Ref - 19 of 22 - 00895075

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

300 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 23, NO.

4, OCTOBER 2000

Modeling of Time-Pressure Fluid Dispensing


Processes
X. B. Chen, G. Schoenau, and W. J. Zhang

Abstract—The process of time-pressure fluid dispensing has


been widely employed in the semi-conductor industry, where the
fluid is applied to boards or substrates. In such a process, the flow
rate of fluid dispensed and the shape of fluid formed on the board
are the two most critical performance indexes, yet extremely
difficult to represent because of their complex behavior. This
paper presents the development of a model for the time-pressure
fluid dispensing process, by which the flow rate and shape can
be established. Experiments have been performed to validate the
model developed.
Index Terms—Dispensing process, equilibrium shape, flow rate, Fig. 1. Time pressure dispensing schematic.
modeling, non-Newtonian fluid.
a dispensing process is very important, yet demanding and
I. INTRODUCTION difficult due to the complexities involved in describing the
fluid rheological behavior and solving the equations governing
F LUID dispensing is a method by which fluid is applied
to boards or substrates in a controlled manner. It has been
widely employed in such processes as advanced integrated
the spreading of fluid on the board which is a free-boundary
problem. Using the system identification method, West et al.
[5] developed an Auto Regressive with External input (ARX)
circuit encapsulation (AICE) and surface mount technology
model to represent the volume of fluid dispensed. Razban [6]
(SMT) in the semi-conductor industry [1]–[4]. Depending
identified the dynamics of the volume dispensed as a single
on the method by which the fluid is forced out of a needle,
lumped parameter dynamics with a pure time delay that is
dispensing can be mainly classified into three types: time-pres-
a function of operating conditions. It is noted that all these
sure, rotary screw, and positive displacement piston. Of these
models were black-box models or empirical models, which
methods, the time-pressure is the most widely used due to its
were based on the measured input and output data. However,
flexibility to different application situations, simple operation
for the proper design of a dispensing process, a physical model
and easy maintenance. A schematic diagram of the time-pres-
or combination of a physical model and empirical models is
sure dispensing method is illustrated in Fig. 1. The air pressure
very useful to investigate the influence of structural parameters
is applied to the top of a syringe to force the fluid through and
on the performance of the process, and to design the appropriate
out of the needle. Then the fluid extruded drops to the board,
controller for the process.
flows and forms a fixed shape on the board.
This paper presents the modeling of the time-pressure fluid
In such a dispensing process, the volume dispensed and
dispensing process. In particular, models for the flow rate of
the shape formed on a board are the two most important
fluid dispensed and the shape of fluid formed on the board are
performance indexes. With the chip or component becoming
developed. As the rheological behavior of the fluid dispensed
smaller and smaller, these two indexes are becoming more and
is critical in a dispensing process and significantly affects the
more critical to successful dispensing. The ever-increasing
performance of this process, the representation of the behavior
demand for more precise volume and shape has been not only
is first addressed, and some models reported in the literature
significantly spurring semi-conductor manufacturers to pursue
are reviewed. A model is developed for the flow rate of fluid
new dispensing methods, but also stimulating research into the
for the time-pressure dispensing process. Based on the equa-
theoretical and experimental investigations into the dispensing
tions governing the spreading of fluid on a board, the shape of
process to improve the performance. In these investigations,
fluid is determined by applying the equilibrium conditions of
the development of a model to represent the performance of
the spreading process. For the validation of the models devel-
oped in this paper, experiments are carried out and the results
Manuscript received August 17, 2000; revised November 10, 2000. This are compared with those of the model simulations.
work was presented in part at the Sixth Triennal International Symposium
on Fluid Control Measurement and Visualization, Sherbrooke, QC, Canada,
August 13–17, 2000. This work was supported by ASM Assembly Automated, II. MODELING OF THE RHEOLOGICAL BEHAVIOR OF FLUIDS
Ltd., Hong Kong, and by the Natural Science and Engineering Research
Council (NSERC) and Atomic Energy Canada Limited (AECL) through a Usually, the dispensing fluids, such as epoxy, adhesive, and
Cooperative Research and Development Grant. encapsulant, are non-Newtonian or rheological fluids. One of
The authors are with the Department of Mechanical Engineering, University
of Saskatchewan, Saskatoon SK S7N 5A9, Canada. the characteristics of the non-Newtonian fluid is that the rela-
Publisher Item Identifier S 1521-334X(00)11628-3. tionship between the shear stress and the shear rate (the ratio of
1521–334X/00$10.00 © 2000 IEEE

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: MODELING OF TIME-PRESSURE FLUID DISPENSING PROCESSES 301

shear stress to shear rate is defined as apparent viscosity) is not


a constant, and could be a function of the shear rate, temper-
ature, and shearing history [7]–[9]. There is not a universally
applicable equation for describing the viscosity of non-New-
tonian fluids. In the literature, there are many empirical equa-
tions available, and a detailed review of them was given by
Holdsworth [10] for the application of these equations in the
food processing industry.
For the modeling of rheological behavior of the fluid in
dispensing, only some of these empirical equations could be
employed. They are Power Law, Generalized Power Law (or
Herschel–Bulkley), or Time-Dependent Herschel–Bulkley
equations, which were given by [10], respectively Fig. 2. Schematic of fluid flowing in time-pressure dispensing process.

Power Law: exit of the needle, respectively, and is the sum of all fric-
tion losses between the air–fluid interface and the exit of the
Generalized Power Law: needle.
From Assumption (4), the term in the left side of
(2) can be neglected as approximately equals to zero, and
Time-Dependent Herschel-Bulkley Law: the friction losses are predominantly those associated with
(1) flowing in the needle. Thus

In (1), and are the shear stress and the yield stress, respec- (3)
tively; the shear rate; the time of shearing; , , represent
the time consistency index with units of ( where is the frictional pressure drop through the needle,
for Power Law and Generalized Power Law equations), the the loss correction factor. The first term is the energy used to
time flow behavior index (dimensionless), and the thixotropic overcome friction in a fully developed flow through the needle,
index (dimensionless), respectively. It is noted that the Power and the second term accounts for the minor losses due to the
Law equation does not include the influence of the yield stress entrance effect, flow-through fitting, etc.
while the Generalized Power Law and Time-Dependent Her- Substituting (3) into (2) and solving for the frictional drop
schel–Bulkley equations do; and that, in the Time-Dependent across the needle for a fully developed flow, one has
Herschel–Bulkley equation, the time-dependent thixotropic
characteristics (decreasing in shear stress with the shear time)
(4)
of fluids is represented by introducing the thixotropic index .
It should be noted here that, for non-Newtonian fluids, the
III. MODELING OF THE FLOW RATE IN THE TIME-PRESSURE evaluation of the kinetic energy correction factor, , and the en-
DISPENSING PROCESS trance loss correction factor, , is very difficult, and not well
It is assumed, for the time-pressure dispensing process, that defined in the literature. Very limited measurements [11], [12]
1) fluid is incompressible and time-independent; on non-Newtonian fluids have indicated that the combined cor-
2) fluid is in steady laminar flow; rection for kinetic energy effects and entrance effects is the same
3) there is no slip between the fluid and the syringe as well as that for Newtonian fluids. So the method of evaluation of
as needle walls; and for Newtonian fluids is adopted here to calculate the fric-
4) internal diameter of the syringe is much greater than that tional pressure drop across the needle. In the case of Newtonian
of the needle. fluids, has the value of 0.5, and is suggested as 0.23 for a
These conditions closely approximate the typical dispensing slightly rounded entrance by Foust et al. [13]. Thus, the expres-
process, and the geometry for this problem is shown in Fig. 2. sion for calculating the frictional pressure drop can be rewritten
The frictional pressure drop is first derived by applying the total as
mechanical energy conservation at the air–fluid interface, where (5)
the pressure is (equal to the air pressure applied to the dis-
penser), as well as the exit of the needle, where the pressure is It is assumed that the relationship between the shear stress
(usually atmospheric pressure). This is illustrated in and the shear rate of the fluid dispensed is well described by
. The flow rate, , through the needle can be established by
(2) using the general equation for laminar flow of time-independent
fluids in a cylindrical tube [7], as

where and ( ) are the average velocity and the ki-


(6)
netic energy correction factor at the air–fluid interface and the

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.
302 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 23, NO. 4, OCTOBER 2000

where is the local velocity at the radial, , and given by


. And is the shear stress at the needle
wall. Note that is dependent on the frictional pressure drop,
, and can be derived, by applying the equilibrium of the
forces acting on the fluid in the needle

(7)

IV. MODELING OF FLUID SHAPE ON A BOARD


Fig. 3. Schematic geometry of fluid spreading on board.
Once the fluid drops onto a board, it begins to spread and
finally forms a “fixed” or equilibrium shape on the board. This
are not concerned with the spreading process, but the outcome of
shape is a very important index in the control of a dispensing
this process, i.e., the final equilibrium shape of the fluid formed
process. This section presents the development of a model to
on the board. Therefore, one can apply the final equilibrium con-
represent the shape of fluid in the dispensing process.
ditions to simplify (8) in order to obtain the final shape. These
Consider that the fluid is spreading on a smooth, horizontal
equilibrium conditions are given by
rigid board located at position and kept at a constant tem-
perature , which is regulated by a temperature controller 1) There is no movement in the fluid
in the dispensing process. The fluid is surrounded by ambient air
at a temperature of , which has a viscosity and thermal con-
ductivity that are relatively small compared to those of the fluid. (9)
The configuration of this problem is shown in Fig. 3. The fluid 2) The fluid reaches the same temperature as the board.
spreading is considered in either two-dimensional Cartesian co-
ordinates where a line is formed with the needle moving (10)
along the axis as shown in Fig. 1, or axisymmetric cylindrical
coordinates where a dot is formed without movement of 3) The interface between the fluid and the air is symmetric
the needle. Both cases are examined and the equation numbers and smooth, and retains an equilibrium shape due to the
denoted by (for line) and (for dot) in the following, respec- surface tension, , and the pressure difference, , be-
tively. The shape of the interface between the spreading fluid tween the fluid and the air, which is governed by the
and the ambient air is denoted by for a line, or Laplace equation [17]
for a dot.
The velocity and thermal fields in the fluid are governed by
the Navier–Stokes, continuity, and energy equations [14]
(11l)

(8)
(11d)
where
T;
T is the velocity vector;
Note that the surface tension, , is nearly linearly depen-
the pressure; dent on the temperature [17]
the temperature of the fluid;
, and are the viscosity, specific heat and the thermal (12)
conductivity of the fluid, respectively.
Given the boundary and initial conditions, one could solve the where is the surface tension at the temperature, , and
above equations as function of time. In general, however, this is the temperature coefficient.
very difficult as what is involved is a moving-boundary problem 4) The volume of the fluid is conserved during the spreading
for the Navier–Stokes equations. The problem can be simplified process as the fluid is assumed incompressible. There-
by using the lubrication approximation which assumes that the fore, the following equations are established:
ratio of the height to the width of a line or the radius of a dot
is very small [14]–[16]. However, this assumption is not appli- (13l)
cable for the case considered in this paper as the height is the
same order of magnitude as the width or the radius in the dis-
(13d)
pensing process. Fortunately, for the control of this process, we

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: MODELING OF TIME-PRESSURE FLUID DISPENSING PROCESSES 303

where
volumetric flow rate of the fluid dispensed;
dispensing time;
velocity of the needle movement;
position of the equilibrium contact line;
height at ;
inverse function of .
5) A boundary condition is still needed at the contact line be-
tween the fluid and the board, which is usually measured
by the contact angle as shown in Fig. 3. The contact
angle at the equilibrium, , is called static contact angle
and found to be a constant depending on the temperature
on the board [18]

(14l)

(14d)

Thus, substituting the equilibrium conditions described in (9)


and (10) into (8) yields

(15)
Solving the previous equation results in the pressure in the
fluid at height

(16)

where is the pressure at . Then, the pressure difference


between the fluid and the air can be obtained by
(17)
where is the pressure in the air around the fluid.
Thus, combining (11), (13), (14) and (17), one can solve and Fig. 4. (a) Viscosity versus temperature measured at the shear rate of 22 1/s
determine the final equilibrium shape for a line or a and (b) Viscosity versus shear rate measured at the temperature of 50 C.
dot numerically.
TABLE I
V. COMPARISON OF THEORETICAL AND EXPERIEMENTAL TABLE OF VALUES OF CURVE-FITTING PARAMETERS
RESULTS
A. Rheological Characterization
The viscosity of the encapsulant considered in this paper was
measured with a rheometer as a function of temperature and foil). In the experiments, the encapsulant used was the same as
shear rate. The measured rheological results are illustrated in the one identified in previous section.
Fig. 4. Fig. 4(a) shows the viscosity versus the temperature, and The first experiment was conducted to measure the flow rate
Fig. 4(b) the viscosity versus the shear rate. of encapsulant dispensed. In the experiment, two needles of dif-
The Power Law model given in (1) is chosen to correlate the ferent sizes (length of 18 mm; and internal diameters of 0.8382
experimental results, in which the index keeps a constant and mm and 1.3462 mm) were used, and the temperatures of the
is related to the temperature, , by needle and the board were controlled at 25 C and 70 C, respec-
(18) tively. By applying air pressures of 3, 4, 5, and 6 bar for a period
of 1 minute on the dispenser, various amounts of the encapsu-
where , , and are curve-fitting parameters, and their lant were dispensed on the board. They were then weighted, and
values listed in Table I. the measured results are shown in Fig. 5.
Another experiment was performed to measure the width and
B. Flow Rate and Shape the height of the encapsulant lines formed on the board. In the
Experiments were performed using a dispensing center (at experiment, two needles of the same sizes as those the first ex-
ASM, Hong Kong), which employed the time-pressure dis- periment were used, and the temperatures of the needle and the
pensing method. The test rig is used to control the air pressure board were controlled at the same temperature of 25 C while
and the temperatures of the needle and the board (aluminum the air pressure was manipulated to keep a constant of 4 bar. The

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.
304 IEEE TRANSACTIONS ON ELECTRONICS PACKAGING MANUFACTURING, VOL. 23, NO. 4, OCTOBER 2000

the static angle (44.1 ) at the temperature of 25 C. The simula-


tion results are illustrated in the solid lines in Figs. 5 and 6. It can
be seen that the simulation results agree with the experimental
results well. It is important to note that values of the rheological
parameters of the encapsulant are critical for the validation of
these models as the simulation results are sensitively affected
by the values of these parameters.

VI. CONCLUSION
Fluid dispensing is widely employed in the semi-conductor
industry today. In this process, the volume of fluid dispensed and
the shape of the fluid formed on the board are two critical perfor-
mance indexes. This paper addressed the modeling of these two
performance indexes for a time-pressure dispensing process.
Beginning with the relationship between the shear stress and
the shear rate of the fluid dispensed, a model for the flow rate
Fig. 5. Comparison of the measured flow rates and the model prediction.
of the fluid has been developed for the time-pressure dispensing
process. The final shape of the fluid on the board has been math-
ematically described and solved numerically, based on the dif-
ferential equations governing the fluid spreading process and
a set of equilibrium conditions that must be satisfied once an
equilibrium shape is formed. These two models have been val-
idated through the experiments performed. The results are very
promising.

REFERENCES
[1] M. Norris, “Dispensing technology,” Surface Mount Technol., pp.
56–58, 1996.
[2] X. B. Chen, G. Schoenau, and W. J. Zhang, “Issues on dispensing in
surface mount technology (SMT),” in Proc. Int. Conf. Adv. Manufact.
Technol., 1999.
[3] D. Dixon, J. Kazalski, F. Murch, and S. Marongelli, “Practical issues
concerning dispensing pump technologies,” Circuits Assembly, pp.
36–40, 1997.
[4] R. Bush, “Matching fluid dispensers to materials for electronics appli-
cations,” Electron. Packag. Prod., vol. 37, pp. 56–62, 1997.
Fig. 6. Comparison of the measured widths and heights to the model [5] A. A. West, D. J. Williams, and C. J. Hinde, “Experience of the applica-
prediction. tion of intelligent control paradigms to real manufacturing processes,”
in Proc. Inst. Mech. Eng., vol. 209, 1995, pp. 293–308.
dispenser was controlled to move along the axis as shown in [6] A. Razban, “Intelligent control of an automated adhesive dispensing
cell,” Ph.D. dissertation, Imperial College, London, UK, 1993.
Fig. 1 at different velocities during the dispensing process. The [7] A. H. P. Skelland, Non-Newtonian Flow and Heat Tranfer. New York:
area of the encapsulant conserved during the spreading process Wiley, 1967, pp. 1–180.
on the board was identified by measuring the net weight of the [8] W. L. Wilkinson, Non-Newtonian Fluids. New York: Pergamon, 1960,
pp. 1–10.
encapsulant dispensed and the length of the line, and using the [9] CRC Handbook of Mechanical, CRC, Orlando, FL, 1998, pp. 114–127.
density of the encapsulant (1780 Kg/m ). The results giving the [10] S. D. Holdsworth, “Rheological models used for the prediction of the
measured width and height of lines versus the areas conserved flow properties of food products: A leterature review,” Trans. Inst.
Chem. Eng., vol. 71-C3, pp. 139–179, 1993.
during the spreading process are shown in Fig. 6. [11] D. C. Bogue, “Entrance effects and prediction of turbulence in non-New-
Simulation studies have been performed using the models de- tonian flow,” Ind. Eng. Chem., vol. 51, p. 874, 1959.
veloped in this paper. For calculating the weight of the encapsu- [12] R. N. Weltmann and T. A. Keller, “N.A.C.A.,” Tech Mem., 1957.
[13] A. S. Foust et al., Principles of Unit Operations. New York: Wiley,
lant dispensed on the board, the relationship between the shear 1960, p. 542.
stress and shear rate described in the Power Law equation is [14] P. Ehrhanrd and S. H. Davis, “Non-isothermal spreading of liquid drops
adopted and substituted into the flow rate model given in (6). on horizontal plates,” J. Fluid Mech., vol. 229, pp. 365–388, 1991.
[15] H. P. Greenspan, “On the motion of a small viscous droplet that wets a
The weight is then determined by multiplying the calculated surface,” J. Fluid Mech., vol. 84, pp. 125–143, 1978.
flow rate with the dispensing time and the density of the encap- [16] P. J. Haley and M. J. Miksis, “The effect of the contact line on droplet
sulant. For calculating the width and height of the line formed on spreading,” J. Fluid Mech., vol. 223, pp. 57–81, 1991.
[17] B. D. Summ, Physical Quantities. Orlando, FL: CRC, 1997, pp.
the board, (11l), (13l), (14l), and (17) are combined and solved 409–419.
numerically. All of these calculations are performed in Matlab, [18] S. Han and K. K. Wang, “Analysis of the flow of encapsulant during un-
and the identified rheological parameters of the encapsulant are derfill encapsulation of flip-chips,” IEEE Trans. Comp., Packag., Man-
ufact. Technol. B, vol. 20, pp. 424–433, Nov. 1997.
used, which include the cure-fitting parameters ( , , , and [19] S. Newman, “Kinetics of wetting of surfaces by polymers; Capillary
) given previously, and the surface tension (0.122 N/m) and flow,” J. Colloid Interface Sci., vol. 26, pp. 209–213, 1968.

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: MODELING OF TIME-PRESSURE FLUID DISPENSING PROCESSES 305

X. B. Chen received the M.S. degree in mechanics and the Ph.D. degree in W. J. Zhang received the Ph.D. degree in the design theory and computer
manufacturing engineering from the Harbin Institute of Technology, China, in aided design of mechanism systems from the Technical University of Delft,
1991 and 1996, respectively, and is currently pursuing a Ph.D. at the University The Netherlands, in 1994.
of Saskatchewan, Saskatoon, SK, Canada. He was then appointed to be an Assistant Professor with the City University
Prior to this, he was a Research Assistant with the City University of Hong of Hong Kong. In June 1998, he was promoted and appointed to be an Asso-
Kong, and a Lecturer with Harbin Institute of Technology, China. His current ciate Professor in the Department of Mechanical Engineering, The University
research interests are on the fluid dispensing process and the semiconductor chip of Saskatchewan, Saskatoon, SK, Canada, and a Chair sponsored by Atomic
encapsulation. Energy Canada, Limited (AECL), to direct the research work at the Advanced
Engineering Design Laboratory (AEDL) of the Department. Under his direc-
tion, AEDL has received annual research funds over $500 000 qCanadian from
NSERC, U.S. Intel Co., HK ASM, Ltd., and some other local industries. He has
G. Schoenau received the B.Sc. and M.Sc. degrees from the University published over 40 technical papers in refereed journals and numerous confer-
of Saskatchewan, Saskatoon, SK, Canada, and the Ph.D. degree from the ence papers. His current research interests include: integrated design and control
University of New Hampshire, Durham, in 1974. of complex engineered systems, information modeling for design and manufac-
He is a Professor of Mechanical Engineering at the University of turing, intelligent production systems, and kinematics and dynamics of multi-
Saskatchewan and was previously Head of that Department for seven years. body systems.
His research interests include fluid power, control systems and building, and Dr. Zhang is a member of ASME and Senior Member of SME.
agricultural thermal systems. He has close to 200 papers in various journal
and conference proceedings in these areas of interest and has supervised or
cosupervised numerous graduate students and other professional staff.
Dr. Shoenau is a Member of ASHRAE, APEGS, and SESCI.

Authorized licensed use limited to: University of Pittsburgh. Downloaded on July 31,2020 at 19:42:54 UTC from IEEE Xplore. Restrictions apply.

You might also like