Proceedings of the ASME 2012 International Mechanical Engineering Congress & Exposition
IMECE2012
November 9-15, 2012, Houston, Texas, USA
IMECE2012-89672
THERMAL CONDUCTIVITY PERFORMANCE MODELING AND CHARACTERIZATION OF A
NOVEL ANISOTROPIC CONDUCTIVE ADHESIVE USED IN LEAD-FREE ELECTRONICS
PACKAGING
Matthew J. Combs
Rochester Institute of Technology
Rochester, NY 14623
Email: [email protected]
Dr. S. Manian Ramkumar
Center for Electronics Manufacturing
and Assembly
Rochester Institute of Technology
78 Lomb Memorial Drive
Rochester, NY 14623
Phone: 585-475-6081
Fax: 585-475-7167
Email: [email protected]
ABSTRACT
The continued desire to utilize an alternative to leadbased solder materials for electrical interconnections has led to
significant research interest in Anisotropic Conductive
Adhesives (ACAs). The use of ACAs in electrical connections
creates bonds using a combination of metal particles and
epoxies to replace solder. The novel ACA discussed in this
paper allows for bonds to be created through aligning columns
of conductive particles along the Z-axis. These columns are
formed by the application of a magnetic field, during the
curing process. The benefit of this novel ACA is that it does
not require precise printing of the adhesive on pads and also
enables the mass curing without creating shorts in the
circuitry.
This paper will present the findings of the thermal
conductivity performance tests using the novel ACA and its
applicability as a thermal interface material and for
assembling bottom termination components, power devices,
etc. The columns that act as electrical conduction paths also
contribute towards the thermal conductivity. The thermal
conductivity of the novel ACA was measured utilizing a
system that is similar to that in ASTM (American Society of
Testing Materials) D5470 standard. The goal was to examine
the influence of Bond Line Thickness (BLT), particle loading
densities, particle diameters and adhesive matrix curing
conditions on the electrical and thermal performance of the
novel ACA. This paper will also present a numerical model to
describe the thermal behavior of the novel ACA.
The novel ACA’s applicability for PCB-level
assembly has also been successfully demonstrated by RIT,
including base material characterization, effect of process
parameters, failures, and long-term reliability. Reliability
testing included the investigation of the assembly performance
Dr. Satish Kandlikar
Department of Mechanical
Engineering
Rochester Institute of Technology
Rochester, NY, 14623
Phone: 585-475-6728
Email: [email protected]
in temperature and humidity aging, thermal aging, air-to-air
thermal cycling, and drop testing.
INTRODUCTION
The use of conductive adhesives as a thermal
interface material has been in existence for many years in
electronics packaging. Generally speaking most conductive
adhesive solutions use some sort of conductive filler particles
that are then cured to create the electrical and thermal bonds.
The most common type of adhesive is the Isotropic
Conductive Adhesive (ICA) [1]. This type of adhesive has
become quite popular due to improvements in the reliability of
bonds and a refined understanding of its curing process [2].
The general concern with ICAs is its conductivity in all
directions, requiring it to be precisely printed, especially for
fine pitch applications. Anisotropic Conductive Adhesives
(ACA), on the other hand, offers the ability to conduct in one
direction, which is crucial for electrical and thermal
conductivity. Therefore, ACAs become more useful when
dealing with fine pitches [3].
In this paper, the performance of a novel ACA will
be examined as a possible thermal interface material. This
ACA is unique, as it uses no pressure during assembly, to
capture the monolayer of particles and create the bonds, as
with traditional ACAs. Instead, the fillers for this ACA are
silver coated ferromagnetic materials, which allow for the
magnetic alignment of the fillers into conductive z-axis
columns, during cure, and provide the anisotropy (Figure 1.).
A complete cure schedule and assembly methodology is
provided in a previously published paper [4]. The three ACA
formulations investigated as part of this experiment include
silver coated ferrite particles, which are 3µm, 8µm and a
combination of 3 and 8µm diameters, in an epoxy matrix.
1
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME
attached samples were assembled using the appropriate solder
paste and reflow profile.
Copper
Blank
Cured ACA
Holes for
Thermocouples
Figure 1. Magnetically Aligned Ferromagnetic Particles
The primary aspect of this novel ACA that will be
examined in this paper is its thermal resistance as compared to
tin-lead and lead free solder. The initial analysis will be based
upon experimentation conducted to measure the thermal
resistance of the novel ACA joint. The experimental analysis
will be utilized to develop a model to analytically calculate the
thermal resistance of the novel ACA. The thermal model will
be based upon the electrical model proposed in a prior
publication by one of the lead authors of this paper for the
novel ACA [4].
Figure 2. Assemblies for Experimentation
After the curing process, the samples were measured
for their thermal resistivity. The methodology used for
measuring the thermal resistivity was similar to that outlined
in ASTM D5470. The measurement setup consisted of a heat
source on one side of the sample and a heat sinking material
on the opposite side. The thermocouples in the holes of the
copper blocks measured the temperature difference between
the two copper blanks [5]. The blocks were heated to
approximately 50°C to determine the general thermal
resistivity at this temperature.
EXPERIMENT
A Design of Experiment (DOE) was conducted in
order to investigate the thermal resistance of the novel ACA.
Experimental factors that have proven to be significant during
previous testing have been the cured bond line thickness and
the particle size. These factors proved to have bearing on the
electrical resistance of the novel ACA [4]. Additionally, the
magnetic field strength, applied to the novel ACA during the
bonding process was also varied (1000 and 1600 Gauss) as a
part of this DOE.
In order to test these factors and its influence on the
thermal resistance (Units: K/W – Kelvin per Watt) of the
novel ACA assembly, uniform copper blanks were used as test
vehicles. These blanks had holes drilled directly in the center
in order for thermocouples to be attached, for temperature
measurement. After being cut to size (approximately 9.6mm x
9.6mm x 4.7mm), one surface of each copper blank was
polished to a uniform finish. The test samples were created by
attaching the two polished surfaces of the copper blanks as
shown in Figure 2, using the novel ACA or solder.
The particle size that was used in the experiment
included 3µm, 8µm and a mixture of the two (denoted as 38 in
the graphs). The novel ACA was applied to approximately
83% of surface of the copper blanks by using screens cut to
the 8mm by 8mm dimensions and placed directly over the
blanks. The reduced print area is to accommodate for the
spread of the ACA when the blanks are bonded to each other.
The ACA was found to spread to the entire area of the copper
blanks. The thickness of the screens determines the bond line
thickness (1, 2, 4 and 5 mils). The mating blank was placed by
a pick and place machine to ensure uniformity of placement.
The blanks were then cured at a temperature of 160°C for 15
minutes in an oven that was capable of providing a constant
magnetic field, in the Z-axis, to align the particles. The
magnetic field within this oven could be varied to provide
different magnetic flux densities for the experiment. Solder
RESULTS
The results of the DOE analysis reflected similar
results seen in other studies with this novel ACA. The particle
size and the bond line thickness were identified as the
significant main effects in this study. These two main effects
were also measured as significant factors during the
measurement of electrical resistance [4]. The magnetic field
strength during the cure process was not a significant factor in
measuring thermal resistivity of the samples.
Main Effects Plot for Thermal Resistance
Data Means
Bond Line Thickness
Particle Size
1.2
Mean
1.1
1.0
0.9
0.8
1
2
4
5
3
8
38
Figure 3. Main Effects Plot for Bond Line Thickness and
Particle Size
Figure 3 reveals the distinct advantage of mixing the
3µm and the 8µm particles in providing the lowest thermal
resistivity. The 3µm formulation provides lower resistance
than the 8µm formulation primarily due to the number of
parallel columns formed, for a given printed area and bond
line thickness of the novel ACA. In addition, as one would
expect, lower bond line thickness provides the lowest thermal
resistance, because of lesser number of particles per column
2
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME
and hence considerably reduced interfacial resistance between
particles.
Particle Resistance
Particle Resistance
20
THERMAL MODEL
The proposed thermal model will take a similar
electrical model [4] and modify it for the thermal domain. This
model basically analyzes the significant factors that affect the
columns within the epoxy. Overall, this model begins at the
particle level resistance and builds up to the entire material
including multiple parallel chains for calculating the overall
thermal resistance of the assembly. In building this model,
first the columnar resistance is determined and then applied to
the entire system of columns within the epoxy matrix.
The model is built around determining the thermal
resistance of a single chain of aligned particles (RChain) as
shown in equation 1. In determining the RChain, constriction
(
, particle
and interfacial
resistances must be
determined. Additionally, the number of particles (M) that
form the column based on the particle size and bond line
thickness must be calculated.
10
y = 2.0912ln(x) - 0.3077
R² = 0.9985
5
0
0
20
3um
40
60
80
Particle Diamer/Contact Diamter
8um
Log. (3um)
100
Log. (8um)
Figure 4. Particle Resistance as a function of Particle
Diameter and Contact Diameter
( )(
)
(3)
Where:
(1)
The initial step is to model the resistance of the
individual particles. Particle resistance is determined by
treating the particles as individual spheres that come into
contact with other spheres providing a contact area for thermal
conductivity. This model ignores the possible agglomeration
of particles during the formation of conductive chains. The
particle resistance is described as:
√
y = 6.1791ln(x) - 2.1436
R² = 0.9777
15
Since it is difficult to determine the constriction
resistance experimentally, between particles, for the 3µm and
the 8 µm formulations, it must be determined mathematically,
based upon the diameter of the contact area. Figure 5 shows
the relationship between the constriction resistance and the
ratio of particle diameter to the contact diameter. The core
material resistivity was considered to be that of iron as the
particles were silver coated iron particles in both formulations.
As the contact diameter decreases, the resistance increases
linearly. As with the particle resistance, the only way to
increase the contact diameter and thereby reduce constriction
resistance is by ensuring adequate compression of particles
against each other. This will not be possible with the novel
ACA, as no pressure is applied during component assembly.
The column formation (automatic stacking of particles one on
top of each other) controls the contact diameter. Figure 5
purely provides the relationship that can be used to
approximate the constriction resistance for a given particle
diameter, assuming a contact diameter.
(2)
Where:
Figure 4 shows the influence of the ratio of particle
diameter to the contact diameter on the particle thermal
resistance, for both the 3µm and 8µm formulations. As the
contact diameter decreases, the resistance increases. The only
way to increase the contact diameter is by ensuring adequate
compression of particles against each other. This will not be
possible with the novel ACA, as no pressure is applied during
component assembly. The column formation (automatic
stacking of particles one on top of each other) controls the
contact diameter. Figure 4 purely provides an empirical
relationship that can be used to approximate the particle
resistance for a given particle diameter, assuming a contact
diameter.
The next step in putting together the thermal model is
to determine the resistance that occurs between the particles.
This specific resistance is referred to as the constriction
resistance. It manifests as a resistance due to the deformities
that occur as the particles are pressed together. The
constriction resistance is mainly determined as a relationship
between the size of the particle and contact area of the
particles as shown in equation 3.
Constriction Resistance
Constriction Resistance
50
y = 1.3263x - 1.3263
R² = 1
40
y = 0.4974x - 0.4974
R² = 1
30
20
10
0
0
20
40
60
80
Particle Diameter/ Contact Diameter
3um
8um
Linear (3um)
100
Linear (8um)
Figure 5. Constriction Resistance as a function of Particle
Diameter and Contact Diameter
3
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME
The final aspect of modeling the column’s thermal
resistance is determining the interfacial resistance. This
resistance is contributed by the insulating film that may be
present between particles. The interfacial resistance is
determined only by empirical means using the resistance
offered by each particle chain, formed during the assembly of
the copper blanks, for varying particle diameters and bond line
thicknesses (Equation 4).
BLT(mils)
1
2
4
5
M(Theoretical)
8µm
3
6
13
16
Table 2. Theoretical Number of Particles per Column
Using the mean values for the surface area for the
copper block assemblies the theoretical total number of
columns was determined to be as shown in Table 3.
(4)
The number of particles (M) that would theoretically be
present in each column is determined by:
3µm
8µm
𝐶
(5)
𝐶
2
= 92
Where:
12945
= 92
2
= 1,190,940 𝐶
(6)
Where:
Thermal Resistance of the Particle Chain
The total thermal resistance of the assemblies was
determined from the experiment. The thermal resistance
provided by the individual particle chains, for the two material
formulations, is derived using the approximations described in
the previous section and the total thermal resistance of the
assemblies. The number of columns per unit area was first
determined as shown in Table 1.
2
8µm
BL
T
Mean
(K/W)
RChain (K/W)
Mean (K/W)
RChain
(K/W)
1
0.9880
1176685
1.0945
183266
2
0.9152
1089981
0.7832
131136
4
0.9224
1098506
0.9434
157957
5
0.8719
1038382
1.0940
183184
3µm
=
/
2
Table 4. Measured Assembly Mean Thermal Resistance
and Calculated Resistance per Chain
8µm
6(0.061)
𝑈
𝜋(.008)2
= 1820
/
1820
= 167,440 𝐶
3µm
Once this value is determined, RChain can be calculated,
(7)
Where:
3µm
2
Table 3. Total Number of Columns per Assembly
Using the experimentally obtained mean values for
the thermal resistance of the assemblies, and the theoretical
values for the constriction and particle resistances, and the
number of particles in the chain, the RChain was determined for
each bond line thickness, as provided in Table 4. In
calculating the appropriate values for both RC and Rp, a small
contact area was selected, to have a ratio of less than 10 for
the particle diameter to contact diameter. The calculation for
interfacial resistances is shown in Table 5 and a plot of Ri
versus the bond line thickness is shown in Figure 6. This
provides a model to determine Ri.
In order to calculate RChain from the experimentally
measured thermal resistance, the theoretical approximate
number of columns formed within the epoxy must be
calculated. Initially, the number of columns per unit area must
be determined and then this value applied to the surface area
that contacts the epoxy [6].
6(0.061)
𝑈 =
𝜋(.003)2
= 12945
3µm
8
17
34
42
2
Table 1. Column Density per Unit Area
The theoretical number of particles in each chain (M)
is shown in Table 2, for each particle diameters and bond line
thickness.
8µm
BLT
RC
RP
RI (K/W)
RC
RP
RI (K/W)
1
2.65
5.105
130733.1
3.4815
4.105
45809.32
2
2.65
5.105
60544.57
3.4815
4.105
18726.09
4
2.65
5.105
31375.82
3.4815
4.105
11274.73
5
2.65
5.105
24138.33
3.4815
4.105
10767.56
Table 5. Calculated Interfacial Resistance between
Particles
4
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME
are eliminated. The authors are investigating the void volume
to identify possible correlation between the void volume and
thermal conductivity.
Interfacial Resistance
140000
y = 128569x-1.035
R² = 0.9987
120000
100000
80000
Actual R (K/W)
y = 41418x-0.906
R² = 0.9617
60000
Particle Size (microns) = 3
2.5
40000
Predicted Resistance Values
Actual R (K/W)
20000
0
0
3um
1
2
8um
3
4
Power (3um)
5
6
Power (8um)
0.92
0.922
0.87
1.5
1.35
1.37
0.99
1.0
Figure 6. Interfacial Resistance for 3µm and 8 µm
particles vs. Bond Line Thickness
Using this model, the theoretical values for the
thermal resistance were determined for different bond line
thicknesses and particle diameters. The predicted thermal
resistances are summarized in Table 6.
0.87
0.77
1.00
0.87
0.94
0.5
1
2
4
5
BLT (mils)
Figure 7. Actual Thermal Resistance for 3 Micron
Particles and BLT
8µm
BLT
RColumn
R
RColumn
R
1
1176665
0.988014
183264.1
1.094506
2
1089939
0.915192
131132.3
0.78316
4
1098423
0.922316
157948.9
0.943316
5
1038279
0.871815
183174
1.093968
Actual R (K/W)
Particle Size (microns) = 8
2.5
Predicted Resistance Values
1.09
0.78
0.94
1.09
2.0
Actual R (K/W)
3µm
2.0
0.988
Table 6. Predicted Thermal Resistances
Figures 4, 5 and 6 reveal the resistance components
(RC, RP and RI) for the 3µm and the 8µm particles as a
function of the particle diameter and contact diameter. These
resistance components are used in calculating individual
column resistances of the 3µm and the 8µm particle chains. As
shown in Table 6 the individual column resistance for the 3µm
particle chain is definitely higher than the 8µm particle chain,
as one would expect. When calculating the overall resistance
within a given area, the 3µm particles provide less resistance
when compared to the 8µm particles, due to the total number
of columns that are formed with the different particle sizes
(Table 1). More parallel columns provide less overall
resistance.
Figure 7 and 8 show the actual thermal resistances
measured for the 3µm and the 8µm particles and the range of
data measured. The figures show the median values for the
resistances and the symbol (black dot) indicates the mean. The
predicted resistance values for both the 3µm and the 8µm
particles were slightly lower than the actual measured values.
Additional experiments will help fine tune the model and
improve its prediction.
1.45
1.5
1.09
1.0
1.18
1.15
1.03
0.93
1.12
0.90
0.5
1
2
4
5
BLT (mils)
Figure 8. Actual Thermal Resistance for 8 Micron
Particles and BLT
Mean
Thermal
Resistance
Lead
Free
Solder
Tin-Pb
Solder
3µm Novel
ACA
8µm
Novel
ACA
0.1570
0.2037
0.9224
0.9434
Table 7. Mean Thermal Resistance Measurement
Comparisons
CONCLUSION
After observing the thermal resistivity of the novel
ACA, it appears to have comparable properties to other
thermal epoxies but not as effective as solder or metal attach
or nano-particle filled materials. Overall, a viable model was
provided to explain the thermal conductivity of the novel
ACA. Although, the current formulations of the novel ACA
may not be suitable for all thermal conductivity applications, it
can serve as one, in applications that do not require high
power dissipation while providing electrical conductivity.
COMPARISON TO SOLDER
As part of this experiment, measurements were also
made as a baseline for the lead-free solder and leaded solder,
assembled using the copper blanks. In comparing the average
measured thermal resistance of all materials (Table 7), it is
evident that the Novel ACA is significantly less thermally
conductive when compared to solder. It was also identified
that void formation was prevalent in solder attachments. The
solder attach performance could be improved further if voids
5
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME
References
1.
2.
3.
4.
5.
6.
Inoue, M., Muta, H., Maekawa, T., Yamanaka, S., Suganuma,
K. 2006. “Thermal conductivity of isotropic conductive
adhesives composed of an epoxy-based binder”. Proceedings of
2006 Conference on High Density Microsystem Design and
Packaging and Components Failure Analysis – 2006. IEEE.
Piscataway, NJ,USA. 6pp.
Yu Tao, Yanping Xia, Hui Wang, Fanghong Gong, Haiping Wu,
Guoliang Tao. 2009. “Novel Isotropical Conductive Adhesives
for Electronic Packaging Application”. IEEE Transactions on
Advanced Packaging. 32. (3). pp 582-92.
Chin, M., Iyer, K.A.,Hu, S.J. 2004. “Prediction of electrical
contact resistance for anisotropic conductive adhesive
assemblies”. IEEE Transactions on Components and Packaging
Technologies. 27. (2). pp 317-26.
Ramkumar, S. M., Srihari, K. 2008. “Modeling and
experimental verification of the contact resistance of a novel
anisotropic conductive adhesive”. ASME International
Mechanical Engineering Congress and Exposition, Proceedings.
ASME. New York, NY. pp 149-157.
Liu, C.H., Huang, H., Wu, Y., Fan, S.S. 2004. “Thermal
conductivity improvement of silicone elastomer with carbon
nanotube loading”. Applied Physics Letters. 84. (21) pp 424850.
Jin, S.,Tiefel, T.N., Wolfe, R. 1992. “Directionally-conductive,
optically-transparent composites by magnetic alignment”. IEEE
Transactions on Magnetics. IEEE. USA. pp.2211-13.
6
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/15/2014 Terms of Use: http://asme.org/terms
Copyright © 2012 by ASME