See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249512792

Optimizing the fluid dispensing process for immersion lithography

Article in Journal of Vacuum Science & Technology B Microelectronics Processing and Phenomena · November 2004
DOI: 10.1116/1.1824065

CITATIONS READS
7 333

7 authors, including:

Amr Y. Abdo Gregory Nellis

IBM University of Wisconsin–Madison
49 PUBLICATIONS 273 CITATIONS 208 PUBLICATIONS 3,314 CITATIONS

SEE PROFILE SEE PROFILE

Roxann Engelstad Steven Brueck

University of Wisconsin–Madison University of New Mexico
261 PUBLICATIONS 1,579 CITATIONS 571 PUBLICATIONS 14,165 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Steven Brueck on 26 December 2014.

The user has requested enhancement of the downloaded file.

Optimizing the fluid dispensing process for immersion lithography
A. Abdo,a) G. Nellis,b) A. Wei, M. El-Morsi, and R. Engelstad
Computational Mechanics Center, University of Wisconsin, 1513 University Avenue, Madison,
Wisconsin 53706
S. R. J. Brueck and A. Neumann
Center for High Technology Materials, University of New Mexico, 1313 Goddard, SE, Albuquerque,
New Mexico 87106
(Received 3 June 2004; accepted 4 October 2004; published 14 December 2004)
The concept behind immersion lithography is the insertion of a high refractive index liquid in the
space between the final projection lens of an exposure system and the device wafer to improve the
overall resolution of the exposure process. Computational fluid dynamics (CFD) simulations were
performed in order to investigate the process of initially filling the lens-wafer gap with immersion
fluid. The CFD models were used to investigate the effects of dispense velocity, gap height, and fluid
dispense angle on the fill process; specifically on the possibility of air entrainment. The simulations
revealed that there is an optimal region in the parameter space of gap height and dispense velocity
for which the gap fills completely. Outside of this region, either excessive inertial or surface tension
forces cause an undesirable, incomplete filling process. The optimal region was found to shift
somewhat based on the fluid dispense angle. Finally, experiments were performed to verify the CFD
models. The CFD simulations and the experimental results were in good agreement, both
qualitatively with regard to the shape and evolution of the free surface and quantitatively with regard
to the velocity of the contact line. © 2004 American Vacuum Society. [DOI: 10.1116/1.1824065]

I. INTRODUCTION quasisteady manner; these problems include fluid heating4,5

and air entrainment due to free surface flows over wafer
Immersion lithography has been proposed as a method for
topology.6 The simulations presented in this work parametri-
extending optical lithography resolution to 45 nm and
cally study the filling process that must occur prior to initia-
below.1 The premise behind immersion lithography is to im-
tion of immersion exposures. The parameters investigated
prove resolution by increasing the index of refraction in the
include the fluid dispense velocity and angle as well as the
space between the final projection lens of an exposure sys-
gap height between the lens and the wafer.
tem and the device wafer by inserting a high index liquid in
An optimal region in the parameter space of fluid velocity
place of the low index air that currently fills the gap. The
and gap height is identified. The region is delineated on one
liquid in the gap is an optical element and therefore must
have a uniform index of refraction. To this end, the gap must side by a critical Reynolds number, above which the dis-
be completely filled with liquid (water for 193-nm lithogra- pense jet impinges on the wafer forming a thin film that does
phy with other fluids being evaluated for 157-nm not completely fill the gap. On the other side, the region is
lithography2) and air bubbles cannot be tolerated. delineated by a critical Weber number, below which a me-
In order for immersion lithography to be practical, the niscus is formed from the dispense port and crawls along the
introduction of liquid into the wafer-lens gap must not ad- lens. The subsequent rupture of the meniscus upon its even-
versely impact the overall manufacturing process. Since the tual contact with the wafer is likely to entrain unacceptable
immersion fluid acts as an optical component during the air bubbles. The fluid dispense angle has an effect on the
lithographic process, it must retain a high and uniform opti- location of this optimal region.
cal quality. It is critical that the fluid management system In order to verify some aspects of the CFD simulations,
reliably and rapidly fill the entire lens-wafer gap,3 maintain the model predictions were compared with an experiment in
the fill under the lens throughout the entire exposure process, which water was used to fill a gap that replicates the lens-to-
and ensure that no bubbles are entrained during filling or wafer gap in an immersion lithography tool. There was good
scanning. Optimization of the fluid dispensing process is es- agreement between the CFD and experimental results, both
sential to the success of immersion lithography; therefore, qualitatively in terms of the shape of the free surface and its
two- and three-dimensional (2D and 3D) computational fluid progression across the gap and quantitatively in terms of the
dynamics (CFD) models have been developed to simulate velocity of the filling process.
the filling process. Several previous studies have used CFD The use of 2D CFD models for the parametric studies and
simulations to investigate potential problems that may arise 3D CFD models for experimental verification was required
when the immersion lithography process is operating in a by practical constraints. It was important to verify that the
CFD models were predictive and therefore some experimen-
a)
Currently with IBM Microelectronics Division, East Fishkill, NY. tal comparison was required; however, the experimental ap-
b)
Electronic mail: [email protected] paratus involved 3D flow and therefore a 3D CFD model

3454 J. Vac. Sci. Technol. B 22(6), Nov/Dec 2004 0734-211X/2004/22(6)/3454/5/$19.00 ©2004 American Vacuum Society 3454
3455 Abdo et al.: Optimizing the fluid dispensing process 3455

FIG. 1. (a) Schematic of the dispensing process. The

behaviors are delineated in (b) which shows the fluid
dispense velocity as a function of gap height with three
characteristic regions identified. The effects of dispense
velocity and gap height on the fill process are shown in
(c) for the behavior just after the meniscus breaks for a
low dispense velocity (note trapped air bubble), in (d)
with the thin, high velocity fluid film associated with a
high dispense velocity, and in (e) for the complete and
rapid filling process that occurs when an acceptable dis-
pense velocity is used.

was necessary to yield quantitative agreement. Unfortu- pending on these process variables, as illustrated in Fig. 1(b).
nately, the use of a 3D CFD model to carry out the paramet- The simulation was conducted with the three gap heights
ric studies would require a prohibitive amount of computa- only; the connecting line shown in the graph is qualitative, to
tional time. Therefore 2D CFD models were used for the illustrate the trend of the simulation results.
parametric studies of the underlying physics. The correlation At low fluid dispense velocity the liquid forms a hanging
between the 2D and 3D results has been observed experi- droplet that does not immediately touch the wafer surface.
mentally; for example, it is possible to experimentally ob- As more liquid is added to the droplet it tends to crawl along
serve the transition between impinging jet and optimal dis- the hydrophilic lens surface; eventually the meniscus touches
pense at a constant gap height by increasing the velocity the wafer and at this point it breaks very rapidly, as shown in
using the experimental test facility described in this paper. In Fig. 1(c). This behavior is undesirable because the sudden
other testing, the transition between meniscus crawl and op- rupture of the meniscus is unpredictable and may entrain
timal dispense behavior has been observed when the dis- macroscopic air bubbles that can be clearly identified in Fig.
pense velocity is reduced. While the experimentally observed 1(c). This air cannot escape from the meniscus to wafer gap
transition velocities are not exactly equal to those shown in during the rupture process and therefore forms small bubbles
the paper, they are nominally consistent with the critical We- under the influence of surface tension forces. At each gap
ber number and Reynolds number identified by the 2D CFD height, a series of simulations at progressively lower dis-
model. pense velocities was used to delineate the transition from a
completely filled gap to this meniscus crawl behavior, as
shown in Fig. 1(b). The specification of this transition is
II. CFD MODELING RESULTS
somewhat qualitative but is based on the extent to which the
Figure 1(a) shows a schematic of the filling process used meniscus has progressed in the direction of the gap at the
in the two-dimensional CFD simulations. Immersion fluid time when it touches the wafer.
(water for these simulations) was dispensed with a uniform At very low fluid dispense velocities, the inertia of the
velocity from a port adjacent to the lens surface. The lens dispensed fluid is insufficient to overcome surface tension
and dispense port are positioned at the same distance from a and break the meniscus. Thus it has the chance to crawl
wafer which is stationary for the simulations and experi- along the lens. The transition into this behavior is therefore
ments described in this article. The progress of the filling governed by a balance between fluid inertia and surface ten-
process is monitored after the initiation of the fluid dispense. sion which constitutes the Weber number 共We兲:
The CFD simulation results showed that the fluid dispense
process is sensitive to a number of parameters including the ␳ v 2D
We = , 共1兲
fluid dispense velocity, the wafer-to-lens gap height, and the ␴
fluid dispense angle. Three different gap heights (i.e., 0.3 where ␳ and ␴ are the density and surface tension of the
mm, 0.6 mm, and 0.9 mm) were examined in a parametric fluid, D is the length scale associated with the dispense port,
study; for each height, the fluid dispense velocity was varied and v is the dispense velocity. Note that the Weber number as
from 0.0 to 3.0 m / s. The parametric studies showed that the defined in this way is independent of gap height which is
dispense process behavior falls into different regimes de- consistent with the results shown in Fig. 1(b), in that the gap

JVST B - Microelectronics and Nanometer Structures

3456 Abdo et al.: Optimizing the fluid dispensing process 3456

FIG. 2. Delineation of the different physical behaviors of the fluid dispense

process based on critical Weber and Reynolds numbers.

height has only a very small effect on the fluid dispense

velocity at the onset of the meniscus crawl behavior.
At very high fluid dispense velocities the fluid inertia is
sufficient to create a thin, high velocity fluid film on the
wafer surface; if the fluid film is sufficiently thin, the gap
does not entirely fill with liquid, as shown in Fig. 1(d). This
behavior is also undesirable. At each gap height, a series of
simulations at progressively higher dispense velocities was
used to delineate the transition from a completely filled gap
to this impinging jet behavior, also shown in Fig. 1(b). The
specification of this transition is more clear than the menis-
cus crawl transition as it is based entirely on whether an
unfilled region persists adjacent to the dispense port.
The thickness and velocity of the fluid layer that is formed
on the wafer can be approximately determined by balancing
fluid momentum and shear in the thin film while enforcing FIG. 3. (a) Schematic illustrating the different dispensing angles investi-
that the mass carried by the film is consistent with the dis- gated, (b) the CFD simulation for a 30° dispense angle with a 3.0 m / s
dispense velocity, (c) the CFD simulation for a 60° dispense angle with a
pensed mass flow rate. If the fluid momentum is large 1.8 m / s dispense velocity, and (d) the effect of changing the dispensing
enough then the thickness of the liquid film is less than the angles and velocities on the optimum dispensing zone: the 0.3 mm gap
gap height and impinging jet behavior occurs. The transition (shown by the dotted line) was used to illustrate the fluid dispensing behav-
into this behavior is therefore governed by a balance between ior for three dispensing angles, the velocities indicated on the graph shows
the new locations for the transition zones (i.e., 0.8 and 0.9 are the meniscus
fluid inertia and viscous shear which constitutes a Reynolds crawl critical velocities for 30° and 60° angles). The arrows show the new
number 共Re兲: potential location of the transition lines as the dispensing angle increases.
␳vh
Re = , 共2兲
␮
with a constant Weber number and Reynolds number. Notice
where h is the gap height and ␮ is the fluid viscosity. Notice that the transition from optimal dispense to impinging behav-
that this Reynolds number is proportional to the gap height, ior is approximately consistent with a transition Reynolds
which is consistent with the results shown in Fig. 1(b) in that number 共Ret兲 of 750 and the transition from optimal dispense
the transition velocity decreases as the gap height increases. to meniscus crawl behavior is approximately consistent with
These two extreme cases bracket a range of acceptable a transition Weber number 共Wet兲 of 1. There is some discrep-
dispensing velocities. An acceptable dispense velocity is ancy between the transitions associated with these critical
characterized as being sufficiently high so that the meniscus values of Weber and Reynolds number and the transitions
immediately breaks directly under the dispense port but predicted by CFD, particularly at larger gap heights. These
small enough so the entire gap is filled, resulting in the rapid discrepancies can be partly explained by the fact that the
and controlled fill process shown in Fig. 1(e). Figure 1(b) transition between dispense behaviors is actually a gradual
illustrates that the range of acceptable dispensing velocities change rather than an abrupt one and therefore the determi-
narrows as the gap height increases. nation of the critical dispense velocities that delineate these
Figure 2 overlays the CFD results onto lines associated regions relies both on judgment and the resolution of the

J. Vac. Sci. Technol. B, Vol. 22, No. 6, Nov/Dec 2004

3457 Abdo et al.: Optimizing the fluid dispensing process 3457

FIG. 5. Photographs showing the progress of the filling water front for (a)
0.5 mm gap, (b) 1.5 mm gap, and the corresponding CFD simulation results
showing similar behavior for (c) 0.5 mm gap, (d) 1.5 mm gap.

shown Fig. 3(d) which contains the original points delineat-

ing the dispense process regimes for the 0° fluid dispense
angle case as well as additional points for a gap height of 0.3
mm and fluid dispense angles of 30° and 60°. Note that for a
30° fluid dispense angle, indicated by the triangles in Fig.
3(d), both the meniscus crawl transition and impinging jet
transition, shown in Fig. 3(b), are shifted to higher velocities.
For a 60° dispense angle, the meniscus crawl transition again
moves to a higher velocity. However, the impinging jet tran-
sition no longer occurs. Instead a new behavior is exhibited
in which the fluid remains attached to the lens and tends to
FIG. 4. (a) Schematic illustrating the details of the model verification experi-
ments, (b) a close-up showing the fixture and glass window replicating the pull air into the gap, as shown in Fig. 3(c).
lens, (c) a schematic of the lens showing the recesses and the fluid dispense
and recover ports, and (d) the CFD half-symmetry model developed to rep-
licate the experiment.
III. EXPERIMENTAL VERIFICATION
In order to verify some aspects of the CFD model, the
simulated dispensing process was compared with an experi-
velocities used in the CFD simulation. Also, the physics of ment in which a metered amount of water was used to fill a
the problem changes as the gap increases; for example, a controlled lens-to-wafer gap. Figure 4(a) shows a schematic
larger gap allows the meniscus that forms on the lens to of the experiment. The progress of the filling process was
expand laterally by a greater amount prior to contacting the monitored using a camera that was mounted over a glass
wafer and as a result the length scale associated with the window (which takes the place of the lens element). Figure
Weber number is no longer directly related to the dispense 4(b) shows a cut-away view of the mounting fixture and
port as given by Eq. (1). indicates passages for the dispense and recovery of water.
Additional CFD simulations showed that the dispensing Figure 4(c) is a solid model of the bottom of the holder
behavior is affected by the fluid dispense angle. Figure 3(a) which shows the recesses in the structure used to manifold
illustrates the three dispensing angles investigated in this the water to the dispense and recovery ports.
work, i.e., 0°, 30°, and 60°, with the positive angles indicat- A three-dimensional CFD model was developed with the
ing that the fluid is directed towards the lens. For a moderate goal of replicating the actual experimental geometry as
dispense angle (e.g., 30°), the same regions of behavior exist closely as possible, as shown in Fig. 4(d). The details of the
but the transitions are shifted. The inertia of the fluid in the recessed dispense port, the recessed recovery port, and the
direction perpendicular to the wafer is reduced when the gap can be identified in the CFD model. The experiment was
fluid is dispensed at an angle. Both the meniscus crawl and carried out at the same flow rate but varying gap heights.
impinging jet transitions are proportional to this fluid inertia Three of these experiments were simulated, including 0.5
and therefore both lines shift upwards; a higher dispense ve- mm, 1.0 mm, and 1.5 mm gap heights. Figures 5(a) and 5(b)
locity is required to yield the same inertia. This effect is show pictures of the fluid front that forms as the immersion

JVST B - Microelectronics and Nanometer Structures

 3458 Abdo et al.: Optimizing the fluid dispensing process 3458

fluid fills a 0.5 mm and 1.5 mm gap, respectively. Notice that

the fluid front formed for the smaller gap is much flatter
across its advancing edge than for the larger gap. This dif-
ference is related to whether the dispense flow rate is suffi-
cient to fill the entire width of the gap at the rate that the
leading edge of the meniscus is drawn through the gap by
surface tension. The CFD simulations for the filling process
for the same gaps, shown in Figs. 5(c) and 5(d), show similar
behavior. Quantitative comparisons between the experiments
and the CFD simulations are shown in Figs. 6(a), 6(b), and
6(c), where the distance between the water front to the dis-
pense port measured along the center of the circular glass
window is shown as a function of time relative to the initia-
tion of the fill process for the 0.5 mm, 1.0 mm, and 1.5 mm
gaps, respectively. The results from the CFD simulations and
the experiments are in very good agreement.

IV. SUMMARY AND CONCLUSIONS

Two-dimensional CFD models were used to investigate
the effects of fluid dispense velocity, gap height, and fluid
dispense angle on the process of filling the lens-wafer gap
prior to immersion lithography. The simulations revealed that
an optimal region exists in the space of fluid dispense veloc-
ity and gap. The position of this region shifts as the fluid
dispense angle is changed.
A three-dimensional CFD model was developed in order
to replicate an experiment in which a controlled fill is moni-
tored through a glass window. The simulation results agreed
well with the experimental observations both quantitatively
in terms of the progression of the fluid front with time and
qualitatively in terms of the shape of the fluid front.

ACKNOWLEDGMENTS
The UW research was funded by DARPA/ARL and the
Semiconductor Research Corporation (SRC). The UNM por-
tion of the work was funded by International SEMATECH
and by the ARO/MURI program in Deep Subwavelength
Optical Nanolithography. Computer support was provided by
the Intel Corporation and Microsoft.
1
M. Switkes and M. Rothschild, J. Microlithog., Microfabr., Microsyst. 1,
225 (2002).
2
R. R. Kunz, M. Switkes, R. F. Sinta, J. E. Curtin, R. H. French, R. C.
Wheland, C.-P. Chai Kao, M. P. Mawn, L. Lin, P. M. Wetmore, V. J.
Krukonis, and K. Williams, J. Microlithog., Microfabr., Microsyst. 3, 73
(2004).
3
A. Wei, G. Nellis, A. Abdo, R. Engelstad, C.-F. Chen, M. Switkes, and M.
Rothschild, J. Microlithog., Microfabr., Microsyst. 3, 28 (2004).
4
A. Wei, A. Abdo, G. Nellis, R. Engelstad, J. Chang, E. Lovell, and W.
Beckman, J. Vac. Sci. Technol. B 21, 2788 (2003).
5
FIG. 6. Distance from the fluid front to the dispense port along the centerline A. Wei, A. Abdo, G. Nellis, R. Engelstad, J. Chang, E. Lovell, and W.
of the glass window measured from experimental results and predicted by Beckman, Microelectron. Eng. 73-74, 29 (2004).
6
the CFD simulations for (a) 0.5 mm gap, (b) 1.0 mm gap, and (c) 1.5 mm A. Wei, M. El-Morsi, G. Nellis, and R. Engelstad, J. Vac. Sci. Technol. B,
gap. these proceedings.

J. Vac. Sci. Technol. B, Vol. 22, No. 6, Nov/Dec 2004

View publication stats