S09_P2_Daniel_Duffy

advertisement
ASSEMBLIES CONTAINING COPPER PILLAR STRUCTURES PROCESSED
USING ONE STEP CHIP ATTACH MATERIALS (OSCA) AND
CONVENTIONAL MASS REFLOW PROCESSING
Daniel Duffy Ph.D., Hemal Bhavsar, Lin Xin, Jean Liu, Bruno Tolla Ph.D.
Kester Inc.
Itasca, IL, USA
dduffy@kester.com, btolla@kester.com
ABSTRACT
One step chip attach materials (OSCA) are
dispensable polymeric materials for flip chip assembly,
which are designed to flux metallic interconnections and
subsequently turn into an underfill upon curing. OSCA
materials enable a drastic simplification of the assembly
process by combining the reflow, flux residue cleaning and
capillary underfilling steps used in traditional die attach
processing into a single step. The key challenge when
designing OSCA materials for conventional mass reflow
processing (OSCA-R) is the timing the sequence of curing,
fluxing and soldering events during reflow processing.
OSCA-R materials must also have a process-friendly
rheological design that integrates seamlessly with standard
dispensing equipment and enables high filler loading levels.
This paper presents silicon die and substrate test vehicles
with copper pillar interconnect structures assembled using
OSCA-R materials for fluxing and underfilling that have
been processed using conventional mass reflow techniques.
Filler particle contributions to OSCA-R design and
performance will also be discussed including; rheology,
thermo-mechanical and thermal transport performance.
Assembly of devices containing SnAg, SAC, will be
discussed in conjunction with the reflow profile and process
windows for OSCA-R materials.
reflow processing using conventional mass reflow ovens.
The use of OSCA-T materials designed for processing by
thermos-compression bonding (TCB) has been discussed
elsewhere [4].
Reflow Processing
Dispensing
Fluxing
OSCA
Controlled Flow
& Wetting
Die
Placement
Solder Wetting &
Interconnection
OSCA Curing
Assembled Device
Filet Formation
Key words: Copper pillar, conventional reflow, flip chip,
no-flow, non-conductive paste, soldering, fluxing underfill
INTRODUCTION
The continuous need in semiconductor assembly
processes for higher throughput in conjunction with
decreasing interconnect size/pitch and increasing IO count
offer an opportunities for development and introduction of
assembly materials and processes to meet these demands
[1,2]. Copper pillar interconnection is becoming an
extremely geometry common for high IO device
interconnection where micro-bumps and C4 are not
preferred [3]. One step chip attach (OSCA) materials
combine the fluxing properties and the underfill properties
of the two respective materials in the traditional assembly
process into one material. OSCA materials are applied
before die placement, flux and enable joint formation during
reflow and then cure to form the underfill simplifying the
assembly process as shown in Figure 1. Materials referred to
here in this paper as OSCA-R materials are intended for
Figure 1. Simplified assembly process using OSCA-R
material that acts as both the flux and the underfill.
Additional assembly and design benefits can be
enabled using OSCA-R materials. One key benefit is the use
of OSCA-R materials to aid in increasing device density in a
design by either placing die closer together or stacking
vertically. Figure 2(a) illustrates the use of OSCA-R
materials for die stacking and Figure 2(b) illustrates close
placement of devices in a multi die package. The use of
OSCA-R materials in the multi die stack application, Figure
2(a) can overcome challenges of cleaning flux residues and
applying the capillary underfill process to these assemblies [
]. In the design suggested in Figure 2(b) there is no need to
keep the distance between die large enough for a dispense
needle from a PDP system or to large enough for the flight
of droplets from jet dispensers. The pre-deposition pattern
and volume of OSCA-R materials can be dialed in to
exactly fill the volume under the die and/or build the correct
filet on the perimeter of the devices.
Dip to Apply Flux
(a)
Place Die
Reflow Processing
Reflow Processing
Wash Flux Residues
Cure Underfill
Capillary Underfill
(b)
Figure 2. Use of OSCA-R materials for assembly of multi
die designs where (a) 3D integration and device stacking or
(b) density and close spacing are required.
The traditional assembly process shown in Figure 2
consists of; applying flux by dipping, die placement, reflow
processing to form interconnections, a wash step to remove
residues, capillary underfill step followed by cure of the
underfill. Several challenges are encountered for the
traditional assembly process as device IO count increases,
feature size becomes smaller and the spacing of features
drops below 100 microns. First is flux residue removal
during the washing step. Removal of the residues enables
the capillary underfill to flow, wet and cure properly leading
to improved reliability by maintaining adhesion and the
mechanical integrity of the joints. Residual flux residues
present after the washing step can contribute to chemical
corrosion and decrease device reliability [6]. The traditional
assembly process involves at least six steps and three
materials (flux, cleaner/water and underfill) and can become
cumbersome for 3D assemblies. The use of OSCA materials
for assembly as illustrated in Figure 1 eliminates the need
for residue cleaning and the time needed for capillary
underfill [7].
The key technical challenge for designing OSCA-R
materials is balancing the flow properties required for
dispensing and during die placement with the fluxing and
curing kinetics during reflow processing [8,9]. Several
articles discuss the successful design of OSCA-R materials
for assembly of traditional C4 device geometries [10-13].
This paper focuses on the assembly of devices using copper
pillar geometries that have been interconnected using
conventional reflow processing techniques such as
convection and conduction reflow tools.
Figure 3. Conventional flip-chip assembly process using
flux and capillary underfill materials.
EXPERIMENTAL
Test vehicles for investigating assembly using
OSCA-R materials were fabricated on 0.75mm thick silicon
wafers (die and substrate). The final die size is 6.35 mm x
6.35mm. The devices are composed of two daisy chained
arrays; the central array is 20 by 20 bumps (400), the
perimeter is 5 bumps wide, 70 bumps to a side (1300) and
there are 128 bumps located in the sparse electrically
unconnected area between the arrays as shown in Figures
4(a) and 5(a). The bump pitch is 80 microns in the dense
regions. The Cu pillars are 40 micron tall with 10 microns
of SnAg solder on top of the pillar for a total height of 50
microns as shown in Fig. 10(a). The substrate has a
matching set of copper pads 10 micron in height configured
in a daisy chain configuration using a Ni metallization layer
with a 2 micron oxide passivation layer. Additional copper
traces to route the daisy chain to test pads located on the
edge of the device as shown in Figures 4(b), 5(b) and 5(c).
Jet dispensing was conducted using a Speedline
Technologies Prodigy platform equipped with a NanoShot
pump system and a 150 micron nozzle. All dispensing trials
were conducted with a substrate temperature of 40C. Figure
6 shows OSCA-R that has been jet dispensed onto one of
the copper pillar test vehicles used in this study. Assembly
of the copper pillar test vehicles was conducted using a
Finetech Pico die bonder system with a conduction heating
stage shown in Figure 7. Die placements were conducted
with a substrate heating temperature of 40C. The alignment
process was done manually and with an accuracy of 5
microns in the presence of dispensed OSCA-R. The
placement speeds and forces used to prepare the assemblies
are presented in Figure 8. Three reflow profiles were
programmed for processing devices using the conduction
stage of the Finetech bonder and an Electrovert 7 zone
convection oven. Figure 9 shows the reflow temperature
profiles used for processing the die and substrate
components shown in Figures 4, 5 and 6.
(a)
Figure 6. Filled OSCA-R jet dispensed onto device in a
cross pattern.
(b)
Figure 7. Die placement using Finetech die bonder system.
Figure 4. Test vehicle configuration; (a) die and (b)
substrate.
Approach
Contact
Apply
Force
Hold
(a)
Retract
Placement
Tool
High Speed
High Speed
(b)
Change Over Distance
0.5 to 2 mm
Low Speed
0.5 to 3mm/sec
(c)
Figure 5. (a) Cu pillars, (b) Cu pads and (c) substrate
illustrating the daisy chain and test trace layout.
Force
1 to 15 Nt
( 10x10 mm die )
Dwell
0.05 to 0.2 Sec
Figure 8. Device placement profiles; approach speeds,
placement forces and dwell times.
300
1
Temperature (oC)
250
2
3
200
150
100
50
0
360
340
320
300
280
260
240
220
200
180
160
140
120
100
80
60
40
20
0
Time (sec)
Figure 9. Reflow profiles implemented on Finetech die
bonder (conduction) and Electrovert 7 zone reflow oven.
RESULTS & DISCUSSION
Flow Design and Voiding Performance
Design of OSCA-R flow properties are critical for
the assembly process to succeed. Table 1 presents a
summary of the flow and cure properties for OSCA-R
materials. The filler type, loading, rheology, cure chemistry
and kinetics can be tuned for a particular device
configuration and target assembly process. A discussion of
the rheology design considerations for highly filled OSCAR materials for jet dispensability, controlled flow after
dispense (no bleed out) and z-profile control to promote
placement void elimination has been published previously
[11]. In summary, the non-Newtonian flow behavior of
OSCA-R materials enables better pattern control when using
jet-dispense tools than materials with Newtonian flow
characteristics.
Table 1. Properties of OSCA-R materials
Property
Technique
Units
Filler Size, D50
micron
Filler Loading
SiO2
Filler Loading
Al2O3
wt%
0.5 to 10
Figure 10 illustrates the z-profile control concept.
The height and curvature of the dispensed pattern can
promote void elimination. On contrary, a flat profile as
shown in 10(b) tends to pin and trap voids during die
placement. The flow of OSCA-R materials around pillar
features and along the die and substrate interface while
during compression when the die is being placed is required
to push voids toward the edge of the die and eliminate them.
Dispensing OSCA-R in a cross pattern as shown in Figure 6
helps create wetting lines that do not impinge and trap voids
during die placement but we have found that having OSCAR materials which develop a taller z-profile after jet
dispensing can readily eliminate placement voids. The test
devices shown in Figure 6 tended to show large voids after
reflow when Newtonian materials were used. Figure 11
shows a comparison of CSAM images for devices after
reflow processing. The device shown in Figure 11(a) was
assembled using OSCA-R materials with non-Newtonian
flow properties ( STI > 1 and yield stress > 0 Pa ) and the
device shown in Figure 11(b) was assembled with a OSCAR material with Newtonian flow properties ( STI = 0 ). The
difference in voiding observed in the assembled devices
illustrates the OSCA-R flow and rheology design.
Some key results regarding the die placement
conditions shown in Figure 8. First is that a rapid approach
velocity helps with throughput and does not appear to
influence the voiding or interconnection process. Reduction
in the velocity to a range of 1 to 5 mm/sec as the placement
head approaches the OSCA-R material interface is
recommended, 3 mm/sec was found to produce acceptable
voiding performance. There is no advantage gained for
using slow speeds < 1 mm/sec in the contact portion of the
placement sequence. Contact forces less than 1 Nt were
found to result in partial interconnection. Placement forces
> 5 Nt are recommended for filled OSCA-R materials.
Dwell time did not appear to have an influence in our
studies as long as sufficient placement force was achieved.
(a)
Up to50
Up to 70
Tg
CTE-1
CTE-2
DSC
TMA
TMA
°C
ppm/K
ppm/K
128
30 to 60
105 to 208
Viscosity @ 1Hz
Viscosity @ 10Hz
STI
Yield Stress
Ea
Rheometer
Rheometer
Rheometer
Rheometer
Rheometer
Pa-s
Pa-s
Ratio
Pa
(K/1000)
2.5 to 60
2.4 to 25
1.0 to 2.4
0 to 8
4.20 to 7.93
(b)
Figure 10. Illustration of dispense profile relationship with
placement voids; (a) tall z-profile leads to flow and void
elimination, (b) flat dispense profile pins voids.
(a)
(b)
OSCA-R materials have been designed to have
minimal polymerization during the initial temperature ramp
and when the temperature exceeds the solder melting
temperature of 220C for SAC alloys. OSCA-R flows and
allows the solder to form an interconnection with the pads
during the approach to the peak temperature in the reflow
profile where temperatures can exceed 240C as illustrated in
Figure 12. The low viscosity of the OSCA-R materials
above the solder melting temperature allows filler particles
to flow out of the joint as it is forming preventing particle
entrapment and blocking. It has been found in this work that
certain bump and pillar designs can help to exclude filler
particle entrapment. Figure 13 shows several designs that
have been explored in this work including both organic
substrates and silicon substrates. The device configurations
shown in 13(a), (b), (g), (h) and (i) were found to favor
particle exclusion from the joints and produce
interconnections without trapped filler particles.
Solder Wetting &
Interconnection
250
OSCA Curing
Figure 11. CSAM images of copper pillar test device
assembled with OSCA-R materials (a) non-Newtonian
rheology design and (b) Newtonian flow properties.
Fluxing and Cure Kinetics
For OSCA-R materials to be used to produce
interconnected assemblies; fluxing, solder melting, joint
formation and polymerization must occur in the proper
order for the specified reflow profile. Figure 12 illustrates a
reflow profile and the sequencing of the steps. During the
initial temperature ramp OSCA-R materials flux and clean
the oxides and OSP from the interface of the copper pillars
(or bumps) and bonding pads. During the temperature ramp
OSCA-R materials are designed to reduce their viscosity
allowing flow and wetting of the joints and to promote fillet
formation. The Arrhenius flow activation energy (Ea), is
related to an OSCA-R material’s ability to thin and flow
with increasing temperature. Larger values of (Ea) reflect a
larger thinning effect with temperature; smaller values
reflect a smaller sensitivity to temperature. For device,
reflow profile and application there is a best value of (Ea).
For example if a large fillet is desired then larger values of
(Ea) will promote more flow during reflow. Alternatively, if
the application required the OSCA-R to stay close to the
edge of the component or flow only within a well-defined
region and not-flow extensively onto the substrate then
OSCA-R materials designed with smaller values of (Ea)
would be preferred.
Temperature ( oC )
200
150
100
Fluxing
Assembled Device
50
0
0
60
120
180
240
Time ( seconds )
Figure 12. Timing of OSCA-R fluxing, solder melting, joint
formation / interconnection and OSCA-R curing during
reflow processing.
Once the joint interconnection has occurred in the
reflow profile the polymerization kinetics of OSCA-R are
timed such that the material will gel and develop a
sufficiently high degree of cure during the later portion of
the reflow profile as shown in Figure 12. Timing the
fluxing, interconnection and curing is not trivial. To
facilitate OSCA-R design the curing kinetics during reflow
processing were studied experimentally using dynamic
scanning calorimetry (DSC) heat flow curves and analysis
using the method described by Borchardt and Daniels [15],
referred to here as the BD analysis. The BD analysis
produces empirical modeling parameters that can be used to
track and understand the influence of reflow profiles on the
cure of OSCA-R materials and conversely design OSCA-R
materials for specified reflow profiles. The BD model also
(a)
Silicon Substrate
300
1
0.9
250
0.8
Temperature
200
0.7
TAL
0.6
150
0.5
0.4
100
(f)
0.3
Conversion
Fractional Conversion
Organic Substrate
Table 2: BD kinetics parameters for silica filled and
unfilled OSCA-R materials.
BD Model
Definition
Units
Unfilled,
Filled,
Parameter
0%
40%
Reaction
N
-0.4
0.6
order
Activation
E
kJ/mol
44
52
Energy
Log ( Z )
Pre Factor
1/min
3.9
4.9
Reaction
J/g
360
235
H
Enthalpy
Temperature (oC)
provides a method to track the response of the curing
kinetics to the filler level, type, surface properties, size and
even dispersion state. The influence on OSCA-R curing
kinetics has been discussed in greater detail elsewhere
[11,12]. Table 2 presents the parameters obtained from the
BD analysis for unfilled and 40% filled OSCA-R
formulations as an illustration of how filler loading can
influence the curing kinetics. OSCA-R materials can be
iteratively adjusted such that they gel at the right time and
temperature during a given reflow profile. The addition of
the filler particles decreases the reaction enthalpy (H) as
the fillers dilute the reactive polymer and reduce the total
enthalpy of reaction. Table 2 also shows an increase in the
reaction order (N) as well as the activation energy (E).
Increasing the reaction order (N) tends to speed up the
reaction rate while increasing E tends to slow down the
reaction rate. The pre factor (Z) can be thought of as a
frequency factor, or probability for reaction events to occur.
Adding filler in this case increases (Z) and thus increases
the overall rate of reaction.
0.2
50
0.1
(b)
(g)
0
0
0
100
200
300
Time ( sec )
(c)
(h)
(d)
(i)
(e)
Figure 13. Overview of device geometries; Organic
Substrate (a) – (e); Silicon Substrate (f) – (i).
An example of how the cure kinetics parameters
can be used is shown in Figure 14. The point of gelation is
designed to occur as the assemblies are past the peak
temperature in the reflow profile and are entering the
cooling zones of the reflow oven. The length of time OSCAR spends in the cooling portion of the profile increases its
ultimate degree of conversion. Using the BD model
parameters application engineers can design the later portion
of the profile to drive toward 100% conversion or use the
residual degree of cure to anticipate the required post-bake
times required.
Figure 14. Modeling the degree of conversion (cure) of
OSCA-R as it is reflow processed using DSC measurements
and the BD kinetic model.
The kinetics analysis allows a wide range of reflow
profiles to be used to process OSCA-R materials. Figure 9
presents three profiles that correspond to different assembly
application needs. Profile 1 in Figure 9 is a fast profile with
a single ramp to the peak temperature. This has been found
useful for processing devices with relatively low thermal
mass where the oven temperature and device temperatures
are in phase and little substrate warpage is expected from
thermal gradients. Profile 2 in Figure 9 has a short “soak” as
the profile approaches the solder melting point. Such
profiles are applicable to systems where the thermal mass is
larger and temperature gradients across the substrates can
cause curvature, die float or tilting. Profile 3 contains a long
soak zone before passing above the solder melting
temperature. Such profiles are found to be relevent for
substrates with larger thermal mass, require time to outgas
or if longer times are required for flux activation.
Assembled Devices
Cross sectional images of the test vehicles shown
in Figures 4 and 5 that have been assembled using a 40%
filled OSCA-R material are shown in Figure 15. The results
illustrate the successful interconnection of the copper pillar
devices using profile 2 in Figure 9 via conduction mode
reflow processing. The utility of OSCA-R materials
demonstrated for assembly of C4 device geometries has
been extended to the assembly of copper pillar based
devices with high IO density. Based on these results the it is
proposed that OSCA-R materials can be used for assembly
of 3D and dense multi die structures suggested in Figure 2.
(a)
(b)
reflow processing of devices containing low melting point
alloys for 3D device assembly; (2) reliability of devices
assembled using OSCA-R materials and (3) assembly of
stacked
ACKNOWLEDGEMENT
Mahesh Desai, Kal Chokshi, Maulik Shah, Chris
Klimaszewski, Jim Lowe and David Eichstadt of Kester
Inc., Itasca IL. Chris Gregory and Alan Huffman of the
Research Triangle Institute, Research Triangle Park, NC.
Neil O’Brian, Finetech, Gilbert, AZ.
REFERENCES
[1] B. Schmaltz, "Packaging materials for 2.5D/3D
technology," IEEE International Symposium on
Advanced Packaging Materials, Irvine, CA, USA, Feb.
27-Mar. 1, 2013 (2013), 93-101;
[2] R. Huemoeller, "Market Demand Readiness for 2.5/3D
TSV Products" IMAPS 2012 Device Packaging
Conference, Scottsdale, AZ. March 5-2012.
Figure 15. Cross sections of a copper pillar assembly
interconnected using OSCA-R material.
Table 3 summarized the progress in reflow process
evaluation to date and interconnection results at the time of
publication. Interconnection has been investigated and
achieved using conduction-mode reflow processing for all
three profiles. Only the fastest profile (profile 1 in Figure 6)
has been investigated so far using a convection mode reflow
oven. The results so far indicate that the process shown in
Figure 1 utilizing OSCA-R materials that act as fluxes and
underfills can be successfully implemented for assembly of
copper pillar devices.
Table 3: Reflow profile result summary; (+) indicates
interconnection has been achieved; refer to Figure 6.
Profile
Conduction
Convection
Mode
Mode
1
+
+
2
+
TBD
3
+
TBD
CONCLUSIONS
Test vehicles with copper pillar geometries have
been successfully assembled using the one step chip attach
approach and filled OSCA-R materials and conventional
reflow processing techniques. The results demonstrate the
potential utility of OSCA-R materials for assembly of multi
die devices, wafer level assembly and stacked devices.
OSCA-R materials can be designed for a range of reflow
processing conditions, reflow tools, dispensing systems and
providing a window of assembly conditions for building
devices. Future research efforts will be focused on; (1)
understanding the applicability of OSCA-R materials to
[3] A. Eitan; K. Hung, “Thermo-compression Bonding for
Fine-pitch Copper-pillar Flip-chip Interconnect Tool
Features as Enablers of Unique Technology”,
Proceedings of the 2015 Electronic Components &
Technology Conference, Pg 460-464, San Diego, CA
[4] D. Duffy, C. Gregory, C. Breach, A. Huffman, "3D and
2.5D Packaging Assembly with Highly Silica filled One
Step Chip Attach Materials for both Thermal
Compression Bonding and Mass Reflow Processes",
Proc. 64 ECTC, Orlando, FL 2014, Pages: 1803 – 1809
[5] B. Tolla, X. Wei, Y. Shi, H. Bhasvar, J. Allen, K.
Loomis, “Reactivity of No-Clean Flux Residues in
Electronic Assemblies: A Systematic Study”, Corrosion
and Flux Residue Impacts of Lead Free Solders; SMTA
International 2015, Rosemont, IL
[6] Y. Orii, K. Toriyama, S. Kohara, H. Noma, K.
Okamoto, D. Toyoshima; K. Uenishi, Transactions of
the Japan Institute of Electronics Packaging (2011),
Micro structure observation and reliability behavior of
peripheral flip chip interconnections with solder-capped
Cu pillar bumps, 4(1), 73-86
[7] Q. Tong, “Underfill adhesive materials for flip chip
applications”, Edited by Alam, M. O.; Bailey, C,
Advanced Adhesives in Electronics (2011), 137-154.
[8] C.P.Wong, S. Shi, "Study of the Fluxing Effects on the
properties of No-Flow Underfill Materials For Flip
Chip Applications", Proc. 48th ECTC, Seattle, WA.
1998
[9] J. Liu, R. Kraszewski, X. Lin, L. Wong, SH Goh, J.
Allen. "New Developments in Single Pass Reflow
Encapsulant for Flip Chip Application", Proc. 51
ECTC, Orlando, FL 2002
[10] D. Duffy, H. Bhavsar, L. Xin, J. Liu, B. Tolla, “One
Step Chip Attach Materials (OSCA) for Conventional
Mass Reflow Processing”, IMAPS 2014 Proceedings
Oct. 13 – 16, 2014, San Diego, CA, Pages: 262 – 267
[11] D. Duffy, H. Bhavsar, L. Xin, B. Tolla “Design of
Filled One Step Chip Attach Materials (OSCA) for
Conventional Mass Reflow Processing: Curing Kinetics
and Solder Reflow Aspects, IMAPS Device Packaging
Conference, March 17th-19th 2015, Scottsdale AZ
[12] D. Duffy, M. Desai, H. Bhavsar, L. Xin J. Liu, and B.
Tolla, Rheology Design Considerations for One Step
Chip Attach Materials (OSCA) used for Conventional
Mass Reflow Processing, Proc. 65th ECTC, San Diego
CA, 2015
[13] D. Duffy, H. Bhavsar, L. Xin, J. Liu, B. Tolla , “Design
of Filled One Step Chip Attach Materials (OSCA) for
Conventional Mass Reflow Processing: Rheology
Considerations for Jet Dispensing and Die Placement”,
IMAPS 2015 Proceedings, Oct. 27 – 29, 2015, Orlando,
FL, Accepted
[14] D. Duffy, A. Xiao, “Rational design of highly filled
reactive resins for electronic material applications with
multiple performance constraints”, Materials Research
Society Symposium Proceedings, Volume968, Adv.
Electronic Packaging, Pages175-180, 2007
[15] ASTM-E2041 "Standard Test Method for Estimating
Kinetic Parameters by Differential Scanning
Calorimeter Using the Borchardt and Daniels Method.
Download