|
Based on the experience acquired with this previous transfer technique,
we would like to expand its power further by exploring the possibility
of using it to distribute devices from one wafer to many wafers
using wafer-level selective device transfer.
Hence, while preserving the advantage of a high level of integration,
it has the potential to be competitive in terms of costs with the
more traditional flip-chip die-to-die integration. Flip-chip bonding
is in general more cost-effective, because it allows the dies to
be tested before being joined, and hence only good dies are joined.
This has a large advantage in terms of yield compared with wafer-level
integration. It also offers greater flexibility because the die
sizes do not need to be the same and because of its suitability
for small-volume production.
The cost inefficiency of wafer-level transfer is dramatic when
the technique is used to transfer a small device on a large receiving
die because, for wafer-level transfer, the pitches of both devices
should be the same and, on the wafer of the smaller device, an artificial
large pitch between devices equal to the pitch of the large device
needs to be implemented (poor filling factor). This is why most
of the wafer-level bonding techniques are developed for use when
the two devices to be integrated are of similar size (e.g. a large
array of transducers with its CMOS electronics, the capping structure
for MEMS device protection).
The aim of the wafer-level selective device transfer technology
is to be able to distribute microdevices from one wafer to many
(as illustrated in Fig. 1) and hence allowing cost-effective
wafer-level device integration comparable to flip-chip technology,
but with a higher degree of integration and a higher performance
level.
To achieve this, the idea is to exploit the laser ablation flood
releasing technique, that we have developed, for local ablation.
Flood releasing, as perform up to now, is done by scanning a laser
with a relatively big beam size (< 1 cm²) across the wafer.
A large beam size is desirable because it increases the throughput
of the releasing. However, this beam can be focused to a much smaller
size, and the laser can be programmed to ablate only zones in which
we would like the devices to be transferred. The devices that are
not transferred stay on the glass wafer and can be aligned to another
wafer in order to transfer another set of devices. This procedure
is illustrated in Fig. 1. With this technique a maximum of
flexibility is kept in terms of the device size, approaching that
of flip-chip bonding, while keeping the integration rate and the
performance gain of wafer-level device transfer.
|
