The Green Monster Stirs

The only thing limiting the crowd was the fact that it was an evening session. OK, a Sunday evening session. Um… ok, SuperBowl Sunday, actually. So...  all things considered, 75 people showing up for a panel on green design on Superbowl Sunday evening is pretty impressive. They even delayed the start for about five minutes to accommodate the dramatic SuperBowl finish only fifteen minutes earlier. Talk about yer hard act to follow… And so, performance anxiety thus having been injected, the ISSCC panel, “Green Electronics: Environmental Impacts, Power, E-Waste,” organized by AMI’s Jan Sevenhans and chaired by Infineon’s Rudolf Koch, got underway.

The panel consisted of a member of Greenpeace to set the context, followed by two representatives of the semiconductor manufacturing and packaging industries. The next obvious industry for a semiconductor audience was solar, followed by a couple presentations on green automotive.

Martin Hojsik spoke for Greenpeace. Now, if you’re like many folks, the view you have of Greenpeace probably comes from news footage, with angry protesters shouting into industry’s face and using other aggressive, provocative, even pugnacious tactics. Zodiaks circling fishing boats; that sort of thing. So it was refreshing to see Mr. Hojsik in concert with the other industry representatives; he wasn’t there to bait them, he wasn’t there for the sake of dramatic polemics; he was the voice of the outsider, the self-admitted non-expert. He was the you and I that look at some of these problems and want to say, “This has to be fixed. I don’t know how to do it, and maybe it’s not easy, but we have to do something.” And not least, then adding, “And I’ll work with you to help find and, more importantly, propagate the solutions.”

Most of the press’s green focus these days is on the climate. Newscasters will try to tie everything from a heavy snowstorm to muggings in Central Park to global warming. A key point made in this session was that there is more to the picture than just carbon emissions, important though they be. The three issues raised by Mr. Hojsik were materials – both resource depletion and waste; product lifecycle and rapid obsolescence; and energy expenditure in both the manufacture and use of electronic equipment. According to Mr. Hojsik, in making and using electronics, we donate a non-trivial 2% contribution to greenhouse gas emissions. But there are also such naïve but obvious questions as, if laptops can be efficient, why can’t desktops? Can we rein in the tendency of software makers – some rather notoriously – to assume that people will blithely buy new computers every couple years to accommodate bloatware that is inflating at post-Big-Bang levels? And the attendant shipment of the old computers into some unseemly heap in a country like China, India, Viet Nam, or Pakistan? We’ve gotten rid of lead; now what about brominated flame retardants (BFRs) and perflourocarbons (PFCs)?

Prof. Duane Boning of MIT took the semiconductor processing topic. He characterized his three main concerns as water, energy, and effluents. His main take-away was that changes have to be considered early on in the R&D process. Once a process is up in a fab, no one wants to touch it, so you have to build green considerations in before a process goes into production. The good news is that, because new process generations come around so frequently, we have many opportunities to bring change.

He provided an interesting example of typical considerations and tradeoffs by showing a three-dimensional IC problem. In a 3-D IC, multiple chips are stacked and interconnected to do more stuff or hold more data in the same footprint as a single chip. The problem is, how do you attach the wafer layers to each other as they stack? Wafers normally have inactive silicon – the bulk layer or substrate – that’s much thicker than the actual useful layers of the wafer, which is fine, because it makes the wafer strong enough to be handled. But if you’re stacking wafers, the upper wafers can’t be as thick as the bottom one. If you simply grind each of the upper wafers before attaching to the base wafer, they become too thin and fragile to handle. So the problem to be solved is, how can you affix a thin target wafer as a second layer over a base wafer?

The incumbent approach is to use a sacrificial “handle wafer” and bond it to the top of the target wafer using a layer of aluminum. The combined handle/target wafer can be handled, ground, and stuck to the base wafer, with the handle wafer providing structure. After the target wafer is securely bonded to the base wafer, the handle wafer is removed by etching away the aluminum layer. The problem, from a materials, energy, and effluent standpoint, is that the layer of aluminum has to be very thick so that the etchant can penetrate and dissolve it all the way to the center of the wafer. This means a 20-µm layer –  forty times thicker than the normal 0.5-µm layer used for simple metallization. This step adds 340 kWh of energy to the 540 kWh required to create the entire base wafer – an energy input increase of 63% just to bond the second wafer to the first. It also increases the water usage by over 50%. And, of course, it uses and then discards a lot of aluminum.

He then took a brief look at three alternatives, including a “Smart Cut,” where hydrogen is implanted in the handle wafer for thermal release, creating channels between the wafers (including channels in the active wafer) to allow better migration of the chemical etchants, and an “oxide release layer” that also uses channels cut more extensively into the handle wafer. Now if you’re expecting that through this process a white knight rode in and saved the day, you will be disappointed, since in the real world, there are rarely easy answers or obvious solutions. In fact, the Smart Cut approach had the best environmental impact but was far too expensive, so the oxide release layer proved the best combination, having improved environmental characteristics and being inexpensive.

The main point of the example was that during the development of a new technology, environmental impact can be considered alongside performance and cost, and the sum of such tiny decisions can add up to continued reduction in the negative effects of building a chip. But these kinds of improvements could never be made on a process already being used.

Packaging was the topic taken on by Amkor’s Gary Hamming. He noted some barriers to green adoption, among them cost (of course), reluctance to requalify old devices with new package materials, regulation exemptions or delays for a few industries, and reliability concerns. He further makes the point that to date there isn’t an industry-wide definition of what constitutes a “green” package. So he used Amkor’s definitions to illustrate progress, which came at three levels.

The most basic level is lead-free, which has more or less been achieved, barring the exemptions. The next level comes from the Reduction of Hazardous Substances rules (RoHS, pronounced like “Rojas” in Spanish). RoHS adds limits to the amounts of mercury, cadmium, a chromium ion, and a couple specific organic compounds (polybrominated biphenyl and polybrominated diphenyl ether… yum…). “Fully green” is the third level, and it further limits the amounts of bromine, chlorine, and antimony. They also toss in the ability to reflow at 260°C. Amkor sees itself making continued progress, with 90% green molding compounds this year, leadframes making good progress towards green, and the only barrier being halogen-free substrates in laminate packages due to some specific technical concerns that need to be sorted through.

The remaining presentations were on green aspects of making photovoltaics (Robert Mertens of IMEC) and automotive applications (Marcos Laraia of AMI and Justin Ward of Toyota). Much of the subject matter of these discussions was outside the scope of semiconductors, but the common underlying theme was that increasingly sophisticated inversion, sensing, and control circuitry is required for optimizing the operation of systems, be they solar generation or automotive.

One subtext to many of the discussions was the fact that much, if not all, of the progress made so far has come about because companies have been forced by government regulations to make changes. It’s easy to assume that regulations are viewed as evil by industry, but a large part of that view comes courtesy of an auto industry that seems to have a knee-jerk reaction to any regulation and would probably spend millions of dollars vigorously opposing any attempt to legislate that wheels must be round. In fact, even Greenpeace’s Mr. Hojsik opined that a number of companies are actually getting ahead of regulations (and others spoke up to note that Toyota was an exception to the anti-green-regulation efforts by the auto industry – although it should also be noted that Toyota joined the Detroit 3 in opposing the new CAFE standards). But rather than everyone going it alone, regulations set a level playing field: if there’s not an obvious payoff within the quarter to going green, some companies might try to get short-term advantage by lagging – if not outright stalling– their efforts to go green. Regulations prevent this, and in fact a number of companies are actually asking for regulations. Regulations also help establish common targets; it’s easier to achieve a goal if you know what the goal is. All in all, regulations seem to have won a surprising level of acceptance.

Back in the world of IC design, all of this combines with much higher scrutiny on reductions in power consumption, so it’s fair to expect that this overall topic will creep into more of our working lives as green becomes yet another domain that must be optimized, alongside those that we already have.

by Bryon Moyer , IC Design and Verification Journal

February 12, 2008

[back to top]