Boffins melt defects from chips
Nano state
BOFFINS AT PRINCETON UNIVERSITY have come up with an innovative way of melting away microchip defects, in a move which they reckon will drastically improve chip quality, without raising costs.
As companies strive to make smaller and smaller chips, miniscule defects are becoming bigger and bigger problems. Rough edges, breaks in lines, and other tiny flaws can all damage chip performance, causing voltage leaks, interference and worse. But now electrical engineer Stephen Chou, a Princeton Professor of Engineering, along with his graduate student, Qiangfei Xia, think they have come up with something to perfect chip making technique. Because microchips perform best when all their structures are straight, thin and tall to exact proportion, Chou and Xia knew that they had to discover a better and more precise way for companies to shape the microchip components.
"What we propose is a paradigm shift: Rather than struggle to improve fabrication methods, we could simply fix the defects after fabrication," said Chou, who’s findings were published yesterday in Nature Nanotechnology.
Chou’s discovery, which is dubbed Self-Perfection by Liquefaction (SPEL), works by melting all the structures on a chip for a tiny fraction of a second, in order to take advantage of surface tension, which lets the molten liquid re-solidify in more precise, and geometrically accurate shapes.
"We are able to achieve a precision and improvement far beyond what was previously thought achievable," gushed Chou, who added that he believed the process of fixing the defects could even be made automatic, or “a process of self-perfection”, as he puts it.
Currently, chip flaws are fixed by having every single defect painstakingly measured for size and shape, in order for an expensive tailor made repair to be hacked together. But the process is a slow one, as well as being expensive, which makes Chou’s melting system a very attractive alternative indeed.
In order to melt only the top layer of the chip, and not damage the underlying structure, Chou and his team used a light pulse from an excimer laser, which is quite similar to the lasers used in corrective eye surgery. The light pulse melts down the semiconductor and metal structures for a fraction of a millionth of a second, letting the molten metal flow and change shape in an instant.
The method is perfected by placing a tiny plate over the melting chip surface, which helps the liquid flow in the right way, and stops the whole structure from expanding or warping in shape. Another advantage of the guiding plate is that, when it is placed a tiny bit above the molten structure, the melted metal rises up to touch the plate, which makes any line structures on the chip taller and narrower, which is excellent in terms of chip design. The semiconductor factories are already salivating at the prospect.
Chou notes that "regardless of the shape of each defect, it always gets fixed precisely and with no need for individual shape measurement or tailored correction".
Chou says that he and his team will now work towards demonstrating the new melting technique on big (8-inch) wafers. µ
L’Inq
Nanotechwire

Comments
cool
Good for them, i hope they get compensated well for this discovery. Smart guys tooYeah?
These guys really have no clue about the wide variety of defects that kill die at the fab. There idea sounds great from the aspect of a science fiction novel, but in reality it wont work. There are TOO many different mechansims that can't be corrected by this technique, in fact it will probably cause more problems..sounds clever
Does it work on strained silicon as well?Credit goes to Professors
One mustn't be a Genius to find that.We all knew, that melting the chips for a short time can increase their quality and overclockability.
We did this since the early K6 II days. Overclockers used to "burn in" their Chips. Very slow but continuous heating up to the maximum allowed Temperature. The most important is the sloe heating and the slow falling of the Temperature.
Thats comparable with what u do to plastics after u milled something in it, u temper it that it looses stresses inside.
It worked wuite well with the Athlons and gave aproxximately a 10% increase in maximum clockspeed, while it didn't help a lot with Voltage. U still needed more Voltage for the higher clock, but hey, that was ok, because most cpus at that time had under 50Watts even overclocked and who cared about the electricity bill that time?
Why is it always professors who don't know that what they just found out was common knowledge with regular geeks? This guy will probably earn a lot of money with technology that every hacker used over 10years ago at home to overclock his K6 II oder Pentium Chip even higher.
Steal a Technology from the people, maybe alter it a little and allow it for massproduction, - then tout it as a groundbreaking Innovation!
Thats what AMD/ATI did with Overdrive, thats what Intel had in mind with the new Voltage regulation for Cores (in fact it were geeks who undervoltet their processors to funny voltage number like 1.175, while intel used only numbers like 1.35).
But most importantly:
Make money and get famous for inventing something!
Mr. Silicon
Mr. Silicon is obviously born of the crowd that said in the early sixties "They will never put a man on the moon the complications and engineering difficulties are far to great to overcome".Rather amusing to read comments like those of Mr. Silicon, "never" or "can't" are words you should rarely use in the technology industry, unless of course your primary skill is talking out of your *rse, in which case use them all you like...
RE:
So can anyone comment on how you can correct a tungsten contact short by heating up the wafer? How about an extra poly pattern creating a parasitic transistor? Well you cant, since tungsten melts at a really high value. Secondly in the low temp layers like the copper backend wiring, melting this layer wouldn't cause the defect site to "heal". The copper wiring on an ic is plated in a wet process, hence it has already "flowed" into it's final shape and then polished. If you had a scratch and created a short between two adjacent lines, reflowing the copper would still result in the scratch being filled with copper and shorting the two lines!Lastly, the failure mechanisms which can be fixed by heat (i.e silicon dislocations) are generally identified and fixed during process development. It's a well understood phenomenon that most IC makers use in that realm. It's not an uncommon practice to bake out wafers which have implant issues (thus dislocations).
Sorry, but it's my job and career to understand failures on wafers, and I haven't seen anything compelling here to believe the end is near.
Overclockers
Overclockers burned in CPUs in the old days to cure the thermal contact between the CPU and the heatsink.In the old days most if not all heatsink are bonded via thermal patches, and those requires burn in time to melt, spread and thinned. Running CPU at high temp would make the process faster. It also filters out those that can't run stably at high heat.
Thermal grease largely nolonger requires high temp burn in period, but there are still manufacturer advising burn in to reach optimal heat transfer (Arctic Silver asked for 200hrs burn in time). The difference can sometimes be a few degrees.
Silicon melts at 1000+ C and not something geeks can perform own their own anyway. Thus the process described isn't the same burn ins every day geeks perform.
Mr. Silicon
Well, from my experience at the coal face, was that thermal insult was not only very effective in remedying defects, it could also be used to over come a new "tape out" redesign requirement of the chip, by allowing the device design to be modified in the metal layers. Effectively saving weeks of the redesign cycle, and that was being achieved over ten years ago.I see great potential for the application of this technology. Your inference that they have no clue is born of ignorance of what they are actually doing, and is almost certainly an emotional response, born from a either a blinkered or conventional view on what actually can be achieved.
Sure it may seem impossible with the knowledge and experience you possess, but given a big enough technology gap, things once society take for granted would seem to be magic to another less advanced society.
If you did know exactly what they were doing and how it was being done, you wouldn't be diss'ing it here, and you would be on the payroll instead.