posted by Bryon Moyer
The Synopsys Users’ Group scheduled a panel session on FinFETs at their recent session. This is consistent with pretty much every EDA company providing FinFET content for their users; they’re the latest hot topic, and represent a non-trivial change.
But the popularity of the topic was driven home rather starkly. They used the auditorium at the Santa Clara Convention Center, and even so, it was standing-room only, and many, many people had to be turned away when there wasn’t more room for standing. So interest was obviously quite keen. Which reflect the fact that FinFETs are in transition: once the far-out concept for some future technology node, this stuff is becoming real for lots of real engineers.
And, for the most part, the panel said the usual things you might expect about why FinFETs are necessary or good and what’s different about using them. The Synopsys speaker talked about the tools, of course, but it was still too early to see the results of the lead lots they had put through.
And it was refreshing to have Cavium on the panel as a user, with the candor to discuss what they felt was the biggest drawback about FinFETs: parasitic capacitances. As in, there are too many of them. So they were having trouble keeping dynamic power down. They requested EDA help both in optimizing dynamic power consumption and with EM analysis.
Not that they thought that FinFETs were a bad idea in general; they agreed with the litany of benefits that is typically trotted out. But their willingness to say, “They’re good but…” helped give a real-world feel to the kind of panel that can too often be simply an echo chamber for laudatory talking points.
posted by Bryon Moyer
It’s funny that, as all things silicon shrink and we look forward to alternatives, we’re getting used to hearing about carbon as a regular material – but always in the form of graphene or tubes. As diamond, it seems completely foreign. Yet it’s basically the same stuff.
Then I saw a press release about Element Six ramping up production on diamonds for use in EUV. How a diamond factors into EUV was completely non-obvious to me (well, until I had a discussion with them and then reread the release). Part of my confusion was the fact that diamonds (at least good ones) are transparent, and yet EUV systems don’t use lenses or any other transmissive optics; the EUV energy is channeled by reflection – it’s all mirrors.
So where could a diamond figure in? And why something so “expensive”? (Gonna bypass the whole market control issue here…) Those of us not involved also might think of artificial industrial diamonds as made by simulating the conditions under which natural diamonds are created: high temperature and pressure.
But there are a lot more options than that. Element Six (no, they don’t compete with DeBeers: they are a part of DeBeers) has numerous ways of making diamonds for different applications. Yes, large single-crystal pieces using the high-temp/pressure method, but also grits and purer polycrystalline diamonds.
The ones used in EUV are grown using CVD by creating a plasma out of methane and hydrogen, with the methane breaking down to release carbon. And these guys are particularly pure. Their important properties for EUV, according to Element Six, are the fact that they have the highest thermal conductivity of any room-temp solid and that they can pass a wide range of frequencies without lensing or aberrations.
So where does this fit into the EUV picture? No, not in the actual EUV delivery portion: in the CO2 laser itself for a laser-produced plasma (LPP) system. It forms the windows through which the laser travels, including the output coupler from the lasing cavity, along the transport path, and up to and including the final window in the subsystem. Other materials tend to break down around 8 KW or so; diamond can handle megawatts, so it’s the only material that can realistically be used in this application.
An increase in production would presumably reflect confidence that EUV systems will hit their stride and be produced in “volume” quantities…
And if you wish to see more about their specific announcement, you can find it in their release.
posted by Bryon Moyer
MEMS elements are delicate. They sit there in their little cavities, expecting to operate in some sort of controlled environment – perhaps a particular gas or pressure (or lack of it). And if they’re collocated with CMOS circuitry, then they need to be protected from any further processing steps. In other words, they need to be sealed off from the rest of the world. And wafer bonding is a common way to do that: bring another wafer (perhaps with etched features) face-to-face with the working wafer and get them to bond.
Covalent molecular bonds are the strongest; if you bring two silicon wafers together, for example, the ideal is to have the silicon atoms at the surface of each wafer bond covalently with their counterparts on the other wafer so that the whole thing starts to look like a continuous crystal. That’s the ideal.
Doing this isn’t trivial, of course, since the surfaces are likely to have imperfections and contaminants. So surface preparation has been an important part of the wafer bonding process. It has also involved intermediaries like water that establish a preliminary bond; an anneal then precipitates the reactions that result in the appropriate covalent bonds and out-diffusion of any extraneous elements.
Initially, high temperatures were required for the annealing. But, of course, anything over 450 °C won’t sit well with any CMOS that might be in place, so various surface preparation techniques have been devised to get the anneal temps down below that threshold.
But even these temperatures can be an issue for bonding unlike materials, or for wafers that have unlike materials in the stack, where stresses can result from differing rates of thermal expansion during the anneal process.
EVG has recently announced a new way of preparing the surface so that covalent bonding occurs immediately, at room temperature. To be clear, they have announced that they have this new process; they haven’t announced what it is; they’re still being coy on that. This eliminates the annealing step completely, and therefore the thermal expansion issue as well.
Equipment using this new technique should ship sometime this year. You can find out more in their release.