for a Ball Stitch on Bump Wire Bonding Process_图文

Simulation and Experimental Analysis for a Ball Stitch on Bump Wire Bonding Process

above a Laminate Substrate

Yong Liu, Howard Allen, Timwah Luk and Scott Irving

Fairchild Semiconductor Corp.

82 Running Hill Road, Mail Stop 35-2E, South Portland, ME 04106

Email: yliu@; Tel: (207) 761-3155; Fax: (207) 761-6339

experimentation. The goals of our study are to determine the

stress and deformation mechanism of the bonding process on

a laminate substrate and to understand the impact of different

wire bonding parameters to the stress balance and

deformation of a bond pad with partial support at the bottom

of laminate. The simulation will consider both the ultrasonic

transient dynamic bonding process and the stress wave

transferred to the interface between bond structure and

laminate substrate. The model considers the bonder capillary

as a rigid body due to its high hardness, thus the rigid and

elastic plastic contact pair between capillary and FAB is

introduced. The contact surfaces between the FAB and bond

pad are a non-linear contact pair with consideration of the

dynamic friction. The Pierce strain rate dependent model is

utilized to model the wire bonding stain behavior. Different

laminate material parameters are studied to understand their

impact on the bond pad structure. Different ultrasonic

parameters such as bonding force and frequency are studied

and discussed for the effects of the bonding process on

Introduction

Currently there is a substantial volume of modeling work

laminate substrate structures with partial supports.

on standard ball and wedge wire bonding process, most of

Experimental test work includes a DOE study with different

which consider pure mechanical bonding loads with static or

parameters of ultrasonic power and bonding force. Ball shear

quasi dynamics methodology to simulate the free air ball

strength is used for the DOE test response. Finally, the trend

(FAB), compressive bonding process on silicon and wedge

comparison and discussion of modeling and experimental

bonding on lead frame substrate [1-3]. In 2004, we presented

results are presented.

a complete transient dynamic modeling work for the wire

bonding process of both FAB and the bond pad metallization

Problem Definition, Material Properties and assumptions

system at ECTC54 [4]. In recent years, there have been

The basic bond pad structure of the laminate substrate,

studies of wire bonding for bond pad structure with low K

shown in Fig.1, is created using Cu, Ni and Au layers plated

film above silicon substrate [5-6]. However, little work has

onto the laminate material. The wire bonding area is located

been reported about a new wire bonding process called BSOB

near the via, which is also very close to the edge of the die.

and its use with laminate-based substrate packages. In

Furthermore, due to the substrate design the bottom is only

standard wirebonding the first bond is a ball bond to a

partially supported. This increases the difficulty of BSOB

bondpad on the die followed by a wedge bond to a bondfinger

wire bonding. Before starting the actual assembly BSOB wire

on the substrate. The BSOB process is different in that the

bonding process, carefully simulation and analysis are

first ball bond is attached to the die and the wire to the ball is

necessary for cost savings. In order to conduct an effective

then broken, leaving a gold bump. The next bond is a second

simulation, the following assumptions are made: (1) The

ball which is bonded to the bondfinger on the substrate. From

temperature of FAB is the same as substrate (in reality, there

this ball the wire is played out back to the first ball upon

is some difference due to the transient temperature cooling

which a wedge bond is made, completing the bonding cycle.

from FAB forming and moving to contact bond pad). (2)

BSOB wire bonding is used where exceptionally short wires

FAB is rate dependent elastic plastic material during bonding

are needed such as in very thin packages where loop height

process. The bond pad and other metal layers are treated as

control is extremely critical and the distance between wire to

elastic plastic material. All the other materials are considered

to be linear elastic. (3) The contact intermetallic effect,

die edge is very small.

It is known that in reality, wire bonding is a complicated,

diffusion in bond formation due to ultrasonic energy and heat

multiple physical transient dynamic process and is completed

induced by friction, will not be considered in this paper. (4)

within a very short time. The dynamic impact of wirebonding

The capillary is a rigid body due to much higher Young’s

to both devices and the substrate is critical and significant.

modulus and hardness. The inertia force from capillary

This study will evaluate the BSOB wire bonding process on a

transferred to FAB is not considered in this paper. The related

laminate substrate by use of both modeling and

material properties are listed in Table 1.

Abstract

This study will focus on a ball stitch on bump (BSOB)

wire bonding process above a laminate substrate by modeling

and experiment. The goals of our study are: (1) to determine

the stress and deformation mechanism of BSOB wire bonding

process on laminate substrate; (2) to understand the impact of

wire bonding parameters. The simulation will include the

ultrasonic transient dynamic bonding process, and the stress

wave transferred to the interface between bond structure and

laminate substrate. Different laminate material parameters are

studied for the optimized solution. Different ultrasonic

parameters of bonding force and frequency are studied and

discussed for the effects of bonding process on laminate

substrate structures with partial supports. Experimental test

work includes a DOE study with different parameters of

ultrasonic power and bonding force. Finally, the comparison

of modeling and experimental results is provided.

1-4244-0152-6/06/$20.00 ©2006 IEEE19182006 Electronic Components and Technology Conference

Ikeda et al indicated [2]: a gold ball is impacted by a

capillary at the loading speed of 0.98 N/sec, which may result

in the strain rate of the gold ball more than 1000 1/s locally.

Based on the Hopkinson impact bar tests by Ikeda, the yield

stress of FAB with strain rate dependent Pierce model can be

approximated by [4]:

&

pl

⎤⎡

ε

σ

s

=

⎢

1

+

⎥

σ

0

(1)

γ

⎦⎣

where

σ

0

=0.0327GPa , m=1 and γ=561.4 (1/s)

A general finite element code, ANSYS, is used in the

modeling. A non-linear large deformation and transient

dynamics implicit algorithm with the above rate dependent

Peirce model (1) is selected. Since the bonder capillary is

considered as a rigid body due to its high hardness value, this

leads to the rigid and elastic plastic contact pair between

capillary and FAB. While the contact surfaces between FAB

and bond pad are a non-linear contact pair with consideration

of the dynamic friction. The capillary moves down a certain

height (bonding height) to press the FAB with a high speed

and different frequency. Fig.2 illustrates the capillary on a

FAB before compression. Fig.3 shows the local (with half

via) deformed meshes of FAB and bond pad system in the

wire bonding process in which the yellow area is space/air,

light blue area is laminate and dark blue area is the copper..

m

Fig.2 Ultrasonic capillary on a FAB of a second

BSOB above laminate

Au

Ni

Cu

BT

Air

Fig.3. Meshes of deformed FAB and bond pad

model for a BSOB wire bonding (second bond)

with 15018 elements

Modeling Results and Discussion

(1) Impact of Wire Bonding Force

The results of impact of wire bonding force are showed in

Fig.4 - Fig.10. These results are obtained under a fixed

ultrasonic frequency 128 KHz.

Table 1 Materials parameters

Material Modulus Poisson Yield stress

ratio (GPa)

(GPa)

Laminate 20.5 0.39

Ni 205 0.3

Cu 110 0.3

Al(Cu) 70.0 0.35 0.2 (25C)

0.05 (450C)

Au(FAB) 60.0 0.44 0.0327(200C)

CU/NI/AU BONDFINGER

MOLD COMPOUND

SUBSTRATE

EXTERIOR PACKAGE

TERMINAL

Area with no vertical support

during wirebond.

Area with full

vertical

support during

bonding.

This surface is the support

during wirebonding

Fig.1 Bond pad structure and laminate of a BSOB

system with partial support at bottom.

Fig.4. von-Mises stress distribution of a BSOB

under wire bonding force 650 mN

19192006 Electronic Components and Technology Conference