It’s common knowledge among those working feverishly to bring 3D TSVs to market that one of the areas still being ironed out in the back-end-of-line processes is thin wafer handling. Research institutes along with materials and equipment suppliers (IMEC, Léti, Sematech, SUSS MicroTec, EV Group, 3M, Brewer Science, T-MAT, Dupont all come to mind) have come up with an assortment of viable and cost effective methods for temporary bonding and debonding of a support carrier wafer. The sticking point (no pun intended) appears to lie in the debonding process, especially when working with 300mm wafers.
At IMAPS International, I heard about an alternative solution in development at Institute for Microelectronics (IME) in Singapore, in a project lead by Xiaowu Zhang, Ph.D. Working with Brewer Science’ Waferbond material and and EV Group bonder to form the temporary bond, the team experimented with the debond process by using a perforated support wafer. After backside processing, thinning to 50µm, the wafers are soaked in Waferbond remover solution at 90°C. The “release holes” in the support wafer measured 1mm in diameter and were set at a pitch of 3mm, however this didn’t allow enough of the chemical solution through to dissolve the adhesive, so breakage of the processed device wafer was still occurring. But according to Zhang’s paper, when the pitch of the holes was reduced to 2mm, along with an increased number of wafer edge perforations, the chemical solution penetrated the bonding adhesive more uniformly, allowing for successful debond of the wafers.
This caught my eye, because it’s different from the Thermoslide process developed by Brewer Science and EVG, which relies on heating the support wafer to soften the adhesive before sliding them apart. However since I merely follow these technologies and report on this stuff, I’d like to hear opinions on this from the engineers who have been working on solutions to the debond process. What do you think? Comments are encouraged!
The simplest thing to do is not bond to a support wafer. This means you have to have total support from backgrinder through dicing.
Regarding #1
Steve –
If you don’t perform a temporary bond, how do you secure the device wafer to the the support wafer? Can you expand on this comment?
Francoise
You are assuming you have to use a support wafer. Some companies are working around the support wafer problems.
Regarding #3
Steve –
It’s my understanding that for TSV processes, a support wafer is necessary to handle the wafers being thinned to <50µm. Are the companies you're referring handling wafers that thin, and if so are they using TSV as the method of interconnect?
This is not really a new technique. Brewer Science released a material called WaferBond years ago – it was actually their very first temporary bonding material, and was specifically planned as a solvent release system (as opposed to the materials used for the “Thermoslide” process you mention, which are the WaferBond HT family of materials, including HT250 and HT10.10). Solvent release temporary bonding has existed since at least 2001 – my first experiences with it were with the General Chemical “GenTak” materials. As mentioned in the post, perforated carriers are a requirement, and this brings a few negatives, as well.
The three greatest percieved issues were:
A) The lack of support for the wafer over the perforations would frequently lead to an inverted “dimple” profile on the wafer – pressure from the grind/polish wheel would press the wafers down into the perforation and create a slightly thicker profile.
B) Care and handling of the freed thin wafer in the liquid solution bath. It is difficult enough to securely handle a 50 um thick wafer in air, where turbulent flow is only achieved at high flow rates. In liquid, turbulent flow is almost guaranteed, even just from the action of removing a boat from a bath, which puts the wafers at severe risk.
C) Handling of the perforated carrier is not trivial. Having a pattern of 1mm perforations with only 1mm of solid material between them creates a substrate that makes Swiss cheese look robust – the carrier itself is very fragile, and typically needs to be made from an amourphous material like glass – with crystalline materials like silicon, your carrier becomes about as fragile as the thin wafer it is trying to support. Additionally, a lot of systems may throw fits attempting to deal with a wafer that highly perforated. Any vacuum chucking system would put a stress on the 50um “membrane” at the top of the 1mm hole (although silicon is pretty strong).
Even in the 2001 time frame, there was some movement away from the solvent release materials, initially with the use of thermal release tapes. The tapes allowed bonding and debonding using a solid carrier – typically just a plain, mechanical silicon wafer. Their primary limitation was maximum processing temperature (150’C), which led to the currently used slide-off adhesives.
Chad –
TOK presented this process last week at the RTI Conference. I’m told by one source that it is what has been used at TSMC, which might explain why TSMC’s Douglas Chen-Hua Yu noted the following challenges with the temporary bonding/debonding:
“Debonding induces damage. The industry
doesn’t have a suitable solution yet.
Especially when we want to debond, damage to the chip such as glue
leaving a residue, missing bumps, damage to interconnect layers, bubble
residue, out-gassing, thermal stability
and yield impact. He said tool suppliers need to address these issues.