[labnetwork] More on wafer bonding techniques

[Robert M. HAMILTON]

Labnetwork Colleagues,
I am forwarding comments from one of the Marvell NanoLab process engineers
versed in temporary wafer bonds, Ryan Rivers. I have confidence in his
information and hope it is a useful addition to the current dialogue.
Bob Hamilton


Robert Hamilton
University of CA, Berkeley
Marvell NanoLab Equipment Manager
Rm 520 Sutardja Dai Hall, MC 1754
Berkeley, CA 94720
Phone 510-809-8618 (desk - preferred)
Mobile 510-325-7557 (my personal mobile)
E-mail preferred: bob at eecs.berkeley.edu
http://nanolab.berkeley.edu/



---------- Forwarded message ----------
From: Ryan Rivers <rdrivers at eecs.berkeley.edu>
Date: Wed, Aug 19, 2015 at 2:22 PM
Subject: Labnetwork Response
To: Bob Hamilton <bob at eecs.berkeley.edu>


Hi Bob,

Here's my take on the labnetwork thread:

Hi Savitha,

I wrote up a fairly comprehensive wafer bonding tutorial for the UC
Berkeley Marvell Nanolab. It is available at:

http://nanolab.berkeley.edu/process_manual/chap2/2.16waferbonding.pdf

This document covers all of our wafer bonding methods. We have experimented
with Fomblin and found Santovac V to be superior in all processing
respects, as well as available at significantly reduced cost. As a general
purpose bonding agent, I tend towards Cool Grease and Santovac V. The phase
changes experienced by crystalbond during DRIE can lead to wafer bow from
significant stresses at the thermal interface. This tends to increase
helium leak rates leading to dechucked wafers and burnt masks. I would
avoid thermal release tape for DRIE as it does not hold up well to the
interface temperatures of the etch. Photoresist is best applied when
interface temperature contact is not a consideration (I.E. Rough cut deep
etching). It holds up very well, but no heat transfer of note is going to
occur.

The single most important aspect of wafer bonding is simply *practice*. You
need to practice the bond and make sure there are no air bubbles. No matter
what material you use, an interface with air bubbles is not only bad for
heat transfer, but a wafer damage risk in a DRIE chamber. Those bubbles are
at 760 Torr, you're at ~0.010-0.035 Torr in that chamber. A full 400-600
micron thick wafer will hold that back, but a thin membrane from a deep
etch is not going to succeed, and the bubbles will burst through your
channel and usually destroy your device.

Long story short - I recommend Santovac 5. It's just plain easier.

Thanks,
Ryan Rivers
R&D Engineer 3 - Process Staff
UC Berkeley Marvell Nanolab

-- 
Ryan Rivers
R&D Engineer 3 - Process Staff
UC Berkeley Marvell Nanolab
510-809-8627
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mtl.mit.edu/pipermail/labnetwork/attachments/20150819/6b21b7b9/attachment.html>