ShowDaily Day3 - Flipbook - Page 23
sponsored by
AI
continued from page 22
Lam’s DirectDrive® technology,
a key innovation in its Akara® etch
tool, replaces traditional RF generators with solid-state plasma sources.
“The timing control is a hundred
times better… we are able to deconstruct [etch] into complex waveforms
and achieve unprecedented control
for scaling in the angstrom era.”
Dry resist technology, delivered
through Lam’s Striker® system, uses
CVD instead of spin-on methods.
“Mechanically, the film that you
deposit is superior, and CVD has a
lot more process knobs… for compositional control, grading, and thermal
quaternary systems,” Varadarajan
said. This enables smaller features
and better variability control.
Packaging innovation is equally
critical. Lam’s SABRE3D® and
VECTOR TEOS 3D® address
challenges like current crowding and
gap-fill in heterogeneous packages.
“You would think it’s easier, but
the reality is it’s so big that stress
management, cracking, edge delamination, and chipping present many
challenges… critical in terms of
controlling the yield of the entire
package itself.”
managing director of
Semiverse Solutions.
Unlike traditional
Pareto-based methods,
Fabtex enables engineers
to address multiple yield
detractors simultaneously,
accelerating variability
propagation and uncertainty quantification across
complex process steps.
A graph shown during the keynote
compared two yield trajectories:
• The yellow curve (“Wafer statistics
only”) shows gradual improvement.
• The cyan dashed curve (“Wafer
statistics + Fabtex Yield Optimizer”) shows a steeper rise, reaching
higher yield faster.
Upward arrows along the Fabtex
curve highlight cost and yield
improvements and efficiency gains.
In one case, a leading device manufacturer used Fabtex to correct
systematic defects during a new
product introduction (NPI) tape-out,
saving weeks and reducing costs.
Collaboration and talent:
Building the future workforce
“This much disruptive innovation
requires collaboration,” Varadarajan
Software innovation: Fabtex
said. “We practice what we preach.
Yield Optimizer and the
Our collaboration starts with our
digital twin revolution
customers, but beyond that, we
Lam unveiled Fabtex Yield Optimizer, have supplier partners, consortium
a Semiverse Solutions software tool
partners, and universities as well.”
designed to improve yield in HVM.
“There’s a huge talent gap,” he
It uses virtual silicon digital twins,
acknowledged. “Each of us actually
inline fab data, and AI/ML to identify has a responsibility to think about
process variability and recommend
what talent shortage means in our
new metrology targets.
sphere of influence and how we
“If you can adjust your process
can go about solving it.” Traditional
window virtually without spending
classroom and fab-based training
months running sequential expermodels are no longer scalable. “If we
iments on physical wafers, you
don’t bring expertise to the talent,
can save time to yield and improve
we risk missing out on opportunities
time to market,” said Joseph Ervin,
of bringing and recruiting our future
www.semiconductordigest.com
innovators to our industry.”
Lam’s solution is to use AI to solve
the very talent shortage that AI is
creating. “We have to become efficient consumers of AI and advanced
compute to solve the very same
talent shortage that it is creating,”
Varadarajan said.
“By moving our interaction with
the tools to a digital environment,
we solve two fundamental issues,”
he explained. “Number one, this is
available anytime, anywhere. The
second one is these digital platforms
are very agile.” SEMulator3D®
allows students to simulate process
flows and build structures like
gate-all-around transistors virtually.
“Now a researcher or an engineer
or a student rather anywhere in the
world can effectively answer the
question: So how do you actually
build a gate-all-around structure?”
Lam has digitally reconstructed
its major tools, enabling students to
manipulate equipment and run procedures in a safe, virtual space. This
immersive training compresses the
time needed to prepare a ready-now
workforce and eliminates safety risks.
“We are in a virtuous cycle
where we are plowing all our innovation towards accelerating AI,”
Varadarajan said. “And in turn, the
systems that we enable… we end
up using that back to power our
Semiverse engine.”
Thursday, October 9 | 23
