editor's blog
Subscribe Now

Dry Etch Edges Wet

Two announcements have come out recently regarding dry etch systems. Now… dry etch is nothing new. Although it is newer than wet etch, which is still being used. And, as they say, therein lies the rub.

The first announcement came late January regarding a new system shipping from Memsstar. Their focus was on MEMS, and, in particular, on reducing yield failures due to stiction. There are two pieces to this move.

First, they note that wet etch processes must be followed by a wash to clean out all of the etchant and resulting groddy bits. This involves water, and it’s very easy for bits of water to get “caught” in small spaces due to capillary action. It’s generally not a good idea to have residues like that any place in the circuit, but if it happens to be touching a moving MEMS element, then it may cause the element to bind (or, at the very least, not move with the same force/acceleration relationships as were designed for). This is a form of stiction.

Dry etch processes do not involve a wash, and so simply moving away from wet etch might seem sufficient. But in particular with an HF dry etch, one of the products of the reaction is water. It’s normally in gaseous form, where it won’t cause stiction issues. But, according to Memsstar, if the reaction isn’t properly controlled, you can get excess water, some of which may then liquefy, in which case you’re back to dealing with stiction.

So part two of this move is their claim of having the only system with the process monitoring that can ensure this doesn’t happen. So you go from a two-step wet etch process (etch, wash) with stiction to a one-step dry etch process without stiction.

The other announcement was more recent; it was from Nanoplas. They actually got their start with a MEMS-oriented “high-density radical flux” (HDRF) etcher that, due to a source that provided a concentration of species 1000 times higher than a typical source, they could be much more efficient when cleaning out residues from irregular MEMS configurations, especially cavities and high-aspect-ratio features.

But their latest announcement is for what they call “atomic layer downstream etching,” or ALDE. And they can’t really say a lot about the details now due to patent filing issues. But their current focus for it is CMOS and the problems being encountered at 28, 20, and 14-nm nodes using traditional phosphoric acid wet etching.

Etching normally involves a measure of selectivity. You want to etch things you’re trying to remove without touching anything else. Unfortunately, nothing is perfect, so you always get some impact on the stuff you’re trying to keep; a measure of the selectivity is the ratio of how much you etch what you’re trying to remove vs. how much you etch what you’re trying to keep. Rates these days can be 10:1 or even 50:1.

But, according to Nanoplas, from a process control standpoint, etch rate and selectivity have been tightly bound together. You can be more selective if you etch slower, so it’s hard to optimize for quality and throughput. Their claim is again about control: they say that, with their new source, they can independently control selectivity and etch rate.

Typically, there’s one thing you’re trying to etch, and everything else is what shouldn’t be etched. Which means there may be more than one material that needs preserving. But even if it’s just one (and Nanoplas says that it’s typically one or two), it could be oxide or silicon or resist or nitride or poly or metal or… you get the picture.

But they say that the control they provide allows the etch to be optimized for any of them. Right now, they’re focusing on SiN etch for a specific client that’s dissatisfied with the performance of traditional wet-etch approaches for spacer technology. But, going forward, they’ll be able to optimize for any other etches as well. Oxide and poly will be their next areas of work.

They won’t actually be selling a complete system, but rather a process module that OEMs can integrate into their platforms.

At such time as I can get more info on how the Nanoplas system works, I’ll follow up with those details.

 

You can find out more about the Memsstar system in their release; likewise, there’s some more detail on the Nanoplas announcement in their release.

Leave a Reply

featured blogs
Nov 12, 2024
The release of Matter 1.4 brings feature updates like long idle time, Matter-certified HRAP devices, improved ecosystem support, and new Matter device types....
Nov 7, 2024
I don't know about you, but I would LOVE to build one of those rock, paper, scissors-playing robots....

featured video

Introducing FPGAi – Innovations Unlocked by AI-enabled FPGAs

Sponsored by Intel

Altera Innovators Day presentation by Ilya Ganusov showing the advantages of FPGAs for implementing AI-based Systems. See additional videos on AI and other Altera Innovators Day in Altera’s YouTube channel playlists.

Learn more about FPGAs for Artificial Intelligence here

featured paper

Quantized Neural Networks for FPGA Inference

Sponsored by Intel

Implementing a low precision network in FPGA hardware for efficient inferencing provides numerous advantages when it comes to meeting demanding specifications. The increased flexibility allows optimization of throughput, overall power consumption, resource usage, device size, TOPs/watt, and deterministic latency. These are important benefits where scaling and efficiency are inherent requirements of the application.

Click to read more

featured chalk talk

Accelerating Tapeouts with Synopsys Cloud and AI
Sponsored by Synopsys
In this episode of Chalk Talk, Amelia Dalton and Vikram Bhatia from Synopsys explore how you can accelerate your next tapeout with Synopsys Cloud and AI. They also discuss new enhancements and customer use cases that leverage AI with hybrid cloud deployment scenarios, and how this platform can help CAD managers and engineers reduce licensing overheads and seamlessly run complex EDA design flows through Synopsys Cloud.
Jul 8, 2024
33,764 views