Breakthrough discovery could speed up computer processors simply by conducting more heat

Scientists have demonstrated a new material that conducts heat 150% more efficiently than conventional materials used in advanced chip technologies.

The device – an ultra-thin silicon nanowire – could enable smaller, faster microelectronics with heat transfer efficiency that surpasses current technologies. Electronic devices powered by microchips that dissipate heat efficiently would in turn consume less energy – an improvement that could help mitigate energy consumption produced by burning carbon-rich fossil fuels that have contributed to global warming.

“By overcoming the natural limitations of silicon in its ability to conduct heat, our discovery addresses a hurdle in microchip engineering,” said Junqiao Wu, the scientist who led the Physical examination letters study reporting the new device. Wu is a faculty scientist in the Division of Materials Science and a professor of materials science and engineering at UC Berkeley.

The slow flow of heat through silicon

Our electronics are relatively affordable because silicon – the material of choice for computer chips – is cheap and plentiful. But while silicon is a good conductor of electricity, it’s not a good conductor of heat when reduced to very small sizes – and when it comes to fast computing, that’s a big problem. for tiny chips.

Within each microchip are tens of billions of silicon transistors that direct the flow of electrons in and out of memory cells, encoding bits of data as ones and zeros, the binary language of computers. Electric currents flow between these hard-working transistors, and these currents inevitably generate heat.

Heat naturally flows from a hot object to a cold object. But heat flow becomes tricky in silicon.

In its natural form, silicon is made up of three different isotopes – forms of a chemical element containing an equal number of protons but a different number of neutrons (hence a different mass) in their nuclei.

About 92% of silicon is made up of the isotope silicon-28, which has 14 protons and 14 neutrons; about 5% is silicon-29, weighing 14 protons and 15 neutrons; and only 3% is silicon-30, a relatively heavy weight with 14 protons and 16 neutrons, explained co-author Joel Ager, who holds the titles of senior scientist in the Berkeley Laboratory’s Materials Science Division and professor assistant in materials science and engineering at UC Berkeley.

As phonons, the waves of atomic vibration that carry heat, meander through the crystal structure of silicon, their direction changes when they strike silicon-29 or silicon-30, whose different atomic masses “confuse” the phonons, slowing them down.

“The phonons eventually get the idea and find their way to the cold end to cool the silicon material,” but that indirect path allows waste heat to build up, which also slows down your computer, Ager said. .

A big step towards faster and denser microelectronics

For many decades, researchers have hypothesized that pure silicon 28 chips would exceed the thermal conductivity limit of silicon and therefore improve the processing speeds of smaller, denser microelectronics.

But purifying silicon into a single isotope requires intense energy levels that few facilities can deliver — let alone specialize in making market-ready isotopes, Ager said.

Fortunately, an international project in the early 2000s enabled Ager and semiconductor materials expert Eugene Haller to source silicon tetrafluoride gas – the starting material for isotopically purified silicon – from an ancient Soviet era isotope manufacturing plant. (Haller founded the Berkeley Lab’s DOE-funded Electronic Materials Program in 1984 and was a senior researcher in Berkeley Lab’s Materials Science Division and a professor of materials science and mineral engineering at UC Berkeley. He died in 2018.)

This led to a series of pioneering experiments, including a 2006 study published in Nature, in which Ager and Haller shaped silicon-28 into single crystals, which they used to demonstrate quantum memory storing information as quantum bits or qubits, units of data stored simultaneously as a one and a zero in the spin of an electron.

Subsequently, semiconductor thin films and single crystals made with Ager and Haller’s silicon isotope material were shown to have 10% higher thermal conductivity than natural silicon – an improvement, but from the point from the computer industry’s perspective, probably not enough to justify spending a thousand times more money to build a computer from isotopically pure silicon, Ager said.

But Ager knew that silicon isotopic materials had scientific importance beyond quantum computing. So he kept what was left in a safe place in the Berkeley lab, just in case other scientists needed it, because few people have the resources to make or even buy isotopically pure silicon, he said. reasoned.

A path to cooler technology with silicon-28

About three years ago, Wu and his graduate student Penghong Ci were trying to find new ways to improve the rate of heat transfer in silicon chips.

One strategy for making more efficient transistors is to use a type of nanowire called a Gate-All-Around field-effect transistor. In these devices, silicon nanowires are stacked together to conduct electricity and heat is generated simultaneously, Wu explained. resounding fire in a large building without an evacuation plan,” he said.

But heat transport is even worse in silicon nanowires, because their rough surfaces — scars from chemical processing — scatter or “confuse” phonons even further, he explained.

“And then one day we asked ourselves, ‘What would happen if we made a nanowire out of isotopically pure silicon-28? ‘” Wu said.

Silicon isotopes aren’t something you can easily buy on the open market, and Ager was said to still have silicon isotope crystals in storage at the Berkeley lab – not a lot, but still enough to share “if anyone has a good idea on how to use it,” Ager said. “And Junqiao’s new study was such a case.”

A surprising big reveal with nano tests

“We are really lucky that Joel had the isotope-enriched silicon material ready to use for the study,” Wu said.

Using Ager’s silicon isotope materials, the Wu team tested the thermal conductivity in 1 millimeter bulk silicon-28 crystals compared to natural silicon – and again their experiment confirmed this. that Ager and his collaborators discovered years ago – that bulk silicon-28 conducts heat only 10% better than natural silicon.

Now let’s move on to the nano test. Using a technique called electroless etching, Ci fabricated natural silicon and silicon-28 nanowires just 90 nanometers (billionths of a meter) in diameter, about a thousand times thinner than a single human hair.

To measure thermal conductivity, Ci suspended each nanowire between two microheating pads fitted with electrodes and platinum thermometers, then applied an electric current to the electrode to generate heat on a pad that flows outwards. other via the nanowire.

“We expected to see only an additional benefit – something like 20% – from using an isotopically pure material for thermal conduction of the nanowires,” Wu said.

But Ci’s measurements surprised them all. Si-28 nanowires conducted heat not 10% or even 20%, but 150% better than natural silicon nanowires with the same diameter and surface roughness.

It defied anything they expected to see, Wu said. The rough surface of a nanowire typically slows down phonons. So what was going on?

High-resolution transmission electron microscopy (TEM) images of the material captured by Matthew R. Jones and Muhua Sun of Rice University have revealed the first clue: a glass-like layer of silicon dioxide on the surface of the silicon nanowire -28.

Computer simulation experiments at the University of Massachusetts Amherst led by Zlatan Aksamija, a leading expert on the thermal conductivity of nanowires, revealed that the absence of isotopic “defects” – silicon-29 and silicon-30 – prevented the phonons from escaping to the surface, where the layer of silicon dioxide would slow the phonons down considerably. This in turn kept the phonons on track in the direction of the heat flow – and therefore less “confused” – inside the “core” of the silicon-28 nanowire. (Aksamija is currently an associate professor of materials science and engineering at the University of Utah.)

“It was really unexpected. Discovering that two distinct phonon blocking mechanisms – surface versus isotopes, previously thought to be independent of each other – now work synergistically to our advantage in thermal conduction is very surprising but also very rewarding,” Wu said.

“Junqiao and the team discovered a new physical phenomenon,” Ager said. “It’s a real triumph for curiosity-driven science. It’s quite exciting.”

Wu said the team then plans to take their discovery to the next step: investigating how to “control, rather than just measure, heat conduction in these materials.”

Researchers from Rice University; the University of Massachusetts-Amherst; Shenzhen University and Tsinghua University participated in the study.

This work was supported by the DOE Office of Science.

About Mark A. Tomlin

Check Also

UMass Amherst Receives Second Prize in NSF Taking Action: COVID-19 Diversity, Equity, and Inclusion Challenge: UMass Amherst

For its efforts to support faculty during the pandemic, UMass Amherst received second place in …