FSI International, Inc.
Chaska, Minnesota
Carryover Layer
The success of aqueous-based cleaning of silicon wafers
depends on the complete removal of both the residual
contaminants and the cleaning solutions by a rinse process.
The primary component to be rinsed is the "carryover" layer,
the film of process chemical that clings to the wafer as it is
transferred from the process tank to the rinse tank. Spearow
et al. gives the thickness of the carryover film h in microns as
10
20
30
40
50
60
ExtractionVelocity (cmlsec)
0.05
0.1
0.15
0.2
0.01
0.02
0.03
0.04
0.05
4.1
0.1
0.2
0.3
0.4
0.5
0.6
Ionic Diffusion
An analysis by Tonti indicates that after 15 min in an
overflow rinse tank, 0.88 percent of the initial surface
contamination is still carried over into the subsequent process
bath or dry sequence in the 20-pm layer of water on the wafer
carried over from the rinse tank.4
This poor rinsing performance is caused by a nearly
stagnant boundary layer near the surface of the wafer where
there is little bulk motion of the fluid. Contaminants must
diffuse through this boundary layer to reach the flowing rinse
water and be swept away.
The problem can be treated as diffusion in one dimension
from a finite source (i.e., the total amount of contaminant is
constant). Fick's First Law applied to these boundary
conditions gives a solution of the form
M
[&I
Velocity Profile
Rinse tanks are normally not stagnant, but have average
flow velocities parallel to the surface of the wafers near
1 cmls. The fluid motion in these "overflow" rinse tanks acts
to sweep away contaminants that have diffused into the flow
stream from the surface of the wafers. Figure 3 shows the
velocity distribution of the water between two 200-mm
wafers with 0.25-in. pitch and 1 c d s average water velocity.
The distribution is parabolic with a velocity profile
where uOVe
is the average velocity of the fluid, d is the distance
between the wafers, and h is the distance from the midpoint
between the wafer^.^ The maximum velocity is 1.5 times the
average velocity and occurs midway between the wafers.
Fluid dynamic boundary conditions fix the flow at zero at the
surfaces of the wafers. The interstitial space between the
wafers is rinsed very effectively.Unfortunately, at the surfaces
of the wafers where the contaminants are concentrated, there
is no bulk fluid motion.
6E+05
-$
5E+05
-g 4EM5
52 3E+05
U
2E+05
1EM5
1
0
120
240
360
480
600
Time (sec)
OE+OO
0
0.005
0.01
0.015
0.02
Distance Fmm Surface (cm)
0.025
0.03
contaminants.
Figure 6 shows the concentration profile for 0.3-pm
particles for diffusion times of 60,300,600, and 900 sec. Less
than half of the particles would diffuse 0.0033 cm into the
flow streamduring a 10-minoverflow rinse. Overflow rinsing
is a very inefficient means of removing particles.
Dump Rinsing
"Dumping" and refilling the rinse tank periodically during
the rinse cycle allows contaminants that have diffused some
distance into the bulk of the boundary layer to drain away.
Only the contaminants in the "carryover film" remain on the
wafer.
Figure 7 shows the thickness of the carryover layer in
comparison to the diffusion profiles for ionic contaminants.
Like the carryover layer from the process to the rinse tank, the
thickness of the carryover layer between rinse cycles is
primarily determined by the bath drain time (extraction
velocity), the viscosity and density of the water, and the force
of g r a ~ i t y . ~
-6OSeconds
-3WSeconds
,600Seconds
900Seconds
300
600 Seconds
,900Seconds
c
.O 0.05
LL
0
0
0.005
0.01
0.015
0.02
-g
IE*
6- lEW
+DlvnpRinsa
R IE+OZ
-Rmped
4:
OverRowRi~
Rinse
~ ' l Iondad
e
Initial Areal Density in dmp rinse
9lO' l a d u d lnilialAma! Density in ramp rinse
lodadlec Difhasii Coefficient
lE+Ol
1EaO
1E-01
0
50
100
150
Time (sec)
0.002
0.004
0.006
0.008
0.01
200
Ramped Rinsing
The most critical parameter in determining the efficiency
of the dump rinse sequence is the thickness of the carryover
layer. Figure 9 shows the fraction of contaminants remaining
in the carryover layer as a function of diffusion time and layer
thickness. For a60-sec diffusion time, reducing the carryover
layer from 20 mm to 10 pm would reduce the remaining
contaminants by a factor of two. After four such cycles, the
net reduction in contamination remaining on the wafer would
be 24 or 16 fold.
The thickness of the carryover layer can be reduced
greatly by supplementing the force of gravity with centrifugal
force by spinning the wafers either about their axis as, in a
spin-rinse drier, or off axis, as in a multiple-position spray
processor. The carryover thickness after a 10-secramp at 500
rpm in a multiple-position spray processor was measured to
be less than 2 pm. At 500 rpm, the fluids are removed by a
centrifugal force approximately 50 times as strong as gravity.
Reducing the thickness of the carryover layer from 20 pm
to 2 pm results in areduction in the residual contamination in
the carryover film from 5 percent to 0.5 percent-a 10-fold
reductionforevery rinse cycle. Figure 7 shows the contaminant
distribution in the rinse water after 60 sec. The 2-pm (0.0002cm) layer that remains after the 500-rpm ramp is approximately
equal to the thickness of the y-axis.
Ramped rinsing is also effective in removing particles.
While only 12 percent of 0.3-pm particles diffuse through the
20-pm carryover layer in a dump rinse, 88 percent diffuse
through the 2-pm carryover layer left after a ramp. Thus, 88
JOURNAL OF THE INSTITUTE OF ENVIRONMENTAL SCIENCES
percent of the particles that are dislodged from the surface are
removed in each 1-min ramped rinsing cycle. Ramped rinsing
has a strong, positive effect on particle removal efficiency.
The bottom terraced line of Figure 8 shows the theoretical
performance of ramped rinsing. As with the dump rinsing
data, the model uses a diffusion time of 40 sec followed by a
20-sec ramp period in which no diffusion takes place. The
initial surface concentration for the ramp rinse data is 10
times less than that of the dump rinse data because the
carryover of process chemicals is 10 times less in the ramped
case. After three I-min rinse cycles, the ramped wafers are
1000 times cleaner thun the dump rinse wafers.
A further benefit of ramped rinsing over dump or overflow
rinsing is reduced water usage. As seen in Figure 2, water
farther than 0.05 cm from the surface of the wafer plays a
small role in the diffusion process and therefore the rinse
process. In immersion, 80 percent of the water in the 0.5-cm
gap between the wafers is wasted. If the water is dispensed by
spray, ramped rinsing achieves a five-fold reduction in water
usage by only wetting the top and bottom surfaces of the
wafer.
Experimental Performance
Figure 10 shows a comparison of the experimental
performance of ramped and overflow r i n ~ i n g . ~
In both cases, challenge wafers were prepared by dipping
wafers in a saturated KC1 solution and spinning them dry.
Residual potassium was measured with TXRF. The
contamination is reduced in the overflow rinser by only 1
LRV each 4 min. A clear potassium signal is still detectable
after a 10-min rinse.
The simple theoretical model in Figure 8 predicts a 2.3
LRVImin for ramped rinsing. Experimentally,ramped rinsing
reduced the signal by 3.5 LRV in 30 sec, three times faster
than the theory and 28 times faster than the overflow rinser.
No potassium signal is detectable after 30 sec of ramped
rinsing. There are two reasons that ramped rinsing performs
so much better than the prediction of the simple model:
Diffusion continues during the 20-sec ramp-off period.
If rinse water is dispensed during the ramp, a thin layer of
clean water above the 2-pm carryover layer will be
refreshed constantly, as if there were an infinite sink for
contaminants a few microns from the wafer surface.
While these mechanisms can also be active in dump
rinsing, the 50-fold ratio in the centrifugal force in the ramp
to the gravitational force in the dump renders them far less
significant in dump rinsing.
RINSING IN TRENCHES
120
240
360
480
600
Secondsof Rnsing
Cmax'c
(t = 0)
0.27
0.058
5.7E-4
2.5E-7
5.1E-14
3.
4.
5.
6.
7.
8.
9.
10.
1 I.
12.
13.
ACKNOWLEDGMENTS
The author thanks Brent Carlson, Jim Oikari, and Moshe Olim
of FSI International for helpful discussions and significant
laboratory andfield work in testing these ideas.
AUTHOR BIOGRAPHY
Dr. Kurt K. Christenson, a senior process physicist, joined FSI
International in 1990. He works in process development for the
centrifugal spray processor product line. He received his B.A. in
Physics in 1980 at Bethel College, St. Paul, Minnesota, and his M.S.
andPh.D. in Physics from theuniversity of Illinois aturbanain 1983
and 1988,respectively.His postdoctoral workinE-beam lithography
was completed at the IBM T.J. Watson Research Center in 1990. He
has written over 45 papers on microanalysis, E-beam lithography,
and semiconductor wafer cleaning.
50
Journal Reviewers
The editors of the Journal of the Institute of Environmental Sciences wish to express their sincere appreciation to the following reviewers for their dedicated
efforts in reviewing manuscripts submitted for publication. As a result, their helpful comments to the authora
have resulted in works that are a credit to both the
authors and our Journal.
Albaugh, Kevin
Allen, Hillary
Anderson, Robert
Bateman, Vesta
Bevirt, David
Blackford, David
Borson, Eugene
Bugni, David
Busnaina, Ahmed
Bzik, Thomas
Chenoweth, Halsey
Chu, Anthony
Connon, William
De Pinillos, J. Martinez
De Vecchi, Franco
Dillenbeck, Keith
Dixon, Anne Marie
Djordjevic, Walter
Donovan, Robert
Dryden, Richard
Ensor, David
Fitzpatrick, Michael
Fosnight, William
Francis, Terry
Geminder, Robert
Giroux, Robert
Gotlinsky, Barry
Greenberg, Barry
Haider, Asad
Himelblau, Harry
Hope, David
Hu, Jimmy
Hubach, Frank
Jensen, David
Krasich, Milena
Kuehn, Thomas
Lambert, Ronald
Lee, Aleck
Lieberman, Alvin
Lipp, Louie
Lloyd, J.
Lunde, Albert
Ma, Ce
Mattina, Charles
McAndrew, James
McDonald, Bruce
Merritt, Ronald
Mielke, Robert
Miller, Robert
Moore, Donald
Nagarajan, R.
O'Hanlon, J.
Patterson, Douglas
Pederson, Paul
Piersol, Allan
Popolo, Joseph
Ramstorp, M.
Scialdone, John
Sem, Gilmore
Shadman, Farhang
Sherf, Zeev
Silver, William
Simpson, Alda
Smallwood, David
Smith, Harold
Steakley, Joyce
Swinehart, David
Thomas, Valerie
Tousi, Saied
Van Sickle, Penny
Walter, A.
Wasz, Glenn
Wei, Jan
Wilkinson, Bud
Yang, Michael
Zahka, Joseph