Project Implementation
The low-cost flip chip technology and redistribution will be developed
at HKUST mostly in the Microelectronic Fabrication Facility (MFF). Dissemination and
technology transfer will be on-going as engineers from companies will be invited to
work on-site. Flip-chip modules will be designed, fabricated, redistributed, bumped,
bonded and tested at the request of companies. Below are the main steps to implement
this project.
Design
The project team members will work with local companies to establish
test chip design layouts. The designs will be drawn using the Cadence
chip design program.
For companies with wafers from their customers, redistribution is necessary
but not a requirement depending on the bond pad pitch. We will design
the redistribution routing for the bump pads. Currently, almost 100%
of wafers produced in the world are intented for wirebonding. The pads
are usually placed in close proximity around the periphery of the die.
Redistribution will allow bumps to be arranged uniformly in a grid array
or peripherally. The end result is wider pitches reducing the potential
of bump-to-bump bridging later on during solder reflow. Several metal
and photoresist masks must be designed for this process. Therefore,
redistribution to "fan-in" the pads such as shown in the following schematic
is needed for bump deposition on most wafers. It should be noted that
not all wafers need to be (or can be) redistributed.
 |
Bond Pads |
 |
Bump Pads |
| Redistribution for wafer level packaging |
Fabrication
Companies can select and approve their designs for us to build the test chips in the
Microelectronics Fabrication Facility (MFF). The fabrication of test chips involves
semiconductor fabrication steps to deposit and make patterns of very small feature
sizes on the wafer. Engineering resources will be focused here so that wafer yields are high.
Custom-designed masks can be directly fabricated from the Cadence Design files.
These masks will be used to transfer patterns to the insulator or the conductor
thin films on the wafer. Patterning of insulator or conductor films on wafers uses
photolithography technology. Basically, a photosensitive film (photoresist) is
spun on the wafer to a very uniform thin layer prior to soft bake. A mask is aligned
over this wafer which is then exposed to light which develops the exposed area on the wafer.
Areas covered by the opaque portion of the mask are not developed and can then be removed by
a solvent. The exposed underlying insulator or conductor can be etched either by wet acid
etch or dry etch revealing the desired pattern. The remaining photoresist is removed and
this process is repeated for each layer of the wafer. For wafer redistribution, the
photosensitive material will be left on the wafer as part of the building block.
Such materials and their processes will be developed and optimised by the group as part
of this project.
Under Bump Metallisation
The aluminum pads for the bumps must be pretreated with underbump metal
(UBM). HKUST has been using a metal sputtering process, which is not
low cost, due it is low throughput and high machine cost. A non-sputter
process that will be developed as part of this project will use electroless
plating solution supplied by one of our sponsors. The electroless chemical
baths must be well understood and controlled to avoid uneven or non
plating.
Bumping
Stenciling solder for fine pitch or high density interconnect on wafers
is a challenging new technology that will be developed as part of this
project. Feature sizes of 0.10 mm have to be stenciled on, requiring
extremely thin specially made stencils with 0.08mm thickness. The stencil
printing technique allows flexibility in changing from lead solder to
lead-free solder by a change in solder paste material. For each type
of solder paste, the process will be optimized to obtain uniform bump
size with high strength. The potential solders to be evaluated are:
| Standard solder: |
Sn/Pb-63%/37% (183 oC melting point) |
| Lead-free solder: |
Sn/Bi/Cu - 90.5/8.5/0.5 (198 oC)
Sn/Ag - 96.5/3.5 (221oC)
Sn/Cu -99/1 (227 oC)
|
Chip Attach
During this part of the project, companies can have the bumped chip
bonded or attached to their product (glass displays, PCB, flex circuits,
etc.). Bonding or chip attach is quite critical and the machine doing
this requires high placement accuracy. Misalignment and cold solder
joints can be observed under the X-ray system. Rework at this point
is still possible. After attach, a liquid underfill material will be
dispensed into the 0.05mm gap between the chip and the substrate. The
underfill material improves the strength of the flip-chip-on-board.
However, this process may be the bottleneck in production due to the
extended cure times, which limit the throughput. Snap cure underfills
and no underfill systems will be evaluated for reliability as part of
this project. Optimization of the underfilling process is critical to
the reliability of the end-product. Voids, filler settling and delamination
can be observed under a scanning acoustic microscope as part of the
inspection process. These defects must be minimized for good reliability.
Reworkability after underfilling becomes much more complex but is still
possible and will be evaluated as part of this study.
Reliability
Prototypes will be submitted for reliability testing, mechanical testing or
electrical testing. This can be done either at the user¡¦s location, or at
any an institution such as HKPC. HKUST will assist in the product evaluation
and reliability testing of the flip-chip on the applied substrate. Electrical
testing includes monitoring of ohmic degradation during humidity testing or
thermal cycling. Industrial test standards as established by IPC or JEDEC
will be followed.
Prototypes with lead-free solders having higher melting points will be a
reliability concern due to the higher reflow temperatures. Issues such as
moisture induced delamination and pop-corning will be more severe for
lead-free flip-chip packages due to the higher temperature reflow profiles.
Team members from another Industry Department funded project at HKUST on
delamination prediction can provide support.
The next stage of this project will be to sustain a supply of test chip
systems to local companies and continuous dissemination of know-how to local industries.