fulltext.pdf

Journal of ELECTRONIC MATERIALS, Vol. 36, No. 3, 2007

Letter

DOI: 10.1007/s11664-007-0089-5

2007 TMS

Failure Morphology after the Drop Impact Test of the Ball Grid

Array Package with Lead-Free Sn-3.8Ag-0.7Cu on Cu and Ni

Under-Bump Metallurgies

J.W. JANG, 1,2

J.K. LIN, 1 and D.R. FREAR 1

1.—Freescale Semiconductor, Inc, Tempe, AZ 85284, USA. 2.—e-mail: j.jang@freescale.com

High strain-rate drop impact tests were performed on ball grid array (BGA)

packages with lead-free Sn-3.8Ag-0.7Cu solder (in wt.%). Plated Ni and Cu

under-bump metallurgies (UBMs) were used on the device side, and their drop

test performances were compared. Failure occurred at the device side,

exhibiting brittle interfacial fracture. For Ni UBM, failure occurred along the

Ni/(Cu,Ni) 6 Sn 5 interface, while the Cu UBM case showed failure along the

interface between two intermetallics, Cu 6 Sn 5 /Cu 3 Sn. However, the damage

across the package varied. For Cu UBM, only a few solder balls failed at the

device edge, whereas on Ni UBM, many more solder bumps failed. The dif-

ference in the failure behavior is due to the adhesion of the UBM and inter-

metallics rather than the intermetallic thickness. The better adhesion of Cu

UBM is due to a more active soldering reaction than Ni, leading to stronger

chemical bonding between intermetallics and UBM. In our reﬂow condition,

the soldering reaction rate was about 4 times faster on Cu UBM than on Ni

UBM.

Key words: Drop impact test, ball grid array (BGA), lead-free solder,

intermetallics

INTRODUCTION

The miniaturization of electronic components has

driven the evolution of portable devices, such as cell

phones, digital cameras, and personal music players.

The use of portable devices has created the need for a

new reliability criterion of drop impact tests because

of the tendency to accidentally drop these devices. 1

Under drop conditions, failure typically occurs

through the fracture of solder joints. Ball grid array

(BGA) packages are more prone to damage after drop

impact, because the package is larger and the solder

joints are the only source of a mechanical adhesion

(no underﬁll is used in BGA). 2,3 Also, as solder alloys

move to lead-free compositions, the increased

strength of these alloys, especially in the Sn-Ag based

solders, may lead to drop concerns.

In the study of the drop impact test, much

attention was focused on numerical analysis and

simulation. 1–8 However, very limited work has been

reported on the metallurgical response. 9,10 Drop

impact leads to highly accelerated strain/stress

conditions, so the failure morphology of the solder

joint may differ from that of a convention mechan-

ical test. In the present study, we compare the drop

test response of the lead-free solder joints for BGA

packages with two different under-bump metallur-

gies (UBMs), Cu and Ni. The solder alloy composi-

tion was Sn-3.8Ag-0.7Cu (in weight percent and

abbreviated as SnAgCu). The purpose of this work is

to develop a relationship between the drop perfor-

mance of a given BGA package as a function of

UBM structure and the resultant interfacial inter-

metallics.

EXPERIMENTAL PROCEDURE

A BGA package with a size of 10 mm · 10 mm

was used in this study. Two UBMs were studied:

plated Cu and plated Ni, respectively. Solder balls

(200 lm in diameter), with a composition of

Sn-3.8Ag-0.7Cu, were reﬂowed and wetted on the

UBM at the peak temperature of 260C for 1 min

(Received August 14, 2006; accepted December 16, 2006;

published online April 13, 2007)

207

208

Jang, Lin, and Frear

above the melting temperature of the solder alloy.

The BGA package was then assembled to a board

(with Cu metallization) by an additional reﬂow

process at the same temperature and time as the

initial reﬂow.

Drop impact testing was performed to make the

entire board plane drop to the ground without tilt-

ing. Impact time and acceleration of the dropped

package were controlled in order to meet the JE-

DEC drop test criterion (1,500 g within 0.5 ms). 11

To monitor the drop condition, an accelerometer

was attached on the hitting stage and a strain gage

was glued to the board, close to the edge of the

device. For every drop test, the acceleration and

board strain were recorded. Five drop tests were

performed on each assembled sample. After the drop

impact test, the packages were metallurgically cross

sectioned and polished for microstructural analysis.

Optical microscopy and scanning electron micros-

copy (SEM) were used to observe the failure mor-

phology and locations for drop impact damage.

the drop test. Failure occurred at the solder inter-

face of the device side for both cases, similar to that

observed by others. 5,6 From the images, it is evident

that solder has no plastic deformation. For conven-

tional tensile tests, SnAg-based solders show

extensive deformation. However, brittle interfacial

failure is observed at the high strain rates. This

may be due to the hard Ag 3 Sn intermetallic pre-

cipitates that prevent solder deformation. As a re-

sult, the failure occurs at the interface at a higher

strain rate in SnAg solder, unlike eutectic SnPb

solder. A void is observed in the BGA bump, as

shown in Fig. 1c. This should not affect the behavior

after drop test because the crack location is mostly

along the intermetallics/UBM interface, but the

void makes a small contact area with this interface.

Figure 2a shows the failure location of the BGA

solder joint on Ni UBM with a brittle failure along

the interface between Ni UBM and (Cu,Ni) 6 Sn 5

IMC. The intermetallic morphology before failure is

shown in Fig. 2a. This is the initiation site for fail-

ure, because the intermetallics are brittle, compared

to the other materials in the interconnect. The BGA

package with Cu UBM shows more interesting

behavior. Two interfacial failures are possible be-

cause two interfacial intermetallics (Cu 3 Sn and

Cu 6 Sn 5 ) on Cu UBM exist. Figure 3a shows the

solder ball at the edge of the device, and the crack

appears to be along the Cu UBM/Cu-Sn intermet-

RESULTS AND DISCUSSION

Figure 1a shows a low-magniﬁcation optical im-

age of the BGA assembly. The cross section was

performed through an edge row of the solder balls.

Figure 1b and c show the cross-sectional images of

the BGA bumps with Ni and Cu metallizations after

Fig. 1. (a) Overall optical image of the cross section for the BGA package in this study. A BGA bump after the drop test for (b) Ni UBM and (c) Cu

UBM. The entire cracking around the device side is shown for both cases.

Failure Morphology after the Drop Impact Test of the Ball Grid Array Package with Lead-Free

Sn-3.8Ag-0.7Cu on Cu and Ni Under-Bump Metallurgies

209

Fig. 2. (a) Representative failure morphology of the BGA bump with Ni UBM after the drop test. Failure along Ni/(Cu,Ni) 6 Sn 5 is clear. (b) BGA

solder joint morphology on Ni UBM before the drop test.

allics interface. However, high magniﬁcation

(Fig. 3b) reveals a thin intermetallic layer on the

Cu side, and the failure is through the Cu 3 Sn/

Cu 6 Sn 5 interface. This morphology is much clearer

in the third bump from the edge, shown in Fig. 3c.

After initial reﬂow, Cu 3 Sn intermetallics are thin

(less than a micron thick), as shown in Fig. 3d,

unless the time of reﬂow extended or performed

multiple times. In spite of somewhat different

failure locations for Ni and Cu UBMs, both are

typical of a brittle failure because crack propaga-

tion is rapid with a planar morphology. This type

of brittle interfacial failure usually occurs when

bulk solder does not accommodate the impact

strain, and so fracture occurs through the inter-

metallic interface.

Fig. 3. (a) Representative failure morphology of the BGA bump with Cu UBM after the drop test. (b) Magniﬁed image for the rectangular region in

(a), showing that the failure location is actually the Cu 3 Sn/Cu 6 Sn 5 interface. (c) Hairline cracking showing failure along the Cu 3 Sn/Cu 6 Sn 5

interface more closely. (d) BGA solder joint morphology on Cu UBM before the drop test.

210

Jang, Lin, and Frear

Fig. 4. SEM images of the BGA bump with Ni UBM showing the damage from the edge to the center after the drop impact test. (a) Corner bump

(BGA ball ‘‘1’’ of Fig. 1a), (b) second bump (BGA ball ‘‘2’’ of Fig. 1a), (c) third bump (BGA ball ‘‘3’’ of Fig. 1a), and (d) fourth bump (BGA ball ‘‘4’’ of

Fig. 3).

For the individual BGA bumps, both Ni and Cu

UBM exhibited a brittle failure mode. However, the

extent of damage differs signiﬁcantly. Figure 4

shows a series of SEM images of failed BGA inter-

connects with Ni UBM. Four consecutive bumps

from the device edge failed through the intermet-

allics/Ni UBM interface. On the other hand, as

shown in Fig. 5, only two bumps on Cu UBM failed;

the third bump had a partial failure with hairline

cracking (Fig. 5c), and no damage was observed in

the fourth bump (Fig. 5d). We veriﬁed the same

trend of failure by repeating the analysis with

several samples. Another interesting point for Cu

UBM is that the crack stopped inside the solder

bump (Fig. 5c). This suggests that the crack growth

on Cu UBM is not as fast as on Ni UBM, because a

slowly growing crack has a greater potential of

stopping inside the solder bump, indicating a

stronger interface between Cu UBM and the Cu-Sn

intermetallics.

These results indicate that Cu UBM has a higher

drop impact resistance. Mattila et al. characterized

the drop test performance of lead-free SnAgCu

solder on Cu and electroless NiP UBMs. 9 They

reported a better drop test reliability of Cu UBM

than NiP UBM through Weibull statistics, rather

than the solder bump damage analysis on the entire

sample, as shown in this study. The failure mode

was not signiﬁcantly different from our data, but

the NiP case had a complicated failure mode due to

Ni 3 P intermetallics. Chiu et al. reported that the

interfacial failure in drop becomes much worse

after solid-state aging due to excessive formation of

Kirkendall voids. 12

Under the drop test condition, the normal stress

component is more critical than the shear compo-

nent. During the upward and downward bending of

the board after drop, tensile and compressive

strains are alternately imposed on the BGA balls, 4

so the adhesion between the UBM and intermetal-

lics is more critical. If brittleness (or low fracture

toughness) of intermetallics is the primary factor in

drop test performance, the fracture of the inter-

metallics would have been observed. However, only

planar interfacial failure is observed. Therefore, the

role of intermetallics in the drop condition is dif-

ferent from thermomechanical fatigue or shear

deformation, where the intermetallic size/shape

(i.e., growth kinetic) is a prevailing factor. A number

of papers demonstrated that with the thicker in-

termetallics, the solder joints tend to have a brittle

failure mode, dependent on the detailed solder

alloys composition. 13–16 Excessive growth of inter-

metallics can deteriorate solder joint ductility. For

Failure Morphology after the Drop Impact Test of the Ball Grid Array Package with Lead-Free

Sn-3.8Ag-0.7Cu on Cu and Ni Under-Bump Metallurgies

211

Fig. 5. SEM images of the BGA bump with Cu UBM showing the damage from the edge to the center after the drop impact test. (a) Corner bump

(BGA ball ‘‘1’’ of Fig. 1a), (b) second bump (BGA ball ‘‘2’’ of Fig. 1a), (c) third bump (BGA ball ‘‘3’’ of Fig. 1a), and (d) fourth bump (BGA ball ‘‘4’’ of

Fig. 3).

example, in thermal fatigue damage, where shear

deformation dominates as a main factor, there are

several failure modes, including intermetallics/sol-

der failure and bulk solder failure, dependent on

solder alloy composition. 13 The schematics shown in

Fig. 6 represent the state of stress in solder inter-

connect under two extreme conditions. For the drop

test condition (Fig. 6a), the strain rate is extremely

high and the duration time is extremely short (less

than 1 ms). Therefore, failure occurs over a very

short period of time. Brittle fracture is expected and

the failure has a planar morphology because crack

propagation is extremely fast. On the other hand,

for the thermomechanical fatigue condition, the

strain rate is very slow (typically, 10 ) 3 –10 ) 5 s ) 1 )

and the duration is much longer (several hours to

many days). This condition allows the shear stress

to act on the entire solder joint, rather than the

most brittle location. Figure 6b demonstrates the

simple geometry of how the intermetallic size is

more sensitive to shear stress, which is related to

the intermetallic size. Also, a very slow strain rate

would allow sufﬁcient time for the intermetallic

layer to response to the shear stress.

From the above results (Figs. 4 and 5), Cu-Sn

intermetallics have a stronger adhesion on Cu than

Ni-Cu-Sn intermetallics do n Ni. Why does Cu

have better adhesion to its interfacial intermetal-

lics than Ni UBMs? No study has been reported

regarding this adhesion, but it can be related to

the chemical reaction between the intermetallics

and the UBMs. Intermetallics form and grow be-

tween the UBM and solder by means of wetting

reaction and solid-state aging. After the initial

wetting reaction to create the BGA assembly, solid-

state aging begins and the intermetallics continue

to grow. Even at a relatively low temperature (less

than 70C), intermetallic growth is still observed. 17

This means that chemical species (UBM materials

such as Cu or Ni) diffuse through the interface

between the UBM and intermetallics at all times,

providing chemical bonding, and contributes to the

adhesion. This also implies that a more active

soldering reaction leads to more atomic movement

through the interface, enhancing the chemical

adhesion. Recently, one study showed the correla-

tion between the chemical bonding of the UBM/

intermetallic interface and diffusion through the

interface. 18 During the solid-state annealing of the

high-lead solder on Cu UBM, the Sn-Cu intermet-

allics spalled off from Cu UBM. This was attrib-

uted to the loss of the chemical bonding between

Cu UBM and Cu-Sn intermetallics due to no

additional chemical reaction.

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: