When it comes to computer chips, smaller is better. Manufacturers are continually trying to shrink their size and pack as many transistors as possible onto the same chip.
More power, less space is the mantra – Moore’s Law suggests that the number of transistors on a chip doubles every two years.
But that creates its own complications – in particular heat generation and how to control it.
As chips get hotter they create more errors in their processing, and protective slow-downs start to occur. So cooling systems are designed; but usually electronic technologies, designed by electrical engineers, and cooling systems, designed by mechanical engineers, have been done independently and separately.
Now, like a warlord, a team from Switzerland’s EPFL has united both factions and created a microchip that has its own embedded cooling systems . The details have been published in the journal Nature.
The work, led by Remco Van Erp, involved etching microfluidic channels into the structure of the semiconductor chip itself, turning the silicon from a low-cost carrier into a high-performance heat sink. These internal channels allowed the team to target hotspots within the chips, letting them deliver cooling where it was needed most.
The team fabricated chips with integrated channels measuring from 100 micrometres to just 20 micrometres wide. These microchannels are fed by larger channels closer to the surface. Cold coolant enters from one end of the chip, extracts the heat, and exits the other end.
The power required to pump the fluid through the channels is very low, while extraction of heat was very high, the researchers say.
“We placed microfluidic channels very close to the transistor’s hot spots, with a straightforward and integrated fabrication process, so that we could extract the heat in exactly the right place and prevent it from spreading throughout the device,” says Elison Matioli, who oversaw the research.
In an editorial, Stanford University’s Tiwei Wei writes that, because such integration greatly increases the complexity of chip fabrication, it could potentially increase their cost. However, the EPFL team found a way to manufacture the microfluid channels cheaply and simply.
Narrow slits are etched into a silicon substrate coated with a layer of the semiconductor gallium nitride, with the slits then enlarged and connected. Any openings in the gallium nitride layer are capped with copper. The gallium-silicon layer, with channels already in place, is then ready to be used to fabricate an electronic chip.
With the cooling channels embedded in the chip itself, there is no need for bonding or interfaces between the electronic chip and a cooling layer or manifold, which helps increase its efficiency, say the researchers.
The cooling liquid they used was deionized water, which doesn’t conduct electricity.
“We chose this liquid for our experiments, but we’re already testing other, more effective liquids so that we can extract even more heat out of the transistor,” says Van Erp.
That heat-extraction efficiency and low power requirements should slash energy costs for cooling large-scale computing. Data centres in the United States alone use 24 terawatt-hours of electricity and 100 billion litres of water for cooling every year. That’s roughly equivalent to the needs of a city the size of Adelaide.
The approach taken by Van Erp has the potential to drop power requirements for cooling from 30% of a data centre’s power, to 0.01%, he says.
“We’ve eliminated the need for large external heat sinks and shown that it’s possible to create ultra-compact power converters in a single chip. This will prove useful as society becomes increasingly reliant on electronics,’ adds Matioli.
“[This] work is a big step towards low-cost, ultra-compact and energy efficient cooling systems for power electronics,” writes Wei.