Physicists develop a cooling system for the processors of the future

January 19, 2016
Temperature distribution in an active plasmonic waveguide on an optoelectronic chip with a cooling system

Researchers from MIPT have found a solution to the problem of the overheating of active plasmonic components for high-speed data transfer in the optoelectronic microprocessors of the future. These will be able to function tens of thousands of times faster than the microprocessors currently in use today. In the paper, published in ACS Photonics, the researchers have demonstrated how to efficiently cool optoelectronic chips using industry-standard heatsinks despite the high heat generation in active plasmonic components.

The speed of multicore and manycore microprocessors, which are already used in high-performance computer systems, depends not so much on the speed of an individual core, but rather on the time it takes for data to be transferred between the cores. The electrical copper interconnects used in microprocessors today are fundamentally limited in bandwidth, and cannot maintain the advancement of processor performance. In other words, doubling the number of cores will not double the processing power.

Leading companies in the semiconductor industry, such as IBM, Oracle, Intel, and HP, envision switching from electronics to photonics, and they are currently investing billions of dollars into this. Replacing electrons with photons will mean that large amounts of data can be transferred between processor cores almost instantly, which will mean that the processor performance will be nearly proportional to the number of cores.

However, due to diffraction, photonic components are not as easy to scale down as electronic components. Their dimensions cannot be smaller than the size approximately equal to the light wavelength (~1 micrometer or 1000 nanometers), but transistors will soon be as small as 10 nanometers. This fundamental problem can be solved by switching from bulk waves to surface waves, which are known as (SPPs). This will confine light on the nanoscale. Along with the leading research centers of industrial companies and the laboratories of leading universities, Russian scientists from the Laboratory of Nanooptics and Plasmonics of MIPT's Center of Nanoscale Optoelectronics are also making good progress in this field.

The main difficulty that scientists face is the fact that SPPs are absorbed by metal, which is a key material in plasmonics. This effect is similar to resistance in electronics, where the energy of electrons is lost and converted into heat when current passes through a resistor. The SPP loss can be compensated by pumping additional energy into the SPPs. However, this pumping will produce additional heat, which in turn will cause an increase in temperature in the plasmonic components and in the processor as a whole. The higher absorption in the metal, the greater the loss, and the stronger pumping will be required. This raises the temperature, which again causes a loss increase and makes it more difficult to create optical gain, which is required to compensate for the loss, and this means that more powerful pumping is required. A cycle is formed in which the temperature can rise to such an extent that a processor chip simply burns out. This is no surprise, since the heating power per surface unit of the active plasmonic waveguide with loss compensation exceeds 10 kW/cm2, which is twice as high as the intensity of solar radiation at the surface of the sun.

Dmitry Fedyanin and Andrey Vyshnevyy, researchers at MIPT's Laboratory of Nanooptics and Plasmonics, have found a solution to this problem. They have demonstrated that high-performance optoelectronic chips can be cooled withthermal grease—high-performance thermal interfaces consisting of layers of thermally conductive materials placed between the chip and the cooling system.

Based on the results of numerical simulations, Fedyanin and Vyshnevyy concluded that if an optoelectronic chip with active plasmonic waveguides is placed in air, its temperature will increase by several hundred degrees Celsius, which will cause the device to malfunction. Multi-layered thermal interfaces of nano- and micrometer thickness, combined with simple cooling systems, can reduce the temperature of the chip from several hundred degrees to approximately ten degrees with respect to the ambient temperature. This opens the prospects for the implementation of high-performance optoelectronic microprocessors in a wide range of applications, including supercomputers and compact electronic devices.

Explore further: Researchers clear the way for fast plasmonic chips

More information: Andrey A. Vyshnevyy et al. Self-Heating and Cooling of Active Plasmonic Waveguides, ACS Photonics (2015). DOI: 10.1021/acsphotonics.5b00449

Related Stories

Researchers clear the way for fast plasmonic chips

August 4, 2015

Researchers from the Laboratory of Nanooptics and Plasmonics at the MIPT Center of Nanoscale Optoelectronics have developed a new method for optical communication on a chip, which will could decrease the size of optical and ...

Recommended for you

High-precision magnetic field sensing

December 2, 2016

Scientists have developed a highly sensitive sensor to detect tiny changes in strong magnetic fields. The sensor may find widespread use in medicine and other areas.

8 comments

Adjust slider to filter visible comments by rank

Display comments: newest first

Osiris1
not rated yet Jan 20, 2016
Who says the Russians can not do really gooood science?
dirk_bruere
not rated yet Jan 20, 2016
"They have demonstrated that high-performance optoelectronic chips can be cooled withthermal grease—high-performance thermal interfaces consisting of layers of thermally conductive materials placed between the chip and the cooling system."

Maybe it should be called "heatsink compound".
Is this a joke article?

andervish
not rated yet Jan 20, 2016
It supposed to be "They have demonstrated that using high-performance thermal interfaces, i.e. layers of thermally conductive materials placed between the chip and the cooling system to ensure efficient heat removal from the chip, (thermal grease is a popular type of thermal interface, although it is not very efficient) high-performance optoelectronic chips can be cooled using conventional cooling systems."
Eikka
not rated yet Jan 21, 2016
Is this a joke article?


It certainly seems to be lacking in any relevant information as to why it was ever in doubt that you can just slap a heatsink on the chip.

The solution seems trivial and obvious.
andervish
not rated yet Jan 22, 2016
With that much of heat generation even using the best heat-sink one could imagine (T=T_ambient at a chip boundary), we may still have a very hot spot at the waveguide location, due to thermal spreading resistance. Luckily, it is not the case and all we should care about is to find approriate heat-sink. We are being lucky twice, since we should not go to extremes, like invent and use CVD-diamond microchannel heat-sinks - a common air-fan/heat-sink assembly soldered to the chip would suffice.
Hyperfuzzy
not rated yet Jan 22, 2016
Future optical chips will not need cooling. They will select a state that does what we want, either stay or oscillate or return the same wave. i.e an inverter, transmitter, buffer. Then add multi-state optical logic. The non-linear action of this medium need not produce heat, i.e. no loss.

Atomic phonoms that produce heat need not be applied, i.e.unnecessary molecular motion. Such alternately polarizing and unpolarizing the entire atom or molecule. We will use selective energy.
andervish
not rated yet Jan 26, 2016
I suppose fully optical chips are still farther in the future than optoelectronic counterparts, which are based on attempts to "marry" traditional CMOS electronics with photonics (and plasmonics) and are considered here. One of reasons behind this are huge investments necessary to establish new manufacturing facilities, thus semiconductor industy is reluctant to part with CMOS process they had for decades.
In case of (silicon) photonic interconnects lasers and modulators are the sources of heat, while the waveguide itself is lossless and does not generate heat. Plasmonics allows considerable miniaturization, but waveguides generate heat as hell. Despite that, they can be kept cool.
Hyperfuzzy
not rated yet Jan 27, 2016
...
In case of (silicon) photonic interconnects lasers and modulators are the sources of heat, while the wave-guide itself is loss-less and does not generate heat. Plasmonics allows considerable miniaturization, but wave-guides generate heat as well. Despite that, they can be kept cool.

Suggest optical chips are totally different from current or voltage based chips. Waves are produced by particle motion but waves may select a particular particle state and motion. i.e. the phonom of molecules and atoms are totally different from electrons in orbit. Known physics! Expect only a few years, 2025. Different frequency for molecular and atomic polarization and motion rather than an optical response of an electron, any electron. From QM to actuality.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.