A chemorheological analysis of a no-flow underfill was conducted using curing kinetics through isothermal and dynamic differential scanning calorimetry, viscosity measurement, and solder (Sn/27In/54Bi, melting temperature of 86 °C) wetting observations. The analysis used an epoxy system with an anhydride curing agent and carboxyl fluxing capability to remove oxide on the surface of a metal filler. A curing kinetic of the no-flow underfill with a processing temperature of 130 °C was successfully completed using phenomenological models such as autocatalytic and nth-order models. Temperature- dependent kinetic parameters were identified within a temperature range of 125 °C to 135 °C. The phenomenon of solder wetting was visually observed using an optical microscope, and the conversion and viscosity at the moment of solder wetting were quantitatively investigated.
It is expected that the curing kinetics and rheological property of a no-flow underfill can be adopted in arbitrary processing applications.
Keywords: Flexible packaging, Underfill, Chemorheological analysis, Curing kinetics.
Manuscript received Jan. 27, 2016; revised Aug. 1, 2016; accepted Aug. 18, 2016.
Yong-Sung Eom ([email protected]), Ji-Hye Son ([email protected]), Hyun-Cheol Bae ([email protected]), Kwang-Seong Choi ([email protected]) and Jin-Ho Lee (leejinho
@etri.re.kr) are with the ICT Material & Components Research Laboratory, ETRI, Daejeon, Rep. of Korea.
I. Introduction
In electronic packaging, the flip-chip bonding process is a key technology for electrical and mechanical interconnections including high input/output (I/O) counts [1]. To enhance the reliability of the packaging after the flip-chip bonding process, an underfill process is conducted to fill the gap between the device and the substrate using capillary force. However, this conventional underfill process has various problems, including a very slow flow rate in a resin/filler system and incomplete filling between gaps. In addition, these problems are becoming serious owing to increases in device sizes and I/O counts, as well as decreases in the gaps between the device and substrate.
To manage the problems of a conventional underfill, a no-flow underfill process was introduced [2].
A no-flow underfill based on an epoxy system has been developed [3]–[10]. In the no-flow underfill process, the underfill resin is dispensed onto the substrate before device displacement. With an increase in processing temperature, the solder bump covered with the no-flow underfill is melted, and the oxide on the surface of the solder is simultaneously removed by the fluxing capability of the no-flow underfill.
With an increase in the chemical reaction, the viscosity of the no-flow underfill based on thermosetting resin is suddenly increased during the gelation phenomenon [11].
Winter reported that, using the dynamic mechanical measurement at a given frequency, the gelation point of a cross-linking polymer was detected when the storage modulus (G) and the loss modulus (G) crossed each other [12].
Gelation is defined as a phenomenon in which a dramatic
Curing Kinetics and Chemorheological Behavior of No-flow Underfill for Sn/In/Bi Solder in
Flexible Packaging Applications
Yong-Sung Eom, Ji-Hye Son, Hyun-Cheol Bae, Kwang-Seong Choi, and Jin-Ho Lee
increase in viscosity occurs owing to initial network formation.
Therefore, the coalescence and wetting of solder can be prevented by the gelation of a no-flow underfill. For a perfect joint between the solder bump and contact pad, the solder wetting should be completed before the gelation of the no-flow underfill occurs during curing. If the viscosity of the no-flow underfill is too high because of the gelation when the solder is melted, the solder joint may be hindered from wetting.
However, if the solder joint is completely formed, the liquid state of the no-flow underfill should be changed to the solid state as soon as possible.
During the interconnection process with the no-flow underfill, the resin is fully or partially cured. If the no-flow underfill is not fully cured, post-curing is required to obtain good adhesion between the device and the substrate. Therefore, an understanding of the chemorheological behavior of a no- flow underfill is very important for predicting conversion and viscosity during the curing process.
In the present research, a no-flow underfill, which has a fluxing capability for Sn/27In/54Bi solder with a melting temperature of 86 °C, was developed for a processing temperature of 130 °C in a flexible packaging application. To optimize the temperature cycle of the no-flow underfill during processing, an exact understanding of its chemorheological behavior is required to predict the conversion trajectory and viscosity as functions of temperature and time. Figure 1 shows a schematic of the conversion and viscosity according to an arbitrary temperature cycle. Curing kinetics is a method for predicting the conversion of a curing resin system through a numerical prediction using some phenomenological model equations [13]–[16]. In addition, the relationship between the conversion and rheological behavior is identified to optimize the best temperature cycle of the no-flow underfill.
The curing kinetics of thermosetting resin have been widely studied by many researchers to predict a conversion as a function of temperature T and time t, as shown in (1):
( , ).
f T t
(1) There are typically two phenomenological models: nth-order and autocatalytic reaction [17]. These are based on an isothermal analysis, as shown in Fig. 2. For kinetic modeling, the rate of conversion d/dt is expressed as a function of conversion and temperature T. For the nth-order reaction of the kinetic model, shown in Fig. 2 (a), the rate of conversion is described as
1 0
0
d (1 ) with = exp ,
d
n E
K K K
t RT
(2) where K is the rate constant as a function of temperature, n1 is the reaction order, K0 is the pre-exponential factor, E0 is the
Fig. 1. Schematic of conversion and viscosity according to an arbitrary temperature cycle.
Time (t)
Temperature
Arbitrary cycle
Conversion,Viscosity
Conversion () Viscosity () Temperature (T)
Fig. 2. Kinetic models: (a) nth-order reaction and (b) autocatalytic reaction.
Conversion, d
dt
Conversion, d
dt
(a)
(b)
activation energy, T is the isothermal curing temperature, and R is the ideal gas constant. As shown in Figs. 2(a) and (2), it is clear that the maximum reaction rate is defined at = 0. On the other hand, the generalized autocatalytic reaction model proposed by Kamal [18], shown in Fig. 2(b), is described through such relations as
1 2
d ( )(1 ) ,
d
m n
K K
t
(3)
01 02
1 01exp E , 2 02exp E ,
K K K K
RT RT
(4) where K1 and K2 are the rate constants, m and n are the reaction orders, K01 and K02 are the temperature-independent pre- exponential factors, and E01 and E02 are the activation energies.
According to this generalized model, that is, (3) and (4), it is obvious that the initial rate of conversion d/dt is K1, which is not zero or the maximum value, as shown in Fig. 2.
In the present research, the no-flow underfill, an epoxy-based
thermosetting material, shows a single exothermic peak, and nth-order and generalized autocatalytic models are used. The parameters (reaction orders, activation energies, and pre- exponential factors) of the kinetic models are investigated through a phenomenological analysis based on isothermal differential scanning calorimetry (DSC).
II. Materials and Experimental 1. Materials
The no-flow underfill used in the present research is a four- component system composed of an epoxy, an anhydride as a curing agent, a reductant, and a latent catalyst. For the fluxing capability of removing metal oxide on the surface of solder balls (Sn/27In/54Bi, melting temperature of 86 °C), a reductant based on carboxyl acid was used, and a latent catalyst was applied to control the viscosity stability at room temperature.
The mixing procedure was as follows. The given amounts of epoxy, curing agent, and reductant were mixed at 100 °C for 10 min and cooled down to room temperature. A latent catalyst was then added to the mixture and stirred at room temperature for 10 min.
2. Experiment
To measure the curing reaction of the no-flow underfill, a DSC device (TA Instrument Model Q20) was applied with isothermal and dynamic temperature cycles. Approximately 10 mg of resin was placed in a hermetic aluminum pan, and DSC experiments were conducted in a nitrogen environment at the given temperature conditions. Heating rates of 2.5 °C/min, 5 °C/min, and 10 °C/min were used for the dynamic DSC, and constant temperatures of 125 °C, 130 °C, and 135 °C were used for the isothermal DSC. Viscoelastic properties were measured using a torsional parallel plate rheometer (HAAKE MARS III, Thermo Scientific, Inc.) at various curing conditions. The viscosity of the no-flow underfill placed between fixed and torsional parallel plates (with a diameter of 20 mm) was measured in dynamic and isothermal temperature conditions at a frequency of 1 Hz, as shown in Fig. 3.
To observe the solder wetting, a solder ball (Sn/27In/54Bi) with a diameter of 140 μm was placed on a Cu plate and covered by the no-flow underfill, shown in Fig. 4(a), prior to the curing process. The phenomenon of solder wetting after the curing process, shown in Fig. 4(b), was observed in a reflow chamber with an oxygen concentration of 1,050 ppm using an surface mounting technology (SMT) reflow device (SK-5000, Sanyo). The contact angle of the wetted solder (Sn/27In/54Bi) was measured using an optical microscope.
Fig. 3. Schematic of torsional parallel plates.
Fixed plate Torsional plate Imposed strain &
measurement stress
No-flow underfill Zero normal
force
Fig. 4. Schematic to observe solder wetting in no-flow underfill on Cu plate (a) before and (b) after curing processes.
Cu plate Solder ball (Sn/27In/54Bi)
Uncured no-flow underfill
(a)
Wetted solder (Sn/27In/54Bi)
Cured no-flow underfill
(b)
III. Results and Discussion 1. Curing Kinetics
To measure the total heat of the reaction for the no-flow underfill, dynamic DSC experiments were conducted at heating rates of 2.5 °C/min, 5 °C/min, and 10 °C/min, as shown in Fig. 5. Isothermal DSC experiments were also conducted at isothermal temperatures of 125 °C, 130 °C, and 135 °C, as shown in Fig. 6. The exothermic peaks from the dynamic and isothermal DSCs were observed with an increase in temperature and time, as shown in Figs. 5 and 6, respectively.
The baselines of the dynamic DSC with a 2.5 °C/min heating rate were defined as tangential linear, as shown in Fig. 5, whereas those of the isothermal DSC with a 125 °C isothermal temperature were determined according to the ending heat flow shown in Fig. 6. From the dynamic and isothermal DSC experiments, the total heat of the reactions HT was calculated through (5):
T 0
d d , d
tf Q
H t
t
(5) where tf is the time required to complete the reaction, and dQ/dt is the rate of heat generation.From the dynamic and isothermal DSCs, the total heat reactions shown in the shaded areas of Figs. 5 and 6 were calculated through (5), as shown in Fig. 7. The average total
Fig. 5. Dynamic DSC curves at heating rates of 2.5 °C/min, 5 °C/min, and 10 °C/min.
–0.4 –0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
60 80 100 120 140 160 180 200 220
Temperature, T (C) Baseline
2.5 C/min 5.0 C/min Heating rate
10 C/min
Heat flow, dQ/dt (W/g)
Fig. 6. Isothermal DSC curves at constant temperatures of 125 °C, 130 °C, and 135 °C.
–0.2 0.0 0.2 0.4 0.6 0.8
0 10 20 30 40 Isothermal curing temperature
Time, t (min) Baseline
125C 130 C 135 C
Heat flow, dQ/dt (W/g)
heat of the reaction from the dynamic and isothermal DSC measurements was 284 J/g and 288 J/g, respectively. It is usually difficult to analyze the curing kinetics of a thermoset material, especially beyond the gelation and vitrification phenomena [17]. Along with the resin curing processes and the resin crosslinks with a chemical-controlled reaction, the glass transition temperature also increases. When the glass transition temperature approaches the curing temperature, the resin changes from a rubbery state to a glassy state.
At this moment, the mobility of the reacting radicals is hindered, and the resin reaction is controlled by diffusion rather than a chemical reaction. Therefore, the total heat of the reaction from the dynamic DSC is usually higher than that from the isothermal DSC because there is a greater likelihood of a diffusion-controlled reaction in an isothermal DSC than in a dynamic DSC. In the present research, the total heat of the reaction of the dynamic DSC (284 J/g) is very similar to that of
Fig. 7. Total heat of reaction calculated from dynamic and isothermal DSC experiments.
282 283 287 291 283 291
0 50 100 150 200 250 300 350 400
2.5 5 10 125 130 135
Heating rate (C/min) Constant temperature (C/min) Dynamic DSC
Isothermal DSC
Total heat of reaction, ΔHT (J/g)
the isothermal DSC (288 J/g) at the given constant temperatures.
Therefore, it can be inferred that the glass transition temperature of a fully cured no-flow underfill is lower than the constant temperatures for the isothermal DSC. In addition, it is believed that the vitrification effect caused by a diffusion- controlled reaction will be very weak in the isothermal DSC results. It was determined that the total heat of the reaction for the no-flow underfill in the present study was 286 J/g.
T
( ), H t
H
(6)
T
d 1 d .
d d
Q
t H t
(7) Figures 8 and 9 show the calculated conversions of (6) from the dynamic and isothermal DSC experiments, respectively.
The conversion at a low heating rate increased more rapidly than at a high heating rate in the dynamic DSC, and that at a high constant curing temperature increased more dramatically than at a low curing temperature in the isothermal DSC. For the curing kinetics, the rate of conversion d/dt, as shown in Fig. 10, was derived from Fig. 9 by (7), where dQ/dt was provided by the isothermal DSC measurement. It was observed that the phenomenological shape of the curves in Fig. 10 is very similar to the generalized autocatalytic model introduced in Fig. 2(b).
At the beginning of the curing reaction, = 0, (3) is simplified as
1 0
d ,
d K
t
(8) where K1 is obtained directly from the reaction rate of the conversion shown in Fig. 10. The autocatalytic model in (3)
can be changed into a logarithmic form, as shown in (9):
Fig. 8. Calculated conversion versus temperature from dynamic DSC experiment.
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
60 80 100 120 140 160 180 200 220 Temperature, T (C)
Heating rate 2.5 C/min
10 C/min 5.0 C/min
Conversion,
0
Fig. 9. Calculated conversion versus time from isothermal DSC experiment.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 5 10 15 20 25 30
Isothermal curing temperature
125 C
Conversion,
130 C 135 C
Time, t (min)
Fig. 10. Reaction rate of conversion as a function of the conversion at given constant curing temperatures.
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030
0.0 0.2 0.4 0.6 0.8 1.0
Conversion, Isothermal curing temperature
125 C 130 C 135 C
Rate of conversion, d/dt (s–1)
1 2
In d In(1 ) In( ).
d
n K K m
t
(9)
From (9), n is considered a tangent of a linear equation, which is a function of ln(1 − ). When parameters K1 and n are determined through (8) and (9), K2 and m can also be calculated using (10):
1 2
In d /d In( ) In( ).
(1 )
t K m K
(10)
In (10), m and ln(K2) are also considered as the tangent and y- intercept of a linear equation, which is a function of ln().
From ln(K2), the rate constant K2 is obtained. The rate constants K1 and K2, mentioned in (3) with Arrhenius equations, are expressed in logarithmic form in (11) and (12):
01
1 01
In( ) E 1 In( ),
K K
R T
(11)
02
2 02
In( ) E 1 In( ).
K K
R T
(12) It was also recognized that (−E01/R) and (−E02/R) are tangents, whereas ln(K01) and ln(K02) are y-intercepts, of linear equations (11) and (12), which are functions of 1/T. From (−E01/R), (−E02/R), ln(K01), and ln(K02), the activation energies (E01 and E02) and pre-exponential factors (K01 and K02) are obtained, respectively.
As shown in Fig. 11, the linear relation between (1/T) and ln(K1) or ln(K2), which were obtained by (8) and (10) based on Fig. 10, were confirmed, and are defined by (11) and (12).
From the linear equations in (9) and (10) based on Fig. 10, the reaction orders m and n were obtained as shown in Fig. 12.
Table 1 summarizes kinetic parameters such as the activation
Fig. 11. Arrhenius plot of rate constants at different curing temperatures for no-flow underfill using autocatalytic model in (11) and (12).
–7.5 –7.0 –6.5 –6.0 –5.5 –5.0 –4.5
0.00244 0.00248 0.00252
1/Temperature, 1/T (K–1)
In(K1), In(K2) In(K2)
In(K1)
Fig. 12. Reaction orders n and m at different curing temperatures for no-flow underfill using autocatalytic model in (9) and (10).
0.4 0.6 0.8 1.0 1.2 1.4
120 125 130 135 140 Temperature (C)
m
n
m, n
Table 1. Kinetic parameters calculated from isothermal experiments based autocatalytic model.
Model Parameter
E01 (J/mol) 7.20 × 104 E02 (J/mol) 5.68 × 104 K01 (1/sec) 3.06 × 106 K02 (1/sec) 6.85 × 104
m –0.017 × T (C) + 2.86 Autocatalytic
(conversion: 0.0–0.6)
n 1.10
energies (E01, E02), the pre-exponential factors (K01, K02), and the reaction orders (m, n) in the conversion range of 0.0 to 0.6, calculated from isothermal DSC experiments.
All of the kinetic parameters were estimated using a regression technique in Microsoft Excel, and were taken as the average values over all isothermal curing temperatures.
Specifically, it was shown that m was a constant, while n in the present epoxy-anhydride system decreased as the curing temperature increased. For epoxy-amine systems, Mijovic and others [19] observed that m increased as temperature increased, whereas Ryan and coworkers [20] reported that m decreases as temperature increases.
Figures 13 and 14 compare the rate of conversions d/dt and
as a function of conversion and time, respectively, as obtained from the autocatalytic model in (3) using isothermally determined kinetic parameters and the measurement results from the isothermal DSC at different curing temperatures for the no-flow underfill. Figure 14 shows a comparison of conversions () as a function of temperature using the dynamic DSC results and the model calculations. Excellent agreements
Fig. 13. Comparison of rate of conversion d/dt calculated using autocatalytic model, and isothermal DSC at different curing temperatures for no-flow underfill.
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030
0.0 0.2 0.4 0.6 0.8 1.0
Conversion,
Rate of conversion, d/dt (s–1) 135 135
130 130
125 125
Isothermal temperature (C) Measured Calculated
Fig. 14. Comparison of conversion calculated by autocatalytic model and isothermal DSC at different curing temperatures for no-flow underfill.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 5 10 15 20 25 30
Conversion,
135 135
130 130
125 125
Isothermal temperature (C) Measured Calculated
Time, t (min)
were observed within the conversion range of 0.0 to 0.6 for isothermal and dynamic DSC scanning. It was determined that the autocatalytic kinetic model and its parameters could be used within a conversion range of 0.0 to 0.6 for isothermal and dynamic heating conditions.
On the other hand, it was observed that they were somewhat different between the calculated and measured values in the conversion range from 0.6 to 1.0, as shown in the isothermal DSC (125 °C and 135 °C in Figs. 13 and 14) and the dynamic DSC (2.5 °C/min, 5.0 °C/min, and 10 °C/min in Fig. 15). The calculated rates of conversion (Fig. 13) and the conversions (Fig. 14) were higher than the measured results within the conversion range of 0.6 to 1.0. It was inferred that these phenomena were caused by the diffusion-controlled reaction around the gelation or vitrification [11].
In Fig. 15, the calculated conversions are greater than the
0.0
Fig. 15. Comparison of conversion calculated by autocatalytic model and dynamic DSC at different heating rates for no-flow underfill.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
50 70 90 110 130 150 170 190 Temperature, T (C)
Conversion,
2.5 2.5 5.0 5.0
10 10
Heating rate (C/min) Measured Calculated
Fig. 16. Reaction order n1 at different curing temperatures for no- flow underfill using nth-order model in (13).
0.8 0.9 1.0 1.1 1.2 1.3
124 126 128 130 132 134 136 Temperature (C)
n1
measured results under heating rates of 2.5 °C/min and 5.0 °C/min, and were lower than the measured value under a heating rate of 10 °C/min, within the conversion range of 0.7 to 1.0. It is generally known that the rate of a chemical reaction is more significantly affected at a lower heating rate than at a higher heating rate [21], which corresponds to the present results in Fig. 15. Therefore, it was concluded that another kinetic model was required in the conversion range from 0.6 to 1.0.
For an epoxy anhydride system, Rabearison and coworkers [22] made a serious attempt to express the diffusion phenomenon through kinetic modeling using a diffusion parameter within a high conversion range. For a conversion range of greater than 0.6, as shown in Fig. 10, the rate of conversion continuously decreases from the maximum value at a conversion of 0.6, which allows the nth-order kinetics in (2) to be used within this conversion range. The nth-order model in
Fig. 17. Arrhenius plot of rate constant at different curing temperatures for no-flow underfill using nth-order model in (14).
–7.0 –6.5 –6.0 –5.5 –5.0 –4.5
0.00244 0.00246 0.00248 0.00250 0.00252 1/Temperature (K–1)
In(K)
Table 2. Kinetic parameters calculated from isothermal experiments based on nth-order model.
Model Parameter
E0 (J/mol) 1.00 × 105 K0 (1/sec) 3.71 × 1010 nth-order
(conversion: 0.6–1.0)
n1 0.014 × T (C) – 0.708
(2) can be changed into a logarithmic form, as shown in (13):
1
In d In(1 ) In( ).
d n K
t
(13)
From (13), n1 is considered the tangent of a linear equation, which is a function of ln(1 − ), as shown in Fig. 16. In addition, the rate constant K in (2) can be changed into a logarithmic form, as shown in (14):
0
0
In( ) E 1 In( ).
K K
R T
(14) It is also recognized that (−E0/R) is a tangent, whereas ln(K0) is a y-intercept, of the linear equation in (14), which is a function of 1/T, as shown in Fig. 17. From (−E0/R) and ln(K0), the activation energy and pre-exponential factor are obtained, respectively.
Table 2 summarizes the kinetic parameters of the nth-order model such as the activation energy (E0), pre-exponential factor (K0), and reaction order (n1) within the conversion range of 0.6 to 1.0, based on the isothermal DSC experimental results.
Figure 18 shows a comparison of the rate of conversion as predicted from the autocatalytic (conversion range of 0.0 to 0.6), nth-order (conversion range of 0.0 to 0.6) models and
Fig. 18. Comparison of rate of conversion d/dt predicted by autocatalytic (conversion range of 0.0 to 0.6), nth-order (conversion range of 0.6 to 1.0) models, and measured isothermal DSC results at different curing temperatures for no-flow underfill.
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030
0.0 0.2 0.4 0.6 0.8 1.0
Conversion,
Rate of conversion, d/dt (s–1) 135 135
130 130
125 125
Isothermal temperature (C) Measured Calculated
Fig. 19. Comparison of conversion predicted by autocatalytic (conversion range of 0.0 to 0.6), nth-order (conversion range of 0.6 to 1.0) models, and measured isothermal DSC at different curing temperatures for no-flow underfill.
0 0.2 0.4 0.6 0.8 1.0
0 5 10 15 20 25 30
Time, t (min)
Conversion,
135 135
130 130
125 125
Isothermal temperature (C) Measured Calculated
measured isothermal DSC results at different curing temperatures, as a function of the conversion. The agreement was excellent for the no-flow underfill within a curing temperature range of 125 °C to 135 °C. Based on the kinetic parameters in Tables 1 and 2, the predicted conversions showed good agreement with the measured values as a function of time at the given isothermal curing temperatures, as shown in Fig. 19.
2. Chemorheological Behavior and Solder Wetting
The viscosity of the no-flow underfill was measured using a torsional parallel plate with a frequency of 1 Hz, as shown in
Fig. 20. Conversion predicted through kinetic models, and measured viscosity from torsional parallel plates shown in Fig. 3, under the given temperature cycle.
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 10 20 30 40 50 60
Time (min) Predicted conversion
Measured viscosity Temp.: 130 °C Heating rate:
6 °C/min
= 0.26
Viscosity =3.7 Pa.s
Predicted conversion, Measured viscosity (Pa.s)
1.E+05 1.E+06
Fig. 21. Measured storage and loss moduli of the no-flow underfill by torsional parallel plates under the given temperature cycle.
30 50 70 90 150
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
10 20 30 40 50 60
Time (min) Frequency: 1 Hz
Storage modulus (G')
Loss modulus (G") Temperature cycle
Gelation point : G' = G"
Temperature (C)
0
130
110
Shear modulus, G',G'' (Pa)
Fig. 3, under the given temperature cycle with a 6 °C/min heating rate between 25 °C and 130 °C. It was held for 43 min, as shown in Fig. 20. When the temperature reached 130 °C, the measured viscosity and predicted conversion were 3.7 Pa.s and 0.26, respectively. Figure 21 shows the gelation point defined by the storage modulus (G) and the loss modulus (G) crossover at an isothermal cure temperature of 130 °C and a frequency of 1 Hz. The conversions predicted by the kinetic models used in the present research were 0.85 at the gelation points based on the temperature cycle shown in Fig. 21.
Figure 22 shows the measured viscosity as a function of the predicted conversion, which was deduced from Fig. 20. The viscosity is 3.7 Pa.s at a conversion of 0.26 with a curing temperature of 130 °C, and began to increase slightly at a conversion of approximately 0.8. The viscosity was rapidly increased at a conversion of approximately 0.9 because of the
Fig. 22. Measured viscosity by torsional parallel plates as a function of conversion predicted through kinetic models with given temperature cycle shown in Fig. 20.
1 10 100 1,000 10,000 100,000
0.0 0.2 0.4 0.6 0.8 1.0 Predicted conversion,
Frequency : 1 Hz
Temperature reached at 130 C
Measured viscosity (Pa.s)
Fig. 23. Predicted conversion by kinetic models with given temperature cycle.
0 0.2 0.4 0.6 0.8
Temperature (C)
0 40 80 120 160 200
0 1 2 3 4 5
Time (min) Heating rate : 60 C
Point A
(Time = 2.2 min) Point B (Time = 3.5 min)
Predicted conversion,
= 0.21
= 0.06
1.0
cross-linking of epoxy even though the curing temperature was kept at 130 °C. If the given curing temperature is lower than 130 °C, it is inferred that the rapid increase in viscosity is observed around a gelation point between 0.4 and 0.6.
However, the curing temperature used for the no-flow underfill was determined to satisfy the melting temperature of Sn/27In/54Bi solder in the present research. Therefore, it is believed that the wetting and liquidity of the melted solder are unhindered because the viscosity of the no-flow underfill is sufficiently low within the conversion range of 0.0 to 0.8.
To observe the solder wetting in the no-flow underfill, a solder ball with a diameter of 140 μm was placed as shown in Fig. 4. The solder ball within the no-flow underfill was heated through the temperature cycle shown in Fig. 23, and the conversion was predicted using the kinetic model. Images of the solder wetting during the temperature cycle, shown in Fig. 24, were captured from recorded video at time points A
Fig. 24.Captured images of solder with given temperature cycle shown in Fig. 21 at time points (a) A (2.2 min) and (b) B (3.5 min).
(a) (b)
Not wetted solder Wetted solder
200 μm 200 μm
Fig. 25.Photograph of wetted solder on Cu substrate with wetting angle of 42°.
Wetting angle : 42°
Cu substrate
50 μm
and B. In particular, the image in Fig. 24(b) was captured at the moment (time of 3.5 min) when the solder ball was suddenly wetted on the Cu substrate. As shown in Fig. 23, the predicted conversion and measured viscosity at a time of 3.5 min were 0.21 and approximately 3.7 Pa.s, respectively.
As mentioned in Fig. 21, it is clearly shown that the conversion of 0.21 at the moment of solder wetting is lower than the conversion of 0.85 at the point of gelation with an isothermal curing temperature of 130 °C. Therefore, it is believed that solder wetting can be easily achieved because the viscosity of the no-flow underfill is sufficiently low when the solder is melted. The wetting angle of solder on the Cu substrate was 42°, as shown in Fig. 25. Therefore, it was concluded that the conversion of the no-flow underfill according to the arbitrary temperature cycle can be easily predicted from the kinetic model, and the temperature profile can be effectively designed to optimize the solder wetting and no-flow underfill curing.
IV. Conclusion
With the increase in human-friendly electronic equipment such as the Internet of Things (IOT), the processing temperature of flexible interconnection technologies with polymer material is limited because of the stable temperature range of given substrates. In the present research, no-flow underfill having a fluxing capability for Sn/27In/54Bi and a
melting temperature of 86 °C was investigated for a processing temperature of 130 °C in flexible packaging applications.
Curing kinetics was successively applied to predict the conversion according to an arbitrary temperature cycle using autocatalytic and nth-order models. The measured viscosity was analyzed as a function of the predicted conversion under the given temperature cycle.
Based on the curing kinetics and viscosity measurement, the phenomenon of solder wetting on a Cu substrate was observed at a no-flow underfill conversion of 0.26 and viscosity of 3.7 Pa.s. Therefore, the processing cycle of no-flow underfill aiming for a low-temperature interconnection can be easily optimized by given processing conditions including temperature and time. This is because the history of the conversion and viscosity of no-flow underfill considering the mechanism of solder interconnection can be simulated by curing kinetics.
In conclusion, the chemorheological behavior of the no-flow underfill was successively identified by using curing kinetics and viscosity measurement during an arbitrary temperature cycle. It is believed that this analytic technology can be effectively used in low-temperature packaging applications in the near future.
Acknowledgements
This work was supported by the R&D program of ISTK (development of an image-based, real-time inspection and isolation system for hyperfine faults), the Nano-Convergence Foundation (www.nanotech2020.org) funded by the Ministry of Science, ICT and Future Planning (MSIP, Korea) and the Ministry of Trade, Industry and Energy (MOTIE, Korea) (commercialization of 100-Gbps optical receiver and transmitter modules based on nano Ag-coated Cu paste), the KEIT IT R&D program of the Daeduck Innoplis Foundation (commercialization of Ag-coated Cu paste material with low melting point), and the Electronics and Telecommunications Research Institute (ETRI). The authors would like to thank Iseul Jung and Aeson Oh for their support with sample preparation and measurements.
References
[1] S.C. Johnson, “Flip-chip Packaging Becomes Competitive,”
Semiconductor Int., May 2009.
[2] C.P. Wong, S.H. Shi, and G. Jefferson, “High Performance No- flow Underfills for Low-Cost Flip-Chip Applications: Material Characterization,” IEEE Trans. Compon., Packag., Manuf.
Technol.: A, vol. 21, no. 3 Sept. 1998, pp. 450–458.
[3] Y.-S. Eom et al., “Characterization of Fluxing and Hybrid
Underfills with Micro-Encapsulated Catalyst for Long Pot Life,”
ETRI J., vol. 36, no. 3, June 2014, pp. 343–351.
[4] Y.-S. Eom et al., “Electrical Interconnection with a Smart ACA Composed of Fluxing Polymer and Solder Powder,” ETRI J., vol.
32, no. 3, June 2010, pp. 414–420.
[5] Y.-S. Eom et al., “Characterization of Polymer Matrix and Low Melting Point Solder for Anisotropic Conductive Film,”
Microelectron. Eng., vol. 85, no. 2, Feb. 2008, pp. 327–331.
[6] Y.-S. Eom et al., “Electrical and Mechanical Characterization of an Anisotropic Conductive Adhesive with a Low Melting Point Solder,” Microelectron. Eng., vol. 85, no. 11, Nov. 2008, pp.
2202–2206.
[7] J.-W. Baek et al., “Chemo-Rheological Characteristic of a Self- Assembling Anisotropic Conductive Adhesive System Containing a Low-Melting Point Solder,” Microelectron. Eng., vol. 87, no. 10, Oct. 2010, pp. 1968–1972.
[8] K.-S. Jang et al., “Catalytic Behavior of Sn/Bi Metal Powder in Anhydride-based Epoxy Curing,” J. Nanosci. Nanotechnol., vol.
9, no. 12, Dec. 2009, pp. 7461–7466.
[9] Y.-S. Eom et al., “Characterization of a Hybrid Cu Paste as an Isotropic Conductive Adhesive,” ETRI J., vol. 33, no. 6, Dec.
2011, pp. 864–870.
[10] Y.-S. Eom et al., “Development of Conductive Adhesive with Ag- Coated Copper and Solder for Microelectronic Packaging Applications,” J. Electron. Mater., submitted, 2015.
[11] K.P. Pang and J.K. Gillham, “Competition Between Cure and Thermal Degradation in a High Tg Epoxy System: Effect of Time and Temperature of Isothermal Cure on the Glass Transition Temperature,” J. Appl. Polymer Sci., vol. 39, no. 4, Feb. 1990, pp.
909–933.
[12] H.H. Winter, “Can the Gel Point of a Cross-Linking Polymer be Detected by G - G Crossover?,” Polymer Eng. Sci., vol. 27, no.
22, Dec. 1987, pp. 1698–1702.
[13] Y.-S. Eom et al., “Time-Cure-Temperature Superposition for the Prediction of Instantaneous Viscoelastic Properties During Cure,”
Polymer Eng. Sci., vol. 40, no. 6, June 2000, p. 1281.
[14] Y.-S. Eom et al., “Stress-Initiated Void Formation During Cure of a Three-Dimensionally Constrained Thermoset Resin,” Polymer Eng. Sci., vol. 41, no. 3, Mar. 2001, p. 492.
[15] Y.-S. Eom et al., “A Structure and Property based Process Window for Void Free Thermoset Composites,” Polymer Composites, vol. 22, no. 1, Feb. 2001, p. 22.
[16] Y.-S. Eom et al., “Internal Stress Control in Epoxy Resins and their Composites by Material and Process Tailoring,” Polymer Composites, vol. 23, no. 6, Dec. 2002, p. 1044.
[17] R.B. Prime, “Chapter 5–Thermosets,” in Thermal Characterization of Polymeric Materials, New York, USA:
Academic Press, 1981, p. 435.
[18] M.R. Kamal, “Thermoset Characterization for Moldability Analysis,” Polym. Eng. Sci., vol. 14, no. 3, 1974, pp. 231–239.
[19] J. Mijovic, J. Kim, and J. Slaby, “Cure Kinetics of Epoxy Formulations of the Type Used in Advanced Composites,” J.
Appl. Polymer Sci., vol. 29, no. 4, Apr. 1984, pp. 1449–1462.
[20] M.E. Ryan and A. Dutta, “Kinetics of Epoxy Cure: a Rapid Technique for Kinetic Parameter Estimation,” Polymer, vol. 20, no. 2, Feb. 1979, pp. 203–206.
[21] Z. Zhang and C.P. Wong, “Modeling of the Cure Kinetics of No- flow Underfill in Flip-Chip Applications,” IEEE Trans. Compon.
Packag. Technol., vol. 27, no. 2, June 2004, pp. 383–390.
[22] N. Rabearison, C. Jochum, and J.C. Grandidier, “A Cure Kinecic, Diffusion Controlled and Temperature Dependent Identification of the Araldite LY556 Epoxy,” J. Mater. Sci., vol. 46, no. 3, Feb.
2011, pp. 787–796.
Yong-Sung Eom received his BS in aeronautical engineering from Korea Aerospace University, Hwajeon, Rep. of Korea and his MS in aeronautical engineering from the Department of Aerospace Engineering at the Korea Advance Institute of Science and Technology, Seoul, Rep. of Korea, in 1988 and 1991, respectively. He worked at the Korea Institute of Aeronautical Technology, Korean Air Ltd., Seoul, Rep. of Korea, as a design and process engineer for the composite materials of the MD-11 Aircraft Spoiler from 1991 to 1995. In 1999, he received his PhD in material engineering from the Department of Material Science Engineering at École Polytechnique Fédérale de Lausanne, Switzerland. After returning to Korea, he worked at Hynix Semiconductor Ltd., Incheon, Rep. of Korea, as a packaging engineer for memory devices from 2000 to 2001. Since 2001, he has been with ETRI Daejeon, Rep. of Korea, where he has been working as a packaging engineer. His research activities include the development of interconnection materials based on the polymer for an electronic packaging and process design for 3D- IC and MEMS packaging.
Ji-Hye Son received her BS and MS degrees in chemical engineering from the Department of Advanced Organic Materials and Textile System Engineering at Chungnam National University, Daejeon, Rep. of Korea, in 2010 and 2012, respectively. Since 2012, she has been working for ETRI. She has researched the design and synthesis of organic polymers and the characterization of advanced polymers.
Hyun-Cheol Bae received his BS and MS in electrical engineering from Dongguk University, Seoul, Rep. of Korea, in 1999 and 2001, respectively. He joined the SiGe research team of ETRI, in 2001 and worked as a design and process engineer for MMIC and passive devices. He moved to the packaging research group of ETRI in 2007 and has been working as a packaging engineer.
In 2009, he received his PhD in electrical engineering from Chungnam National University. His research interests include the design and fabrication of integrated passive devices, 3D stacked chip packaging using TSV, and wafer-level packaging for MEMS devices.
Kwang-Seong Choi received his BS from Hanyang University, Seoul, Rep. of Korea, and MS and PhD from the KAIST. From 1995 to 2001, he developed chip scale packages and PoP packages and designed high-speed electronic packages for DDR, Rambus, and RF devices for Hynix Semiconductor. He has been developing high-speed packaging technologies for optical devices, such as modulators and receivers, since 2001 at ETRI Currently, his research areas include the development of the materials and processes for 3D ICs with through silicon via (TSV), next-generation displays, and silicon photonics.
Jin-Ho Lee received the BS degree in physics from Kyungpook National University, Daegu, in 1980 and the MS degree from Korea University, Seoul, in 1982. He received the PhD degree from Kyungpook National University, in 1998. He joined ETRI in 1982. He has been working on the power electronic devices, thin film transistor technology and flexible devices. At present, Dr. Lee is the managing director of the IT Components and Materials Industry Technology Research Department at ETRI.
