Previous Page  31 / 52 Next Page
Information
Show Menu
Previous Page 31 / 52 Next Page
Page Background

29

Chip Scale Review November • December • 2017

[ChipScaleReview.com]

Core capabilities of thermocompression bonding

By Hugo Pristauz, Alastair Attard, Andreas Mayr

[Besi Austria GmbH]

Part 1 of this article was published March April 2017 (

http://fbs.advantageinc.com/chipscale/mar-apr_2017/#16)

and focused on an

overview of different thermocompression approaches and the technical challenges that can be solved by this technology. Understanding

the crucial requirements led us to four core capabilities of a thermocompression bonder, which are discussed in detail in this second part.

o review the points made in part

1, the four core capabilities of

a thermocompression bonder

(TCB) are: 1) accuracy, 2) coplanarity, 3)

bond control, and 4) temperature uniformity.

Accuracy

For a better understanding of the error

chain of placement accuracy, the following

pragmatic approach, illustrated in

Figure 1

,

is useful. It starts with a listing of independent

challenges to be mastered for achieving the

accuracy targets. In this example, a target is set

at 2µm@3σ for a TC bonding process at the

chip-to-wafer (C2W) level with matrix-based

bond locations, including temperature ramps,

with considerable bond force.

These challenges are listed in the right-most

(5

th

) column in

Figure 1

, which stands for

target process requirements. Subsequently, a

sequence of simplified reference processes is

defined wherein one challenge is eliminated in

each step. Because the accuracy target should

be easier to achieve after each elimination step,

a more challenging accuracy specification

can be claimed until the simplest reference

application is reached on the left-most side (1

st

column) with the most challenging accuracy

specification.

Following the approach described above,

the first simplification is to move from the

real product die application (5

th

column) to a

glass die on glass substrate placement (GoG

– 4

th

column), which eliminates material

variations and vision errors, thereby giving

more precise post-bond metrology. As a kind

of “compensation” for this simplification,

the accuracy requirement is enhanced to

2µm@4σ. In a similar manner (3

rd

column),

by moving from high (e.g., 250N) to low

bond force (e.g., 30N), the accuracy demand

is increased to 2µm@5σ. This scheme can be

continued by transitioning from hot to cold

process (2

nd

column), demanding 2µm@6σ,

and finally ending at a so-called basic machine

capability test (BMC), repeating single-

position glass die placements with low bond

force without utilization of thermal profiles (1

st

column). Our demand for BMC placement

accuracy is 2µm@7σ, which is equivalent to

0.86µm@3σ.

Having split up the accuracy chain in

this way, one can cross-check whether the

formulated requirements can actually be

met (

Figure 2

). For the reference process

of column 4, the specification of 2µm@4σ

can be slightly out-performed (the actual

accuracy reached is 2µm@5σ), but for the

easy BMC test with a 2µm@7σ specification,

we can achieve significant out-performance

of 2µm@11σ, which relates to 0.55µm@3σ.

Obviously, some of the reference accuracy

processes are easier to achieve, while others

come with significant challenges.

Exercising the kinds of studies described

above extensively, most of the attention is

ultimately paid to one particular step: moving

from a cold (or constant temperature process)

to a temperature ramping process because it

brings about most of the placement inaccuracy.

In this case, it is possible to encounter the

abnormal accuracy behavior shown in

Figure

3

. In this example, placement accuracy is

close to staying within specification for half

of the observation time, but then the process

suddenly runs out of control in one direction.

It is interesting to understand the root cause of

such behavior. Because of the coefficient of

thermal expansion (CTE) mismatch between

the tool holder (bright green part) and the

nozzle (yellow part), it can happen that the

nozzle makes relative

movements during

t he t empe r a t u r e

ramps, which can

e x c e e d r a t e s o f

300°C/s.

B e c a u s e T C

bonders typically use

a method to align a

die on a bond head

nozzle relative to

the substrate in a

cold or moderate

temperature level,

any misalignment

of the die relative

to the bond head

during temperature

ramping has to be

avoided. This is not

T

Figure 2:

Accuracy performance results.

Figure 1:

Breaking up the accuracy error chain.