flipchip_assembly development via modified reflowable underfill proc

Flip-Chip Assembly Development via Modified Reflowable Underfill Process

Ping Miao, Yixin Chew, Tie Wang and Louis Foo

Questech Solutions Pte Ltd, Singapore

33 Marsiling Industrial Estate, Road 3, #03-01, Singapore 739256

e-mail: pmiao@

Tel: (65)-3684822, Fax: (65)-3685277

Abstract II. Formulations and testing procedures

This paper presents two flip-chip assembly processes that Three different formulations were evaluated during the

enable an underfill with higher filler loading to be incorporated

into the package. The first process includes dispensing the

underfill containing higher filler loading on substrate surface,

followed by chip placement and solder reflow under thermal

compression. Apart from this, the second approach is virtually

the modification of standard reflowable underfill process. The

underfill with higher filler loading was spin-coated onto a

bumped wafer surface and then cured. Subsequently, the top

portion of bumps was exposed by laser treatment prior to wafer

dicing. The diced chips with low CTE coating on surface already

were assembled via standard reflowable underfill process that

includes dispensing reflowable underfill, chip placement and

solder reflow through reflow oven.

I. Introduction

Flip-chip technology on an organic substrate has been used in

vast applications in the microelectronic industry. When

assembled on the organic substrate, the large CTE mismatch

between the die and substrate necessitates the use of underfill

materials to enhance the thermo-mechanical reliability of the

interconnections [1]. Currently, underfilling is carried out by the

conventional capillary flow process and reflowable underfill

process (no-flow underfill process). With the increase in die size,

I/O numbers, and decrease of bump pitch, the capillary flow

method has become increasing difficult and time consuming.

However, the reflowable underfill process provides the solutions

through compression flow of underfill that significantly reduces

underfill flow time [2,3]. In addition to the high throughput, this

process also has less process complexity and capital equipment

requirements by eliminating some assembly steps.

Unfortunately, using solid filler to reduce the CTE of underfill,

has become a limitation of this process. The filler particles are

easily trapped between the solder bump and bump pads, hence

preventing the formation of solder joints. Because of this, the

reflowable underfill has lower filler loading or no filler at all and

therefore CTE of this material turns out to be around 70~90

ppm/

o

C that is higher than commonly expected value. Currently,

the main technical challenge lies in how to reduce CTE

mismatch in flip-chip assembly, in particular for packages

having larger die.

This paper describes two reflowable underfill processes that

enable underfill with higher filler loading to be incorporated into

the package, namely thermo-compression reflowable (TCR)

process and wafer-level coating (WLC) process. In order to

prepare proper underfill for these two processes, DOE (Design

of Experiments) was used to discover what factors affect the

properties of reflowable underfill.

course of this work. Underfill I (UI) was formulated for

thermo-compression reflowable process studies while

Underfill II (UII) and underfill III (UIII) was for wafer-

coating studies. All formulations were epoxy resin/

anhydride based although the cure chemistry and resin

functionality were varied to give different final material

properties. Metal acetylacetonate was used as curing

catalyst for UI and UIII while imidazole compound was

used to catalyze the curing reaction of UII. UI and UII were

filled with spherical silica filler at 55% weight loading. The

underfill samples were produced by dispersing the silica

particles in epoxy matrix using a special disperser.

DSC studies were carried out via a modulated DSC (TA

Instruments, Model 2920). About 10 mg sample in hermetic

DSC sample pan was heated in the DSC cell at 10°C /min to

300°C. CTE and Tg were measured on a

Thermomechanical Analyzer (TA Instruments, Model

2940). The specimen was prepared by fully curing the

sample at 150°C for 1.5 hour. After curing, the sample was

ground to a round tablet with a diameter of 5 mm and

thickness of 4-5 mm and heated in the TMA instruments

from room temperature to 200 °C at a rate of 10°C/min. The

viscosity was measured by a Programmable Rheometer

(Brookfield, Model DV-111) at room temperature (25°C).

A Dage 2400 shear tool is utilized to perform die shear.

Fig. 1 Schematic of shear testing

The adhesion of underfill to SiN passivation (U/SiN) and

solder mask (PSR 4000-AUS303, U/SM) has been tested as

the response for DOE experiments. A small drop of

underfill was placed in the center of testing tools (Figure 1)

and the sample was cured at 150 °C for 1.5 hours.

Shear Strength (S, PSI) = Force (Kg) / Area (mm

2

)

S = 4F / πd

2

6-inch silicon wafer bumped with eutectic (63: 37 Sn/Pb)

solder was used in this TCR and WLC process study. Table 1

lists the specification of test wafer and die.

Table 1. Specification of test vehicle

Wafer size 6’

Chip number 140

Chip size

10.30 × 10.30mm

Configuration Full array

I/O 1600

Pitch

250 µm

Bump height

100 µm

Substrate BT laminate

Surface finish Ni/Au

Solder mask opening

170 µm

S/M thickness

40 µm

III. Process Description

IIIa. Thermo-compression reflow process

In TCR process, the underfill with high silica loading was

dispensed on the surface of organic substrate. In order to

eliminate the voids and reduce the viscosity of underfill

materials, the substrate was pre-heated to ~ 90 ºC. The die was

picked and placed onto the substrate and held at high

temperature under certain force for 20 seconds to allow the

formation of solder joints. Finally, the assembly units were sent

to cure at 150°C for 1.5 h.

The yield of this TCR process is strongly dependable on the

properties of underfill that will be discussed later in this paper.

In addition to the underfill properties, the force and temperature

as well as their ramping rate also play key roles in this study.

Using the normal oven reflow process, whereby no holding force

is applied during solder reflow, less than 10% of solder bump

were found to form the solder joints even when the silica loading

of underfill is only 25 wt %. Most solder bumps did not make

contact with Ni/Au pad and there was underfill material trapped

between the solder and pad. When the force was increased to 50

N, more than 99% solders reflowed to form good solder joints.

These few failed joints are due to the trapping of silica fillers

between the bump and bump pad. On the other hand, when the

force increased to 100 N, all solders reflowed and no dry joints

were found after cross-section analysis. The average standoff of

solder joints is 60 µm. After optimizing the process, the holding

force of 75 N was used to assemble 20 units for reliability test.

In order to eliminate the trapping of silica particles inside

solder joints (Figure 3), the bonding force can not be too high

before and during solder melting. Thus the force and

temperature ramping must be carefully controlled. In addition,

fast ramping of temperature will generate plenty of voids. High

reflow temperature is beneficial for solder reflow. However, this

will result in serious outgassing from underfill.

Figure 4 illustrates the optimized force and temperature

program that is used for TCR process. Initial bonding head

temperature and placing force were set at 150ºC and 20N

respectively. Then it took 10s to ramp to 230ºC and 75N. The

whole assembly was subsequently kept for another 15s to reflow

bumps. This bonding program leads to a package with good

solder joint and acceptable void level (Figure 5).

Figure 2. The solder joints of TCR process, F=75N

Fig. 3 The silica was trapped inside the solder bump when

the force and temperature was too high during solder

melting.

Fig. 4 The ramping profile of temperature and force

Fig. 5 C-Sam image of TCR package

Fig. 6 Schematics of TCR process with (left) and without (right)

applying holding force.

The schematic diagram of thermo-compression reflowable

process is shown in Figure 6. In the first step, the solder bump

contacted with Ni/Au pad at a point. This contact point is critical

for the following solder wetting process. When the solder

melted, it starts to wet the Ni/Au surface from the contact point.

The underfill was pushed away from the surface of Ni/Au by the

“wetting force” of melten solder. To keep the contact point for

wetting, certain level force is required during reflow process.

Otherwise, the underfill flowing at elevated temperature could

result in a gap between the solder and pad, which prevented the

formation of solder joints. In addition, the contact point must be

formed before the solder melt. Otherwise, the filler particles

could be easily trapped inside the solder joints.

IIIb. Wafer-level coating process

Figure 7 shows the WLC process. First, the underfill (UII)

was spin-coated onto the surface of wafer (Figure 7a) and cured

at 125°C for 20 minutes. Second, the top layer of coating

material was removed by laser etching (Figure 7b). The wafer

was then singulated (Figure 7c). The single chip was assembled

on an organic substrate using common reflowable underfill

process, i.e. dispensing second layer of underfill (UIII) onto the

substrate, picking and placing the coated chips onto the substrate

(Figure 7d), reflowing the units in reflow oven and fully curing

materials. (Figure 7e). This process provides another way to

introduce silica filler into reflowable underfill process without

using high force and sacrificing the throughput.

The underfill for coating studies was a solvent-free,

silica filler loaded material. The quality of coating film was

dependent strongly on film formability of underfill. Thus, a

surfactant was used to improve the surface tension of

coating material. This surfactant can enhance the wetting of

underfill to wafer surface and prevent underfill contract

from silicone surface during thermo-curing of underfill.

Coating can be done by screen printing, spin coating or

extrusion approaches. In this study, a spin- coat machine

was used. According to the packaging structure (Fig. 7), the

coating thickness (after fully curing) should be controlled to

~ 60% of bump height. Thus, the spin-coating program can

be optimized according to the viscosity of material and the

coating thickness.

Figure 8 shows the cross-section image of one bump

from coated wafer. The thickness of coating in the area

between bumps (H) were around 60 µm. The underfill was

thicker around the bump and there was a thin layer of

underfill covering the top of the bump (Figure 9). This layer

will hinder the formation of solder joints.

Fig. 7 Schematic diagram of WLC process

Fig. 8 Cross-section image of a coated wafer bump before

and after laser etching

Several methods can be used to remove this layer of

underfill, such as a B-stage approach [4], mechanical cutting

[5] and laser or plasma etching. This paper describes the

laser etching method that can remove the upper layer

underfill efficiently. A KrF excimer laser with wavelength

of 248 nm was used as a light source and 400-600 mJ/cm

2

was applied to etch out fully cured epoxy resin. The etched

residue was sucked up by a vacuum system. Figure 10

shows coated wafer surface before and after etching. After

etching, the bump was exposed and the surface was free

from underfill and silica particles. The surface of the underfill

became rougher after etching. This rough surface enhances the

adhesion of second layer underfill to first layer underfill. It is

interesting to find that the material surrounding the bump is

easier to be etched out than other part. Thus, the thickness of

coated material between bumps is not significantly affected

(Figure 8).

Fig. 9 The top layer of bump before etching. This top part of

bump has been flattened by pressing the bump.

In order to form the solder joints, a second layer material

(UIII) which was free from solid filler was used via common

reflowable process. The solder joints are shown in Figure 11 and

12. The shape of the solder joints is dependent on the exposed

height of bump that is, in turn, determined by the coating

thickness and etching degree. Figure 11 shows the solder joints

that are formed from less etched chips while Figure 12 was the

joint from heavy etched chips. The joint of Figure 11 looks like

“vase” with a “wrest” with a standoff of 75 µm which was

higher than that of Figure 12. Figure 13 exhibits a two-layer

underfill structure of a packaging. The layer on chip side is a

high silica loading underfill (UII) while the very thin layer on

substrate side is a solid filler free material (UIII). By using this

method, sufficient fillet is obtained which, in some degree,

ensure the good reliability. C-SAM image (Figure 14) shows

that the number and size of voids are acceptable.

Fig. 10 The surface of a bump before and after etching.

Fig. 11 The Solder joint of WLC process.

Fig. 12 Solder joint of WLC process.

Fig 13. The two-layer structure of WLC packaging

Fig. 14. C-SAM image of WLC packaging

IV. Reflowable underfills

The key of the reflowable underfill process is the

selection of underfill materials. Underfill for this study

should have proper curing properties, strong adhesion to chip

passivation and solder mask of substrate, proper rheology for

processing, low CTE to match the components and reasonable

modules and glass translation temperature. Figure 15 shows the

DSC analysis of UI, UII and UIII. The curing profiles of UI and

UIII must match the solder reflow profile, as such, UI and UIII

should not gel before solder bump reflow. These two materials

have sufficient fluxing ability to remove the oxide of eutectic

solders. UII, which was formulated for spin coating, was a fast

cure material without self-fluxing capability.

Fig. 15 Curing profile of underfillUI, UII and UIII.

CTE mismatch is a great concern in this study. Figure 16

exhibits the TMA results of theses three materials. The silica

loading of UI and UII was in the range of 50-55%. The CTE (α

1

)

of these two materials were around 40 ppm/ºC. CTE of epoxy

type underfill depended on the silica loading, filler size and the

additives of the formulation. When the silica loading reaches

50%, there is a great reduction of CTE. Thus, normally, the

silica loading was above 55%. The size of silica particles also

affect the CTE of materials. At same loading level, the smaller

size silica resulted in lower CTE. For example, at 55% loading,

5 µm silica loaded underfill has CTE of 33 ppm/ºC while 10 µm

silica loaded underfill has CTE of 37 ppm/ºC. In addition, the

additives also play a role in CTE issue. Those elastomer

additives may increase the CTE and reduce the Tg of underfills.

Fig. 16 CTE and Tg of underfill UI, UII and UIII

Fig.17. Effect of Silica loading on CTE

Another important property of underfill is its adhesive

strength on the surface of other package components. Poor

adhesion of die passivation/underfill and underfill/solder

mask will resulted in the delamination of the interface and

therefore poor reliability of package. Many factors affect

the adhesion strength of materials [6], including the

underfill formulation, passivation and solder mask types,

roughness of these surface as well as the type of additives

and fillers in underfill. As discussed above, several additives

were added to improve the properties of underfills,

including silica particles, wetting enhancing agent, coupling

agent and fluxing agent, etc. Some additives can enhance

the adhesion while others have negative effects on adhesion.

A set of DOE experiments was designed to study the effects

of those additives. Table 2 lists the experiments of DOE.

Wetting agent (siloxane), silica, and coupling agent (silane)

were selected as the main factors and the shear strength

(Shear 1: U/SiN, Shear 2: U/SM)) was used to measure the

adhesion of materials. (response of DOE)

Table 2. DOE experiments

SilicaSiloxane Silane Shear1 Shear2

(%) (phr) (phr) (psi) (psi)

1 25 0.25 1.0 8600 5560

2 25 0.25 1.75 7020 4150

3 25 1.0 1.0 7850 3960

4 25 1.0 1.75 6970 5960

5 55 0.25 1.0 6450 3560

6 55 0.25 1.75 6640 3100

7 55 1.0 1.0 4490 3230

8 55 1.0 1.75 5060 3130

The results of DOE experiments are shown in Table 3

and Figure 18–20. In the case of die passivation, the

significant effects are the loading amount of silica and

siloxane. Among them, silica loading is the most significant

factor influencing adhesion strength. High silica loading

will significantly reduce the shear strength due to the

reduction of the resin amount in underfill system. The more

significant factor is siloxane which also has a negative effect

on the adhesion strength. Silane is the least significant.

Increasing the silane amount did not significantly increasing

the adhesion. In contrast, too much of silane may reduce the

adhesion slightly. The Coefficient of Determination (r

2

) of

shear 1 is ~ 99%, indicating that the model as fitted explains

99% of the variability in shear 1.

Table 3. Estimated effects for shear 1 and shear 2

Factors Shear 1 Shear 2

A. silica -1995.0 -1650.0

B. siloxane -1085.0 -25.0

C. silane -435.0 5.0

AB -680.0 -125.0

AC 810.0 -285.0

BC 270.0 940.0

r

2

99.9119 86.5085

Table 2 also indicates that the adhesion of underfill to solder

mask is much weaker than that of SiN passivation. As in the

case of the passivation, the silica loading has the main effect and

the silane has the least effect on the adhesion strength of U/SM

(Figure 20). However, unlike the case of U/SiN, siloxane did not

result in a significant change in U/SM.

Based on the above results of DOE, the conclusions on the

adhesion strength of silica filled underfill are as follow.

1. The adhesion strength is reduced by loading silica fillers.

For 55% loading of silica, the die shear strength is only

about half of that of the non-filled underfill.

2. Adhesion on SiN is stronger than the adhesion on solder

mask. Thus, future works should be focused more on the

underfill/SM interface.

3. Addition of wetting agents can significantly reduce the

adhesion of underfill/SiN, but have small effect on the

adhesion of underfill/SM. A more effective wetting agent is

needed to reduce the amount of siloxane in formulations.

4. More coupling agent (silane compounds) is not useful for

the adhesion of underfill/SM and underfill/SiN. In contrast,

it may produce out-gassing problem due to its relatively low

boiling point.

Fig. 18 The effects of silica, silane and siloxane on the adhesion

of U/SiN.

Fig. 19 The response plots of main effects (silica and

siloxane) on the adhesion of U/SiN

Fig. 20 The effects of silica, silane and siloxane on the

adhesion of U/SM

V. Conclusions

Two reflowable processes were proposed to apply the

low CTE underfill. By thermo-compression reflow process,

the silica was introduced into the package directly. After

process optimization, solder joints, which are free of trapped

fillers between bump and pad are obtained by applying

underfill which has filler loading up to 55%. Alternatively,

the low CTE underfills can also be applied by wafer

coating-etching approach. For this reflowable underfill

process, two-layer structure was clearly demonstrated with

one thick layer of low CTE underfill at the chip side and

another thin layer of high CTE reflowable underfill at the

substrate side. The keys to achieving high packaging yield

and better reliability is to making underfill which has low

CTE, yet with strong adhesion, sufficient fluxing ability.

Studies of package reliability, evaluation of other coating

and etching methods are in process.

Acknowledgement

Finally, we are deeply grateful to Dr Lu Yongfeng and

Dr Ren Zhongmin in Data Storage Institute, Singapore who

have been so helpful in conducting the laser etching

experiment.

References

1. D. Suryanarayana and D. S. Farquhar, “Underfill

encapsulant for flip chip applications” in Chip on Board, J

H. Lau, Ed. New York:Van Nostrand Reinhold, p. 504,

1994

2. C. P. Wong, D. F. Baldwin, M. B. Vincent, B. Fennel, L. J.

Wang, and S. H. Shi, “Characterization of a no-flow

underfill encapsulent during the solder reflow process,” in

Proceedings of the 48

th

Electronic Components and

Technology Conference ’98, p.1253, 1998

3. N. W Pascarella and D. F. Baldwin, “Compression flow

modeling encapsulants for low cost flip-chip assembly,”

IEEE Transactions on Components, Packaging, and

Manufacturing Technology-Part C, Vol. 21 (4), p. 325,

1998

4. S. H. Shi, T. Yamashita, and C. P. Wong, “Development of

the wafer level compressive-flow underfill process and its

involved materials,” IEEE Transactions on Electronics

Packaging Manufacturing, 22(4), p.274, 1999

5. M. Topper, J. auersperg, V. Glaw, K. Kaskoun, E. Prrack,

B. Keser, P. Coskina, D. Jager, D. petter, O. Ehrmann, K.

Samulewicz, C. Meinherz, S. Fehlberg, C. Karduck, H.

Reichl, “Fab Integrated Packaging (FIP): A new concept for

high reliability wafer level chip size packaging,” in

Proceedings of the 48

th

Electronic Components and

Technology Conference ’2000, p. 74, 2000

6. Q. Yao, J. M. Qu, J. L. Wu and C. P. Wong,

“Characterization of underfill/substrate interfacial

toughhness enhancement by silane additives,” IEEE

Transactions on Components, Packaging, and

Manufacturing Technology-Part C, Vol. 22 (4), p. 264,

1999