The presentation plotted how chip temperatures had changed with clock speed and extrapolated that relationship. The conclusion was that if things continued as they were, it wouldn’t be long before chip temperatures were hotter than the surface of the Sun.
That was obviously something that couldn’t happen, so the industry – and Intel in particular – took the decision to stop pursing ever faster clock speeds in favour of distributing the processing load across multiple cores.
Yet the problem of cooling processors – and other electronic components – remains a challenge in many applications. Traditional solutions have included heatsinks and chip mounted fans. But more recent research is looking at physical effects to keep chips cool.
Dr Alexandre Cuenat is a researcher in the National Physical Laboratory’s Thermoelectric and Electrothermal metrology group (NPL). “Typically,” he said, “the issue of thermal management is big in the electronics industry.”
The NPL has two strands of research into thermal materials. Dr Cuenat explained: “We’re interested in how materials will change temperature when they interact with electric fields. We’re investigating two kinds of processes: electrocaloric, which is a very efficient process, but which only moves a relatively small amount of heat; and thermoelectric, which is less efficient, but which has a higher heat flux.”
Dr Cuenat said the electrocaloric effect saw a big push in the 1960s because it was thought it could be used efficiently. “In the last 10 years, researchers have discovered new materials which improve the technique’s performance and, now, it’s all the rage in terms of research. But the applications are not there.”
The electrocaloric effect is being researched at Penn State University in the US. The university explains the effect. ‘Turn on an electric field and a standard electrocaloric material will eject heat to its surroundings. Do the same thing and a negative electrocaloric material will absorb heat’.
“The advantage of the electrocaloric material is its very high efficiency, compared with other solid state coolers, such as the thermoelectric cooler,” said Xiaoshi Qian, the primary investigator with the Penn State project. According to Qian this can be attached to a chip in need of on demand cooling.
However, for most electrocaloric materials, the electric field needs to remain active in order for the effect to work; something which could end up heating a material which is being cooled.
The team’s work is based on ferroelectric polymers and recently the Penn State scientists have developed a ferroelectric polymer blend that can retain heat when the electric field is switched off – and it believes this approach could be adapted to provide cooling on demand for small scale systems, including electronic components.
Qian’s team says the electrocaloric effect – the addition or removal of heat – is produced by dipoles within the material reordering themselves.
The material under examination at Penn State is described as a hybrid ferroelectric polyvinylidene fluoride-trifluoroethylene copolymer, with the addition of ferroelectric chlorofluoroethylene (CFE) as a relaxer. According to Qian, the addition of CFE introduces defects into the copolymer’s molecular chain and this results in the material exhibiting what is called ‘dipolar randomness’, instead of the ordering seen in the copolymer.
Fig 1: When an electric pulse is applied to the electrocaloric polymer, the molecules align in the same direction. However, a second pulse sees the molecules assume a random alignment and to retain that state, allowing the material to maintain its cooling effect. Image: Penn State
When an electric pulse is applied to the enhanced polymer, the molecules will align in the same direction. However, a second pulse sees the molecules assume a random alignment and to retain that state. This, says the research team, allows the ferroelectric material to maintain a large cooling effect once an applied voltage has been removed.
“We would like to improve the electrocaloric materials in the future so that the cooling generated upon an electric pulse in the electrocaloric material can be much larger,” Qian said. “This study is the first step toward that direction.”
The NPL’s Dr Cuenat is not as enthusiastic about electrocaloric cooling, however, noting the approach is not on the thermal technology road map produced by International Electronics Manufacturing Initiative (iNEMI), due to its lack of cooling power.
“People are claiming efficiencies which aren’t reproducible or which can’t be quantified,” he contended. “And another problem is that some approaches are based on lead, which isn’t ideal.” Despite the reservations, NPL is working with Imperial College to develop electrocaloric cooling for large scale refrigeration applications.
According to iNEMI, the industry needs to develop a range of techniques to deal with thermal issues. Amongst these are: low cost, higher thermal conductivity packaging materials; advanced cooling technology; and advanced 3D packaging techniques that can remove heat effectively.
And it is in this latter area that Dr Cuenat is working. However, rather than developing thermal management solutions themselves, Dr Cuenat is looking at the metrology aspects. “The issue for electrocaloric and thermoelectric cooling solutions is that measuring heat flow is difficult; it’s not like measuring electrical power.”
NPL is part of the EMPIR initiative; a three year collaborative project involving eight European bodies, investigating the issues involved with 3D stacked semiconductors. EMPIR’s work package 1 (WP1) aims to develop methods to measure accurately the electrical and thermal transport properties of nanostructured copper through silicon vias in order to establish traceable measurements of electrical conductivity and temperature change in copper as a function of the current density. Modelling of thermal transport in those structures, says WP1, will help to identify the various thermal scattering mechanisms in nanostructured copper grains.
“We’re looking to measure the change in temperature of a chip,” Dr Cuenat explained. “People usually measure this temperature change directly, but we believe that’s the wrong approach; we’re looking to measure as precisely as possible how cool the chip is and its real temperature change.
“We’re working on 3D ICs as part of this. Because of their power dissipation, heat management is more challenging and we’re not only trying to find out how to detect hotspots within the devices, we’re also working on materials that can be used to cool the chips.”
The challenge with this approach is the increasingly smaller features of modern devices. “Not only do people want to see results online and as close to production as possible, they don’t want chips destroyed in the process.”
Dr Cuenat gave an indication of the scale of the challenge. “We are talking about measuring the temperature of an area which could be as small as 3 x 3µm – and this area could be 25°C hotter than the surroundings. But, because everything around the target area looks hot, we are looking at a resolution of 1µm.”
As Dr Cuenat noted, this work needs to proceed on a non-destructive basis. “We’re measuring temperature through the packaging. The area being examined might be 25K hotter, but it’s not in contact with the cooling side, so we also need to find out how this change in temperature propagates through the chip.
“Our objective is to provide the industry with tools that will enable temperature changes across the whole chip to be measured to mK accuracy,” he concluded.