EFFICIENT MANUFACTURING OF FLIP CHIP PACKAGES
FOR SMART CARDS AND RFID PRODUCTS
JIM CLAYTON, POLYMER FLIP CHIP CORPORATION
MARK KEENE, POLYMER FLIP CHIP CORPORATION
ZAK FATHI, LAMBDA TECHNOLOGIES
The demand for RFID and smart card products is projected to provide one of the
fastest growing markets over the next few years. To be competitive in this arena,
technology companies and manufacturers of flex circuits and assemblies are
being forced to look at increased production rates and lower cost manufacturing
techniques. Polymer Flip Chip (PFC), of Billerica Mass, has successfully
addressed this requirement with the selection of a new process technique. By
implementing a Variable Frequency Microwave (VFM) curing process, PFC found
the potential to dramatically increase process throughput capability, permit use of
lower cost substrates, while simultaneously reducing inventory cost, utility cost,
fixture expense, and space on the clean room assembly floor. VFM was initially
chosen for curing because it selectively heats and enables the use of cost
effective, low deformation temperature materials. When using VFM, the Flip
Chip modules are selectively heated above the temperature of the substrate by
25 to 50°C. Furthermore, the VFM process significantly accelerates the cure of
both conductive and non-conductive adhesives, enabling faster throughput and
continuous flow cure processes. A technical review of process development,
comparing convection with the VFM cure results will be presented. Optical
microscopy, DSC, IR imaging and electrical functionality were evaluated during
the qualification process.
Conductive polymer inks continue to gain widespread interest as a viable high-
volume solution for low-cost, flip chip interconnection. The processing steps
required for both wafer bumping and chip assembly are simple, yet reliable.
However, one significant disadvantage has been the longer work-in-process
(WIP) times needed to fully cure both conductive epoxy bumps and non-
conductive underfill epoxies. Convection oven curing times for polymer
adhesives may range from as little as 30 minutes to longer than 8 hours,
depending on the cure temperature and exothermic reaction of the constituent
resins of the epoxy system.
Conductive epoxy bumps, in particular, appear to be “mass sensitive”, meaning
that it takes an inordinate amount of time to cure the smaller mass associated
with stencil printed bumps (e.g. 100 micron diameter bumps) than the much
larger mass used in DSC (Differential Scanning Calorimeter) sampling when
deriving the recommended curing schedules. This difference is believed to be
attributable to the weaker exothermic reaction contributed by the low percentage
of organic resin mass present. Published curing schedules, therefore, need to be
viewed as generalizations only, when attempting to establish a conventional oven
cure schedule for bump interconnections made with conductive polymers.
Recent development work, based on VFM (Variable Frequency Microwave) oven
curing, has demonstrated that epoxy bump cure rates may be reduced to as little
as 3 minutes (and even faster) for the same epoxy systems which previously
required eight or more hours to cure. VFM curing through microwave energy
provides, in general, faster processing and greater control over the curing
process than that achieved using conventional heating methods, such as
convection or IR ovens. In particular, the VFM technique enables very selective
heating of materials while avoiding the typical problems associated with
electrical-arcing of metallic components or circuitry. The MicroCure™ VFM
product line, developed by Lambda, provides a remarkable new tool for rapid
curing of polymer conductive adhesives inks and liquid encapsulants within the
electronic packaging industry.
Two primary types of Smart Card and RFID inlets or flex circuits are presently
being fabricated at Polymer Flip Chip Corporation and are discussed in this
paper: (a) screen-printed conductive-ink antennas on ABS-plastic and (b) etched-
copper antennas on PET. ABS-plastic is the least expensive material for these
types of products and is primarily used for ISO-type Smart Card inlets assembled
with 8 to16-bit read-write chips. Copper-etched PET flex circuits are used
primarily for RFID applications. These circuits provide better line density,
required when routing traces beneath the smaller 8-bit chips, and lower
resistance needed for higher-Q antennas.
Since both of these flex materials deform easily when exposed to temperatures
exceeding 90 deg-C, the epoxy materials chosen for assembly must cure below
this temperature to be compatible. Unfortunately, low-temperature curing
epoxies typically require several hours to fully cure and have a shorter, 1 to 2-
hour, processing window for screen printing before the hardening agents begin to
increase the ink viscosity. A typical batch process flow using conventional oven
curing is illustrated in Figure 1. As a result of longer cycle-times required for low
temperature curing, the total process time often exceeds 9 hours, excluding time
required for loading and unloading panels from the vertical carts. In contrast, a
total in-line process flow can be substituted using VFM curing, as illustrated in
Antenna Printing &
Batch Curing Line
(not required for
antennas) Screen Print Load Panels Cure & Cool Un-load Panels
onto Vertical Panels in from Vertical
Carts Walk-in Oven Carts
Chip Placement &
Batch Curing Line
Load Panels onto Cure & Cool Un-load Panels
Screen Print Pick & Place
Vertical Carts Panels in from Vertical
Antenna Bumps Chips
Walk-in Oven Carts
Chip Underfill and
Batch Curing Line
Underfill Load Panels Cure & Cool Un-load Panels
Chips onto Vertical Panels in from Vertical
Carts Walk-in Oven Carts
Not required for etched- FIGURE 2
copper antenna process line
Screen Print VFM-cure Antenna Screen Print Pick & Place VFM-cure Antenna Underfill VFM-cure Chip
Antenna Traces Traces Antenna Bumps Chips Bumps Chips Underfill
HOW VFM-CURING WORKS
VFM-curing uses a variable swept frequency to apply microwave energy into a
closed chamber for curing. Polymer resins, containing polarized molecules, will
couple and absorb with the microwave energy, causing rapid polymerization of
the resins as the energy is dissipated in the form of heat. Different materials and
shapes respond in different ways to microwave irradiation, either to absorb or
reflect the energy, and these differences can be engineered into an optimum
combination of materials and pallet designs. Silicon and carbon, for instance, act
as good absorbers for microwave energy and heat quickly. FR-4, ABS and PET
materials, however, are fairly transparent to the same energy and will heat more
slowly. Selecting FR-4 material for pallet construction will insure that the inlet
substrates remain 50-60 degrees cooler than silicon RFID or Smart Card chips.
Energy for curing either the epoxy bumps or underfill comes from both molecular
excitation within the resins themselves and from thermal conduction supplied by
the overlying silicon chips.
Since the frequency is rapidly swept, arcing problems associated with metal or
semiconductor components is eliminated. Arcing is the result of excessive
charge build-up in metallic materials in the presence of standing wave patterns.
In the VFM technique, the electric fields are electronically stirred and the
microwave energy is not focused at any given location more than a fraction of a
second. The dynamics of charge build-up that leads to arcing are never
achieved. Hence, bare semiconductor devices are not adversely affected by the
microwave energy, resulting in defect-free processing.
In the course of this evaluation, it became apparent that different epoxy materials
exhibited dramatic differences in VFM-curing properties. Due to the close
association between Polymer Flip Chip Corporation and Epoxy Technology, a
new class of polymer adhesives, both electrically conductive and non-conductive
underfill epoxies, are being formulated specifically for rapid microwave curing.
The objective of this work was to evaluate the VFM process for curing silver-filled
conductive polymers (Epo-Tek No’s. E2101-PFC and EE162-5) and non-
conductive underfill epoxy (Epo-Tek No. U-300) on ABS and PET substrates.
Primary goals for this project were to significantly reduce the existing process
cycle time, ensure equivalent or enhanced throughput (line balance), equivalent
or improved material and product properties using minimum equipment floor
Three different applications were tested:
1. Curing of U-300 Underfill on PFC smart card and RFID assemblies.
2. Curing of EE162-5 printed antennas (5 turn coil) on ABS substrates.
3. Curing of E2101-PFC flip chip bumps on etched-copper PET substrates.
A summary of the results are as follows:
1. The curing of screen-printed antennas was performed using a special
microwave fixture. Five-turn ISO standard antennas were cured within 15
minutes as opposed to 120 minutes at 90-95°C. Recent formulation
adjustments in the conductive ink are expected to be further reduce the total
process cycle in the range of 7 minutes or less.
2. The curing of conductive epoxy bumps on PET with etched copper traces was
accomplished in as little as 3 minutes, a process currently carried out at 80°C
for 8.5 hours.
3. The two-part Epo-Tek underfill was cured as rapidly as 30 seconds, a process
currently carried out at 75°C for 30 minutes followed by 95°C for 30 minutes.
1. UNDERFILL CURING
The underfill dispense was performed by hand using a pneumatic EDF
dispenser. The ABS substrates were placed on a hot plate pre-set at 75°C.
Dispense pattern followed an “L” shape.
The variable frequency processing for the PFC underfill curing trials was
performed in a VFM cure station (Lambda Technologies, Inc. model MicroCure
5100). The frequency range utilized during heating was 6.425 ± 0.575 GHz with
a sweep rate of 0.1 seconds. A Raytek infrared pyrometer (refer to Figure 3) is
incorporated in the MicroCure 5100 control system and was used to control the
surface temperature of one of the flip chip die. A Nortech fiber optic contact
probe is also incorporated in the MicroCure 5100 control system and was used to
monitor the temperature of the substrate during the flip chip underfill curing. The
power was pre-programmed to produce a multi-step profile summarized in Table
Table 1. VFM profiles used for processing of PFC parts
VFM Time (min) Time (min) Time (min) Time (min) Total Time
Profile at 60 °C at 70 °C at 80 °C at 90 °C (min)
VFM P1 - 1 - 4.5 5.7
VFM P2 - 1 - 4 5.2
VFM P3 - 1 - 2.5 3.7
VFM P4 - 1 - 2 3.2
VFM P5 - 1 - 1 2.2
VFM P6 1 - 1 - 2.2
Figure 3. Illustrates the IR non-contact monitoring system. The Raytek
pyrometer is mounted on a gimbal that facilitates the movement of the
device and enables the operator to point it to the die of interest.
Sample Prep and Process Conditions
The PFC smart card assemblies were printed onto 12” x 8.5” flexible polymeric
sheets (ABS). These sheets were then cut into individual circuits for the underfill
trials and were processed one at a time. For these trials, all of the die on the test
vehicle were 8-bit, 13.56 MHz chips, underfilled using pre-mixed Epo-Tek U300.
Figure 4 shows two of the test vehicles placed on the conveyor rails inside the
VFM oven. A red laser spot illuminates the chip on the right sample and is used
for accurate alignment of the Raytek infrared pyrometer. A total of 4 test vehicles
were used as set-up parts to determine both dispense and VFM process profiles.
Fourteen test vehicles were then processed using VFM profiles summarized in
Table 1. These profiles are multi-stepped to obtain a gel state of the underfill
before proceeding to fully cure the epoxy. This technique can be used with
epoxy materials that tend to excessively out-gas when rapidly heated. VFM
power levels ranged from just 60 watts up to a maximum of 80 watts for these
profiles. The temperatures for the step profiles were based on the temperatures
used conventionally. All the VFM processed samples were fully cured.
Figure 4. Digital image of representative PFC test vehicle. The ABS sheets
were cut to provide a large number of trial samples. The individual
antennas were placed on a microwave transparent (FR-4) substrate.
Post Cure Analysis
Following cure of the test vehicles, visual inspection was performed and all die
sites appeared cured. The clear underfill turns to an amber color upon curing in
both VFM and convection heating.
To evaluate the underfill cure with respect to voids generated during VFM-curing,
a Lambda Technologies test vehicle, with 3 glass die per board, was dispensed
and cured. The glass die were then inspected under a microscope to check for
the presence of voids. Figure 5 displays an image, which shows the Lambda
Technologies test vehicles adopted for visual inspection after cure. No voids are
exhibited after the accelerated cure of the VFM process.
Epo-Tek U300 underfill
dispensed and VFM-cured
under glass die for visual
Figure 5. Digital image of a Lambda Technologies test vehicle.
No voids were visible under the glass die after the rapid VFM cure.
DSC, Extent of Cure Analysis
To confirm cure extent with the profiles used in the PFC process trials, the Epo-
Tek U300 epoxy was dispensed on a test die and cured with the same process
cycles described in Table 1. A similar test die was cured in a convection oven
using the conventional 60-minute cure cycle. Figure 6 shows the wet adhesive
DSC curve. As indicated in the figure, the reaction temperature range is between
70 and 160C. The measured heat flow (J/g) was found to be (437.77). The VFM
processed samples were also run through DSC. No peak exotherm was
collected after cure. Cure extent was calculated to be in excess of 96%. Heat
flow release (J/g) after VFM cure was in the range of (0.7J/g). In summary, the
rapid VFM process cycle leads to full, complete curing.
Heat Flow (Arbitrary)
40 60 80 100 120 140 160 180 200 220 240
Figure 6. Overlay of Differential Scanning Calorimeter
(DSC) scansFor fresh adhesive and VFM-cured adhesive.
During the VFM process cycle, temperature is monitored by a non-contact IR
system, focused on one die in the test vehicle. The VFM process inherently
ensures a uniform distribution of the microwave energy over the various die.
MicroCure systems are also provided with a surface contact, fiber-optic
temperature measurement system for profiling and set-up. An ABS PFC test
vehicle was subjected to VFM profile # 5 (refer to Table 1). The temperature of
the substrate and that of the die during cure was measured to evaluate the
temperature gradient between the die and the ABS substrate; hence, the
selectivity of VFM heating. Figure 7 illustrates the locations where temperature
The temperature of the die was monitored by the Raytek system with an
emissivity value pre-set at 0.77, the value recommended when referencing from
silicon. The emissivity was kept at this value for subsequent processing. The die
temperature followed the thermal profile described in VFM profile #5, while the
temperature of the substrate directly beneath the die lagged behind by 25°C.
This gradient in temperature is beneficial in this process because it leads to less
risk of substrate deformation (beyond its viscoelastic threshold) during process.
Furthermore, the substrate temperature away from the die, as illustrated in
Figure 7, remained at room temperature during the entire process. Again, the
selectivity of VFM heating is beneficial for this application.
substrate Nortech contact Nortech contact
Figure 7. Schematic representation of temperature
measurements made on the die and substrate.
The thermal gradient was measured to be 25C.
Figure 8. Illustrates an IR image of a VFM heated ABS test vehicle. As
illustrated in this image, the thermal energy distribution over the test vehicle is
selective and focused by the underfill/die area. This same phenomenon is to be
expected for curing liquid encapsulant (glob-top) curing for chip on board
applications used in smart card assemblies.
Figure 8. IR-image illustrating selective
heating of the smart card components.
Line Balance and Equipment Considerations
The target RDIF-tag throughput defined by PFC is 2000 units per hour. Based
on the results of these preliminary trials, the standard MicroCure 5100 should be
capable of achieving this goal. The MicroCure 5100 can host up to five (5)
pallets at a time, or as many as 100 RFID devices per cycle, depending on the
width of the pallets and size and spacing between individual antennas. Based on
the 2 minute total process cycle and assuming a one-minute combined load and
unload time, or a total cycle time of three (3) minutes, this would provide for 20
cycles per hour, or a throughput of 2000 UPH.
In the case of the underfill process, the antennas can potentially be stacked in a
specially designed cassette (microwave transparent - made of either FR4, G10,
or G11). Up to 4 sheets could potentially be stacked at a time. With the
assumption that 5 cassettes are loaded with 4 sheets each, the UPH in this case
can be quadrupled. Stacking would permit a throughput up to 8000 UPH from
the single Microcure 5100 VFM system.
The MicroCure 5100 is SMEMA compatible and operates in-line with dispense
and buffer modules, as desired. Figure 9 shows an example of pallets carrying
48 X 78mm ISO antennas that have been screen printed on ABS plastic film.
Figure 9. Photograph of screen printed antenna sheets on the
edge-rail conveyor as viewed from service access door.
The feasibility for rapidly curing flip chip underfill epoxies for smart card
applications has been demonstrated. Several VFM profiles were determined to
be satisfactory for the product and pallet configuration and the process is directly
compatible with the standard MicroCure 5100 VFM system. The major benefits
of VFM processing for this application that directly respond to PFC objectives
include the following:
• Rapid cure permits in-line and just-in-time line balance versus the existing
• Work in process is reduced by 95%.
• Liability for defects occurring between QC stations is therefore
In addition to the above direct benefits, the following factors are also significant in
comparing VFM to existing batch process methods:
• Reduced floor space for curing process (50 inches and in-line).
• Reduced energy and heat discharge into manufacturing environment.
• The MicroCure system only heats the epoxy so the oven exhaust is
minimized and the prime power is only 3-4 kilowatts.
2. CURING OF CONDUCTIVE INK PRINTED ANTENNAS (5 TURN COIL) ON ABS
The major objective for this particular application is curing the thick films (inks)
without substrate deformation. There are two routes that can be taken to achieve
this objective: 1- the substrate material needs to be made of a material that can
withstand the cure temperature without deformation and 2- use a cure process
that is selective in nature. The substrate material choices are dictated by the
economics involved. The VFM was investigated as an enabling technology for
the curing method compatible with the economically viable materials. The
following paragraphs describe different configurations and the resistance
measurements that were performed to confirm the degree of curing. In general,
the lower the antenna resistance, when measured end-to-end, the better the
The curing of conductive inks (antenna printed with E162-5 epoxy) was
performed using special microwave fixtures. These fixtures placed silicon wafers
or metal plates above and/or beneath the sheets of ABS and aided in providing
radiant heat to the antenna ink. In some instances the bottom fixtures contacted
the sheets, in others the sheets were suspended above the surface of the bottom
fixtures using a sheet of Kapton (polyimide) film. The curing of antenna was
carried out using different configurations as illustrated in Figure 10. Full curing of
the 5 turn coils was accomplished in 15 minutes as opposed to 120 minutes at
90-95°C. Recent testing of conductive inks specially formulated for VFM curing
has shown promise that this process can be further optimized to yield a total
process cycle in the range of 7 minutes or less.
Figure 10. The various configurations used for curing antenna on ABS
The top fixture was silicon, the bottom fixture was metal. A first test vehicle was
heated to 95°C and held at temperature for 15 minutes. The resistance was
found to be 6.5 Ω. A second test vehicle was similarly heated to 95C and held
for 20 minutes and yielded 3.6Ω.
Both top and bottom fixtures were made of silicon. A first test vehicle was heated
to 95°C and held at temperature for 15 minutes. The resistance was found to be
2.0 Ω, and seemed fully cured. A second test vehicle was similarly heated to
95°C and held at temperature for 14 minutes and yielded a resistance of 2.3 Ω.
A third test vehicle was held at 90°C for 12 minutes and yielded a resistance of
Several other configurations were tried and found to have similar results. Curing
within 15 minutes, using standard off-the-shelf ink and without substrate
warpage, is quite possible. Better fixtures that answer the production concerns
and can further reduce the cure times need to be explored. Further study is also
needed to gain a better understanding of which curing profiles work best for this
type of VFM application.
3. CURING OF FLIP CHIP BUMPS ON ETCHED-COPPER PET SUBSTRATES:
PET sheets, containing etched-copper antenna, are primarily used for RFID inlet
assembly. A typical antenna, designed for a 4-bit, 13.56 MHz chip, is illustrated
in Figure 11. The inside and outside loops of the antenna terminate at two small
pads which are connected to the two active pads of the RFID-chip, located in
opposite diagonal corners of the chip. Before the chips are diced on the wafer,
conductive epoxy bumps are stencil printed and cured in the corner of each chip
of the wafer. An example of conductive bumps on a chip, are shown in Figure
12. In addition to the two active pads on the antenna, two non-functional pads
are situated between the antenna loops, and are used to physically balance the
chip during flip chip placement.
Figure 11. Example of a copper-etched RFID antenna showing the
chip placement pads.
Figure 12. Philips I-Code RFID chip showing the polymer conductive
bumps in each corner. Only the upper-left and lower right bumps are
connected to “active” pads of the chip. These bumps are stencil printed
and cured before the wafers are diced.
The first process step in assembling these tags is to stencil print two small
conductive bumps on the active pads for each antenna etched on the PET sheet.
Figure 13 shows the appearance of the wet epoxy bumps following this first step.
Then, using automated pick and place equipment, the active bumps of the RFID
chips are aligned and pressed into these wet epoxy bumps. When assembled,
the RFID chip bridges across the antenna traces, as shown in Figure 14.
Because the resulting “bump-to-bump” connections are electrically conductive at
this point, the antennas can be functionally tested both before curing and again
after curing, allowing the VFM-curing process yields to be easily identified and
Figure 13. During assembly, conductive bumps are stencil
printed onto the antenna pads. The chips are then aligned and
placed to establish “bump-to-bump” interconnections.
Figure 14. Appearance of the RFID-inlet after flip chip placement.
Hoping to eliminate the long, 8.5-hour bump cure-cycle mentioned earlier, VFM-
curing was viewed as being very strategic to lowering WIP, handling, and
ultimately unit cost. The through-put goal for this process was also 2000 UPH.
A variety of cure profiles were tested in combination with differing pallet
materials. To maintain flatness of the PET sheets during chip placement, the
processing pallets were designed as separate top and bottom panels that held
the sheets in the middle. Cut-outs machined in the top panels provide access to
the antenna pads where the chips are placed.
Consistent curing results were obtained by automatically ramping to110 deg-C,
at a rate of 1.0-2.0 degrees per second, and holding at this temperature for
approximately 1 minute. Since the temperature is referenced off the top of one of
the silicon chips on the antenna sheet, there is a thermal lag of approximately 25-
50 degrees between the PET sheet and silicon chips. The total process time
varies, dependent upon the total number of pallets with antenna sheets being
simultaneously cured, but typically ranges from 3-4 minutes. This through-put
falls short at 1500 UPH, assuming four panels with 25 antennas per sheet and a
4 minute cycle time, but recent process enhancements have shown that the cure
cycle can be reduced to approximately 2 minutes.
After substituting the VFM-curing process in place of the conventional oven-
curing process, PFCC saw an immediate improvement in assembly yields.
Production yields now average a solid 99% compared to fluctuating yields
between 85-95% using the older process. These results are surprising in light of
the fact that this production line has only been running about 12 weeks and is not
under statistical process control yet. Some of the yield increase may be
attributed to reduced handling, but some previous types of failures have now
As a result of the successful qualification of VFM curing for in-line production of
smart card and RFID inlet products, Polymer Flip Chip Corp. purchased a
LAMBDA Model #5100 VFM oven and intends to add a second unit to their RFID
assembly line shortly. Figure 15 shows the VFM unit installed at PFCC’s
manufacturing facility in Billerica, Massachusetts.
Figure 15. LAMBDA Microcure Model 5100 as installed on the RFID-inlet
manufacturing line at Polymer Flip Chip Corporation.