Prototype delivers current density 50 times greater than copper interconnects; paves way for smaller ICs

By Brian Santo, contributing writer

Image source:
University of California Riverside

The
semiconductor industry is rapidly approaching the physical limits not only of
silicon, but also of copper, commonly used for IC interconnects. Engineers at
the University of California at Riverside have demonstrated prototype devices using zirconium tritelluride
(ZrTe3)
that can conduct a current density 50 times greater than conventional
copper interconnects.

Ideally, the
use of some exotic material that can sustain high current densities – perhaps
ZrTe3 – will enable the semiconductor industry to continue with
silicon for a few additional device generations.

IC
manufacturers relied mostly on aluminum for interconnects until about 20 years
ago, when the industry moved to copper. Since copper is a superior conductor, less
material was needed for the interconnects. As a practical matter, swapping the
materials helped the industry continue to scale its circuitry. The switch also necessitated
the development of new production equipment and processes. It was difficult and
expensive, but since it enabled the industry to keep scaling silicon for
another two decades, it was unquestionably worth the investment.

At today’s
processing nodes copper is running up against limits in current density — the
amount of electrical current per cross-sectional area at a given point. As the
industry attempts to continue scaling ICs down in size, transistors need higher
and higher current densities to perform at the desired level. Copper (and most
other conventional electrical conductors) tends to break due to overheating or
other factors at high current densities, presenting a barrier to creating
increasingly smaller components, UC Riverside researchers explained.

So, the
industry has been exploring other materials as potential replacements for
copper, just as copper once replaced aluminum.

Some
companies are already using alternatives to copper at the smallest processing
nodes. It’s been reported that both Intel and Global Foundries
are using cobalt for some interconnects at the 10-nm node. Cobalt is better
than copper, but not by much.

The industry
is also looking at more exotic alternatives that can provide significantly
higher current densities. Notable among them is graphene, which was recently discovered to be superconducting. Graphene is considered a
two-dimensional (2D) material, in that it can be created in sheets that are a single
atom-layer thick. Researchers have been able to create what they’re calling
nanoribbons of graphene down to a few atoms wide.

But if 2D is
good, wouldn’t 1D be better? The question sparked an investigation into the
potential use of materials that can be fabricated as single-atom strands. They’re
not technically 1D, but they are as close as you can get. When researchers are
being scrupulous about terminology, they call single-strand materials
“quasi-one-dimensional.”  

The research
at UC Riverside is led by Alexander A. Balandin, a professor of electrical and
computer engineering. He and his team have discovered that ZrTe3 has
an exceptionally high current density that far exceeds that of any conventional
metals like copper.

Copper has a
current density of 2 MA/cm2 to 3 MA/cm2. Zirconium tritelluride
has a current density of about 100 MA/cm2, according
to a UC Riverside research paper. 

Thus far, UC
Riverside researchers have been cutting nanoribbons of ZrTe3 from
sheets of the material, but the team is confident that it will be possible to
process the material as 1D strands. ZrTe3 nanoribbons could be made
into either nanometer-scale local interconnects or device channels for
components of the tiniest devices, UC Riverside said.

“Conventional
metals are polycrystalline. They have grain boundaries and surface roughness,
which scatter electrons,” Balandin said. “Quasi-one-dimensional materials such
as ZrTe3 consist of single-crystal atomic chains in one direction.
They do not have grain boundaries and often have atomically smooth surfaces
after exfoliation. We attributed the exceptionally high current density in ZrTe3
to the single-crystal nature of quasi-1D materials.”

While the
bulk resistivity of the 1D is higher than that of copper, Balandin told Electronic Products, it is expected that
it will not degrade as fast as the resistivity of copper does with a decreasing
cross-section area.

Thus far, ZrTe3
quantum wires exhibit current densities higher than reported for any metals or
other 1D materials – and almost reach the current density in carbon nanotubes
and graphene.

Balandin said
his team has tried several other materials and that tantalum triselenide
(TaSe3) also has promise as a semiconductor interconnect.

The
team has produced a prototype that demonstrates the ability to create ZrTe3
nanoribbons on a silicon/silicon-dioxide substrate, but Balandin told Electronic Products that he and his team
have not yet made a functional device. 

If the
current density of ZrTe3 is sufficient for the next few nodes of
semiconductor manufacturing, being 1D might be more of an advantage than being
superconducting. The practicality of using ZrTe3 would also be
predicated on how well it can be added to the silicon manufacturing process.
The UC Riverside researchers report they are developing a process to grow ZrTe3
nanoribbons directly on silicon wafers.

When asked
about the prospect of using ZrTe3 with standard CMOS production equipment,
Balandin said, “the project is funded by the Semiconductor Research Corporation
(SRC) with industrial liaisons from Intel Corporation. At this point, industry
is open for any material which can deliver high-current-density in small
cross-sections, and with acceptable resistance. Our materials have shown record
high-current-density.”